NUCLEAR MAGNETIC RESONANCE STUDIES OF
BIOFILM – POROUS MEDIA SYSTEMS
by
Catherine Mullinnix Kirkland
A dissertation submitted in partial fulfillment
of the requirements for the degree
of
Doctor of Philosophy
in
Engineering
MONTANA STATE UNIVERSITY
Bozeman, Montana
July 2017
©COPYRIGHT
by
Catherine Mullinnix Kirkland
2017
All Rights Reserved
ii
DEDICATION
To Tim, Fergus, and Liam
iii
ACKNOWLEDGMENTS
This material is based upon work supported, in part by the US Department of
Energy, Office of Science under Award DE-SC0006376 and DE-FE0024296. Any
opinions, findings and conclusions or recommendations expressed in this material are
those of the author(s) and do not necessarily reflect the views of the Department of
Energy.
This work is supported by the National Science Foundation (NSF) Graduate
Research Fellowship Program under Grant No. DGE-1049562. The work comprising
Chapter 9 was also supported by the Netherlands Organisation for Scientific Research
(NWO) in conjunction with NSF as part of the Graduate Research Opportunities
Worldwide program. Experiments conducted on the 22T (950 MHz) NMR instrument
were supported by uNMR-NL, an NWO-funded National Roadmap Large-Scale Facility
of the Netherlands (project 184.032.207).
iv
TABLE OF CONTENTS
1. INTRODUCTION ...........................................................................................................1
Low-field NMR ...............................................................................................................2
High-field NMR and MRI ...............................................................................................4
Outline..............................................................................................................................5
References .......................................................................................................................8
2. INTRODUCTION TO NUCLEAR MAGNETIC RESONANCE ................................11
Quantum Mechanics NMR Theory ................................................................................11
Spin Angular Momentum ......................................................................................12
Spin Magnetism and the Zeeman Interaction ........................................................14
Classical Mechanics NMR Theory ................................................................................18
Excitation ...............................................................................................................18
Reference Frames...................................................................................................20
Relaxation ..............................................................................................................22
Auto-correlation Functions ........................................................................22
Spectral Density Functions ........................................................................24
Experimental Background .............................................................................................28
Experimental Equipment .......................................................................................28
Signal Detection .....................................................................................................29
Basic Pulse Sequences ...........................................................................................33
Inversion Recovery ....................................................................................34
Hahn Spin Echo .........................................................................................36
CPMG Echo Train .....................................................................................37
Stimulated Echo .........................................................................................38
Phase Cycling.............................................................................................39
Introduction to Magnetic Resonance Imaging ...............................................................40
Gradients and k-space ............................................................................................40
Signal Encoding .....................................................................................................46
Selective Excitation ...............................................................................................49
Introduction to Molecular Motion Measurements .........................................................52
Normalized Echo Amplitude and q-space .............................................................53
Propagators ............................................................................................................54
References .....................................................................................................................58
3. ADVANCED NMR CONCEPTS .................................................................................59
Introduction ...................................................................................................................59
Time Varying Magnetic Fields and Phase Factors ........................................................60
Coherent Flow and Moments of the Gradient................................................................61
v
TABLE OF CONTENTS – CONTINUED
Bloch-Torrey Equations for Diffusion and Flow ..........................................................63
The Stejskal – Tanner Experiment .................................................................................64
Generalized Translational Motion .................................................................................67
Brownstein – Tarr Theory ..............................................................................................69
Multi-dimensional PFG and Relaxation NMR ..............................................................73
References ......................................................................................................................79
4. LOW-FIELD BOREHOLE NMR APPLICATIONS IN
THE NEAR-SURFACE ENVIRONMENT ..................................................................80
Contribution of Authors and Co-Authors ......................................................................80
Manuscript Information .................................................................................................81
Abstract .........................................................................................................................82
Introduction ...................................................................................................................83
Theory ....................................................................................................................87
Measurement of Soil Moisture ......................................................................................94
Characterization of Unconsolidated Aquifers ...............................................................97
Detection of Subsurface Biogeochemical Processes ...................................................100
Outlook and Conclusions ............................................................................................106
References ...................................................................................................................108
5. BIOFILM DETECTION IN A MODEL WELL-BORE
ENVIRONMENT USING LOW-FIELD NMR ..........................................................113
Contribution of Authors and Co-Authors ....................................................................113
Manuscript Information ...............................................................................................115
Abstract .......................................................................................................................116
Introduction .................................................................................................................117
Theory .........................................................................................................................120
T2 Relaxation ........................................................................................................120
Materials and Methods ................................................................................................123
Bioreactor Design and Construction ....................................................................123
Bacterial Culture ..................................................................................................125
NMR Data Acquisition ........................................................................................127
Sampling and Imaging .........................................................................................129
Results and Discussion ................................................................................................130
References ...................................................................................................................142
vi
TABLE OF CONTENTS – CONTINUED
6. IN-SITU DETECTION OF SUBSURFACE BIOFILM
USING LOW-FIELD NMR – A FIELD STUDY ......................................................146
Contribution of Authors and Co-Authors ....................................................................146
Manuscript Information ...............................................................................................148
Abstract .......................................................................................................................149
Introduction .................................................................................................................150
Methods .......................................................................................................................152
Results and Discussion ................................................................................................159
Microbiological Data and Water Chemical Analysis ..........................................166
Supporting Information ...............................................................................................172
Site Preparation ....................................................................................................172
Bacterial Culture ..................................................................................................173
NMR Measurements ............................................................................................175
Microbiological and Water Chemical Analysis ...................................................175
Results and Discussion ........................................................................................176
References ...................................................................................................................178
7. DETECTING MICROBIALLY-INDUCED CALCITE
PRECIPITATION (MICP) IN A MODEL WELL-BORE
USING DOWNHOLE LOW-FIELD NMR ................................................................182
Contribution of Authors and Co-Authors ....................................................................182
Manuscript Information ...............................................................................................183
Abstract .......................................................................................................................184
Introduction .................................................................................................................185
Background ..........................................................................................................186
NMR Theory ........................................................................................................188
Materials and Methods ................................................................................................191
Bioreactor .............................................................................................................191
Media and Injection Strategy ...............................................................................192
Bacterial Culture ..................................................................................................193
NMR Measurements ............................................................................................194
Sampling ..............................................................................................................195
Results and Discussion ................................................................................................196
Water Content and Porosity .................................................................................198
Relaxation ............................................................................................................202
References ...................................................................................................................205
vii
TABLE OF CONTENTS – CONTINUED
8. NMR INVESTIGATION OF WATER DIFFUSION
IN DIFFERENT BIOFILM STRUCTURES ..............................................................208
Contribution of Authors and Co-Authors ....................................................................208
Manuscript Information ...............................................................................................210
Abstract .......................................................................................................................211
Introduction .................................................................................................................212
Materials and Methods ................................................................................................215
Biofilm Sample Preparation .................................................................................215
MRI and PFG-NMR ............................................................................................217
MRI ..........................................................................................................217
Diffusion Measurements ..........................................................................218
Data Processing ....................................................................................................220
(Bi)-exponential Model ............................................................................220
Gamma Distribution Model .....................................................................221
2D Inverse Laplace Transform ................................................................221
Results and Discussion ................................................................................................222
Biofilm Characterization: Comparison of Common Quantities and Images .......222
Diffusion of Water in the Presence of Biomass ...................................................225
Influence of Biofilm Structure on Water Dynamics ............................................228
Influence of Diffusion Time on Water Dynamics ...............................................232
Correlation of Diffusion and Transverse Relaxation ...........................................235
Conclusions .................................................................................................................237
References ...................................................................................................................238
9. STRUCTURE AND DIFFUSION OF AEROBIC GRANULAR
SLUDGE USING MAGNETIC RESONANCE ........................................................243
Abstract ........................................................................................................................243
Introduction .................................................................................................................243
Materials and Methods ................................................................................................247
Sample Collection and Preparation ......................................................................247
NMR and MRI Measurements .............................................................................249
Data Analysis .......................................................................................................251
T2 Maps ....................................................................................................251
Diffusion Images ......................................................................................251
Multidimensional Correlation and Exchange Experiments .....................252
Results and Discussion ................................................................................................252
MRI of Granule Internal Structure .......................................................................252
T2 Maps ....................................................................................................255
viii
TABLE OF CONTENTS – CONTINUED
Boundary Layer .......................................................................................256
PFG and Multidimensional NMR ........................................................................259
Future Work ................................................................................................................266
Acknowledgements .....................................................................................................267
References ...................................................................................................................268
10. CONCLUSIONS .......................................................................................................271
REFERENCES CITED ....................................................................................................275
ix
LIST OF TABLES
Table Page
5.1 NMR experiments with well-logging probe ................................................128
6.1 Experiment overview for biofilm detection field study ...............................155
6.2 NMR experimental parameters for field study ............................................159
8.1 Characterization of biofilms used in diffusion studies ................................216
8.2 Acquisition parameters used for diffusion measurements ...........................219
8.3 Stereoscopic and MRI images of biofilms ...................................................224
8.4 Summary of diffusion coefficients measured at Δ of 200 ms ......................228
9.1 MRI measurement parameters ....................................................................250
x
LIST OF FIGURES
Figure Page
2.1 Schematic of the Zeeman interaction energy .................................................15
2.2 Excitation schematic ......................................................................................20
2.3 Rotating reference frame schematic...............................................................21
2.4 Relaxation times in relation to the Larmor frequency ...................................26
2.5 Free Induction Decay .....................................................................................32
2.6 Inversion recovery pulse sequence ................................................................34
2.7 Inversion recovery magnetization evolution ..................................................34
2.8 Evolution of Mz in inversion recovery ...........................................................35
2.9 Hahn echo pulse sequence and magnetization evolution ...............................36
2.10 CPMG pulse sequence .................................................................................38
2.11 Stimulated echo pulse sequence...................................................................39
2.12 Schematic of a magnetic field gradient ........................................................41
2.13 Phase evolution under the influence of a field gradient ...............................44
2.14 Gradient echo and spin echo phase evolution ..............................................45
2.15 Phase encoding to traverse k-space .............................................................47
2.16 Frequency encoding to traverse k-space ......................................................48
2.17 Fourier transform of rectangular pulses .......................................................50
2.18 Fourier transform of soft pulses ...................................................................51
2.19 Schematic of selective excitation .................................................................52
2.20 PGSE pulse sequence ...................................................................................56
xi
LIST OF FIGURES – CONTINUED
Figure Page
3.1 PGSE pulse sequence .....................................................................................65
3.2 PGStE pulse sequence....................................................................................66
3.3 The Stejskal-Tanner plot ................................................................................67
3.4 Multi-dimensional PFG and relaxation NMR pulse sequences .....................77
4.1 Schematic of borehole NMR logging tool .....................................................88
4.2 CPMG pulse sequence ...................................................................................89
4.3 Temporal evolution of saturation in a column ...............................................95
4.4 Soil saturation profile .....................................................................................96
4.5 NMR well logs from the Massachusetts Military Reservation ......................98
4.6 Comparison of NMR logging data and DP permeameter data ......................99
4.7 Temporal evolution of CPMG signal decay and T2 distributions ................102
4.8 Temporal evolution of T2ML data due to biofilm growth ............................103
4.9 Temporal evolution of CPMG signal decay and T2 distributions ................105
5.1 Model well-bore bioreactor..........................................................................124
5.2 Temporal evolution of CPMG signal decay and T2 distributions ................132
5.3 Square of echoes reduction ..........................................................................134
5.4 Heterotrophic plate counts ...........................................................................136
5.5 Microscopy images of biofilm and porous media........................................137
6.1 Diversion disk attachment on the low-field NMR logging tool ..................158
6.2 Evolution of T2ML during field study ...........................................................160
xii
LIST OF FIGURES – CONTINUED
Figure Page
6.3 T2 distributions for the higher-frequency and lower-frequency wells .........161
6.4 Well profiles measured by NMR logging tool .............................................165
6.5 Heterotrophic plate counts and pH data .......................................................167
6.6 Square of Echoes data from field study .......................................................177
7.1 Model well-bore bioreactor..........................................................................192
7.2 Biomineralized sand annulus prior during destructive sampling .................196
7.3 Evolution of CPMG signal decay curves and T2 distributions ....................197
7.4 Temporal evolution of NMR-measured water content ................................199
7.5 SEM micrographs of calcite encrusted and control sand .............................200
8.1 Diffusion measurements at 200 MHz ..........................................................227
8.2 D1 and Dmean for various biofilm structures .................................................230
8.3 Time-dependent diffusion coefficients ........................................................233
8.4 Distribution of D using Γ distribution .........................................................235
8.5 D-T2 correlation for biofilm carriers ............................................................236
9.1 Schematic of aerobic granular sludge ..........................................................244
9.2 Sample collection and preparation ...............................................................248
9.3 Image of granular sludge samples in NMR tubes ........................................249
9.4 T1- and T2-weighted image of fresh Utrecht granule ...................................253
9.5 T1-weighted images of VFA and Garmerwolde granules ............................254
9.6 T2 maps of fresh and aged Garmerwolde granules .......................................256
xiii
LIST OF FIGURES – CONTINUED
Figure Page
9.7 TEM of Garmerwolde granule surface ........................................................258
9.8 TEM of Garmerwolde granule interior ........................................................259
9.9 1D diffusion image of alginate beads and aged Garmerwolde granule .......260
9.10 Apparent diffusion coefficient map of anammox granules........................261
9.11 D-T2 correlation of aged Garmerwolde granules .......................................262
9.12 1D T2 distribution and T2 map of Garmerwolde granule ...........................263
9.13 T1-T2 correlation of aged Garmerwolde granules ......................................264
9.14 T2-T2 correlation of aged Garmerwolde granules ......................................265
xiv
ABSTRACT
Nuclear magnetic resonance (NMR) allows for in-situ non-invasive studies of
opaque systems over a wide range of length and time scales, making the method uniquely
suited to studies of biofilms and porous media. The research comprising this thesis uses
NMR to explore biophysical, chemical, and transport properties within heterogeneous
porous media systems at both a macro- and micro-scale. The macro-scale projects
validate a low-field borehole NMR instrument to monitor field-scale environmental
engineering applications like subsurface biofilms and microbially-induced calcite
precipitation (MICP). Subsurface biofilms are central to bioremediation of chemical
contaminants in soil and groundwater whereby micro-organisms degrade or sequester
environmental pollutants like nitrate, hydrocarbons, chlorinated solvents and heavy
metals. When composed of ureolytic microbes, subsurface biofilms can also induce
calcite precipitation. MICP has engineering applications that include soil stabilization
and subsurface barriers, as well as sealing of cap rocks and well-bore regions for carbon
dioxide sequestration. To meet the design goals of these beneficial applications,
subsurface biofilms and MICP must be monitored over space and time – a challenging
task with traditional methods. The low-field borehole NMR tool recorded changes in the
T2 relaxation distribution where enhanced relaxation indicated biofilm accumulation in a
sand bioreactor and in subsurface soil. Additionally, the tool was able to detect MICP in
a sand bioreactor. The changed mineral surface of the sand lead to an increase in T2
relaxation times. The complementary high-field NMR project investigated micro-scale
internal structures and mass transport within biofilm granules used for wastewater
treatment. Granular sludge, composed of spherical aggregates of biofilm grown without
a carrier, is an innovative biological treatment method with the potential to vastly reduce
the cost of wastewater treatment without sacrificing efficiency. Large gaps remain,
however, in our understanding of the fundamental formation mechanisms and the factors
that control granule activity and stability. Magnetic resonance imaging (MRI) identified
heterogeneous internal structures within aerobic granular sludge where relaxation rates
and diffusion coefficients vary. Ultimately, these results will help improve modeling for
optimization of granular sludge wastewater treatment process design.
1
INTRODUCTION
Nuclear Magnetic Resonance (NMR) is sensitive over varying length and time
scales to the physical and chemical environments, as well as translational motion, that
hydrogen-bearing species experience. NMR is non-invasive, non-destructive, and can be
applied to opaque samples in diverse applications including subsurface biogeochemical
investigations and high-resolution imaging of micro-scale structures within biofilms.
Especially in these applications, magnetic resonance can measure physical features or
characteristic behaviors that are challenging to directly observe with other methods.
The experiments described in this thesis use NMR or Magnetic Resonance
Imaging (MRI) to explore biofilm – porous media systems. The experiments can be
broadly classified into two categories, of which the first represents the vast majority of
this thesis: 1) low-field NMR detection of macro-scale biofilm accumulation and
biomineralization in porous media for in-situ bioremediation applications, and 2) high-
field NMR and MRI measurements of granular biofilms used in wastewater treatment to
identify micro-scale internal structural features and their influence on reactive transport
within the biofilm. These two categories rely on the same method, NMR, but use
different hardware configurations and field strengths. The two categories also focus on
biofilm systems, though the biofilms of interest in the two cases are cultivated for
different applications.
Biofilms are mixed microbial aggregates, typically adhered to a surface with a
sticky hydrogel called extracellular polymeric substance, or EPS [1, 2]. EPS consists of
polysaccharides, proteins, DNA, and other entangled macro-molecules [3, 4]. While
2
some biofilms pose significant challenges to human health [5] or infrastructure [6],
beneficial biofilms can be cultivated to effect biogeochemical conversions or
hydrodynamic changes in the subsurface as part of a bioremediation project [7, 8].
Biofilms can also be used to induce mineral precipitation in the subsurface with
applications including well-bore leakage mitigation [9] and geotechnical engineering [10,
11]. In the granular sludge system studied in the final chapter of this thesis, the biofilm
granules are microbial aggregates which self-assemble under specific conditions without
a support material. In this case, the granule is both the biofilm and the porous media.
The remainder of this introduction puts the two types of NMR measurements in to
historical context and outlines the organization of the thesis.
Low-field NMR
In natural materials such as soils and rock formations, the use of high-field NMR,
typically defined as greater than 200 MHz, poses experimental challenges. Natural
porous media, in contrast to model media like glass beads, contains minerals with a wide
range of magnetic susceptibilities. When placed in a background magnetic field, these
contrasts in magnetic susceptibility generate internal gradients in the local, pore-scale
magnetic field that broaden the spectral peaks and cause rapid signal attenuation.
Reducing the strength of the applied magnetic field reduces the influence of these local
gradients, allowing signal to be captured from natural materials, though with reduced
signal to noise and spectral resolution than high-field methods [12]. Low-field devices
typically operate in the range between several hundred kHz to 10MHz. Low-field NMR
3
devices include well-logging tools used in the oil and gas industry to measure the
distribution of pore sizes in the formation and quantify fuel-bearing reservoirs [13, 14],
surface probes that measure soil moisture or water intrusion into concrete surfaces [15-
18], and small diameter probes, like the Vista Clara Javelin, to measure aquifer
characteristics in the near subsurface [19].
The advent of low-cost and portable low-field NMR instrumentation in the last
decade has led to an expansion of potential applications for the technology. Thus far,
low-field NMR devices have not been applied in practice to monitor in-situ permeability
reduction due to biofilm accumulation and biomineralization. The research within this
thesis demonstrates that the Javelin low-field NMR logging tool can provide spatio-
temporally resolved data regarding changing physico-chemical conditions in the near
subsurface to improve monitoring capabilities and inform critical decision-making.
Specifically, the device detects changes in signal relaxation response indicative of biofilm
accumulation or microbially-induced calcite precipitation (MICP). Current methods to
determine the extent of these processes in situ are limited and indirect. Using a NMR
probe as a monitoring method may improve the efficacy of subsurface engineering
projects involving biofilms and or MICP by allowing for simpler and more cost-effective
assessment of their integrity.
4
High-field NMR and MRI
High-field NMR has a rich history of use in laboratory biofilm systems. For
more than two decades, high-field NMR has been used to study mass transport and
hydrodynamics in a variety of model bioreactors using relaxometry and displacement
measurements [20-26]. Potter et al. used diffusion-weighted NMR to detect bacterial
cells and measure cell density within porous media [27]. MRI allows researchers to
spatially resolve biofilm accumulation in porous media or capillaries through mapping
regions with differing relaxation rates, where faster relaxation indicates the presence of
biofilm [28, 29]. MRI can identify internal structural features of spherical biofilm
aggregates, called granular sludge or biofilm granules [30, 31].
The high-field NMR experiments move the research from macro-scale field
applications of commercially available low-field NMR technology to micro-scale
exploration of fundamental processes related to formation of, and reactive mass transport
within, granular sludge. By their very nature as spherical biofilm aggregates, granules
are composed of a variety of microniches—aerobic, anoxic, anaerobic —where diverse
bio-chemical conversions can occur simultaneously within the same granule [32, 33]. As
a result, granular sludge treatment processes have the potential to vastly reduce the costs
of cleaning wastewater by minimizing the capital costs associated with construction of
separate process reactors and clarifiers, as well as reducing ongoing operations costs
associated with aeration [34, 35]. The ultimate goal of the high-field NMR experiments
on biofilm granules is to fill the gaps in our understanding of fundamental formation
5
mechanisms and conversion processes so that full exploitation of the technology for
wastewater treatment can be realized.
The high-field NMR research explores the structure and function of aerobic
biofilm granules using pulsed field gradient (PFG) NMR and MRI by measuring
spatially-resolved relaxation rates and diffusion coefficients of single granules in a static
test tube. Ongoing experiments with flow-through cells will explore the role of
wastewater particulates and particulate substrate in the formation, stability, and activity
of the granules by measuring the propagator, or probability of displacement, of tracer
particles. The results of these experiments will be used to refine the models used to
simulate granule formation and substrate removal. These models are instrumental in
process design optimization and hydraulic modeling of full-scale granular sludge
wastewater treatment plants.
Outline
First, this thesis introduces NMR theory in Chapter 2, including discussions of the
quantum and classical mechanical underpinnings of the technology, as well as excitation,
relaxation, signal detection, basic pulse sequences, and an introduction to imaging and
measurement of motion. Chapter 3 addresses more advanced NMR concepts related to
encoding for diffusive motion and coherent flow, Brownstein and Tarr’s [13]
characterization of signal relaxation behavior in various pore geometries and multi-
dimensional methods to measure molecular motion.
6
Chapters 4 – 7 describe the low-field NMR experiments conducted during this
thesis, beginning with a review of low-field borehole NMR applications in the near
subsurface. This invited manuscript is under review for publication in the Vadose Zone
Journal. Chapter 5 presents results of a laboratory experiment using the low-field NMR
logging tool to detect biofilm accumulation in a model well-bore reactor. T2 relaxation
distributions were measured over time while biofilm was cultivated in the reactor,
resulting in a shortening of the mean log T2 relaxation time. Biofilm growth was
confirmed with microscopy and microbiological methods. The manuscript was published
in Groundwater Monitoring and Remediation. Following the successful laboratory
demonstration, the biofilm detection experiments were conducted in a field study at an
engineered field-testing facility in Butte, Montana. Again, the low-field NMR logging
tool detected accumulation and removal of biofilm in the soil surrounding the test wells
by recording changes in the measured T2 relaxation distributions. The manuscript
comprising Chapter 6 is published in Environmental Science and Technology. Chapter 7
describes a second laboratory experiment using the low-field NMR logging tool, this time
applied to the detection of calcite precipitation in a sand-filled model well-bore reactor.
T2 relaxation distributions recorded during the biomineralization process showed an
increase in the mean log T2 time and a bifurcation of the initial single relaxation mode
into a small population with very fast relaxation and a larger population with slower
relaxation. This manuscript is also published in Environmental Science and Technology.
Chapters 8 and 9 shift the focus from low-field borehole NMR experiments in
porous media to high field NMR and MRI experiments on wastewater biofilms. Chapter
7
8 describes research the author collaborated on with Dr. Maria Pia Herrling, a summer
visitor to the Montana State University Magnetic Resonance Lab from Karlsruhe Institute
of Technology in Karlsruhe, Germany. As part of her research into water diffusion in
various biofilm structures, including floccular sludge, granular sludge, and biofilm grown
on plastic carriers, Dr. Herrling measured diffusion-relaxation correlations at MSU with
the author. The manuscript has been submitted for publication in Biotechnology and
Bioengineering. Finally, Chapter 9 describes ongoing experimental work on the structure
and diffusion properties of aerobic granular sludge biofilms sampled from full-scale
wastewater treatment plants in the Netherlands. High-field MRI was used to image the
complex and heterogeneous internal structure of the granules. These data were collected
in the Netherlands at Wageningen University and Research and at the national NMR
facility at the University of Utrecht under the supervision of Dr. Henk Van As as part of
an international collaboration with Dr. Merle de Kreuk at Delft University of
Technology. Additional high field PFG NMR experiments related to diffusion and
transport within the granules are currently being conducted at MSU using samples from
Dutch treatment plants. As this chapter includes work still in progress, the data and
results presented are preliminary and require further validation prior to publication in
peer reviewed journals. This research will be presented at the International Water
Association’s 10th International Conference on Biofilm Reactors in Dublin, Ireland, in
May 2017.
8
References
1. Characklis, W.G. and K.C. Marshall, Eds., Biofilms. 1990, New York: John Wiley
& Sons, Inc.
2. Stoodley, P., et al., Biofilms as complex differentiated communities. Annual
Review of Microbiology, 2002. 56: p. 187-209.
3. Allison, D.G., The biofilm matrix. Biofouling, 2003. 19(2): p. 139-150.
4. Sutherland, I.W., Biofilm exopolysaccharides: a strong and sticky framework.
Microbiology, 2001. 147: p. 3-9.
5. Costerton, J.W., P.S. Stewart, and E.P. Greenberg, Bacterial biofilms: A common
cause of persistent infections. Science, 1999. 284(5418): p. 1318-1322.
6. Camper, A.K., Organic matter, pipe materials, disinfectants and biofilms in
distribution systems. 2014, IWA Publishing. p. 73-94.
7. Cunningham, A.B., et al., Subsurface biofilm barriers for the containment and
remediation of contaminated groundwater. Bioremediation Journal, 2003. 7(3-4):
p. 151-164.
8. Taylor, S.W. and P.R. Jaffe, Enhanced in-situ biodegradation and aquifer
permeability reduction. Journal of Environmental Engineering-ASCE, 1991.
117(1): p. 25-46.
9. Cunningham, A.B., et al., Abandoned well CO2 leakage mitigation using
biologically induced mineralization: current progress and future directions.
Greenhouse Gases-Science and Technology, 2013. 3(1): p. 40-49.
10. Dejong, J.T., et al., Biogeochemical processes and geotechnical applications:
progress, opportunities and challenges. Geotechnique, 2013. 63(4): p. 287-301.
11. Phillips, A.J., et al., Engineered applications of ureolytic biomineralization: a
review. Biofouling, 2013. 29(6): p. 715-733.
12. Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media
using a low-field magnetic resonance system. Environmental Science &
Technology, 2013. 47(2): p. 987-992.
13. Brownstein, K.R. and C.E. Tarr, Importance of classical diffusion in NMR studies
of water in biological cells. Physical Review A, 1979. 19(6): p. 2446-2453.
9
14. Timur, A., Pulsed nuclear magnetic resonance studies of porosity, movable fluid,
and permeability of sandstones. Journal of Petroleum Technology, 1969.
21(JUN): p. 775-&.
15. Blümich, B., F. Casanova, and S. Appelt, NMR at low magnetic fields. Chemical
Physics Letters, 2009. 477(4–6): p. 231-240.
16. Fukushima, E., Nuclear magnetic resonance as a tool to study flow. Annual
Review of Fluid Mechanics, 1999. 31(1): p. 95.
17. Marble, A.E., et al., A constant gradient unilateral magnet for near-surface MRI
profiling. Journal of Magnetic Resonance, 2006. 183(2): p. 228-234.
18. Sucre, O., et al., Low-field NMR logging sensor for measuring hydraulic
parameters of model soils. Journal of Hydrology, 2011. 406(1–2): p. 30-38.
19. Walsh, D., et al., A small-diameter NMR logging tool for groundwater
investigations. Groundwater, 2013. 51(6): p. 914-926.
20. Gjersing, E.L., et al., Magnetic resonance microscopy analysis of advective
transport in a biofilm reactor. Biotechnology and Bioengineering, 2005. 89(7): p.
822-834.
21. Hornemann, J.A., et al., Biopolymer and water dynamics in microbial biofilm
extracellular polymeric substance. Biomacromolecules, 2008. 9(9): p. 2322-2328.
22. Seymour, J.D., et al., Anomalous fluid transport in porous media induced by
biofilm growth. Physical Review Letters, 2004. 93(19).
23. Seymour, J.D., et al., Magnetic resonance microscopy of biofouling induced scale
dependent transport in porous media. Advances in Water Resources, 2007. 30(6-
7): p. 1408-1420.
24. Wagner, M., et al., Online assessment of biofilm development, sloughing and
forced detachment in tube reactor by means of magnetic resonance microscopy.
Biotechnology and Bioengineering, 2010. 107(1): p. 172-181.
25. Lewandowski, Z., et al., NMR imaging of hydrodynamics near microbially
colonized surfaces. Water Science and Technology, 1992. 26(3-4): p. 577-584.
26. Van As, H. and P. Lens, Use of H-1 NMR to study transport processes in porous
biosystems. Journal of Industrial Microbiology & Biotechnology, 2001. 26(1-2):
p. 43-52.
27. Potter, K., et al., Assay for bacteria in porous media by diffusion-weighted NMR.
Journal of Magnetic Resonance. Series B, 1996. 113: p. 9-15.
10
28. Hoskins, B.C., et al., Selective imaging of biofilms in porous media by NMR
relaxation. Journal of Magnetic Resonance, 1999. 139(1): p. 67-73.
29. Seymour, J.D., et al., Magnetic resonance microscopy of biofilm structure and
impact on transport in a capillary bioreactor. Journal of Magnetic Resonance,
2004. 167(322-327).
30. Gonzalez-Gil, G., et al., Cluster structure of anaerobic aggregates of an expanded
granular sludge bed reactor. Applied and Environmental Microbiology, 2001.
67(8): p. 3683-3692.
31. Lens, P.N.L., et al., Diffusional properties of methanogenic granular sludge: H-1
NMR characterization. Applied and Environmental Microbiology, 2003. 69(11):
p. 6644-6649.
32. Da-Wen, G. and T. Yu, Versatility and application of anaerobic ammonium-
oxidizing bacteria. Applied Microbiology & Biotechnology, 2011. 91(4): p. 887-
894.
33. de Kreuk, M., J.J. Heijnen, and M.C.M. van Loosdrecht, Simultaneous COD,
nitrogen, and phosphate removal by aerobic granular sludge. Biotechnology and
Bioengineering, 2005. 90(6): p. 761-769.
34. de Kreuk, M.K., N. Kishida, and M.C.M. van Loosdrecht, Aerobic granular
sludge - state of the art. Water Science and Technology, 2007. 55(8-9): p. 75-81.
35. Lotti, T., et al., Pilot-scale evaluation of anammox-based mainstream nitrogen
removal from municipal wastewater. Environmental technology, 2015. 36(9): p.
1167-77.
11
INTRODUCTION TO NUCLEAR MAGNETIC RESONANCE
Nuclear magnetic resonance (NMR) relies on the quantum physical property of
angular momentum intrinsic to a single nucleus and the response of that nucleus to a
magnetic field to study the behavior of macroscopic systems consisting of large
ensembles of nuclei over varying length and time scales. The discussion in this thesis
will begin with the quantum and classical mechanical basis for NMR measurements
before moving into a description of basic NMR techniques for imaging and measurement
of molecular motion. This chapter relies heavily on reference works by Paul
Callaghan[1, 2]. Other primary original sources are referenced in the text where
applicable.
Quantum Mechanics NMR Theory
For any single nucleus, there is a discrete set of energy states defined by quantum
mechanics. Since we are observing the behavior of a sample composed of nuclei on the
order of Avogadro’s number (1023), however, the range of possibilities for the ensemble
appears continuous. Because of its fundamental connection to the subject of NMR, it is
necessary to first discuss the quantum mechanical basis of nuclear angular momentum,
often called spin angular momentum.
12
Spin Angular Momentum
The spin state of a nuclear isotope is described by the angular momentum, or spin,
quantum number I, which is a fixed integer or half-integer value and characterizes the
nucleus in its stable ground state. The spin quantum number I exists in a basis set of
discrete values of the angular momentum, m, measured along the z-axis, where m = –I, (-I
+ 1), (I -1), I. For example, the 1H proton and 13C both have I = ½ . This means that 1H
and 13C have two discrete energy states possible during a measurement, m = - ½ and m =
+ ½ , referred to as ‘spin down’ and ‘spin up’, respectively.
The effect of ‘spin’ is that each nucleus has a magnetic dipole moment, μ, where
μ = γs. The constant of proportionality, γ, is called the gyromagnetic ratio and is defined
as the ratio of the magnetic dipole moment (μ) to the nuclear angular momentum (s). The
gyromagnetic ratio for 1H is γ = 2.675 x 108 rad/(T s) and is among the highest of all
nuclear isotopes. This intrinsic property of the hydrogen proton, as well as its abundance
in nature, explains the predominance of 1H NMR within the larger field of magnetic
resonance research and applications.
NMR measurements are conducted in order to explore the system energy state. In
other words, we measure an observable quantity to describe the energy state of the
system we are interested in and which informs us about the molecular environment.
When we make a measurement, we affect the quantum state of the system. In fact, we
must affect the quantum state of the system in order to make a measurement. The
quantum mechanical nature of a spin system is given by the eigenvalue equation
𝐴𝐴|𝑎𝑎⟩ = 𝑎𝑎|𝑎𝑎⟩ (2.1)
13
where A is the observable of interest, the eigenvector |𝑎𝑎⟩ is a basis state of A, and the
eigenvalue 𝑎𝑎 is the observed value of measurement A. The eigenvalue is a complex
number, containing both amplitude and phase information.
To interpret Equation (2.1), recall that in making quantum mechanical
measurements, we are faced with the Heisenberg uncertainty principle. That is, we
cannot simultaneously know both the position and momentum of an atomic nucleus; we
can only know the probability of a nucleus having a given position and momentum. If a
measurement of A is made while the system is in state|𝑎𝑎⟩, then the observed value, or
amplitude, of the measurement must be 𝑎𝑎 with certainty. If the system, however, is in a
general admixed state resulting from superposition, denoted by |𝛹𝛹⟩, then the eigenvalue
is returned with the probability, |𝑎𝑎𝑚𝑚|2.
To apply these concepts to the spin quantum number, I, consider the case where
the observable of interest is the component of angular momentum about a particular axis,
𝐼𝐼 = 〈𝐼𝐼𝑥𝑥〉𝐢𝐢 + 〈𝐼𝐼𝑦𝑦〉𝐣𝐣 + 〈𝐼𝐼𝑧𝑧〉𝐤𝐤 where i, j, and k are unit vectors along the x, y, and z-axes,
respectively. In the case of the operator, 𝐼𝐼𝑧𝑧, the equation
⟨𝛹𝛹|𝐼𝐼𝑧𝑧|𝛹𝛹⟩ = � |𝑎𝑎𝑚𝑚|2 𝑚𝑚
𝑚𝑚
(2.2)
gives the probability that the measurement 𝐼𝐼𝑧𝑧 resulted in a change of energy state from
some initial basis state to a final basis state. Other important operators include the raising
and lowering operators, 𝐼𝐼+ = 𝐼𝐼𝑥𝑥 + 𝑖𝑖𝐼𝐼𝑦𝑦 and 𝐼𝐼− = 𝐼𝐼𝑥𝑥 − 𝑖𝑖𝐼𝐼𝑦𝑦, respectively. These operators
convert |−1
2
� a basis state to a | + 1
2
� state, and vice versa.
14
Since position and momentum of Heisenberg’s uncertainty principle are also
components of energy, as potential and kinetic energy, respectively, it is constructive to
express changes in the quantum energy state of a system in terms of the Hamiltonian, or
total energy operator, H. The dynamics of the spin system are described by the
Schrödinger equation
𝑖𝑖ℏ
𝜕𝜕
𝜕𝜕𝜕𝜕
|𝛹𝛹(𝜕𝜕)⟩ = 𝐻𝐻|𝛹𝛹(𝜕𝜕)⟩ (2.3)
where ℏ is Planck’s constant. The Schrödinger equation shows that the change in the
energy state with respect to time is related to the total energy of the system. For a
Hamiltonian constant in time, the evolution of the energy state with time is
|𝛹𝛹(𝜕𝜕)⟩ = 𝑒𝑒𝑒𝑒𝑒𝑒 �−𝑖𝑖𝐻𝐻 𝜕𝜕
ℏ
� |𝛹𝛹(0)⟩ (2.4)
Spin Magnetism and the Zeeman Interaction
Outside of an applied magnetic field, nuclear magnetic dipoles are randomly
oriented and no net magnetization occurs. However, in the presence of a static magnetic
field, B0, we observe the Zeeman effect where the magnetic dipole moments reside in one
of two quantum energy states. The nuclear magnetic dipole moments align themselves
either parallel or anti-parallel to the B0 field, with a slight preference for parallel
alignment. The parallel ‘spin up’ alignment corresponds to the lower energy state| + 1
2
�,
while anti-parallel ‘spin down’ alignment is the higher energy state|−1
2
�. For a proton in
a static magnetic field, B0, along the z-axis, the Zeeman Hamiltonian is
15
𝐻𝐻𝑍𝑍𝑍𝑍𝑍𝑍𝑚𝑚𝑍𝑍𝑍𝑍 = −𝛾𝛾ℏ𝐵𝐵0𝐼𝐼𝑧𝑧 (2.5)
This means that the difference in energy between the two possible spin states is equal to
𝛥𝛥𝛥𝛥 = 𝛾𝛾ℏ𝐵𝐵𝑜𝑜 and that the spins precess about the z-axis (Figure 2.1).
Figure 2.1. The two quantum spin states are separated by an energy difference of
𝜸𝜸ℏ𝑩𝑩𝒐𝒐which is also the resonant frequency of spins within the system. The slight
preference for the lower energy state yields a net magnetization on the order of parts per
million.
Writing Equation (2.5) in terms of angular frequency and dropping Planck’s
constant, it is apparent that the intrinsic frequency of the system is equal to 𝛾𝛾𝐵𝐵0, called
the Larmor frequency, 𝜔𝜔0 . The Larmor frequency is both the precession frequency of all
the spin states around the z-axis as seen in the Zeeman Hamiltonian, but also the
frequency of spins ‘flipping’ between the two possible energy states, m = - ½ and m = +
½.
For a large number of nuclei, quantum properties can be expressed using classical
statistics over the ensemble. The equation
⟨𝛹𝛹|𝐼𝐼𝑧𝑧|𝛹𝛹⟩ = 12 ��𝑎𝑎12�2 − �𝑎𝑎−12�2� (2.6)
demonstrates that the ensemble average energy state resulting from the observation
𝐼𝐼𝑧𝑧, ⟨𝛹𝛹|𝐼𝐼𝑧𝑧|𝛹𝛹⟩ , is the difference between the populations in the spin up and spin down
m = + ½
m = - ½
16
states, expressed by �𝑎𝑎1
2
�
2 and �𝑎𝑎−1
2
�
2
, respectively. The Boltzmann energy distribution is
derived from statistical mechanics and governs the relative sizes of the two populations.
The probability of a particular energy state, Pn, within a distribution of energy states is
defined as
𝑃𝑃𝑍𝑍 = exp �−∆𝛥𝛥 𝑘𝑘𝐵𝐵𝑇𝑇� �
∑ exp �−∆𝛥𝛥 𝑘𝑘𝐵𝐵𝑇𝑇� �𝑍𝑍 (2.7)
where ΔE is the energy associated with each state, kB is the Boltzmann constant, and T is
the absolute temperature. Because of the slight preference of spins at thermal equilibrium
for the lower energy state| + 1
2
�, a polarization, 𝐼𝐼𝑧𝑧, aligned with the static magnetic field
develops within a sample.
Since the energy difference between the two states, 𝛾𝛾ℏ𝐵𝐵𝑜𝑜, is orders of magnitude
less than 𝑘𝑘𝐵𝐵𝑇𝑇, the net polarization within a sample is typically in the part per million
range. Moreover, there is no phase coherence between the spins, so it is not possible to
directly measure 𝐼𝐼𝑧𝑧 .
Measurement of the x-component of angular momentum, however, yields
⟨𝛹𝛹|𝐼𝐼𝑥𝑥|𝛹𝛹⟩ = 12 �𝑎𝑎12∗𝑎𝑎−12 + 𝑎𝑎−12∗ 𝑎𝑎12� (2.8)
where 𝑎𝑎±1
2
∗ is the complex conjugate of a. This equation shows that the ensemble average
energy state resulting from the observation 𝐼𝐼𝑥𝑥 is the sum of the off-diagonal matrix
components and reflects the phase coherence between the + ½ and – ½ spin states, where
the relative phase, rather than the absolute phase, is significant.
17
To handle both the polarization and phase coherence, it is useful to introduce the
spin density matrix 𝜌𝜌, a hermitian quantum mechanical operator that evolves according to
the Schrödinger equation and directly provides the ensemble averaged expectation value
for measurement A
〈𝛹𝛹|𝐴𝐴|𝛹𝛹〉 = 𝑇𝑇𝑇𝑇|𝐴𝐴𝜌𝜌| = 〈𝐴𝐴〉 = �𝑃𝑃𝛹𝛹
𝛹𝛹
〈𝛹𝛹|𝐴𝐴|𝛹𝛹〉 (2.9)
where 𝑇𝑇𝑇𝑇|𝜌𝜌| is the trace of 𝜌𝜌—the sum of all state probabilities—and is therefore equal to
1. When I is the observable of interest for spin ½ nuclei, the spin density matrix is
𝜌𝜌 =
⎣
⎢
⎢
⎢
⎡ �𝑎𝑎1
2
�
2
𝑎𝑎−1
2
∗ 𝑎𝑎1
2
𝑎𝑎1
2
∗𝑎𝑎
−
1
2
�𝑎𝑎
−
1
2
�
2
⎦
⎥
⎥
⎥
⎤ = � 12 + 〈𝐼𝐼𝑧𝑧〉 〈𝐼𝐼𝑥𝑥 − 𝑖𝑖𝐼𝐼𝑦𝑦〉
〈𝐼𝐼𝑥𝑥 + 𝑖𝑖𝐼𝐼𝑦𝑦〉 12 − 〈𝐼𝐼𝑧𝑧〉 � (2.10)
The major implication of the spin density matrix is that specifying the vector
components of the spin quantum number 𝐈𝐈 = 〈𝐼𝐼𝑥𝑥〉𝐢𝐢 + 〈𝐼𝐼𝑦𝑦〉𝐣𝐣 + 〈𝐼𝐼𝑧𝑧〉𝐤𝐤 allows full
specification of the four matrix elements and therefore defines all the energy states of an
ensemble of spin ½ nuclei. Recall, these energy states are the results of a direct NMR
measurement of Ix, Iy, or Iz. On a macroscopic level, the net magnetization of a sample in
the laboratory reference frame can be defined as
𝐌𝐌 = 𝑁𝑁𝛾𝛾ℏ �〈𝐼𝐼𝑥𝑥〉𝐢𝐢 + 〈𝐼𝐼𝑦𝑦〉𝐣𝐣 + 〈𝐼𝐼𝑧𝑧〉𝐤𝐤� (2.11)
where N is the number of spins per unit volume. Here, then, we see the connection
between the macroscopic magnetization in the laboratory reference frame and the
quantum energy state of a single nucleus in our sample. Since we define the direction of
18
the B0 field as the z-axis, we can directly detect the quantum states 𝐼𝐼𝑥𝑥 and 𝐼𝐼𝑦𝑦 via an NMR
measurement, where 𝐼𝐼𝑥𝑥 is the real component and 𝐼𝐼𝑦𝑦 is the imaginary component.
Classical Mechanics NMR Theory
Moving from the scale of a single quantum mechanical nucleus to the macro-scale
net magnetization allows use of the language of classical mechanics to describe NMR
theory and practice.
Excitation
When a pulse of oscillating current in the radiofrequency (rf) range is introduced
along the y-axis in the plane orthogonal to B0, a magnetic field B1 is formed along the x-
axis. The B1 field generates a torque, tipping the net magnetization, M, into the
transverse plane, where it precesses around the z-axis. Analogous to quantum mechanical
theory where the magnetic dipole moment produces precession in a static magnetic field,
on a macroscopic level the Zeeman effect also produces precession of M around B0 as the
magnetic field exerts a torque on the magnetization vector according to
𝑑𝑑𝑴𝑴
𝑑𝑑𝜕𝜕
= 𝛾𝛾𝑴𝑴 × 𝑩𝑩 (2.12)
When the system is ‘on resonance’, both the oscillation of B1 and the precession
of M occur at the Larmor frequency, 𝜔𝜔0 = 𝛾𝛾𝐵𝐵0, and the rf field is defined as
𝑩𝑩1(𝜕𝜕) = 𝐵𝐵1 cos(𝜔𝜔0𝜕𝜕) 𝐢𝐢 − 𝐵𝐵1 sin(𝜔𝜔0𝜕𝜕) 𝐣𝐣 (2.13)
Performing the cross-product produces the following expressions for the time evolution
of the magnetization vector in the laboratory reference frame following excitation:
19
𝑑𝑑𝑀𝑀𝑥𝑥
𝑑𝑑𝜕𝜕
= 𝛾𝛾�𝑀𝑀𝑦𝑦𝐵𝐵0 + 𝑀𝑀𝑧𝑧𝐵𝐵1𝑠𝑠𝑖𝑖𝑠𝑠𝜔𝜔0𝜕𝜕�
𝑑𝑑𝑀𝑀𝑦𝑦
𝑑𝑑𝜕𝜕
= 𝛾𝛾[𝑀𝑀𝑧𝑧𝐵𝐵1𝑐𝑐𝑐𝑐𝑠𝑠𝜔𝜔0𝜕𝜕 − 𝑀𝑀𝑥𝑥𝐵𝐵0] (2.14)
𝑑𝑑𝑀𝑀𝑧𝑧
𝑑𝑑𝜕𝜕
= 𝛾𝛾�−𝑀𝑀𝑥𝑥𝐵𝐵1𝑠𝑠𝑖𝑖𝑠𝑠𝜔𝜔0𝜕𝜕 − 𝑀𝑀𝑦𝑦𝐵𝐵1𝑐𝑐𝑐𝑐𝑠𝑠𝜔𝜔0𝜕𝜕�
Solving Equations (2.14) for the initial condition M(t) = M0k gives
𝑀𝑀𝑥𝑥 = 𝑀𝑀0𝑠𝑠𝑖𝑖𝑠𝑠(𝜔𝜔1𝜕𝜕)𝑠𝑠𝑖𝑖𝑠𝑠(𝜔𝜔0𝜕𝜕)
𝑀𝑀𝑦𝑦 = 𝑀𝑀0𝑠𝑠𝑖𝑖𝑠𝑠(𝜔𝜔1𝜕𝜕)𝑐𝑐𝑐𝑐𝑠𝑠(𝜔𝜔0𝜕𝜕) (2.15)
𝑀𝑀𝑧𝑧 = 𝑀𝑀0𝑐𝑐𝑐𝑐𝑠𝑠(𝜔𝜔1𝜕𝜕)
where 𝜔𝜔1 = 𝛾𝛾𝐵𝐵1. This solution shows that in the laboratory reference frame, the
precession of M is a spiral, as the magnetization vector precesses around the B0 field at
ω0, and about the B1 field at ω1 (Figure 2.2).
Figure 2.2 Following a 90° rf excitation pulse, the net magnetization vector, M,
precesses around B0 at ω0 and around B1 at ω1 in the laboratory reference frame.
x
y
z
B1
ω0t
B0
ω1t
M
20
The tip angle, α, is equal to ω1t. By controlling the duration of the rf pulse, t, and
the amplitude of the B1 field, the net magnetization vector will lie in the transverse plane
(α =
𝜋𝜋
2
) where the induced signal is maximized. This rf pulse is called a 90°, or 𝜋𝜋
2
pulse.
Similarly, an rf pulse of duration 2t will result in M aligned with the –z-axis, called a
180°, or π pulse. The precession of M as it returns to equilibrium induces a current that
is measured as a decaying signal by the rf coil.
Reference Frames
To simplify further descriptions of the spin dynamics that occur following
excitation with a resonant rf field, it is typical to use a rotating frame of reference with
angular frequency ω. Recall the Zeeman Hamiltonian for a static magnetic field, B0,
𝐻𝐻𝑍𝑍𝑍𝑍𝑍𝑍𝑚𝑚𝑍𝑍𝑍𝑍 = −𝛾𝛾𝐵𝐵0𝐼𝐼𝑧𝑧 (2.16)
In the rotating reference frame, an additional term is necessary to account for the
frequency of the rotating frame itself, with the result
𝐻𝐻𝑍𝑍𝑍𝑍𝑍𝑍𝑚𝑚𝑍𝑍𝑍𝑍−𝑟𝑟𝑜𝑜𝑟𝑟 = −𝛾𝛾�𝐵𝐵0 − 𝜔𝜔 𝛾𝛾� �𝐼𝐼𝑧𝑧 (2.17)
For a reference frame rotating in the same sense as the precession (clockwise), the
apparent longitudinal magnetic field will be reduced by ω. When an oscillating rf field is
applied in the transverse plane to excite the spins from equilibrium, the Hamiltonian in
the rotating frame becomes
𝐻𝐻𝑟𝑟𝑜𝑜𝑟𝑟 = −𝛾𝛾�𝐵𝐵0 − 𝜔𝜔 𝛾𝛾� �𝐼𝐼𝑧𝑧 − 𝛾𝛾𝐵𝐵1𝐼𝐼𝑥𝑥 (2.18)
When the system is ‘on resonance,’ ω = ω0 and the apparent longitudinal field
disappears, leaving the effective magnetic field along the rotating frame x-axis with the
21
Hamiltonian 𝐻𝐻𝑟𝑟𝑜𝑜𝑟𝑟 = −𝛾𝛾𝐵𝐵1𝐼𝐼𝑥𝑥 (Figure 2.3). Here, the Ix operator is the linear
combination of the raising and lowering operators, 1
2
(𝐼𝐼+ + 𝐼𝐼−), meaning that the energy
change in the system with time corresponds to the flipping of spins between the ‘spin up’
and ‘spin down’ states at a rate of γB1.
Figure 2.3. In the rotating reference frame, the static B0 field is not apparent and M lies
along the -y-axis following excitation with a resonant 90° rf pulse.
x
y
z
B1
ω1t
M
22
Relaxation
The transverse magnetization decays over time following the rf pulse as the spins
return to thermal equilibrium along the z-axis. This process is termed relaxation. The
rate at which the net magnetization decays in the transverse plane and re-forms along the
z-axis is governed by two relaxation mechanisms, T1 and T2. T1 relaxation, also called
spin-lattice relaxation or longitudinal relaxation, is caused by the exchange of energy
between spins and the environment, or lattice, and is related to the timescale for the net
magnetization to return to thermal equilibrium along the z-axis. T1 relaxation typically
occurs on the order of seconds for protons at room temperature.
T2 relaxation, also called spin-spin relaxation or transverse relaxation, is related to
the timescale for the net magnetization to decay in the transverse plane due to dephasing
of spin coherence caused by molecular interactions. Energy exchange between magnetic
dipoles creates micro-scale magnetic fields that dephase the spins within their sphere of
influence. Since some measure of phase coherence is essential for generation and
measurement of the induced signal, a loss of phase coherence leads to signal attenuation
with time. T2 relaxation occurs on the order of seconds for liquids at room temperature
and on the order of milliseconds for biopolymers in porous media.
Auto-correlation Functions. The time-dependence of relaxation processes is due
to molecular motion. A constructive way to describe time-dependent behavior is with the
use of auto-correlation functions. When stochastic variable A is a molecular quantity that
varies as a function of time, then the auto-correlation function of A is
23
𝐺𝐺(𝜕𝜕) = � 𝐴𝐴(𝜕𝜕′)𝐴𝐴(𝜕𝜕′ + 𝜕𝜕)𝑑𝑑𝜕𝜕′∞
0
(2.19)
The auto-correlation function G(t) describes the probability that A(t’) is correlated
to A(t’ + t) at some later time. When the system is stationary as is the case in most NMR
measurements, the origin of time does not matter; only the interval of time over which
measurements are collected is important. Further, in an ensemble of spins where any one
spin is statistically equivalent to any other spin over an appropriate period of time, the
average over time that is implied by Equation (2.19) can also be interpreted as an average
over all particles. Incorporating these concepts and replacing terms into Equation (2.19)
gives
𝐺𝐺(𝜕𝜕) = < 𝐴𝐴(0)𝐴𝐴(𝜕𝜕) > (2.20)
where is the well-defined ensemble average value of A during the interval 0
to t. At time zero, G(t) is equal to the mean squared value of A, . With
increasing time, the relationship between the initial and final values of A decays, as the
molecule ‘loses memory’ of its previous states. The characteristic timescale over which
this process occurs is called the correlation time, τc. When the experimental time, t, is
greater than the correlation time, τc, then there is no correlation with the initial state. The
correlation time can be understood as the time for a molecule (nucleus) to rotate about its
own axis and is defined by the integral
𝜏𝜏𝑐𝑐 = ∫ < 𝐴𝐴(0)𝐴𝐴(𝜕𝜕) > 𝑑𝑑𝜕𝜕∞0 < 𝐴𝐴(0)2 > (2.21)
24
Spectral Density Functions. Relaxation is caused by local spin interactions,
which vary by both magnitude and the rate of fluctuation of the dipolar Hamiltonian, HD.
In a relaxation experiment where A is the fluctuation of dipolar interactions between
spins resulting from molecular tumbling, spin relaxation times are sensitive to the
spectrum of the auto-correlation function, its Fourier transform. This spectrum is called
the spectral density function, J(ω), and describes how the magnetic field fluctuations
resulting from molecular interactions depend on time and the frequency of precession, ω.
For spin-½ nuclei within the Zeeman interaction created by the static magnetic field, B0,
there are three possibilities for behavior between interacting spins. Both interacting spins
may remain in their original energy states, described by J(0)(ω),
𝐽𝐽(0)(𝜔𝜔) = 2415𝑇𝑇𝑖𝑖𝑖𝑖6 𝜏𝜏𝑐𝑐1 + 𝜔𝜔2𝜏𝜏𝑐𝑐2 (2.22)
where rij is the radial distance between spin i and spin j. When one of the two spins
changes its energy state, J(1)(ω) results where
𝐽𝐽(1)(𝜔𝜔) = 415𝑇𝑇𝑖𝑖𝑖𝑖6 𝜏𝜏𝑐𝑐1 + 𝜔𝜔2𝜏𝜏𝑐𝑐2 (2.23)
Finally, if both spins flip to the other energy state, the resulting spectral density is
defined as J(2)(ω),
𝐽𝐽(2)(𝜔𝜔) = 1615𝑇𝑇𝑖𝑖𝑖𝑖6 𝜏𝜏𝑐𝑐1 + 𝜔𝜔2𝜏𝜏𝑐𝑐2 (2.24)
Using these expressions for the spectral density function, the differences between
the mechanisms for T1 relaxation as opposed to T2 relaxation is apparent. The T1
relaxation rate at the Larmor frequency, ω0, is given by the equation
25
1
𝑇𝑇1
= (𝜇𝜇04𝜋𝜋)2 𝛾𝛾4ℏ2 32 𝐼𝐼(𝐼𝐼 + 1)�𝐽𝐽(1)(𝜔𝜔0) + 𝐽𝐽(2)(2𝜔𝜔0)� (2.25)
Qualitatively, this equation shows that T1 relaxation is affected by spins exchanging
energy with the environment, or lattice, as they seek to return to thermal equilibrium
within the Zeeman interaction. There is a contribution from spins precessing at
frequency ω0 as well as a contribution at a frequency of 2ω0. There is no contribution
from spins that remain in the original energy state.
Conversely, the T2 relaxation rate at the Larmor frequency, ω0, is given by the
equation 1
𝑇𝑇2
= (𝜇𝜇04𝜋𝜋)2 𝛾𝛾4ℏ2 32 𝐼𝐼(𝐼𝐼 + 1) �14 𝐽𝐽(0)(0) + 52 𝐽𝐽(1)(𝜔𝜔0) + 14 𝐽𝐽(2)(2𝜔𝜔0)� (2.26)
From this expression, it is evident that T2 relaxation also depends on the zero-frequency
spectral density term, 𝐽𝐽(0)(0), wherein spins remain in their original energy states. This
term implies that there is another energy exchange mechanism aside from the Zeeman
interaction that contributes to T2 relaxation. For this reason, T2 relaxation will occur at
the same rate or faster than T1 relaxation, but never slower. As the precession frequency,
ω0, or the correlation time of the magnetic field fluctuations, τc, increases, the
contribution of the zero-frequency term also increases. Faster precession frequencies and
longer correlation times, therefore, cause a divergence between the T1 and T2 relaxation
times (Figure 2.4).
26
Figure 2.4. T1 and T2 relaxation times are nearly equal in liquids for a given Larmor
frequency. In solids, T1 and T2 diverge with T2 becoming much shorter than T1 because
of long correlation times, τc.
Moving now from quantum energy states and the scale of the molecule to the
macro-scale, relaxation can be described as a process affecting the net magnetization
vector of the sample evolving over time. The process of T1 relaxation, acting only along
the longitudinal axis and governed by the Zeeman interaction, can be described
phenomenologically by the equation
𝑑𝑑𝑀𝑀𝑧𝑧
𝑑𝑑𝜕𝜕
= −𝑀𝑀𝑧𝑧 −𝑀𝑀0
𝑇𝑇1
(2.27)
where 𝑀𝑀0 is the initial equilibrium magnetization. The solution is
𝑀𝑀𝑧𝑧(𝜕𝜕) = 𝑀𝑀𝑧𝑧(0) exp �− 𝜕𝜕𝑇𝑇1� + 𝑀𝑀0 �1 − exp �− 𝜕𝜕𝑇𝑇1�� (2.28)
The phenomenological equation describing T2 relaxation, which occurs only in the
transverse plane, is
𝑑𝑑𝑀𝑀𝑥𝑥,𝑦𝑦
𝑑𝑑𝜕𝜕
= −𝑀𝑀𝑥𝑥,𝑦𝑦
𝑇𝑇2
(2.29)
lo
g
(T
1,
T 2
)
T2
T1
log (1/ω0τc)
solids
1/τc <<ω0
liquids
1/τc >>ω0
ω0τc = 1
27
with the solution for a homogeneous sample
𝑀𝑀𝑥𝑥,𝑦𝑦(𝜕𝜕) = 𝑀𝑀𝑥𝑥,𝑦𝑦(0) exp �− 𝜕𝜕𝑇𝑇2� (2.30)
These equations show that the magnetization decays with time as an exponential
function and applies where the interaction terms related to transverse relaxation are weak,
as in the case of liquid-state molecules. Solids and macromolecules which are
rotationally constrained undergo very slow motions and rapid signal decay that is better
described by more complicated equations.
Combining the expressions for the change in M due to both excitation and
relaxation in the rotating reference frame, we obtain the Bloch equations
𝑑𝑑𝑀𝑀𝑥𝑥
𝑑𝑑𝜕𝜕
= 𝛾𝛾𝑀𝑀𝑦𝑦�𝐵𝐵0 − 𝜔𝜔 𝛾𝛾� � −𝑀𝑀𝑥𝑥𝑇𝑇2
𝑑𝑑𝑀𝑀𝑦𝑦
𝑑𝑑𝜕𝜕
= 𝛾𝛾𝑀𝑀𝑧𝑧𝐵𝐵1 − 𝛾𝛾𝑀𝑀𝑥𝑥�𝐵𝐵0 − 𝜔𝜔 𝛾𝛾� � − 𝑀𝑀𝑦𝑦𝑇𝑇2 (2.31)
𝑑𝑑𝑀𝑀𝑧𝑧
𝑑𝑑𝜕𝜕
= −𝛾𝛾𝑀𝑀𝑦𝑦𝐵𝐵1 − 𝑀𝑀𝑧𝑧 −𝑀𝑀0𝑇𝑇1
These equations describe the precession of the magnetization in the frequency
terms, as well as the decay of the magnetization over time in the relaxation terms.
Similarly, the induced signal contains complex phase and exponential decay terms.
28
Experimental Background
Before moving into a discussion of the practice of NMR—how to manipulate spin
dynamics to extract meaningful information—it is necessary to first describe briefly the
components of NMR hardware that make the measurements possible.
Experimental Equipment
A magnetic resonance system includes, on the simplest level, a magnet to create
the B0 field, a spectrometer to deliver the pulses of current, an rf coil to transmit and
receive the current to and from the sample, and a computer to control the system.
The high-field magnets in the College of Engineering Magnetic Resonance Lab at
MSU (250 MHz and 300MHz) are super-conducting magnets made with a coil of copper
wire several kilometers in length. The coil is maintained at a low temperature with liquid
helium and nitrogen to minimize resistance within the wire, keeping the wire
superconducting. The magnets have a vertical bore that holds samples ranging between
approximately 5—25mm in diameter. The samples are loaded into the magnet within an
rf coil. The coil is composed of a ‘birdcage’ of copper wire positioned so as to create an
orthogonal B1 field and receive the induced signal. Gradient coils are often also included
around the sample to spatially vary the magnitude of the applied magnetic field.
The spectrometer delivers rf pulses of current, gated to the proper pulse design
according to pulse duration and frequency of current oscillation. The same coils transmit
and receive the current so hardware may impose limitations on the time resolution of
29
some NMR measurements, especially where signal relaxation occurs on the timescale of
the conversion between coil modes.
The low-field NMR magnets (245/290 kHz and 425/360 kHz) used in some of the
experiments described later are down-the-borehole solid magnets measuring 4.5 feet long
and 3.5 inches in diameter. These magnets are designed to be lowered into a well casing
and project a vertical B0 field into the soil surrounding the well. The static magnetic field
loses strength with radial distance from the well center, with peak sensitivity between 5—
8 inches from the well center, depending on the operating frequency. The rf coil is
contained within the same probe as the solid magnet. The spectrometer and computer are
housed in a surface station.
Signal Detection
When the net magnetization of a sample is excited from equilibrium with an
infinitesimally short resonant rf pulse, the resulting precession of M at the Larmor
frequency in the laboratory reference frame induces a current in the rf coil according to
Faraday induction. This induced current is the basis for NMR signal detection. The rf
coil, a ‘birdcage’ (though solenoid coils are more efficient), is aligned orthogonal to the
z-axis such that the coil output is an oscillating voltage, 𝑉𝑉(𝜕𝜕) = 𝑉𝑉0 cos(𝜔𝜔0𝜕𝜕). The
strength of the signal is proportional to the Larmor frequency, 𝛾𝛾𝐵𝐵0. Higher field magnets
(large B0) or a high gyromagnetic ratio, γ, increase the measurable signal to noise ratio
and thereby improve sensitivity. Hydrogen protons have the largest gyromagnetic ratio
of any stable nuclear isotope which, together with their natural abundance in materials of
30
interest, helps explain the prevalence of 1H NMR within the larger field of magnetic
resonance.
Since the initial spin density is proportional to the equilibrium polarization,
𝜌𝜌(0)~𝐼𝐼𝑧𝑧, the resulting spin density matrix in the rotating frame following excitation has
proportionality of the form
𝜌𝜌𝑟𝑟𝑜𝑜𝑟𝑟(𝜕𝜕)~𝐼𝐼𝑦𝑦 cos�(𝜔𝜔0 − 𝜔𝜔)𝜕𝜕� + 𝐼𝐼𝑥𝑥 sin�(𝜔𝜔0 − 𝜔𝜔)𝜕𝜕� (2.32)
In order to separately measure the 𝐼𝐼𝑥𝑥 and 𝐼𝐼𝑦𝑦 components of the spin density
matrix, corresponding to the real and imaginary parts of the signal, respectively, the rf
coil uses the process of heterodyne detection. Heterodying with two quadrature detection
channels involves mixing the induced signal voltage with the output from a reference rf
oscillator orthogonal to both the z-axis and the receiving rf coil. Heterodyning is
mathematically equivalent to multiplying the induced voltage by a complex signal, exp(𝑖𝑖𝜔𝜔𝜕𝜕), and filtering out the sum frequency term, with the result that
𝑉𝑉(𝜕𝜕) = 12𝑉𝑉0[cos(𝜔𝜔0𝜕𝜕 − 𝜔𝜔𝜕𝜕) − 𝑖𝑖 sin(𝜔𝜔0𝜕𝜕 − 𝜔𝜔𝜕𝜕)] (2.33)
Thus, we directly measure �𝐼𝐼𝑥𝑥 + 𝑖𝑖𝐼𝐼𝑦𝑦� in the rotating reference frame at the
heterodyne mixing frequency, ω. When ω = ω0, the measured signal consists of two
components 90° out of phase. When ω ≠ ω0, the signal will oscillate at the offset
frequency, ∆𝜔𝜔 = 𝜔𝜔0 − 𝜔𝜔. The resulting phase factor, exp (𝑖𝑖𝑖𝑖), can be removed during
signal processing.
31
The complex signal 𝑆𝑆(𝜕𝜕) is well-suited to analysis with Fourier transforms
between the conjugate variables, ω and t. The Fourier transform of the measured time-
domain signal is expressed as
𝐹𝐹{𝑆𝑆(𝜕𝜕)} = 𝑠𝑠(𝜔𝜔) = � 𝑆𝑆(𝜕𝜕) exp(−𝑖𝑖𝜔𝜔𝜕𝜕) 𝑑𝑑𝜕𝜕∞
−∞
(2.34)
and produces a spectrum in the frequency domain, 𝑠𝑠(𝜔𝜔). The inverse Fourier transform
of the spectrum returns the original signal by
𝐹𝐹−1{𝑠𝑠(𝜔𝜔)} = 𝑆𝑆(𝜕𝜕) = 12𝜋𝜋� 𝑠𝑠(𝑓𝑓) exp(𝑖𝑖𝜔𝜔𝜕𝜕) 𝑑𝑑𝜔𝜔∞−∞ (2.35)
where 𝑓𝑓 = 𝜔𝜔 2𝜋𝜋� and converts between cyclic and angular frequency.
For example, the Free Induction Decay, or FID, is the simplest NMR
measurement and is the result of free precession of spins that induce an oscillating,
decaying signal in the time domain. The FID consists of a single 90° rf pulse followed
immediately by signal acquisition (Figure 2.5). The FID decays at a rate of 1 𝑇𝑇2∗� which
includes effects from an inhomogeneous magnetic field and from molecular interactions
in the transverse plane.
32
The Fourier transform of the FID is, in the real spectrum, a Lorentzian with full-
width-half-maximum (FWHM) of 1 𝜋𝜋𝑇𝑇2∗� , called the absorption spectrum, and an
imaginary spectrum called the dispersion spectrum. The integral of the phase-corrected
absorption spectrum is equal to the amplitude of the NMR signal. The FID produces a
signal
𝑁𝑁𝛾𝛾𝑇𝑇𝑇𝑇 ��𝐼𝐼𝑥𝑥 + 𝑖𝑖𝐼𝐼𝑦𝑦�𝜌𝜌𝑟𝑟𝑜𝑜𝑟𝑟(𝜕𝜕)� = 𝑖𝑖𝑀𝑀0 exp(−𝑖𝑖(𝜔𝜔0 − 𝜔𝜔)𝜕𝜕) (2.36)
which relates the observable 𝑁𝑁𝛾𝛾�𝐼𝐼𝑥𝑥 + 𝑖𝑖𝐼𝐼𝑦𝑦� operating on the spin density matrix in the
rotating frame 𝜌𝜌𝑟𝑟𝑜𝑜𝑟𝑟 to the time and phase evolution of the magnetization.
In practice, the NMR signal is not recorded as a continuous function, but is rather
sampled and digitized for later processing. A finite set of N points are sampled in the
time domain, spaced by the dwell-time interval, T. When Fourier transformed, the data
set in the frequency domain has a spectral width equal to the inverse of the dwell time, 1
𝑇𝑇� , with a digital resolution of
1
𝑁𝑁𝑇𝑇� .
FT
90°
rf pulse
real signal
imaginary
signal
time
t
t
real
spectrum
imaginary
spectrum
ω
ω
Δωω0
ω0
Figure 2.5. The Free Induction Decay (FID) measurement produces complex decaying
signal in the time domain and a spectrum in the frequency domain.
33
Basic Pulse Sequences
NMR measurements consist of orchestrated sequences of rf pulses and applied
magnetic field gradients which manipulate spin dynamics and encode for observables of
interest, as well as wait times and signal acquisition times to allow the evolution and
collection of signal that illuminates some aspect of the physico-chemical system. There
are several measurements fundamental to NMR which will be described here in some
detail. These include inversion recovery, the Hahn spin echo, the Carr-Purcell echo train
and the stimulated echo. Each of these experiments can be conceptualized on the macro-
scale using the net magnetization vector, M, influenced by the Zeeman Hamiltonian and
subject to T1 and T2 relaxation. More complicated measurements take elements of these
basic techniques and re-combine them to provide additional information about the sample
under study. Typically, the same measurement is performed N successive times and the
signal is averaged across all the measurements to improve the signal-to-noise ratio.
Signal adds coherently while noise adds in random phase where its average approaches
zero. The time between experiment repetitions depends on the T1 relaxation time of the
sample since it is typically desirable for the net magnetization to return to its maximum
along the z-axis prior to repeating the experiment.
The timing of NMR experimental elements is displayed in a pulse sequence
diagram, which typically includes the rf pulse timing and sequence on the first row with
time increasing to the right. Applied magnetic field gradients are included in the next
rows, depending on the axis along which they are applied. Acquired signal can be shown
in-line with the rf or gradient sequence or may be included separately on its own line.
34
Inversion Recovery. The inversion recovery experiment measures the T1
relaxation time of a sample by first inverting M onto the –z-axis with a 180° rf pulse
(Figure 2.6). M decays in magnitude as the spins experience spin-lattice relaxation and
begin to return to thermal equilibrium. After some time τ, a 90° rf pulse tips any
remaining magnetization into the transverse plane where it can be measured as it
precesses at the Larmor frequency (Figure 2.7).
Figure 2.6. The inversion recovery pulse sequence measures T1 relaxation by inverting
the magnetization and allowing T1 relaxation to occur before signal acquisition.
Figure 2.7. Evolution of the magnetization vector during the inversion recovery
experiment.
180x°
rf pulse time
90x°
τ
x
y
z
M
x
y
z
M
x
y
z
M
x
y
z
M ω0
35
Repeating the experiment for various τ times produces a curve of equation
𝑀𝑀𝑦𝑦(𝜕𝜕) = 𝑀𝑀0(1 − 2 exp �−𝜕𝜕 𝑇𝑇1� � (2.37)
As seen in Figure 2.8, at τ = 0.6931T1 the measured signal amplitude is zero as the
excited net magnetization crosses the transverse plane on its return to thermal
equilibrium. This result is particularly useful when only one component of a sample is of
interest since it allows for suppression of signal from any spins with a T1 time distinct
from other spins. Applying an rf pulse at an interval τ = 0.6931T1 prior to the
experimental pulse sequence effectively nulls the signal from spins with that particular T1
relaxation time and allows for selective acquisition of signal from spins relaxing at a
different rate. This method is particularly useful in biological tissues where the water
signal can overwhelm signal from other tissue components.
Figure 2.8. The amplitude of the net magnetization is equal to zero at 0.6931T1, allowing
for signal from spins with a specific T1 relaxation time to be nulled in a measurement.
Hahn Spin Echo. Despite the best efforts of hardware manufacturers to design a
NMR magnet with a perfectly homogenous B0 field, some degree of inhomogeneity is
t (s)
Mz
τ = 0.6931T1
M0
-M0
36
unavoidable. Spins that experience a slightly different magnetic field will precess at a
slightly different frequency from other spins in the sample. This phenomenon leads to
dephasing of spins in the transverse plane, as some move either faster or slower than the
Larmor frequency. In 1950, Erwin Hahn [3] demonstrated that this component of signal
attenuation is reversible through the application of a 180° rf pulse in the transverse plane
at some time τ after the 90° excitation pulse (Figure 2.9).
Figure 2.9. The Hahn echo pulse sequence and resulting evolution of the net
magnetization. A 180°y rf pulse refocuses signal to produce an echo at time 2τ following
the 90°x rf excitation pulse. The evolution figures are show below the position on the
pulse sequence where they occur.
The 180° pulse has the effect of refocusing the signal as the faster precessing
spins are placed ‘behind’ the slower moving spins. The spin coherence is maximized at
180y°
rf pulse time
90x°
τ
2τ
x
y
z
M
x
y
z
ω0-
ω0+
x
y
z
ω0-
ω0+
x
y
z
ω0-
ω0+
x
y
z
M
37
time 2τ, when an ‘echo’ in the signal forms. Since the signal is still subject to T2
relaxation during the time 2τ, the echo amplitude is less than the initial amplitude. Only
the attenuation due to magnetic field homogeneity can be re-focused. Relaxation due to
homo-nuclear spin-spin interactions within the sample cannot be reversed.
CPMG Echo Train. Following Hahn’s discovery of the spin echo, Carr and
Purcell in 1954 [4] expanded the concept by performing successive 180° rf pulses to
produce a series of echoes. Meiboom and Gill [5] modified the echo train in 1958 by use
of quadrature 180y° rf pulses to compensate for small turn-angle errors. The resulting
experiment, the CPMG sequence, allows for the direct measurement of T2 relaxation time
in one experiment. The echo envelope decays at a rate of 1 𝑇𝑇2� .
38
Figure 2.10. The CPMG pulse sequence, consisting of a 90° pulse followed by a series of
180° pulses, is used to measure T2 relaxation where the echo envelope decays at a rate of
1/T2.
Stimulated Echo. For materials where the T2 relaxation time is much shorter than
the T1 relaxation time, such as solids, phase coherence expires much faster than the
sample polarization. In other words, signal is no longer measureable because of T2
relaxation though incoherent magnetization remains in the transverse plane. The
stimulated echo can be used to obtain signal from materials with short T2 times,
especially when molecular motion is the observable of interest.
The stimulated echo experiment combines elements from the previously discussed
pulse sequences—inversion recovery and the spin echo. Figure 2.11 shows the pulse
sequence with the induced signal on the rf pulse line. After the initial 90x° excitation
pulse tips M onto the y-axis, the spins relax according to both T1 and T2 relaxation until
the time τ, when another 90x° pulse is applied. The second pulse rotates the y-component
of the magnetization into the longitudinal plane where only T1 relaxation occurs. The
180y°
rf pulse time
90x°
τ 3τ
180y° 180y°
5τ
180y°
7τ
signal
time
39
remaining polarization can be ‘stored’ for a time τ’ (> τe), called fast
exchange, there is a single relaxation time for the system since each spin will experience
70
a different relaxation rate in each domain sampled during the measurement, effectively
averaging the relaxation rates. Intermediate exchange results in non-exponential signal
decay.
A second model relies on a geometric description to relate relaxation and
diffusion in porous media. In this model, we distinguish between the relaxation of the
bulk fluid in the pore space, TiB, and the relaxation of the fluid directly interacting with
the pore surface, TiS. Here, the relaxation time can refer to either spin-lattice (T1) or spin-
spin relaxtion (T2), with the general case denoted as Ti. The solid matrix introduces
surface relaxation sinks that enhance relaxation of the spins nearest to the pore walls.
These sinks can be paramagnetic ions within the mineral surface of the pore walls or a
local hindering of rotational mobility resulting from increased dipole-dipole interactions
between liquid and solid phase molecules. Clearly, the surface to volume ratio of the
pore, S/V, is important in this model and is related to pore size and geometry. There are
also differences in magnetic susceptibility between the solid matrix and the bulk liquid,
resulting in internal magnetic field gradients near the pore walls. Diffusion of spins
within these internal inhomogeneities in the magnetic field further enhance relaxation.
The geometric model was developed in detail by Brownstein and Tarr in 1979.[4]
The authors observed fast and multi-exponential signal decay behavior in the water
within biological cells. They concluded that the multi-exponential behavior was a
consequence of the geometry of the cell and that it is possible to use NMR data to make
inferences about the size and shape of pores (or cells). Brownstein and Tarr approached
the problem by assuming diffusion of the magnetization density, M(r,t), according to
71
Fick’s Laws. They integrated over the pore volume and applied the initial condition,
M(r, 0) = M(0)/V to solve for the general normal mode solution
𝑀𝑀(𝑡𝑡) = 𝑀𝑀(0)�𝐼𝐼 𝑛𝑛 exp �−𝑡𝑡 𝑇𝑇𝑛𝑛� �∞
𝑛𝑛=0
(3.18)
where 𝐼𝐼𝑛𝑛 = 𝑎𝑎𝑛𝑛 ∫𝑑𝑑𝐫𝐫𝑉𝑉 𝑢𝑢𝑛𝑛 and describes the relative intensity. The parameter n is the
mode number, un is an eigenfunction, and the decay time Tn (actually T1n or T2n) is an
eigenvalue. Equation 2.15 shows that the evolution of the net magnetization proceeds as
a sum of decreasing exponential functions in time—resulting in the observed multi-
exponential decay. In and Tn depend on D, the self-diffusion coefficient, a, the
characteristic pore size, and ?̅?𝜌, the surface relaxivity, an empirical parameter describing
the average sink strength of the surface. This solution assumes that there are no
relaxation sinks within the pore volume, only on the active surface of the pore.
Solving for In and Tn for planar, cylindrical and spherical geometries reveals the
importance of the dimensionless sink strength parameter, ?̅?𝜌𝑎𝑎/𝐷𝐷. For each geometry, the
mode amplitude, In, depends on ?̅?𝜌𝑎𝑎/𝐷𝐷, which can be used to assign regimes analogous to
the fast and slow exchange regimes of the first model. When there is fast diffusion and ?̅?𝜌
is weak compared to D such that ?̅?𝜌𝑎𝑎/𝐷𝐷<<1 – the fast exchange regime – there is a single
relaxation mode, I0, for all geometries and the spins will experience single exponential
decay with a relaxation rate, 𝑇𝑇0−1~?̅?𝜌(𝑆𝑆 𝑉𝑉)⁄ . In the slow exchange regime, when ?̅?𝜌 is
strong compared to D and ?̅?𝜌𝑎𝑎/𝐷𝐷>>1, the lowest relaxation mode still dominates but there
is increasing influence of the higher modes. This means that spins will exhibit multi-
exponential decay during the measurement timescale. In the slow exchange regime,
72
therefore, relaxation time, 𝑇𝑇0~ 𝑎𝑎2 𝐷𝐷⁄ , corresponds to the time to diffuse across the pore
since the spin will be lost to relaxation once it interacts with the surface.
These results predict the same relationship between the decay behavior of the
normalized echo amplitude and the relaxation and diffusion properties of a system as the
first, non-geometric model – single exponential decay when the spins can sample the
entire pore, and multi-exponential decay when they cannot. The higher mode decay
rates, 𝑇𝑇𝑛𝑛−1, are visible only in the slow exchange/slow diffusion regime, are nearly
independent of ?̅?𝜌, and are of an order 𝑎𝑎2 𝑙𝑙2𝜋𝜋2𝐷𝐷⁄ . Because of this relationship, we can
estimate the range of pore sizes that may exhibit multi-exponential decay when examined
with PFG NMR. Assuming D = 2.5 x 10-5 cm2/s and 𝑇𝑇𝑛𝑛 is between 1 µs (due to hardware
limitations) and 2 s (the relaxation of bulk water), the pore dimension, a, is in the range
of 1 – 30 µm. This is exactly the size range of biological cells and pore sizes in natural
porous media like sedimentary rock.
While it is difficult to independently and reliably estimate the average surface
relaxivity, ?̅?𝜌, the Brownstein-Tarr model is widely used to obtain pore-size distributions
from multi-exponential relaxation data and complements q-space data as a way to
characterize pore structure and geometry.
Multidimensional PFG and Relaxation NMR
Multidimensional PFG-NMR experiments encode the induced signal in multiple
independent dimensions and can take the form of correlation or exchange experiments.
Correlation experiments involve the measurement of two different parameters, like
73
diffusion (D) and T2, at the same time to see how one parameter is related to the other. In
the case of the D-T2 experiment, for example, we are interested in how molecular
translational mobility (D) is correlated with molecular rotational mobility (T2). Exchange
experiments, on the other hand, encode for one parameter at two different times to probe
molecular migration between domains.
The measured signal is transformed to produce a spectrum using either the Fourier
or Laplace transform. Which type of transform is appropriate for a given signal depends
on the information sought from the sample. Oscillating, decaying signal collected in the
time domain can be Fourier transformed to produce a spectrum of the frequencies
contributing to the signal in the frequency domain. When the signal is collected under
the influence of time-varying gradients and coherent flow, the Fourier transformed data
will produce a spectrum of displacements, the propagator. Diffusive displacements result
in a signal that decays exponentially with respect to q2, the squared gradient pulse area.
A Fourier transform of the diffusion data with respect to q produces a Gaussian
displacement spectrum where the width depends on the diffusion coefficient, D.
If data is collected from a sample exhibiting multi-exponential signal decay due to
a range of relaxation domains or diffusion coefficients, Fourier transformation of the
signal will result in a superposition of Gaussian distributions, making it difficult to
extract the relaxation rates or diffusion coefficients associated with each domain.
Equation 2.19 shows the multi-exponential signal E(q) collected from a sample with
multiple diffusion coefficients. A Fourier transform of the signal produces a spectrum of
displacements, 𝑃𝑃�(𝑍𝑍), rather than a spectrum of diffusion coefficients, 𝑃𝑃(𝐷𝐷).
74
𝐸𝐸(𝑞𝑞) = ∫𝑃𝑃(𝐷𝐷) exp(−𝑞𝑞2𝐷𝐷∆)𝑑𝑑𝐷𝐷 (3.19)
𝑃𝑃�(𝑍𝑍) = �𝑃𝑃(𝐷𝐷)(2𝜋𝜋𝐷𝐷∆)−1 2� exp (−𝑍𝑍2 2𝐷𝐷∆)𝑑𝑑𝐷𝐷⁄ (3.20)
The inverse Laplace transform (ILT), ℒ−1{𝐸𝐸(𝑞𝑞2,∆},with respect to q2 will
produce a spectrum, or distribution, of the probability of having a given diffusion
coefficient, 𝑃𝑃(𝐷𝐷). 𝑃𝑃(𝐷𝐷) = ℒ−1{𝐸𝐸(𝑞𝑞2,∆} (3.21)
𝐸𝐸(𝑞𝑞2,∆) = ℒ{𝑃𝑃(𝐷𝐷)} = � 𝑃𝑃(𝐷𝐷) exp(−𝑞𝑞2𝐷𝐷∆)𝑑𝑑𝐷𝐷 ∞
0
(3.22)
Likewise with measurements involving sub-ensembles with different relaxation
rates, the ILT of relaxation data, ℒ−1{𝑆𝑆(𝑡𝑡)}, returns a distribution of the relaxation decay
rates that contribute to the multi-exponential decay, 𝑓𝑓(𝑅𝑅). 𝑓𝑓(𝑅𝑅) = ℒ−1{𝑆𝑆(𝑡𝑡)} (3.23)
𝑆𝑆(𝑡𝑡) = ℒ{𝑓𝑓(𝑅𝑅)} = � 𝑓𝑓(𝑅𝑅) exp(−𝑅𝑅𝑡𝑡)𝑑𝑑𝑅𝑅∞
0
(3.24)
The forward and reverse Fourier transform are well-defined, symmetric with
bounds between -∞ and ∞, and easy to implement with the fast Fourier Transform (FFT)
algorithm. The ILT, however, is ill-defined, asymmetric with bounds between 0 and ∞,
and not easy to implement. The analytical form of the ILT, shown below for a
distribution of relaxation rates,
𝑓𝑓(𝑅𝑅) = ℒ−1{𝑆𝑆(𝑡𝑡)} = 12𝜋𝜋𝑖𝑖 � 𝑆𝑆(𝑡𝑡) exp(𝑅𝑅𝑡𝑡) 𝑑𝑑𝑡𝑡𝛾𝛾+𝑖𝑖∞𝛾𝛾−𝑖𝑖∞ (3.25)
is a contour integral in a complex plane where 𝛾𝛾 is a vertical contour positioned to the
right of any singularities. This unstable expression can result in exponential divergence
75
when there is noise in S(t), making this analytical form not particularly useful in practice.
Instead, Provencher[5] developed a regularized non-negative least squares method to
perform the ILT in 1982 that uses a discrete form of the signal that allows for noise:
𝑆𝑆(𝑡𝑡𝑖𝑖) = � exp�−𝑡𝑡𝑖𝑖𝑅𝑅𝑗𝑗� 𝑓𝑓�𝑅𝑅𝑗𝑗� + 𝜖𝜖𝑖𝑖𝐷𝐷
𝑗𝑗=1
(3.26)
where 𝜖𝜖𝑖𝑖 is the error or noise in the measurement, i is the discrete time interval, and j is
the discrete relaxation rate domain. A priori knowledge of the system requires that
𝑓𝑓�𝑅𝑅𝑗𝑗�>0. The optimal solution will be as simple as possible and minimize the error, 𝜖𝜖𝑖𝑖,
for a given range of R between Rmax and Rmin, where R is the relaxation rate associated
with each domain.
Provencher used the Tikhonov regularization[6] to ensure parsimony (the simplest
solution) where we seek a minimum value for the solution, 𝑉𝑉(𝛼𝛼), where
𝑉𝑉(𝛼𝛼) = �𝐾𝐾𝑓𝑓 − 𝑆𝑆�2 + 𝛼𝛼2 �Γ𝑓𝑓�2 = 𝑚𝑚𝑖𝑖𝑙𝑙𝑖𝑖𝑚𝑚𝑢𝑢𝑚𝑚 (3.27)
and
�Γ𝑓𝑓�2 = � [𝑓𝑓"(𝑅𝑅)]2𝑑𝑑𝑅𝑅𝑅𝑅𝑚𝑚𝑎𝑎𝑥𝑥
𝑅𝑅𝑚𝑚𝑚𝑚𝑛𝑛
(3.28)
In these expressions, K(t, R) is the kernel, α is regularization parameter, and Γ is a
smoothing operator based on the curvature of f, the distribution, and S is the measured
signal. The expression in Equation 2.27 is discretized such that a solution vector 𝑓𝑓 has
values corresponding to each element, 𝑅𝑅𝑗𝑗, such that
𝑉𝑉(𝛼𝛼) = ��(𝑆𝑆𝑖𝑖 − exp�−𝑡𝑡𝑖𝑖𝑅𝑅𝑗𝑗� 𝑓𝑓𝑗𝑗)2𝐷𝐷
𝑗𝑗
+ 𝛼𝛼2�(2𝑓𝑓𝑗𝑗 − 𝑓𝑓𝑗𝑗+1 − 𝑓𝑓𝑗𝑗−1)2𝐷𝐷
𝑗𝑗
𝑁𝑁
𝑖𝑖
(3.29)
76
The first summation term – the residual – is called χ2 and is minimized by
selecting the value of the regularization parameter,𝛼𝛼, that minimizes the error term in the
second summation. A large value of 𝛼𝛼 implies a high level of confidence in the fine
details of the measurement. Too large a value of 𝛼𝛼 leads to ‘pearling’ of the data where
more detail is shown in the final distribution than exists in the original data. Hence, we
seek the solution that produces the least amount of curvature in the distribution while still
faithfully representing the data.
77
Figure 3.4. Multi-dimensional PFG-NMR correlation and exchange pulse sequences.
The T1-T2 pulse sequence (top) and D-T2 pulse sequence (bottom) correlate two different
parameters in the same time step. The T2-T2 pulse sequence (middle) encodes for
relaxation at two different times and gives insight to exchange between relaxation
domains.
The pulse sequences used to collect multi-dimensional PFG-NMR data and
correlation and exchange measurements are shown in Figure 3.4. The T1-T2 experiment
(top) first encodes for T1 relaxation for a given inversion time, then measures T2 decay.
78
The measurement is repeated for a range of inversion times to correlate the sample’s T1
and T2 relaxation properties.
The T2-T2 exchange pulse sequence (Figure 3.4, middle) encodes for T2 relaxation
at two different time periods, separated by a mixing period, τm, during which spins are
free to migrate between relaxation domains. Magnetization is stored along the z-axis
during the mixing period to preserve the relaxation encoding of the first CPMG train.
The number of echoes produced during the first encoding period is varied to capture both
long and short relaxation components of the first encoding period in the final recorded
signal. The T2-T2 exchange experiment provides insight to the timescale of molecular
transport between physical regions in the sample, i.e. pores, or between relaxation
domains, i.e. bound vs. free water.
The pulse sequence for the diffusion – relaxation, or D-T2, experiment is shown in
Figure 4, bottom. After encoding for diffusion with the PGStE sequence, a CPMG echo
train is collected to measure T2 relaxation. The strength of the pulsed gradient, g, is
varied with each repetition of the sequence to resolve the range of effective diffusion
coefficients present in the sample. Data collected with the D-T2 can be analyzed with the
2D ILT to produce distributions of both diffusion and relaxation, or the diffusion data can
be analyzed with Fourier methods when sufficient positive and negative q-steps are
collected. The Fourier-Laplace analysis produces a T2-resolved propagator.
79
References
1. Callaghan, P.T., Translational Dynamics & Magnetic Resonance: Principles of
Pulsed Gradient Spin Echo NMR. 2011, New York: Oxford University Press.
2. Torrey, H.C., Bloch Equations with Diffusion Terms. Physical Review, 1956.
104(3): p. 563-565.
3. Stejskal, E.O. and J.E. Tanner, Spin diffusion measurements: Spin echoes in the
presence of a time-dependent field gradient. Journal of Chemical Physics, 1965.
42: p. 288.
4. Brownstein, K.R. and C.E. Tarr, Importance of classical diffusion in NMR studies
of water in biological cells. Physical Review A, 1979. 19(6): p. 2446-2453.
5. Provencher, S.W., A constrained regularization method for inverting data
represented by linear algebraic or integral equations Computer Physics
Communications, 1982. 27(3): p. 213-227.
6. Tychonoff, A.N. and V.Y. Arsenin, Solution of ill-posed problems. 1977,
Washington: Winston and Sons.
80
CHAPTER FOUR
LOW-FIELD BOREHOLE NMR APPLICATIONS IN
THE NEAR SUBSURFACE ENVIRONMENT
Contribution of Authors and Co-Authors
Manuscript in Chapter 4
Author: Catherine M. Kirkland
Contributions: Researched and wrote manuscript.
Co-Author: Sarah L. Codd
Contributions: Helped write manuscript. Provided feedback and comments on the
manuscript.
81
Manuscript Information Page
Catherine M. Kirkland, Sarah L. Codd
Vadose Zone Journal
Status of Manuscript:
____ Prepared for submission to a peer-reviewed journal
__X_ Officially submitted to a peer-review journal
____ Accepted by a peer-reviewed journal
_ ___ Published in a peer-reviewed journal
Soil Science Society of America
January, 2017
82
LOW-FIELD BOREHOLE NMR APPLICATIONS
IN THE NEAR-SURFACE ENVIRONMENT
Abstract
The inherent heterogeneity of the near subsurface (< 200 m below the ground
surface) presents challenges for agricultural water management, hydrogeologic
characterization, and engineering, among other fields. Borehole nuclear magnetic
resonance (NMR) has the potential not only to describe this heterogeneity in space non-
destructively, but also to monitor physical and chemical changes in the subsurface over
time. NMR is sensitive to parameters of interest like porosity and permeability,
saturation, fluid viscosity, and formation mineralogy. Borehole NMR tools have been
used to measure soil moisture in model soils and recent advances in low-field borehole
NMR instrumentation allow for estimation of hydraulic properties of unconsolidated
aquifers. We also present results demonstrating the potential for low-field borehole
NMR tools to monitor field relevant biogeochemical processes like biofilm accumulation
and microbially-induced calcite precipitation (MICP) at laboratory and field scales.
Finally, this mini-review addresses some remaining challenges and areas of future
research, as well as other possible applications where borehole NMR could provide
valuable complementary data.
83
Introduction
The inherent heterogeneity of the near subsurface, defined here as< 200 m below
the ground surface, presents challenges for agricultural water management,
hydrogeologic characterization, and engineering, among other fields. Knowledge of the
spatial heterogeneity of unconsolidated sediments is vital to accurately estimate hydraulic
properties related to storage and flow and to locate confining layers. Spatially varying
hydraulic properties can cause preferential flow paths and mixing of subsurface flows
where conductivity and permeability are high, or alternatively, trap contaminants and
impede remediation efforts where conductivity and permeability are low. For example,
the increasing consumptive pressure on groundwater reserves is stimulating managed
aquifer recharge (MAR) and aquifer storage and recovery (ASR) solutions which require
greater sophistication in hydrogeologic investigations and modelling than conventional
wellfields [1]. These methods artificially recharge aquifers with water of a sometimes
differing quality than the native water, such as fresh water stored in a brackish aquifer.
Extraction of the high-quality water later for re-use is maximized only when there is
minimal mixing during storage as would occur in lower transmissivity zones.
Furthermore, recent years have seen increased interest in harnessing the power of
biogeochemical conversions to assist diverse engineering applications, including in-situ
bioremediation of subsurface contaminants [2] and microbially-induced calcite
precipitation (MICP) for fracture sealing [3] or geotechnical applications to strengthen
soils [4, 5]. These methods are intended to alter the physical and chemical environment
84
of the pore space; proper execution of these projects requires knowledge of both temporal
and spatial variations in the subsurface.
The standard reference method for obtaining the volumetric water content of
porous media is gravimetric. A sample of known volume is weighed before and after
oven drying to determine the water content at the time of the measurement. Gravimetric
measurements are necessarily invasive, destructive, and time-consuming to perform,
providing point-scale data at a single time. Neutron thermalization methods, like neutron
scattering, are significantly less invasive than gravimetry and are based on a well-
established, linear correlation between the sediment water content and the ratio of
thermalized neutrons reaching the detector. The radioactive materials which supply the
neutrons, however, impose significant regulatory barriers, making the technology less
attractive as newer methods are developed [6]. Established hydrogeophysical methods
for in-situ characterizations of soil and sediment rely on measurements of geophysical
properties to infer hydrogeological properties, meaning that the methods do not provide
direct measurement of water [6, 7]. For example, Electrical Resistivity Tomography
(ERT), Ground Penetrating Radar (GPR), and seismic methods measure electrical
conductivity or resistivity, acoustic impedance, and velocity and attenuation of seismic
waves, respectively, to provide information about water content and porosity. The
translation of measured geophysical data to hydrogeologic and hydraulic properties of
interest requires non-trivial mathematical inversions and modelling. The many
hydrogeophysical methods available have varying spatial scales and resolution; readers
85
should refer to several recent reviews of geophysical methods for more information [6,
7].
Nuclear magnetic resonance (NMR) methods have the potential to both
characterize subsurface heterogeneity and monitor changes in the geophysical and
biochemical environment over time, suggesting the importance of this technology for
these fields. Low-field NMR instrumentation for field applications can take several
forms – mobile single-sided instruments like the NMR MOUSE, surface NMR
instruments, or borehole NMR probes. Though the focus of this mini-review is borehole
NMR, single-sided NMR and surface NMR tools deserve mention. The NMR MOUSE,
developed by researchers at RWTH – Aachen University in the 1990’s [8], weighs
approximately 1 kg and has been used in applications ranging from biomedicine, cultural
heritage preservation, materials science, and measurement of moisture in soils and
building materials [9, 10]. The depth of investigation for the various single-sided NMR
probes is typically on the order of centimeters. The configuration of the permanent
magnets and coils of wire comprising mobile, single-sided NMR tools vary according to
the application and the goals of the NMR measurement. Readers may refer to the
comprehensive description given by Blumich et al. (2008). Non-invasive explorations of
near surface hydrogeologic and hydraulic properties can also be performed with surface
NMR. Surface NMR uses the Earth’s magnetic field and coils of wire deployed on the
ground surface to collect the NMR measurement. Although surface NMR coils are easy
to deploy and are entirely non-invasive, the technology faces challenges related to low
signal-to-noise, a long instrument dead-time between excitation and detection which
86
means that water in small pores is not measured, and a relatively shallow depth of
investigation (< 100 m) with vertical resolution on the order of 10 m [11]. Several
reviews which provide a description of theory and recent advances in the technology
have been published [11, 12].
Borehole NMR tools have been used in the oil and gas industry to identify and
characterize hydrocarbon reserves since the 1960s [13] due to the sensitivity of NMR to
hydraulic properties of interest, like pore size distribution, permeability and hydraulic
conductivity, and fluid content and viscosity. The mathematics describing the relationship
between the measured NMR signal response and these formation hydraulic properties
developed alongside technical improvements to the instruments, such that now NMR
tools are an indispensable part of the oil and gas industry [14, 15]. Despite the fact that
the same hydraulic parameters are of interest in hydrogeology, geotechnical engineering,
and soil science, the oilfield well-logging tools are too large, expensive, and impractical
for most near-surface applications. As a result, the application of NMR tools to near-
surface investigations in unconsolidated sediments has developed only in the last decade
due to several independent and complimentary factors. Technical innovations have
allowed down-sizing of the hardware, and developments in numerical methods and error
estimation have improved the mathematical inversions needed for data processing [16-
18]. Finally, pressures on global water supplies and the need for improved monitoring
methods drive innovations in borehole NMR technology.
Recently, portable and lower-cost borehole NMR tools have been developed in
both commercial [19] and non-commercial [20] forms. These tools vary significantly in
87
operating frequency, measurement distance from the tool, the size of the probe, vertical
resolution, and the minimum echo spacing, with significant implications for the ease of
use of the instrument and interpretation of the data. This paper describes recent
applications of these tools related to soil water management, hydrogeology, and
biogeochemical engineering in the near subsurface, as well as remaining challenges and
further potential applications for the technology.
Theory
1H NMR measures the response of hydrogen-bearing molecules, typically water,
to perturbations in a magnetic field. As such, NMR is directly sensitive to water in soil,
rock, and unconsolidated subsurface sediments. Borehole NMR tools generally consist
of permanent magnets and radio-frequency (rf) induction coils. The permanent magnets
create a static magnetic field, B0, in the formation. The magnetic field strength in the
formation drops off with the radial distance from the tool as does the resonant frequency
at which the water will respond. The resonant frequency is called the Larmor frequency,
𝜔𝜔0 = 𝛾𝛾B0, where 𝛾𝛾 is the gyromagnetic ratio, a constant with a value of 2.675 x 108
rad/(T s) for hydrogen. Pulses of current are transmitted through the rf coils, exciting
hydrogen at a particular Larmor frequency. The location of the cylindrical excitation
region depends on the tuning of the probe; higher Larmor, i.e. operating, frequencies
produce an excitation - detection shell closer to the wellbore where the magnetic field is
stronger, while lower Larmor frequencies allow measurement deeper into the undisturbed
formation (Figure 4.1). The excited hydrogen protons generate a detectable signal
according to Faraday induction at the Larmor frequency which is received on the rf coils.
88
Figure 4.1. Schematic of a generalized borehole NMR wire-line logging tool with a
cylindrical excitation-detection shell. The magnitude of the static magnetic field, B0,
decays with radial distance from the well. The NMR probe is tuned to excite protons
(water) at a particular Larmor frequency, 𝜔𝜔0 = 𝛾𝛾B0, which corresponds to a particular
radial distance from the well. Lower frequencies produce excitation – detection shells
outside the disturbed soil zone that results from traditional well-drilling methods.
The induced NMR signal amplitude decays in time as the system returns to
equilibrium in a process called relaxation [21]. T1 (spin-lattice) relaxation and T2 (spin-
spin) relaxation both provide information about the local physical and chemical
environment. T1 relaxation is related to the timescale for system to return to thermal
equilibrium as the added energy from the excitation rf pulse dissipates into the
environment, or lattice. T2 relaxation is related to molecular interactions occurring
between the excited water molecules and their interactions with the pore walls.
Measurement of T2 relaxation is robust and faster to perform, making it preferred for low-
field borehole NMR [22]. T2 relaxation is measured with the CPMG pulse sequence [23,
Sensitive zone
Undisturbed formation
Disturbed zone
Well Casing
NMR Probe
𝐁0
89
24] in which a series of re-focusing rf pulses, called 180° pulses, follow the 90° rf
excitation pulse (Figure 4.2). The 180° pulses create a series of signal echoes where the
initial signal amplitude is proportional to the volumetric water content and echo
amplitudes decay at a rate of 𝑇𝑇2−1. The time between echoes is called the echo spacing,
𝑡𝑡𝐸𝐸.
Figure 4.2. The CPMG pulse sequence consists of a 90° excitation pulse and a series of
refocusing 180° pulses. The induced signal consists of a series of echoes in which the
decay curve (thin dashed line) has an initial amplitude proportional to the water content
and an exponential decay rate of 1/T2.
Data acquisition with borehole NMR tools in the field involves lowering the probe
incrementally into the subsurface, either in a borehole well or in an opening made by a
direct push (DP) tool. DP methods drive an instrumented steel rod into the ground using
hydraulic rams and avoid the need to drill boreholes. Aside from the borehole or DP
opening, NMR is non-invasive and non-destructive. An NMR measurement is collected
at each depth increment to create a well log, or soil profile, describing the sediment-water
relationship via T2 relaxation. The vertical resolution of the well log depends on the
180y°
rf pulse time
90x°
𝑡𝑡𝐸𝐸
180y° 180y° 180y°
signal
time
90
dimensions of the rf coil. The logging speed is related to the signal to noise ratio, and
thus, how many measurements are collected and averaged at each depth increment.
Since natural porous media is composed of many pores of varying sizes and
geometries, the signal collected during an NMR measurement is the sum of the signal
from each individual pore. The resulting multi-exponential signal decay curve data is
typically analysed with the inverse Laplace transform (ILT), producing a distribution of
T2 relaxation times. The integrated amplitude of the distribution provides the volume
fraction of water in the measurement region—the porosity if the sediment is saturated or
the water content if it is unsaturated. Such estimates can also be obtained from the initial
amplitude of the signal decay curve [25].
NMR T2 relaxation is sensitive to the physical and chemical properties of
macroscopic porous media systems, including pore size, soil mineralogy, degree of water
saturation, and pore fluid viscosity. The expression for the T2 relaxation rate, typically in
units of ms-1, in a single pore is given by the equation
𝑇𝑇2−1 = 𝑇𝑇2𝐵𝐵−1 + 𝑇𝑇2𝑆𝑆−1 + 𝑇𝑇2𝐷𝐷−1 [1]
where 𝑇𝑇2𝐵𝐵−1 is the bulk fluid relaxation rate, and 𝑇𝑇2𝑆𝑆−1 is the surface relaxation rate. The
final relaxation term is related to diffusion in an inhomogeneous magnetic field and can
influence NMR T2 relaxation measurements where the gradient in the local magnetic field
is high or when the echo spacing, tE, of the measurement is long. Gradients, or
inhomogeneities, in the static magnetic field result when the sediment is composed of
materials with different magnetic susceptibilities or when the strength of the B0 field
rapidly decays in the radial direction [26]. A lower operating frequency reduces the
91
static field gradient. Minimizing tE through instrument design and experimental
parameter selection can reduce the impact of T2D on the NMR signal response.
Many studies of porous media assume T2 relaxation is 1) independent of the self-
diffusion coefficient of water as in the ‘fast diffusion’ regime [27] and 2) dominated by
surface relaxation [28], which can be expressed as 𝑇𝑇2𝑆𝑆−1 = 𝜌𝜌 �𝑆𝑆𝑉𝑉� [2]
where 𝜌𝜌 is the surface relaxivity [cm/s], or the capacity of the pore wall to induce
relaxation, and �𝑆𝑆
𝑉𝑉
� is the surface to volume ratio of the pore [cm], typically modelled as
a sphere. From this expression, it follows generally that T2 is longer in larger pores and
shorter in smaller pores. Also important is the mineral surface of the solid matrix where
paramagnetic species like Fe(III) and Mn lead to faster relaxation (shorter T2) via the
parameter 𝜌𝜌 [29, 30]. In very coarse materials, the ‘slow diffusion’ regime [27]
dominates and Eqn. [2] becomes a function of �𝑆𝑆
𝑉𝑉
�
2
with no dependence on 𝜌𝜌.
The T2 relaxation time distribution can be represented by a single parameter, the
arithmetic mean of log T2, T2ML. This single parameter has been shown to estimate the
mean pore size and permeability with the following equations [31],
𝑇𝑇2ML−1 = 𝜌𝜌 �𝑆𝑆𝑉𝑉� [3] 𝑘𝑘NMR = 𝑏𝑏𝜑𝜑𝑚𝑚(𝑇𝑇2ML)𝑛𝑛 [4]
where �𝑆𝑆
𝑉𝑉
� is the surface to volume ratio of the total pore space, ρ is the surface relaxivity,
kNMR is the estimated permeability from the NMR data [31] with units typically expressed
in millidarcies (1 mD ≈ 10-11 cm2), ϕ is the NMR-determined porosity, and b [mD/ms2],
92
m, and n are empirically determined constants. Eqn. [4] is commonly known as the
Schlumberger-Doll Research (SDR) equation. For consolidated sediments and when
T2ML is in units of milliseconds, m and n are typically 4 and 2, respectively [32, 33]. Eqn.
[4] can also be written in terms of hydraulic conductivity, K, in units of m/s; in this case,
the values and units of the empirical constants account for the inclusion of fluid density
and viscosity terms. For further discussion of the empirical constants, refer to Knight et
al. (2016). From these expressions, it is clear that accurate estimation of k or K depends
on the ability of the instrument to measure T2S without the influence of diffusion
relaxation, T2D, since the influence of T2B is generally insignificant when the pore fluid is
water. Alternatively, diffusion effects must be quantified and decoupled from
measurement of T2 relaxation by encoding the signal also for the effective self-diffusion
of water. This is an ongoing area of research and will be further discussed in the final
section of this article. The influence of the T2D component is reduced by lowering the
tool operating frequency, i.e. static magnetic field strength, and by minimizing tE. Lower
operating frequency also reduces the signal to noise ratio, requiring longer measurement
times.
In saturated sediments, the shape of the T2 distribution estimates the pore size
distribution since T2 relaxation depends linearly on the pore size when ρ is assumed
constant; the distribution yields the relative volume of water in each size pore. Typically,
water in sandstone with a T2 relaxation time greater than 33 ms is considered ‘mobile’,
while water relaxing faster than 33 ms is considered ‘bound’ in capillaries [34]. ‘Clay
bound’ water is characterized by relaxation in less than 3 ms. Water signal decaying
93
faster than the minimum echo spacing is not detected. In the vadose zone, unlike in
deeper consolidated sediments, soils have varying degrees of saturation and hydraulic
conditions are dynamic on timescales ranging from hours to seasons to years. In these
conditions, interpretation of the T2 distribution as a pore size distribution is more
complicated. Soil drying may concentrate dissolved species in the remaining pore water
and influence bulk fluid relaxation, such that the 𝑇𝑇2𝐵𝐵−1 term in Eqn. [1] may become non-
trivial. Additionally, large pores drain before smaller pores. Jaeger et al. (2009) found
that the relaxation time distribution of unsaturated soil samples was strongly correlated to
the soil texture, i.e. relative sand, silt, and clay content, as well as the content of soil
organic matter (SOM).
To date, borehole NMR tools have been applied to the measurement of soil
moisture in laboratory studies and hydrogeologic characterization of aquifers in the field.
Building on those foundations, recent studies have used borehole NMR to detect and
monitor biogeochemical changes in the near subsurface for applications related to
bioremediation.
Measurement of Soil Moisture
Researchers at RWTH – Aachen University developed two non-commercial,
small-scale, and portable borehole NMR devices for measuring soil moisture [20, 35].
Sucre et al. (2011) developed a single-sided borehole NMR probe, 4.8 cm in diameter,
which operated at 11.8 MHz and produced a sensitive zone 4.7 mm from the exterior of
the device on one side. The minimum echo spacing, which controls the fastest-decaying
94
signal the tool can detect, was tE = 70 μs. Subsequently, the probe was re-designed, to
increase the signal to noise ratio and increase the penetration of the sensitive zone, then
again applied to measurement of soil moisture [20]. The optimized probe was 4.1 cm in
diameter, 18.0 cm long, and operated at a frequency of 3.32 MHz. The sensitive zone in
the soil for this instrument, unlike the previous, was a cylindrical shell 220 μm thick
located 4.5 cm from the central axis, or 1.8 cm from the exterior of the casing. The
minimum echo spacing was tE = 250 μs. The optimized probe also significantly reduced
the static field gradient from 24 T/m [35] to 3 T/m [20].
The two probes were tested in a similar manner to demonstrate performance [20,
35]. Each probe was placed in a tube within a larger column of model sandy soil. The
model soil FH31 is a distribution of grain size classes 2% (>0.72mm), 8% (0.71-0.5mm),
30% (0.5-0.355mm), 41% (0.36-0.25mm), 16% (0.25-0.18mm), 3% (<0.18mm). In both
studies, the soil moisture profile was measured first over the full depth of the column
when the soil was fully saturated. Then, the water was allowed to drain from the column
while the probe remained at a fixed depth, measuring the depletion of water over time
from the soil pores. Next, Perlo et al. (2013) repeated these two steps of the experiment
at different depths within the column. Figure 4.3 shows the drying of the soil column
over time for 5 arbitrary and undefined vertical positions. Note that the curves have
similar form and the initial water content in the saturated soil is equal within 1% for all
depths measured [20]. Sucre et al. (2011) instead measured the soil moisture profile over
the full column depth with the single-sided tool when the system reached hydraulic
equilibrium after the draining described above (Figure 4.4).
95
Figure 4.3. Temporal evolution during the draining process for different vertical z-
positions in the column filled with a model soil FH31. The z-positions are not specified in
the original article, but the asymptotic saturation value increases with depth in the
column. Initial saturation for all curves was measured with a precision of 1%. In the
inner box a CPMG-decay with 250 echoes recorded with 58 scans is shown. Reproduced
from Perlo et al. (2013).
96
Figure 4.4. Saturation profile in a model soil FH31 under full saturation conditions
(before outflow) and after the soil has been drained (after outflow). Reproduced from
Sucre et al. (2011).
Sucre et al. (2011) then calculated the soil hydraulic parameters using the NMR
data from the column drainage experiment. Vertical flow was simulated with the
Richards Equation, using the Mualem-van Genuchten model [36, 37] for K(θ), the
hydraulic conductivity, and h(θ), hydraulic head, where both depend on saturation, θ.
The saturated hydraulic conductivity, Ks, and the pore connectivity parameter, τ, were
obtained by an inversion analysis. The simulations generally conformed well to the
experimental results (Figure 4.4), though the rapid dynamic regime of the outflow
experiment produced a Ks of 9 mm/min in the highly conductive model soil FH31,
compared to a literature value of 7 mm/min [35]. The probe was also used to measure
soil moisture profile changes resulting from imbibition and precipitation. It should be
noted here that the relatively high field strength and high static field gradient of these two
97
probes means that especially during soil drying, the surface relaxation component of T2
relaxation, T2S, is strongly coupled to the diffusion relaxation component, T2D [35].
More recently, Vista Clara, Inc. (Mukilteo, WA) has commercialized a backpack
portable system called Dart [38] which is optimized for soil and vadose zone studies
shallower than 50 m. The Dart has an outer diameter of 4.5 cm and operates between
425-475 kHz, producing a sensitive zone that is at a maximum of 5 cm from the tool
surface. The static field gradient is 35 G/ cm (3.5e-5 T/m) at the sensitive zone. The
system is intended to be deployed in holes made with minimal soil disturbance like DP-
installed PVC or hand-augering. Recently the tool has been used to study thawing in
permafrost [39]. The electronics unit running the Dart can also be used to operate a non-
invasive single-sided NMR sensor called Discus, which sits on the soil surface and
measures water at four levels: approximately 5, 10, 15, and 20 cm from the face of the
sensor [38].
Characterization of Unconsolidated Aquifers
These next studies were conducted within the saturated, unconsolidated sediments
below the vadose zone. The first portable borehole NMR device reported in literature
was a commercial probe called the Javelin, by Vista-Clara, Inc. [19]. The Javelin probe
is a slim, borehole logging tool originally developed for hydrogeologic analysis in open
or PVC-cased boreholes as small as 5 cm in diameter [40], and can also be operated in
DP mode. Several versions of the probe have been commercialized, operating in the
range of 250—425 kHz, placing the sensitive region in the undisturbed formation outside
98
the wellbore approximately 11 – 19 cm radially from the center of the probe. The static
field gradient at the sensitive region is approximately 5 G/cm (5e-6 T/m). What is most
significant about this instrument is that the low operating frequency and static field
gradient, together with the minimum tE (on the order of 1 ms), allow a robust
measurement of T2 relaxation without the influence of diffusion relaxation, T2D.
Figure 4.5. Logs from Massachussetts Military Reservation obtained May 2010 (a) NMR
logs acquired in 4-inch well 03GB1060 show the T2 distributions, the NMR measured
water content, and the NMR-derived estimate of K; (b) comparison of water content
measured by NMR log and neutron porosity log acquired in nearby 2-inch well FSW445.
Reproduced from Walsh et al. (2013)
Initial field tests of the Javelin were conducted at the Massachusetts Military
Reservation, near a site of known subsurface contamination from fuel spills and other
activities. The 100 m NMR logging data compared well to a porosity estimation obtained
by a neutron log acquired in nearby well (Figure 4.5 (b)). The T2 distribution, NMR
99
estimation of water content, and NMR derived estimate of K – all measured as a function
of depth – also allowed researchers to identify water content associated with lenses of
low permeability silt which are expected to trap fluids, thereby affecting contaminant fate
and transport (Figure 5 (a)).
Figure 4.6. Comparison of K calculated from NMR logging data using the SDR equation
(Eqn. [4]) and from DP permeameter (DPP) data obtained at the Larned field site. The
figure shows logs from three different locations at Larned (a) Larned E, (b) Larned C, (c)
Larned W. The dashed lines show the uncertainty in KNMR due to the distribution in b
values (±1σ) when fitting Eqn. 4. Reproduced from Knight et al. (2016).
More recently, researchers have used the Javelin tool to characterize hydraulic
conductivity in unconsolidated aquifers at several field sites [41], where the goal of the
research was to identify standard values for the empirical constants in Eqn. [4]. Data
from 3 wells at the Larned Research Site of the Kansas Geological Survey in west-central
100
Kansas is shown in Figure 4.6. The NMR logging data estimation of the hydraulic
conductivity, K, compared very well, generally within an order of magnitude, to the more
established direct-push permeameter (DPP) estimates [42]. The authors found that, like in
consolidated formations [32, 33], there was little variability in the constants between the
different field sites. The value of the empirical constant, b, in Eqn. [4] when written in
terms of hydraulic conductivity, varied less than 50% across 3 field sites. This result
suggests that it may be possible to obtain reliable values for the empirical constants and
significantly reduce the need for site specific calibration to obtain accurate estimations of
K and k from NMR data [41]. Establishment of standard values for the constants will
improve the cost-effectiveness and ease of use of borehole NMR technology. Previous
work by the research group responsible for these measurements includes their initial
demonstration in a high plains aquifer [22], groundtruthing surface NMR with the NMR
logging tool [43], and determining methods to estimate uncertainties [44].
Detection of Subsurface Biogeochemical Processes
As Eqn. [1] shows, T2 relaxation in porous media depends on the properties of the
bulk fluid and the size and mineralogy of the pores. (Diffusion relaxation effects can be
neglected due to the low operating frequency of the borehole tool used in the following
studies.) It therefore follows that changing the properties of the pore fluid, or changing
the pore geometry or mineralogy, should produce a change in the T2 relaxation
distribution from some known initial state. In application of this premise, the research
group at Montana State University has been using the Javelin NMR logging tool to
101
monitor subsurface biogeochemical processes for bioremediation applications [45-47].
The studies included in this section were designed to assess the sensitivity of borehole
NMR to 1) a change in pore fluid – from water to biofilm – both at laboratory and field
scale, and 2) pore structural and mineralogical changes caused by microbially induced
calcite precipitation (MICP) in a laboratory bioreactor. Both biofilm growth and MICP
can be used as part of a bioremediation project.
Bioremediation can be an effective method to contain or degrade chemical
contaminants in the subsurface by exploiting the fundamental biochemical processes of
microbial metabolism. Once established, biofilms have been shown to degrade or contain
contaminants through a variety of mechanisms: utilizing hydrocarbons and other
contaminants directly as a substrate; inducing mineralization to trap contaminants;
transforming heavy metals to insoluble forms; or acting as a bio-barrier to retard the
migration of contamination or re-direct groundwater flow through a treatment zone.
Depending on its physical properties, the biofilm extracellular polymeric substance (EPS)
matrix can change soil pore connectivity, effective pore size, and hydraulic conductivity,
thereby affecting the hydrodynamic properties of the porous media [48] and allowing the
detection of the biofilm state with NMR [49].
Preliminary groundtruthing was demonstrated in a lab-scale bioreactor designed
to model the near-wellbore environment [45]. The bioreactor was constructed with four
concentric PVC pipes, each 0.76 m tall, and filled with 1 mm nominal quartz sand,
creating a combined pore space and reservoir volume of approximately 40 L. Over an 8-
day experimental period, biofilm was cultivated in the reactor sand-pack with a
102
continuously recirculating flow of substrate. Measured NMR T2ML shifted from
approximately 750 to 400 ms, indicating that the pore environment and bulk fluid
properties were changing due to biofilm growth (Figure 4.7). Destructive sampling
employing microbial population analysis and microscopy confirmed biofilm formation.
This experiment demonstrated that the NMR logging tool can detect small to moderate
changes in T2 distribution associated with environmentally relevant quantities of biofilm
in quartz sand.
Figure 4.7. Time evolution of CPMG signal decay curves and T2 distributions. Increased
signal attenuation with biofilm formation produces steeper CPMG signal decay curves
(top panel). The distribution of T2 relaxation times (bottom panel) shifted to faster decay
times as biofilm grew in the reactor. Data collected days 3 and 4 overlap the day 5 curve
and are not shown. Likewise, the day 6 data is obscured by the day 7 curve and is not
shown. Day 8 data was collected from the drained reactor. Reproduced from Kirkland et
al. (2015b).
In 2014, two Javelin toosl were used to measure biofilm accumulation in an
engineered test cell [46]. The test cell is 55 m by 40 m at the surface and is 6 m deep
103
with 2:1 side slopes (x:z). Two measurement wells were used in the study. A Javelin
probe operating at approximately 400 kHz was installed in Well 1 where the soil profile
was slightly more coarse grained than in Well 2. The Javelin probe in Well 2 operated at
approximately 275 kHZ and was the same tool as was used in the previous laboratory
study. Measured T2ML relaxation times were reduced by 43% and 62%, respectively in
Well 1 and Well 2, while biofilm was cultivated in the soil surrounding each well (Figure
4.8).
Figure 4.8. Data showing T2ML measurements from two monitoring wells (triangles –
Well 1 (~400 kHz), squares – Well 2 (~275 kHz)). Inoculation occurred on Day 1.
Growth substrate was injected daily Days 2-10. Days 11-14, the bacteria were starved.
On Day 14, the wells were flushed with high flows of groundwater from the test cell then
a bleach solution was injected to oxidize remaining organics. Day 15 data was collected
after flushing the bleach solution from the wells. Reproduced from Kirkland et al.
(2015a).
104
Differences T2ML between the wells is due to the relatively larger pores in Well 1. The
reduction in T2ML observed in both wells was confirmed to be a result of biofilm
accumulation by bleaching and flushing the wells and observing the NMR signal’s return
to baseline. This result provided evidence of the NMR logging technique as a direct and
non-invasive method to spatio-temporally monitor biofilm accumulation in the
subsurface.
MICP has been widely researched recently due to its relevance for subsurface
engineering applications including sealing leakage pathways and permeability
modification [50, 51]. These applications of MICP are inherently difficult to monitor
non-destructively in time and space. The Javelin probe was used to monitor MICP in the
sand-filled bioreactor described previously, measuring NMR signal amplitude and T2
relaxation over an 8-day experimental period [47]. Following inoculation with the
ureolytic bacteria, Sporosarcina pasteurii, and pulsed injections of urea and calcium
substrate, the NMR measured water content in the reactor decreased to 76% of its initial
value as calcite precipitation displaced pore water. Destructive sampling confirmed final
porosity was approximately 88% of the original value. The overestimation of porosity
reduction by NMR can be attributed primarily to the accumulation of excess CO2 gas in
the reactor as a result of microbial metabolism. Signal decaying faster than the minimum
echo spacing (1.3 ms) may also have reduced the measured NMR water content. T2
relaxation distributions bifurcated from a single mode centered about approximately 650
ms on Day 2 into a fast decaying population (T2 less than 10 ms) and a larger population
with T2 greater than 1000 ms by Day 8 (Figure 4.9). The combination of changes in pore
105
Figure 4.9. Signal decay curves (top) and the corresponding T2 distributions (bottom) are
shown with each curve representing a day. Day 2 occurred during the control period.
Inoculation occurred on Day 3 (not shown). The calcium media injections occurred
between Day 4 – 7. The Day 8 data was collected prior to flushing the reactor with brine
and destructively sampling. Both graphs show fits to the raw data. Reproduced from
Kirkland et al. (2016).
volume and surface mineralogy accounted for the changes in the T2 distributions. In this
system, the change in pore surface mineralogy, from quartz sand containing paramagnetic
impurities to a relatively uniform calcite surface, caused a decrease in the surface
relaxivity, ρ, and lead to the longer T2 relaxation times observed. The authors attribute
the very fast decaying signal to water trapped within the pores of the calcite itself, where
the influence of the surface to volume ratio dominated. These results indicate the low-
field NMR well-logging probe is sensitive to the physical and chemical changes caused
by MICP in a laboratory bioreactor.
106
Outlook and Conclusions
Several technological challenges remain in the further development of low-field
borehole NMR tools. As discussed above, minimizing the echo spacing, tE, is critical for
obtaining an accurate estimation of k and K from NMR data by reducing the influence of
relaxation due to diffusion. Reducing the minimum tE also expands the range of fast-
decaying signals that can be measured by the instrument, a feature which may be
particularly relevant where biogeochemical conversions are concerned.
Pulsed NMR tools have been used in the oil and gas industry since the early
1990’s, allowing for measurements that simultaneously encode the signal for the effective
diffusion coefficient of the pore fluid, D, and T2 relaxation. The correlation of relaxation
behaviour with diffusion properties allows the separation of some of the effects which
combine to influence T2 relaxation [52]. The ability to collect D-T2 correlations
improves the feasibility of using borehole NMR tools to detect degradation of organic
contaminants by allowing the separation of the water and hydrocarbon signal, for
example, since water and hydrocarbons have different self-diffusion coefficients [53].
The latest versions of the Javelin and Dart borehole NMR probes can now encode for
effective self-diffusion and record D-T2 correlations though the corroborating research is
not yet published. While it is often possible to neglect the influence of T2D for simple
relaxation measurements, the addition of diffusion measurements, however, would make
accurate estimation of the internal field gradient strength critical. Measuring the effective
diffusion coefficient requires a long echo spacing to encode for motion. The influence of
107
T2D over this interval must be known. Accurate estimation of these internal field
gradients under field conditions remains a challenge [26].
In addition to the research discussed above, which used borehole NMR probes,
several studies have been completed using traditional benchtop NMR instruments in the
laboratory where the logical future application is in the unconsolidated sediments in the
near surface. One such example is using borehole NMR as a probe of the redox
condition of the aquifer. Injecting oxygen-rich water into a reduced aquifer can cause
leaching of metals into the water, significantly degrading the water quality. Furthermore,
iron (hydr)oxides are highly reactive geochemically and, as such, can be used to
sequester or transform organic or inorganic contaminants. Both of these examples
involve a change in local redox conditions that changes the mineralogical form of iron
compounds [54], or other metals. NMR is sensitive not only to the quantity of iron
present in the measurement region [55], but also to the form of the iron with Fe (III)
producing faster relaxation rates than Fe(II) compounds [30, 56, 57]. Confirmation of
these laboratory results in a complex field environment would broaden the range of
potential applications of borehole NMR technology and vastly improve existing methods
for monitoring changing redox conditions and iron mineralization in the subsurface.
As the diverse studies described here show, low field borehole NMR is a
remarkably versatile technology with great potential not only to describe heterogeneity in
the near subsurface non-destructively, but also to monitor physical and chemical changes
in the pore scale environment over time, of importance in a range of yet unexplored
applications.
108
References
1. Maliva, R.G., E.A. Clayton, and T.M. Missimer, Application of advanced
borehole geophysical logging to managed aquifer recharge investigations.
Hydrogeology Journal, 2009. 17(6): p. 1547-1556.
2. Careghini, A., S. Saponaro, and E. Sezenna, Biobarriers for groundwater
treatment: a review. Water Science and Technology, 2013. 67(3): p. 453-468.
3. Phillips, A.J., et al., Fracture Sealing with Microbially-Induced Calcium
Carbonate Precipitation: A Field Study. Environmental Science & Technology,
2016. 50(7): p. 4111-4117.
4. DeJong, J.T., et al., Bio-mediated soil improvement. Ecological Engineering,
2010. 36(2): p. 197-210.
5. Dejong, J.T., et al., Biogeochemical processes and geotechnical applications:
progress, opportunities and challenges. Geotechnique, 2013. 63(4): p. 287-301.
6. Robinson, D.A., et al., Soil moisture measurement for ecological and
hydrological watershed-scale observatories: A review. Vadose Zone Journal,
2008. 7(1): p. 358-389.
7. Binley, A., et al., The emergence of hydrogeophysics for improved understanding
of subsurface processes over multiple scales. Water Resources Research, 2015.
51(6): p. 3837-3866.
8. Blumich, B., et al., The NMR-mouse: Construction, excitation, and applications.
Magnetic Resonance Imaging, 1998. 16(5-6): p. 479-484.
9. Goga, N.O., et al., Mobile NMR: applications to materials and biomedicine.
Journal of Optoelectronics and Advanced Materials, 2006. 8(4): p. 1430-1434.
10. Blumich, B., J. Perlo, and F. Casanova, Mobile single-sided NMR. Progress in
Nuclear Magnetic Resonance Spectroscopy, 2008. 52(4): p. 197-269.
11. Behroozmand, A.A., K. Keating, and E. Auken, A Review of the Principles and
Applications of the NMR Technique for Near-Surface Characterization. Surveys
in Geophysics, 2015. 36(1): p. 27-85.
12. Hertrich, M., Imaging of groundwater with nuclear magnetic resonance. Progress
in Nuclear Magnetic Resonance Spectroscopy, 2008. 53(4): p. 227-248.13.
Coates, G.R., L. Xiao, and M.G. Prammer, NMR Logging: Principles and
Applications. 1999, Halliburton Energy Services: Houston.
109
14. Kleinberg, R.L., Kenyon, W.E., Mitra, P.P., Mechanism of NMR relaxation of
fluids in rock. Journal of Magnetic Resonance Series A, 1994. 108: p. 206-214.
15. Freedman, R., Advances in NMR Logging. Journal of Petroleum Technology,
2006. 58(1): p. 60-66.
16. Song, Y.Q., et al., T-1-T-2 correlation spectra obtained using a fast two-
dimensional Laplace inversion. Journal of Magnetic Resonance, 2002. 154(2): p.
261-268.
17. Venkataramanan, L., Y.Q. Song, and M.D. Hurlimann, Solving Fredholm
integrals of the first kind with tensor product structure in 2 and 2.5 dimensions.
Ieee Transactions on Signal Processing, 2002. 50(5): p. 1017-1026.
18. Prange, M. and Y.-Q. Song, Quantifying uncertainty in NMR T-2 spectra using
Monte Carlo inversion. Journal of Magnetic Resonance, 2009. 196(1): p. 54-60.
19. Walsh, D.O., et al., Javelin: A slim-hole and microhole NMR logging tool. Fast
Times, 2010. 15(3): p. 67-72.
20. Perlo, J., et al., Optimized slim-line logging NMR tool to measure soil moisture in
situ. Journal of Magnetic Resonance, 2013. 233: p. 74-79.
21. Callaghan, P.T., Translational Dynamics & Magnetic Resonance: Principles of
Pulsed Gradient Spin Echo NMR. 2011, New York: Oxford University Press.
22. Dlubac, K., et al., Use of NMR logging to obtain estimates of hydraulic
conductivity in the High Plains aquifer, Nebraska, USA. Water Resources
Research, 2013. 49(4): p. 1871-1886.
23. Carr, H.Y. and E.M. Purcell, Effects of diffusion on free precession in nuclear
magnetic resonance experiments. Physical Review, 1954. 94(3): p. 630-638.
24. Meiboom, S. and D. Gill, Modified spin-echo method for measuring nuclear
relaxation times. Review of Scientific Instruments, 1958. 29(8): p. 688-691.
25. Jaeger, F., et al., Evaluation of 1H NMR relaxometry for the assessment of pore-
size distribution in soil samples. European Journal of Soil Science, 2009. 60(6): p.
1052-1064.
26. Fay, E.L., R.J. Knight, and Y.Q. Song, Investigating internal magnetic field
gradients in aquifer sediments. Geophysics, 2015. 80(3): p. D281-D294.
27. Brownstein, K.R. and C.E. Tarr, Importance of classical diffusion in NMR studies
of water in biological cells. Physical Review A, 1979. 19(6): p. 2446-2453.
110
28. Kleinberg, R.L. and M.A. Horsfield, Transverse relaxation processes in porous
sedimentary rock. Journal of Magnetic Resonance, 1990. 88(1): p. 9-19.
29. Kenyon, W.E. and J.A. Kolleeny, NMR surface relaxivity of calcite with adsorbed
Mn2+. Journal of Colloid and Interface Science, 1995. 170(2): p. 502-514.
30. Keating, K. and R. Knight, A laboratory study of the effect of Fe(II)-bearing
minerals on nuclear magnetic resonance (NMR) relaxation measurements.
Geophysics, 2010. 75(3): p. F71-F82.
31. Kenyon, W.E., et al., A three-part study of NMR longitudinal relaxation
properties of water-saturated sandstones. SPE Formation Evaluation, 1988. 3(3):
p. 622-636.
32. Straley, C., et al., Core analysis by low-field NMR. The Log Analyst, 1997. 38(2):
p. 84-94.
33. Kenyon, W.E., Petrophysical principles of applications of NMR logging. The Log
Analyst, 1997. 38(2): p. 21-43.
34. Timur, A., Pulsed nuclear magnetic resonance studies of porosity, movable fluid,
and permeability of sandstones. Journal of Petroleum Technology, 1969.
21(JUN): p. 775-&.
35. Sucre, O., et al., Low-field NMR logging sensor for measuring hydraulic
parameters of model soils. Journal of Hydrology, 2011. 406(1–2): p. 30-38.
36. Mualem, Y., Hysteretical models for prediction of hydraulic conductivity of
unsaturated porous-media. Transactions-American Geophysical Union, 1976.
57(8): p. 602-603.
37. Van Genuchten, M.T., A closed form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Science Society of America Journal, 1980.
44(5): p. 892-898.
38. Walsh, D., Environmental Geophysics: Portable NMR tools for measuring and
monitoring soil moisture, in Preview. 2015, CSIRO Publishing. p. 41-42.
39. Minsley, B.J., et al., Evidence for nonuniform permafrost degradation after fire in
boreal landscapes. Journal of Geophysical Research: Earth Surface, 2016. 121(2):
p. 320-335.
40. Walsh, D., et al., A small-diameter NMR logging tool for groundwater
investigations. Groundwater, 2013. 51(6): p. 914-926.
111
41. Knight, R., et al., NMR Logging to Estimate Hydraulic Conductivity in
Unconsolidated Aquifers. Groundwater, 2016. 54(1): p. 104-114.
42. Butler, J.J., Jr., et al., Characterizing hydraulic conductivity with the direct-push
permeameter. Ground Water, 2007. 45(4): p. 409-419.
43. Knight, R., et al., Field experiment provides ground truth for surface nuclear
magnetic resonance measurement. Geophysical Research Letters, 2012. 39.
44. Parsekian, A.D., et al., Bootstrap Calibration and Uncertainty Estimation of
Downhole NMR Hydraulic Conductivity Estimates in an Unconsolidated Aquifer.
Groundwater, 2015. 53(1): p. 111-121.
45. Kirkland, C.M., et al., Biofilm Detection in a Model Well-Bore Environment
Using Low-Field NMR. Ground Water Monitoring and Remediation, 2015. 35(4):
p. 36-44.
46. Kirkland, C.M., et al., In Situ Detection of Subsurface Biofilm Using Low-Field
NMR: A Field Study. Environmental Science & Technology, 2015. 49(18): p.
11045-11052.
47. Kirkland, C.M., et al., Detecting microbially-induced calcite precipitation (MICP)
in a model well-bore using downhole low-field NMR. Environmental Science &
Technology, 2016. 10.1021/acs.est.6b04833.
48. Cunningham, A.B., et al., Influence of biofilm accumulation on porous-media
hydrodynamics. Environmental Science & Technology, 1991. 25(7): p. 1305-
1311.
49. Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media
using a low-field magnetic resonance system. Environmental Science &
Technology, 2013. 47(2): p. 987-992.
50. De Muynck, W., N. De Belie, and W. Verstraete, Microbial carbonate
precipitation in construction materials: A review. Ecological Engineering, 2010.
36(2): p. 118-136.
51. Phillips, A.J., et al., Engineered applications of ureolytic biomineralization: a
review. Biofouling, 2013. 29(6): p. 715-733.
52. Song, Y.Q., Recent Progress of Nuclear Magnetic Resonance Applications in
Sandstones and Carbonate Rocks. Vadose Zone Journal, 2010. 9(4): p. 828-834.
53. Fay, E.L. and R.J. Knight, Detecting and quantifying organic contaminants in
sediments with nuclear magnetic resonance. Geophysics, 2016. 81(6): p. EN87.
112
54. Bryar, T.R. and R.J. Knight, Sensitivity of nuclear magnetic resonance relaxation
measurements to changing soil redox conditions. Geophysical Research Letters,
2002. 29(24).
55. Bryar, T.R., C.J. Daughney, and R.J. Knight, Paramagnetic effects of iron(III)
species on nuclear magnetic relaxation of fluid protons in porous media. Journal
of Magnetic Resonance, 2000. 142(1): p. 74-85.
56. Keating, K. and R. Knight, A laboratory study to determine the effect of iron
oxides on proton NMR measurements. Geophysics, 2007. 72(1): p. E27-E32.
57. Keating, K. and R. Knight, A laboratory study of the effect of magnetite on NMR
relaxation rates. Journal of Applied Geophysics, 2008. 66(3-4): p. 188-196.
113
CHAPTER FIVE
BIOFILM DETECTION IN A MODEL WELL-BORE
ENVIRONMENT USING LOW-FIELD NMR
Contribution of Authors and Co-Authors
Manuscript in Chapter 5
Author: Catherine M. Kirkland
Contributions: Helped conceive and implement study design. Collected and analyzed
data. Wrote manuscript. Provided feedback and comments on the manuscript.
Co-Author: Randy Hiebert
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
Co-Author: Adrienne J. Phillips
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
Co-Author: Elliot Grunewald
Contributions: Helped conceive and implement study design. Provided supervision and
oversight on data collection and analysis. Provided feedback and comments on the
manuscript.
Co-Author: David O. Walsh
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
Co-Author: Joseph D. Seymour
Contributions: Provided feedback and comments on the manuscript.
114
Co-Author: Sarah L. Codd
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
115
Manuscript Information Page
Catherine M. Kirkland, Randy Hiebert, Adrienne Phillips, Elliot Grunewald, David O.
Walsh, Joseph D. Seymour, Sarah L. Codd
Groundwater Monitoring and Remediation
Status of Manuscript:
____ Prepared for submission to a peer-reviewed journal
____ Officially submitted to a peer-review journal
____ Accepted by a peer-reviewed journal
_X___ Published in a peer-reviewed journal
Wiley-Blackwell
Vol. 35, No. 4, Pages 36-44, Fall 2015
116
BIOFILM DETECTION IN A MODEL WELL-BORE
ENVIRONMENT USING LOW-FIELD NMR
Abstract
This research addresses the challenges of the lack of non-invasive methods and
poor spatio-temporal resolution associated with monitoring biogeochemical activity
central to bioremediation of subsurface contaminants. Remediation efforts often include
growth of biofilm to contain or degrade chemical contaminants, such as nitrates,
hydrocarbons, heavy metals, and some chlorinated solvents. Previous research indicates
that nuclear magnetic resonance (NMR) is sensitive to the biogeochemical processes of
biofilm accumulation. The current research focuses on developing methods to use low-
cost NMR technology to support in-situ monitoring of biofilm growth and geochemical
remediation processes in the subsurface. Biofilm was grown in a lab-scale radial flow
bioreactor designed to model the near wellbore subsurface environment. The Vista Clara
Javelin NMR logging device, a slim down-the-borehole probe, collected NMR
measurements over the course of eight days while biofilm was cultivated in the sand-
packed reactor. Measured NMR mean log T2 relaxation times decreased from
approximately 710 to 389ms, indicating that the pore environment and bulk fluid
properties were changing due to biofilm growth. Destructive sampling employing drop
plate microbial population analysis and scanning electron and stereoscopic microscopy
confirmed biofilm formation. Our findings demonstrate that the NMR logging tool can
117
detect small to moderate changes in T2 distribution associated with environmentally
relevant quantities of biofilm in quartz sand.
Introduction
In most environments, the majority of microbial cells exist in complex
communities called biofilms [1]. Bacteria preferentially attach to surfaces, anchoring
themselves with a sticky gel called extracellular polymeric substance (EPS). The EPS,
which consists of proteins, polysaccharides, and DNA, forms a heterogeneous matrix
surrounding the bacterial cells and entrapped organic and inorganic molecules [2]. The
EPS matrix has a complex structure, consisting of hydrated gel-like fibers and
interconnected channels [3]. Depending on its physical properties, the EPS matrix can
change soil pore connectivity, effective pore size, and hydraulic conductivity, thereby
affecting the hydrodynamic properties of the porous media [4-6].
Bioremediation methods exploit these biofilm-induced changes in the soil and can
be effective to contain or degrade chemical contaminants in the subsurface [7]. Once
established, biofilm can consume hydrocarbons and other contaminants directly as a
substrate [8-11], induce mineralization to trap contaminants [12], mediate ion exchange
to remove heavy metals from aqueous solution [13] or transform heavy metals to
insoluble forms [14]. In-situ bioremediation applications often require injection of
appropriate nutrients into the subsurface in order to create redox conditions suitable for
biofilm growth and contaminant transformation. In addition to being an integral part of
the removal of toxins, biofilm can act as a bio-barrier to slow the migration of the
118
contamination or re-direct groundwater flow through a treatment zone [15] by reducing
hydraulic conductivity and restricting advective flow. Furthermore, biofilms better
withstand environmental stresses such as nutrient deprivation, changing pH conditions,
and exposure to biocides or antimicrobial substances compared to suspended cells [16,
17], improving their effectiveness and longevity as bio-barriers.
Numerous computer models have been developed in the last several decades in an
effort to better understand the physical and biochemical processes of biofilm
development. In general, the recent models iteratively solve partially decoupled
differential equations governing conservation of momentum, mass transport, substrate
utilization, and biofilm growth [18-21]. The models help describe the changing
hydrodynamics and biofilm morphology, as well as confirm the overall reduced hydraulic
conductivity of the porous media resulting from biofilm accumulation that is observed
experimentally.
This observed reduction in hydraulic conductivity is the typical method for
determining whether biofilm is present in bioremediation applications [5, 22, 23].
Darcy’s Law, ν = −𝐾𝐾 𝑑𝑑ℎ
𝑑𝑑𝑑𝑑
describes the relationship between the specific discharge, ν
[L/t], and the pressure gradient along a length of porous media, dh/dl [dimensionless],
where the constant of proportionality is the saturated hydraulic conductivity, K [L/t].
Biofilm accumulation in pore spaces has been shown to reduce hydraulic conductivity by
approximately 99% in a field setting and up to 99.99% in laboratory experiments [15,
23].
119
In conjunction with hydraulic conductivity measurements, determination of
biofilm presence is often assessed with heterotrophic plate counts (HPC) to quantify the
number of viable cells in the bulk fluid [24]. Interpretation of this type of population data
is not trivial since the cell count data pertains to suspended bacteria rather than those
attached to surfaces as biofilm. Furthermore, how the number of suspended cells
correlates to the number of cells existing within biofilm colonies is not well understood.
To determine cell populations within the biofilm, samples of the porous media must be
collected and analyzed. Direct destructive sampling of contaminated soils during
remediation can be both costly and potentially hazardous. Moreover, cell count
information in general provides no information regarding the in-situ physical
characteristics of the biofilm, such as the thickness, structure, and density of the EPS
component as this information is destroyed during sampling and plating. While these
methods are useful to describe the hydraulic conductivity or biological community at a
particular place and time, they are less useful for informing us about the spatial and
temporal growth, maturation, and decay of the biofilm [25].
NMR provides an alternative method to monitor in-situ biofilm development,
overcoming several of the limitations described above. The measured signal in 1H NMR
comes from coherently precessing hydrogen protons, or ‘spins,’ making NMR applicable
for the study of materials containing water or organic matter, including biofilm [26, 27].
Previous research indicates that NMR is sensitive to the biogeochemical processes of
biofilm growth [28] and, unlike other methods of study, allows for non-invasive and non-
destructive examination of the relationship between biofilm development and porous
120
media hydrodynamics and mass transport over various time and length scales [29-32]. A
fundamental understanding of these related processes is critical to optimize the
effectiveness of bioremediation applications, making NMR methods a useful addition to
the bioremediation monitoring toolkit.
The Vista Clara Javelin device is a slim, down-the-borehole NMR logging tool
(4.5ft (1.37m) long, 3.5in (8.9cm) diameter) originally developed for hydrogeologic
analysis in open or PVC-cased boreholes between 2 and 8in (5-20cm) in diameter [33].
Expanding from this conventional application, the research reported here uses the Javelin
probe to detect changes in the signal relaxation response over time as a measurement of
biofilm growth in a model sand bioreactor since biofilm accumulation enhances signal
relaxation [28]. The method using a NMR logging tool as described in this article may
provide an improved and non-destructive way to assess the subsurface presence of
biofilm.
Theory
T2 relaxation
Following excitation with a radio-frequency (rf) pulse, the induced NMR signal
decays at a rate governed by two relaxation mechanisms—spin-lattice T1 and spin-spin T2
[34]. T1 relaxation is related to the rate at which the net magnetization grows in the
longitudinal direction and returns to thermal equilibrium. T2 relaxation is related to
dipolar interactions in the transverse plane and loss of phase coherence, resulting in
magnetization decay and signal attenuation.
121
The statistical distribution of T2 relaxation times provides information regarding
the physico-chemical environments the protons experience, since environmental factors
like pore fluid type and pore size control how likely protons are to experience dipolar
interactions. Typically, the signal decays more rapidly in solid-like materials due to T2
relaxation—that is, solids and gels have shorter T2 relaxation times than liquids due to
low rotational mobility enhancing dipolar coupling [34]. The T2 relaxation time for pure
water is on the order of 2-3 seconds in unrestricted environments, whereas it is on the
order of milliseconds in porous materials due to restricted motion and surface
interactions.
While both T1 and T2 relaxation times can provide information about the
molecular and pore-scale environment, T2 measurements require less time to conduct and
are preferred. This experimental study uses measured changes in the T2 distribution to
infer changes in the pore scale environment due to biofilm growth within the bioreactor.
Biofilm formation in porous media leads to shorter T2 relaxation times [25, 27].
Porous media influences the NMR relaxation process in several ways. The
presence of solid surfaces creates relaxation sinks and may introduce paramagnetic
impurities, as well as generating susceptibility-induced magnetic field inhomogeneities.
T2 in a pore is also influenced by the viscosity and chemical properties of the bulk fluid.
Furthermore, diffusion of the liquid molecules between pores, and diffusion within the
pore due to local magnetic field gradients also affect the T2 relaxation behavior of a
porous media system. The expression for the rate of T2 relaxation in porous media
122
includes the relaxation time of the bulk media, T2B, the surface relaxation time, T2S, and
relaxation time due to diffusion, T2D [35]. 1
𝑇𝑇2
= 1
𝑇𝑇2,𝐵𝐵 + 1𝑇𝑇2,𝑆𝑆 + 1𝑇𝑇2,𝐷𝐷 (5.1)
When the bulk fluid is water, 1/T2B is typically very small and is often neglected in
literature on relaxometry in porous materials. However, as the rotational mobility of the
bulk fluid decreases due to biofilm growth in the pore space, 1/T2B increases in
importance to the overall T2 relaxation rate [36]. Surface relaxation, T2S, depends on the
surface relaxivity, ρ, which describes the efficiency of the mineral surface to enhance
relaxation, and the surface to volume ratio, S/V [37].
1
𝑇𝑇2,𝑆𝑆 = 𝜌𝜌 𝑆𝑆𝑉𝑉 (5.2)
Since biofilm is composed of bacterial cells in a hydrated gel matrix, the T2
relaxation time of fluid contained in the biofilm is shorter than for the bulk fluid alone.
As the biofilm grows into the pore spaces, proportionally more protons will be bound in,
or interact with, the EPS matrix, thereby shifting the T2 distribution toward shorter
relaxation times. Enhanced T2 relaxation due to biofilm formation is expected to be
caused by reduced rotational mobility in the pore fluid as the EPS matrix contributes a
gel phase and the cells produce biomacromolecules with shorter relaxation than bulk
water (T2B). Additional relaxation effects may be due to changes in the pore structure and
surface due to the biofilm growth on the grain surfaces (T2S) [38].
T2D is directly related to the echo spacing, tE, according to
123
1
𝑇𝑇2,𝐷𝐷 = 𝐷𝐷 (𝛾𝛾𝛾𝛾𝑡𝑡𝐸𝐸)212 (5.3)
where D is the diffusion coefficient [L2 t-1], γ is the gyromagnetic ratio of hydrogen,
2.675 x 108 rad/(T s), and G is the local effective magnetic field gradient [Gauss L-1]. By
selecting a short tE, this term can be made sufficiently small as to be neglected.
Repeating the experiment with a longer tE indicates whether there are indeed enhanced
relaxation effects due to diffusion to consider.
Materials and Methods
Bioreactor Design and Construction
The lab-scale bioreactor was designed to model the near-wellbore environment
and was constructed inside an aluminum Faraday cage using four concentric PVC pipe
sections (Titan Industries, Paxton, NE), ranging from the inner 8in (20cm) diameter solid
pipe to the outer 20in (51cm) solid pipe (Figure 1). The two middle pipe sections, 10in
(25cm) and 18in (46cm) diameter respectively, were screened over the entire length with
8 rows of 0.020in (0.051cm) wide slots spaced at 0.5in (1.3cm) intervals. All of the pipes
were 2.5ft (76cm) tall. Top and bottom plates were constructed from 0.5in (1.3cm) high
density rigid expanded PVC sheeting with machined 0.125in (0.32cm) concentric
grooves to allow nesting and sealing of the vertical pipe sections.
124
Figure 5.1. Model well-bore bioreactor. The 3.5in (8.9cm) diameter NMR logging probe
was installed in the lab-scale radial flow bioreactor and Faraday cage. The purpose of the
Faraday cage was to reduce external radio frequency noise caused by various
electromagnetic sources in the laboratory.
The 3in (7.6cm) annulus between the slotted pipes was wet-filled with
approximately 2.3ft3 (0.065m3.) of No. 2095 Granusil® silica (quartz) sand (Unimin
Corp., Ottowa, MN) having a nominal diameter of 1 mm and a porosity of approximately
0.37. The combined volume of the pore space and annular reservoirs was approximately
40L.
The reactor was disinfected with two pore volumes (80L) of 10% bleach solution
with 7g/L Tween80 (Fisher BioReagents, Waltham, MA) followed by a flush with 80L
125
sterile de-ionized (DI) water. Two pore volumes (80L) of sterile 2.52g/L sodium
thiosulfate solution was pumped into the reactor to neutralize any remaining bleach. The
sodium thiosulfate solution was flushed from the reactor with 80L of sterile buffer
solution consisting of 3g/L NaNO3, 0.12g/L K2HPO4, and 0.04g/L KH2PO4.
NMR measurements were made in the buffer-filled reactor with the NMR logging
tool to establish a baseline T2 distribution for the reactor prior to any biofilm growth.
Later measurements were compared to this initial condition to identify changes in T2
relaxation time distributions. Experimental constraints prevented the testing of an un-
inoculated control in parallel with measurements of the biofilm reactor. These
constraints are primarily associated with the uncertainties that would have been
introduced by moving the NMR probe between the reactors to conduct measurements.
Moving the probe during the experiment could influence the results since a slightly
different shell within the reactor would be excited and measured for each placement of
the probe. For these reasons, a longitudinal study of a single reactor was the preferred
method for this initial experiment.
Bacterial Culture
One mL frozen stock of Bacillus mojavensis was cultured in 100mL of modified
Brain-Heart Infusion (BHI) broth (36g/L Brain Heart Infusion (BHI) (Becton, Dickinson
and Co., Sparks, MD), 3g/L NaNO3, 0.75g/L NH4Cl, 40g/L NaCl) on a shaker table at
150rpm for 48 hours. Then the 100mL inoculum was added to 10L of the modified BHI
broth and mixed on a stir plate at 1150rpm for 60 hours to a concentration of
126
approximately 7log10 colony-forming units (cfu)/mL. B. mojavensis was selected because
of the abundant mucoid biofilm it produces [16, 28].
The 10L inoculum was pumped into the reactor with two peristaltic pumps (Cole-
Parmer Model 7553-80 with Masterflex Easy-Load 7515-00 heads) at 650-700 mL/min
through eight lines of No. 16 vinyl tubing. The pumps were turned off for approximately
11 hours to allow the bacteria to attach to the sand. Following the attachment period,
radial flow of substrate through the sand promoted bacterial growth and biofilm
accumulation which was monitored by the NMR probe.
No attempt was made to maintain a B. mojavensis monoculture within the reactor
for the duration of the experiment. Following inoculation, subsequent batches of substrate
were mixed and pumped into the reactor in a non-sterile manner. Sterile conditions could
not be maintained due to experimental constraints imposed by the large volumes of
substrate used and waste produced daily.
Following inoculation, 40L of re-circulating substrate was removed daily from the
system and replaced with 40L of fresh substrate. Except during no-flow NMR
measurements, the substrate was re-circulated continuously at a rate of 650-700mL/min
which translates to a specific discharge of approximately 0.04cm/min and a pore velocity
of approximately 0.02mm/s at the radial center of the sand annulus. This flowrate
corresponds to the pore volume of the reactor being replaced every 50-60 minutes.
127
NMR Data Acquisition
T2 relaxation measurements used a CPMG pulse sequence [39, 40] with excitation
pulses of either 245 kHz or 290 kHz corresponding to excitation regions at 19cm and
17cm (7.5in and 7in) from the center of the probe respectively. Each excitation shell is
0.5m high and several millimeters thick. Eight experiments were repeated at the same
time daily (Table 5.1). Each experiment used a repetition time, Tr, of 5s to allow for T1
relaxation.
Observing changes in the T2 measurement as a function of the echo spacing tE can
provide an indication of the significance of NMR diffusion relaxation effects.
Measurements were made under flow and no flow conditions to determine if the fluid
flow affected the T2 measurement. Data was also collected under two probe tuning
protocols. The first held the NMR excitation frequency constant for each experiment; the
second allowed the NMR instrument to auto-tune with an adjustable excitation frequency
periodically within each experiment. This allowed a determination of whether changes in
probe tuning affected the T2 measurement.
128
Exp. No. Averages tE (ms) Flow Excitation Frequency Duration
1 360 6.0 No Auto-tuned 30 min
2 360 6.0 Yes Auto-tuned 30 min
3 360 1.5 No Constant 30 min
4 360 1.5 Yes Constant 30 min
5 360 6.0 No Constant 30 min
6 360 6.0 Yes Constant 30 min
7 360 1.5 No Auto-tuned 30 min
8 2880 1.5 Yes Auto-tuned 4 hours
Table 5.1. NMR experiments with well-logging probe.
The data processing software uses the inverse Laplace transform to generate a T2
distribution for the reactor excitation shell. Because of the generally low signal to noise
ratio that is typical for a low-field NMR device in natural geologic material, the Javelin
data interpretation software also calculates a noise-robust parameter based on the mean
amplitude of the echoes in the record CPMG. The Square of Echoes (SOE) is calculated
as the squared value of the mean of echoes. A reduction in the calculated SOE over the
course of the experimental period indicates that the mean value of the T2 relaxation time
is decreasing as the conductivity of the porous media in the reactor decreases.
129
Sampling and Imaging
A sample of re-circulating substrate was collected daily during Experiment 6
(Table 5.1) and immediately serially diluted and plated [41] to estimate the quantity of
viable cells in the reactor bulk fluid. While the suspended cell counts are not directly
indicative of biofilm formation, or more specifically of EPS production, they provide a
general measure of growth conditions and cell population within the reactor.
On day 8, the reactor was drained and NMR measurements were acquired using
Experiment 7 parameters (Table 5.1). Following the measurements, the bioreactor was
destructively sampled to confirm biofilm accumulation in field-relevant quantities. The
sand was removed from the reactor in layers, with samples collected in sterile, pre-
weighed 15mL Falcon tubes for imaging and cell population analysis. The samples were
collected from the NMR probe’s sensitive region in the annulus center at three depths—
top (6in (15cm) deep from the top of pipe), middle (12-15in (30 – 38 cm) deep) and
bottom (28in (71cm) deep). Six samples were collected for both population analysis
with the drop plate method and imaging with stereoscopic and scanning electron
microscopy.
Ten mL of sterile phosphate buffered saline (PBS) solution was added to each
sample of sand collected for population analysis. The samples were vortexed with a
Thermolyne MaxiMix II Type 37600 mixer for 30 seconds each, then sonicated for 2
minutes (Tutthauer CSU-3). Prior to serial dilution and plating, each sample was again
vortexed for 30 seconds to remove attached bacteria from the sand particles and break up
clumps of cells. Following plating, the buffer solution was poured off and the sand
130
samples were dried at 65°C for until their masses stabilized. The mass of dry sand and
the plate count from each sample indicate cfu/g sand in the reactor.
For stereoscope microscopy analysis, biofouled sand samples from the radial flow
reactor and clean control sand were stained with 300μL of 40μM Syto 9 (Molecular
Probes, NY USA) for 30 minutes. The samples were rinsed with 0.2μm filtered (VWR,
NJ, USA) distilled water and immediately imaged at 470nm excitation with a Niko SMZ
1500 (Nikon, NY, USA) stereoscope equipped with an EXFO Xcite 120 fluorescence
illumination system.
Additionally, images were acquired using a Zeiss Supra 55VP scanning electron
microscope (Zeiss, USA). Sand and biofilm from the radial flow reactor and control sand
samples were sputter coated with iridium and high-resolution images were taken at 1.0kV
at a working distance of 3-4mm.
Results and Discussion
The experimental NMR data shows a shift in the T2 distribution to faster decay
times, indicating that the fluid properties and pore environment changed due to biofilm
growth (Figure 2). No significant relaxation enhancement due to diffusion, flow, or
changes in probe tuning was observed. Therefore, unless otherwise noted, the data
presented was collected during Experiment 8 where the level of background noise was
lowest due to the high number of averages. The mean log T2 relaxation time decreased
from approximately 710ms following the buffer pulse to approximately 390ms on day 7.
131
Figure 2 shows the signal attenuation (top) and T2 distributions (bottom) measured during
the course of testing.
Day 1 measurements were conducted in the disinfected sand pack saturated with a
phosphate buffer solution using Experiment 7 parameters and 4560 averages.
Measurements taken during the inoculation of the reactor with B. mojavensis occurred at
day 1.5. Experiments conducted on day 2 represent the first potential measurement of
attached biofilm within the reactor. The signal decay curves and T2 distributions collected
days 3-5 show significant overlap.The data collected on days 6 and 7 likewise overlap
and are shown in Figure 2 as a thick solid line. The dash-dot curve in both panels shows
the data obtained on day 8 from the drained reactor (Experiment 7).
132
Figure 5.2. Time evolution of CPMG signal decay curves and T2 distributions. Increased
signal attenuation with biofilm formation produces steeper CPMG signal decay curves
(top panel). The distribution of T2 relaxation times (bottom panel) shifted to faster decay
times as biofilm grew in the reactor. Data collected days 3 and 4 overlap the day 5 curve
and are not shown. Likewise, the day 6 data is obscured by the day 7 curve and is not
shown. Day 8 data was collected from the drained reactor.
By day 2, the CPMG signal decay curve is notably steeper and the T2 distribution
shows a narrower peak, centered at approximately 450ms (Figure 2). These data indicate
an increase in the proportion of protons experiencing faster relaxation as would be
expected in the event of biofilm formation. The rate of change of the signal response
decreased after day 2, with little change recorded in the signal days 3-5. The day 6 and 7
data show a decreased signal amplitude in the top panel relative to day 5, indicating a
decrease in the measured reactor water content. The NMR measurement from the drained
reactor on day 8 reflects the presence of biofilm, as seen in the residual peak between
133
200—300ms. The smaller peak at approximately 7ms most likely originates from water
or biofilm that is directly bound on and interacting with the sand surface. The lower
amplitude of the signal from the drained reactor reflects the desaturation of the sandpack.
The measured reactor water content decreased from 37.3% at inoculation to
32.2% on day 7. After draining the reactor on day 8, the measured water content was
7.0% and the mean log T2 relaxation time was 121ms. The decrease in measured water
content during the growth phase is best explained by microbial gas production within the
pores as a metabolic by-product. This biologically-driven reduction in water content of
the reactor sandpack may have contributed to shorter T2 relaxation times by increasing
the effective saturated surface area to fluid volume ratio (S/V) in Eqn. 2. However, the
magnitude of the T2 shift between inoculation and day 7 significantly exceeds that from
between day 7 and day 8 after the reactor was drained (Figure 2). The reduction in mean
log T2 relaxation time by day 7 was approximately 320ms corresponding to a water
content decrease of 5.1 percentage points. The 25 percentage point reduction in water
content after draining the reactor produced only a further 270ms reduction in the mean
log T2 relaxation time. T2 relaxation times decreased nearly six times more on a water
content basis [ms/% water] during the growth phase as compared to when the reactor was
drained. This suggests that decrease in water content in the reactor during the experiment
was likely not the primary mechanism causing the observed shift in the T2 relaxation
distribution. The best explanation for the observed decrease in T2 relaxation times is
biofilm accumulation in the reactor sandpack.
134
Figure 5.3. SOE Reduction. Data for the set of long-average experiments shows a
decrease of approximately 53% in the squared value of the mean of echoes over the
duration of the experiment.
The measured SOE, normalized with respect to water content squared, decreased
53% between day 1 and day 7 (Figure 3). The normalized SOE value shows the same
trends as observed in the signal attenuation and T2 distribution data in Figure 2, where the
initial change is most significant followed by stabilization in the signal response.
HPC of the daily samples show an initial drop in cfu numbers after inoculation,
followed by a steady increase over the duration of the experiment (Figure 4). From an
inoculum concentration on the order of 7log10cfu/mL injected into the reactor on day 1.5,
the day 2 HPC yielded colony counts of approximately 5.5log10cfu/mL. Combined with
the NMR results showing a decrease in T2 relaxation times over the same period, this
135
decline in HPC can be attributed to bacterial attachment and initial biofilm formation on
the sand particles. Planktonic cells increased in number to approximately 6.5log10cfu/mL
on days 3 and 4. These higher colony counts correspond to enhanced signal attenuation
and decreases in the mean log T2 relaxation time, though they are necessarily not a
measurement of the cell numbers in the biofilm. The increase in SOE observed around
day 5 may correspond to a biofilm sloughing event since heterotrophic plate counts from
day 5 were nearly an order of magnitude higher than those from day 4, 7.4log10 versus
6.5log10 cfu/mL, respectively. After day 5, a slower rate of increase in colony numbers
was observed than between days 4 and 5, suggesting a secondary attachment period may
have occurred. This interpretation is supported by the NMR data which shows a decrease
in the normalized SOE over the same period. In general, the HPC data and the diverse
colony morphologies indicate that conditions in the bioreactor were favorable for mixed
population bacterial growth.
136
Figure 5.4. Heterotrophic Plate Counts (HPC). Daily HPC data shows a strong initial
attachment of cells from approximately 7log10cfu/mL in the 10L inoculum to
5.5log10cfu/mL on day 2. Increasing colony numbers resulted in a final population of
approximately 8log10cfu/mL by day 8. The average biofilm HPC from the sand samples
was approximately 8.5log10 cfu/g sand, shown on the secondary axis.
When the bioreactor was drained and destructively sampled, the sand was sticky,
consistent with presence of a biofilm bridging the pore spaces of the sand. Cell
population analysis of sand samples indicates that attached bacterial concentrations in the
bioreactor were approximately 7log10 to 9log10cfu/g dry sand, with an average in the
reactor of 8.5log10cfu/g sand. This concentration is similar to cell counts noted in the
literature for subsurface biofilms used to degrade trace organics, where populations are
typically in the range of 6log10 to 9log10 cfu/g soil [15, 26].
137
Figure 5.5. Porous media microscopy confirmed biofilm accumulation in the region of
the reactor where NMR data was collected. a) Scanning electron microscope (SEM)
image of mixed population bacterial cells and extracellular polymeric substance (EPS) on
a sand particle collected from the middle depth of the reactor. Region 1 shows cocci
bacteria while Region 2 shows rod-like cells. Bacillus mojavensis is a rod-shaped
bacteria. Depth 15 in. Scale bar is 2 µm. b) Stereoscope image of EPS encasing the
bacterial cells and binding grains of 1mm sand, Depth 15 in. 40X magnification.
138
SEM images (Figure 5.5) showed a mixed bacterial population of rod and cocci
cells, with thicker EPS at the bottom of the reactor where the fresh nutrient supply
entered the reactor. Compared to samples collected in the middle and upper regions of
the reactor (Figure 5a), individual cells were less distinguishable on samples collected
from the bottom due to the thicker EPS matrix (not shown). Stereoscopic microscopy
qualitatively confirmed the presence of biofilm on and between sand particles (Figure
5b), where the matrix fluoresces in green and the cells are visible as bright flecks.
Microscopy confirms bacterial growth and EPS production within the reactor.
Population analysis indicates both quantity and diversity of heterotrophic bacteria
present without directly accounting for the presence of EPS. Microscopy of the sand can
confirm the presence of the EPS matrix, but does little to elucidate the in-situ
characteristics and hydrodynamics of the biofouled pore space. For these reasons,
characterizations of observed biofilm accumulation are qualitative and indicative of EPS
formation within the NMR logging tool’s sensitive region but are not intended to quantify
the total amount of EPS formed in the reactor.
Several studies have demonstrated that biofilm growth in soils and rock reduces
the hydraulic conductivity and permeability of the media, often by several orders of
magnitude [5, 15, 42, 43]. Furthermore, biofilm itself is a porous media with an internal
architecture that includes voids and channels [44-46]. The enhanced NMR relaxation
observed in this study could be due to the changing surface properties if the biofilm coats
the porous media, represented as an increase in 1/T2S. However, if the biopolymers are
distributed throughout the pore volume then the faster relaxation could be dominated by
139
exchange of protons between the water and the biopolymers, which would be represented
as an increase in 1/T2B.
NMR studies on T2 decay in porous media typically focus on the influence of T2S
and T2D on the overall relaxation behavior of the medium [47-50]. When the bulk fluid in
the pore space changes, however, as it does during biofilm formation, the influence of T2B
can no longer be neglected. In the bulk phase, the production of polymer-like EPS and
biomacromolecules will enhance spin relaxation as protons bound to the polymers and
dissolved organics rapidly exchange with free protons on the liquid molecules [51]. This
relaxation mechanism produces shorter relaxation times than would be expected in bulk
water. As the content of EPS and biomacromolecules increases within the pore space
relative to the initial condition, the weighted average of ‘bound’ and ‘free’ proton T2
times will shift in the direction of the polymer relaxation time to shorter T2 decay times
[3, 38].
From a macroscopic perspective, biofilm may act as a relatively impermeable
surface, insofar as it is restrictive to convective flow within pores. At the molecular
level, however, biofilm is composed of approximately 97% water [52]. For these
reasons, we attribute the change in relaxation behavior observed in this experiment
primarily to changes in the bulk relaxation of the pore fluid, which includes both a liquid
and polymer-gel phase. In these experiments it is proposed that an increase in 1/T2B is
the best way to conceptualize the observed enhancement in the NMR T2 relaxation.
These experiments demonstrate that a commercially available low-cost NMR
logging tool can detect small to moderate changes in T2 distribution due to biofilm
140
accumulation in quartz sand under laboratory conditions. Mean log T2 relaxation times
decreased from 710ms when the reactor contained a sterile buffer solution to 389ms
following 7 days of biofilm cultivation. Over the same period, the noise-robust SOE
parameter decreased by 53%, indicating an increase in signal attenuation and shorter T2
relaxation times. Normalizing the data with respect to water content squared in the
reactor indicates that the change in T2 is not primarily driven by the saturation state of the
porous media in the reactor, but rather by biofilm accumulation. Heterotrophic plate
counts suggest that planktonic cells in the reactor increased in abundance from
approximately 5.5log10cfu/mL on day 2 to 8log10cfu/mL on day 8. Population data from
destructive sampling confirmed colony counts of attached cells on the order of
8log10cfu/g dry sand. Scanning electron and stereoscope microscopy confirmed the
presence of the biofilm matrix attached to the sand particles compared to control sand.
Our results, therefore, show that the measured shift in the reactor T2 relaxation
distribution toward faster signal decay times is best explained by biofilm accumulation in
the reactor pore spaces.
Our findings have the potential to improve monitoring methods for
bioremediation applications involving the use of bio-barriers to slow the flow of
groundwater or consume chemical contaminants. Deploying an array of the NMR probes
in wells in the bio-barrier zone and recording signal changes over time could provide
another measure of the robustness of the bio-barrier with respect to depth. Preparations
are underway to use the NMR logging tool in situ to detect biofilm formation at an
engineered field site in Butte, Montana. Considerations for this future work include the
141
effects of moving the probe between measurements and potential influences of soil
mineralogy and water chemical parameters that would likely be encountered in a natural
setting.
142
References
1. Sutherland, I.W., Biofilm exopolysaccharides: a strong and sticky framework.
Microbiology, 2001. 147: p. 3-9.
2. Characklis, W.G. and K.C. Marshall, Eds., Biofilms. 1990, New York: John Wiley
& Sons, Inc.
3. Vogt, S.J., et al., Permeability of a growing biofilm in a porous media fluid flow
analyzed by magnetic resonance displacement-relaxation correlations.
Biotechnology and Bioengineering, 2013. 110(5): p. 1366-1375.
4. Seymour, J.D., et al., Anomalous fluid transport in porous media induced by
biofilm growth. Physical Review Letters, 2004. 93(19).
5. Cunningham, A.B., et al., Influence of biofilm accumulation on porous-media
hydrodynamics. Environmental Science & Technology, 1991. 25(7): p. 1305-
1311.
6. Drescher, K., et al., Biofilm streamers cause catastrophic disruption of flow with
consequences for environmental and medical systems. Proceedings of the
National Academy of Sciences of the United States of America, 2013. 110(11): p.
4345-4350.
7. LaGrega, M.D., P.L. Buckingham, and J.C. Evans, Hazardous Waste
Management. 2nd ed. 2001, New York: McGraw-Hill.
8. Kao, C.M., et al., Biobarrier system for remediation of TCE-contaminated
aquifers. Bulletin of Environmental Contamination and Toxicology, 2004. 72(1):
p. 87-93.
9. Kao, C.M., et al., Enhanced PCE dechlorination by biobarrier systems under
different redox conditions. Water Research, 2003. 37(20): p. 4885-4894.
10. Kim, G., S. Lee, and Y. Kim, Subsurface biobarrier formation by microorganism
injection for contaminant plume control. Journal of Bioscience and
Bioengineering, 2006. 101(2): p. 142-148.
11. Komlos, J., et al., Biofilm barriers to contain and degrade dissolved
trichloroethylene. Environmental Progress, 2004. 23(1): p. 69-77.
12. Phillips, A.J., et al., Engineered applications of ureolytic biomineralization: a
review. Biofouling, 2013. 29(6): p. 715-733.
143
13. von der Schulenburg, D.A.G., et al., Spatially resolved quantification of metal ion
concentration in a biofilm-mediated ion exchanger. Biotechnology and
Bioengineering, 2008. 99(4): p. 821-829.
14. Pal, A. and A.K. Paul, Microbial extracellular polymeric substances: central
elements in heavy metal bioremediation. Indian Journal of Microbiology, 2008.
48(1): p. 49-64.
15. Cunningham, A.B., et al., Subsurface biofilm barriers for the containment and
remediation of contaminated groundwater. Bioremediation Journal, 2003. 7(3-4):
p. 151-164.
16. Mitchell, A.C., et al., Resilience of planktonic and biofilm cultures to
supercritical CO2. Journal of Supercritical Fluids, 2008. 47(2): p. 318-325.
17. Petrova, O.E. and K. Sauer, Sticky Situations: Key Components That Control
Bacterial Surface Attachment. Journal of Bacteriology, 2012. 194(10): p. 2413-
2425.
18. Kapellos, G.E., T.S. Alexiou, and A.C. Payatakes, Hierarchical simulator of
biofilm growth and dynamics in granular porous materials. Advances in Water
Resources, 2007. 30(6–7): p. 1648-1667.
19. Pintelon, T.R.R., et al., The effect of biofilm permeability on bio-clogging of
porous media. Biotechnology and Bioengineering, 2012. 109(4): p. 1031-1042.
20. von der Schulenburg, D.A.G., et al., Three-Dimensional Simulations of Biofilm
Growth in Porous Media. AIChE Journal, 2009. 55(2): p. 494-504.
21. Thullner, M. and P. Baveye, Computational pore network modeling of the
influence of biofilm permeability on bioclogging in porous media. Biotechnology
and Bioengineering, 2008. 99(6): p. 1337-1351.
22. Cunningham, A.B., E.J. Bouwer, and W.G. Characklis, Biofilms in porous media,
in Biofilms, W.G. Characklis and K.C. Marshall, Editors. 1990, John Wiley &
Sons, Inc.: New York. p. 697-732.
23. Stewart, T.L. and H.S. Fogler, Biomass plug development and propagation in
porous media. Biotechnology and Bioengineering, 2001. 72(3): p. 353-363.
24. Lazarova, V. and J. Manem, Biofilm characterization and activity analysis in
water and waste-water treatment. Water Research, 1995. 29(10): p. 2227-2245.
25. Bayer, J.V., F. Jaeger, and G.E. Schaumann, Proton Nuclear Magnetic Resonance
(NMR) Relaxometry in Soil Science Applications. Open Magnetic Resonance
Journal, 2010. 3: p. 15-26.
144
26. Potter, K., et al., Assay for bacteria in porous media by diffusion-weighted NMR.
Journal of Magnetic Resonance. Series B, 1996. 113: p. 9-15.
27. Hoskins, B.C., et al., Selective imaging of biofilms in porous media by NMR
relaxation. Journal of Magnetic Resonance, 1999. 139(1): p. 67-73.
28. Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media
using a low-field magnetic resonance system. Environmental Science &
Technology, 2013. 47(2): p. 987-992.
29. Seymour, J.D., et al., Magnetic resonance microscopy of biofouling induced scale
dependent transport in porous media. Advances in Water Resources, 2007. 30(6-
7): p. 1408-1420.
30. Seymour, J.D., et al., Magnetic resonance microscopy of biofilm structure and
impact on transport in a capillary bioreactor. Journal of Magnetic Resonance,
2004. 167(322-327).
31. Van As, H. and P. Lens, Use of H-1 NMR to study transport processes in porous
biosystems. Journal of Industrial Microbiology & Biotechnology, 2001. 26(1-2):
p. 43-52.
32. Gjersing, E.L., et al., Magnetic resonance microscopy analysis of advective
transport in a biofilm reactor. Biotechnology and Bioengineering, 2005. 89(7): p.
822-834.
33. Walsh, D., et al., A small-diameter NMR logging tool for groundwater
investigations. Groundwater, 2013. 51(6): p. 914-926.
34. Callaghan, P.T., Principles of Nuclear Magnetic Resonance Microscopy. 1991,
New York: Oxford University Press.
35. Grunewald, E. and R. Knight, A laboratory study of NMR relaxation times in
unconsolidated heterogeneous sediments. Geophysics, 2011. 76(4): p. G73-G83.
36. Kleinberg, R.L. and M.A. Horsfield, Transverse relaxation processes in porous
sedimentary rock. Journal of Magnetic Resonance, 1990. 88(1): p. 9-19.
37. Brownstein, K.R. and C.E. Tarr, Importance of classical diffusion in NMR studies
of water in biological cells. Physical Review A, 1979. 19(6): p. 2446-2453.
38. Codd, S.L., et al., NMR relaxation measurements of biofouling in model and
geological porous media. Organic Geochemistry, 2011. 42(8): p. 965-971.
39. Carr, H.Y. and E.M. Purcell, Effects of diffusion on free precession in nuclear
magnetic resonance experiments. Physical Review, 1954. 94(3): p. 630-638.
145
40. Meiboom, S. and D. Gill, Modified spin-echo method for measuring nuclear
relaxation times Review of Scientific Instruments, 1958. 29(8): p. 688-691.
41. Herigstad, B., M. Hamilton, and J. Heersink, How to optimize the drop plate
method for enumerating bacteria. Journal of Microbiological Methods, 2001.
44(2): p. 121-129.
42. Taylor, S.W. and P.R. Jaffe, Biofilm growth and the related changes in the
physical properties of a porous medium. 1. Experimental investigation. Water
Resources Research, 1990. 26(9): p. 2153-2159.
43. Taylor, S.W. and P.R. Jaffe, Enhanced in-situ biodegradation and aquifer
permeability reduction. Journal of Environmental Engineering-ASCE, 1991.
117(1): p. 25-46.
44. Allison, D.G., The biofilm matrix. Biofouling, 2003. 19(2): p. 139-150.
45. Stoodley, P., D. Debeer, and Z. Lewandowski, Liquid flow in biofilm systems.
Applied and Environmental Microbiology, 1994. 60(8): p. 2711-2716.
46. Stoodley, P., et al., Biofilms as complex differentiated communities. Annual
Review of Microbiology, 2002. 56: p. 187-209.
47. Anand, V., Hirasaki, George J., Paramagnetic relaxation in sandstones:
Distinguishing T1 and T2 dependence on surface relaxation, internal gradients
and dependence on echo spacing. Journal of Magnetic Resonance, 2008. 190: p.
68-85.
48. Godefroy, S., Korb, J.-P., Fleury, M., Bryant, R.G., Surface nuclear magnetic
relaxation and dynamics of water and oil in macroporous media. Physical Review
E, 2001. 64: p. 021605-1-13.
49. Kleinberg, R.L., Kenyon, W.E., Mitra, P.P., Mechanism of NMR relaxation of
fluids in rock. Journal of Magnetic Resonance Series A, 1994. 108: p. 206-214.
50. Kleinberg, R.L., S.A. Farooqui, and M.A. Horsfield, T(1)/T(2) ratio and
frequency-dependence of NMR relaxation in porous sedimentary rocks. Journal of
Colloid and Interface Science, 1993. 158(1): p. 195-198.
51. Hills, B.P., C. Cano, and P.S. Belton, Proton NMR relaxation studies of aqueous
polysaccharide systems. Macromolecules, 1991. 24(10): p. 2944-2950.
52. Hornemann, J.A., et al., Biopolymer and water dynamics in microbial biofilm
extracellular polymeric substance. Biomacromolecules, 2008. 9(9): p. 2322-2328.
146
CHAPTER SIX
IN-SITU DETECTION OF SUBSURFACE BIOFILM USING
LOW-FIELD NMR – A FIELD STUDY
Contribution of Authors and Co-Authors
Manuscript in Chapter 6
Author: Catherine M. Kirkland
Contributions: Helped conceive and implement study design. Collected and analyzed
data. Wrote manuscript.
Co-Author: Maria P. Herrling
Contributions: Helped conceive and implement study design. Collected and analyzed
data. Provided feedback and comments on the manuscript.
Co-Author: Randy Hiebert
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
Co-Author: Andrew T. Bender
Contributions: Collected and analyzed data.
Co-Author: Elliot Grunewald
Contributions: Helped conceive and implement study design. Provided supervision and
oversight on data collection and analysis. Provided feedback and comments on the
manuscript.
Co-Author: David O. Walsh
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
147
Co-Author: Sarah L. Codd
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
148
Manuscript Information Page
Catherine M. Kirkland, Maria P. Herrling, Randy Hiebert, Andrew T. Bender, Elliot
Grunewald, David O. Walsh, Sarah L. Codd
Environmental Science and Technology
Status of Manuscript:
____ Prepared for submission to a peer-reviewed journal
____ Officially submitted to a peer-review journal
____ Accepted by a peer-reviewed journal
_X___ Published in a peer-reviewed journal
American Chemical Society
Vol. 49, No. 18, Pages 11045 - 11052, September 2015
149
IN-SITU DETECTION OF SUBSURFACE BIOFILM USING
LOW-FIELD NMR – A FIELD STUDY
Abstract
Subsurface biofilms are central to bioremediation of chemical contaminants in
soil and groundwater whereby micro-organisms degrade or sequester environmental
pollutants like nitrate, hydrocarbons, chlorinated solvents and heavy metals. Current
methods to monitor subsurface biofilm growth in-situ are indirect. Previous laboratory
research conducted at MSU has indicated that low-field nuclear magnetic resonance
(NMR) is sensitive to biofilm growth in porous media, where biofilm contributes a
polymer gel-like phase and enhances T2 relaxation. Here we show that a small diameter
NMR well logging tool can detect biofilm accumulation in the subsurface using the
change in T2 relaxation behavior over time. T2 relaxation distributions were measured
over an 18 day experimental period by two NMR probes, operating at approximately 275
kHz and 400 kHz, installed in 10.2 cm wells in an engineered field testing site. The mean
log T2 relaxation times were reduced by 62% and 43%, respectively, while biofilm was
cultivated in the soil surrounding each well. Biofilm growth was confirmed by bleaching
and flushing the wells and observing the NMR signal’s return to baseline. This result
provides a direct and non-invasive method to spatio-temporally monitor biofilm
accumulation in the subsurface.
150
Introduction
Reactive biofilm barriers, composed of attached microbial cells, extracellular
polymeric substances (EPS), and entangled organic and inorganic molecules, can be
cultivated in the subsurface to degrade and immobilize chemical contaminants as part of
a bioremediation project [1]. Contaminants in the groundwater, including nitrates, heavy
metals, hydrocarbons, and some chlorinated solvents, are consumed or transformed by
bacteria attached to soil surfaces while the EPS matrix slows the groundwater flow by
reducing the soil’s hydraulic conductivity [2-7]. While biobarrier technology is well-
established to clean contaminated groundwater [8, 9], it is challenging with current
technology to monitor the in-situ accumulation and maturation of the biofilm
nondestructively. Conventional methods to infer biofilm growth in the subsurface
include reduced saturated hydraulic conductivity [10-12]. While hydraulic conductivity
measurements can be useful for describing how easily the soil conveys water, they are
less useful for monitoring the growth, maturation and decay of the biobarrier with spatial
and temporal resolution.
We present results indicating a small diameter nuclear magnetic resonance
(NMR) well-logging tool provides additional means of monitoring biofilm accumulation
and condition over time in the subsurface by detecting changes in the NMR signal
response. Chemical and physical changes in a sample can impact a wide range of NMR
parameters, including NMR frequency shifts, NMR relaxation times T1 and T2, and
diffusion coefficients [13]. NMR frequency shifts are typically caused by chemical
changes in the fluid sample. However at very low magnetic fields (<1 MHz) even large
151
proton chemical shifts such as those between oil and water are not detectable. Diffusion
coefficients require additional magnetic field gradient hardware for detection, but can be
used to distinguish different viscosity fluids in a sample or determine the average pore
size of a material by measuring the timescale of the restricted diffusion [9]. NMR
relaxation, T1 and T2, is caused by the correlation time of the diffusing fluid with respect
to the resonant frequency of the NMR signal, as well as paramagnetic molecules the fluid
comes into contact with. For these reasons, NMR relaxation times are affected by pore
size distributions, fluid viscosity, and chemical changes in the mineralogy of any solid
matrix or dissolved ions. Measuring T2 is significantly faster than measuring T1. T2
measurements are generally considered the most robust low field measurement
considering acquisition times and signal-to-noise.
In a T2 measurement, the induced NMR signal echoes decay at a rate of 1/T2,
where T2 is the spin-spin relaxation time associated with interactions in the fluid-pore
environment. Analyzing the signal decay curve with the inverse Laplace transform
produces a statistical distribution of T2 relaxation times that reflect the variety of pore-
scale environments occupied by hydrogen protons in the excitation shell. The T2
relaxation time distribution in heterogeneous porous media is used, both in hydrogeologic
analysis [14] and in oil and gas exploration [15-18], to estimate the pore size distribution
in the formation and for fluid typing [9]. Biofilm accumulation in pores causes T2
relaxation times to decrease compared to unclogged pores as the biofilm EPS contributes
a biopolymer gel-like phase [19, 20]. Water in the EPS matrix contributes a reduced
NMR relaxation signal [21] and demonstrates restricted diffusion [22].
152
The NMR well-logging tool used in this study measures T2 relaxation and was
designed for use in small diameter ‘slim-line’ boreholes for groundwater exploration and
aquifer characterization [14]. In preliminary work, it was shown that this NMR logging
tool can detect biofilm growth [20]. The tool was tested in a laboratory-scale model well-
bore bioreactor filled with silica sand and inoculated with Bacillus mojavensis. Biofilm
was cultivated in the reactor over 8 days, during which the measured mean log T2
relaxation time decreased by 45%, from 710 ms to 389 ms [20]. Biofilm accumulation
was confirmed with destructive sampling and subsequent heterotrophic plate counts and
microscopy. The current research significantly advances previous work by showing that
the same logging tool can monitor biofilm growth at field scale, 6 m underground where
complex water chemical interactions exist. The NMR tool could be deployed to
bioremediation sites in parallel with currently used methods to enhance the understanding
of biofilm growth in-situ.
Methods
This study used two 8.9 cm diameter, 1.37 m long NMR well-logging probes
(JP350 Javelin by Vista Clara, Inc., Mukilteo, WA) to monitor T2 relaxation distributions
over time as an indication of biofilm accumulation in the soil in an engineered test cell.
The test cell is 55 m by 40 m at the surface and is 6m deep with 2:1 side slopes (x:z). The
cell is lined with a 30 mm polyvinylchloride (PVC) liner and contains water up to 1 - 1.2
m beneath the ground surface. The test cell contains a number of boreholes of varying
dimensions that were prepared for a previous experiment [7]. Two of the 10.2 cm
153
boreholes were used for measurements of T2 relaxation over an 18-day experimental
period.
A bleach solution consisting of 2 lb (907 g) sodium dichloro-s-triazinetrione
dihydrate (C3H4Cl2N3NaO5) (Spa Guard Chlorinating Concentrate, Bio-Lab Inc.,
Lawrenceville, GA) was pumped into the soil around the two wells at the beginning and
end of the experiment. A 30 L inoculum of Pseudomonas fluorescens strain CPC211a
was injected into each well’s measurement zone and was cultivated with a pulsed flow of
molasses-based substrate (10 g/L molasses (Aunt Patty’s Blackstrap Molasses, Eugene,
OR), 3 g/L sodium nitrate (SQM industrial grade prills, 98%, SQM North America,
Atlanta, GA), 1 g/L yeast extract (Acros Organics, Geel, Belgium), 0.12 g/L potassium
phosphate dibasic, and 0.04 g/L potassium phosphate monobasic (Thermo-Fisher
Scientific, Waltham, MA)) for 10 days. A 4-day starvation period followed the growth
phase. T2 measurements were conducted at the well bottom daily during the growth
phase, Days 1-10, and twice during the starvation phase, Days 12 and 14. Water samples
were collected prior to T2 experiments for subsequent pH measurement and microbial
analysis using the drop plate method [23]. T2 relaxation measurements were also
recorded over the bottom 3-3.5 m of well depth twice during the experiment; wells were
logged before inoculation and during the starvation period. At the end of the experiment,
the sodium dichlor bleach solution and high groundwater flows were applied to the
experimental region as a validation of biofilm growth, since these stress tests would be
expected to remove EPS formed during the growth phase. The experiment was designed
to distinguish normal variation in signal response due to environmental noise and probe
154
placement from those changes resulting from biological activity and biofilm growth. The
experimental sequence of events for a single well is shown in Table 6.1; both wells
received identical treatment. Additional details regarding experimental methods are
provided in Supporting Information.
The two NMR probes were tuned to different frequencies, corresponding to
excitation shells at different radii from the well centers. The lower frequency probe (LF)
was placed in one well, called the LF well, and was tuned to 245 and 290 kHz. Two
excitation shells 0.5 m in height and a few mm thick were located 17-19 cm from the LF
well center. The higher frequency probe (HF), installed in the HF well, was tuned to 360
and 425 kHz, producing two excitation shells located 11-13 cm from the HF well center.
The probes are identical low-field NMR tools other than the tuning frequency.
155
Day Experimental Task Purpose
(-2)-( -1) Site Preparation Create a 'biofilm-free'
initial condition
Injection of 75 – 115 L
sodium dichlor concentrate
solution
Flush with test cell
groundwater at 45.5 L/min
0 Well logs Measure baseline T2
distribution as a function
of depth
T2 distributions measured
in 0.5 m increments over
the bottom 3-3.5 m
saturated well depth
1 Inoculation Promote attachment and
biofilm formation by target
organism
Injection of 30 L inoculum
culture of Pseudomonas
fluorescens CPC211a into
measurement zone
surrounding the NMR
probe at well bottom
T2 measurement in fixed
location at well bottom
Measure baseline T2
distribution at the well
bottom
2-10 Biofilm Growth Phase Promote and measure
biofilm accumulation
Injection of 380-415 L
molasses-based substrate
daily at 1.2 L/min
Measurements of T2
relaxation distributions
daily in fixed location at
well bottom
156
Day Experimental Task Purpose
11-14 Starvation Phase Monitor biofilm resilience
Establish 'initial condition'
to assess signal variation
due to probe re-location
Measurements of T2
relaxation distributions Day
12 and 14 at the well bottom
12 Well logs Measure T2 distribution as
a function of depth after
biofilm growth
Evaluate signal variation
due to probe re-location
T2 distributions measured
in 0.5 m increments over
the bottom 3-3.5 m
saturated well depth
14 Biofilm Stress Tests Monitor signal response to
stresses intended to destroy
biofilm
Flush with test cell
groundwater at 45.5 L/min
Injection of 75-115 L
sodium dichlor concentrate
solution
Measurements of T2
relaxation distributions in
fixed location at well
bottom
15 Measurement of final conditions Verify biofilm
accumulation by signal
restoration after
detachment and removal of
biofilm
Flush with test cell
groundwater at 45.5 L/min
157
Measurements of T2
relaxation distributions in
fixed location at well bottom
Table 6.1. Experiment Overview
A diverter disk attachment was connected to the top of each probe to direct the
substrate and high flows of water into the soil around the well-bore, rather than into the
water column above the probe. A millimeter-scale gap was left between the disk and the
well casing to ensure that the probe could be removed from the well. The disk was
machined with an attachment for a standard garden hose to accommodate high volume
flushing and four 0.1 cm holes for low-flow nutrient tubing (Figure 6.1). Pore velocity
during the substrate injection was approximately 0.2 cm/min in the LF well and 0.4
cm/min in the HF well each probe’s sensitive zone. The high volume flush created a pore
velocity of approximately 9.2 cm/min in the LF well and 14.8 cm/min in the HF well.
158
Figure 6.1. A diverter disk attachment on the top of each NMR logging tool directed
substrate and high flow groundwater flushes into the biofilm growth region in the tool’s
sensitive zone in the soil and reduced backflow up the well casing.
For each probe, Experiment 1 and 2 (Table 6.2) constitute one standard Carr-
Purcell-Meiboom-Gill (CPMG)[24, 25] measurement of approximately 27 minutes. This
standard measurement was conducted 4 times in each well for a typical well-bottom
measurement of approximately 2 hours Days 0-15, excluding Days 11 and 13. The data
was averaged to obtain a single measurement for each day for each well. This standard
measurement was also used at each 0.5 m depth increment when vertically logging the
wells before and after biofilm growth on Day -1 and Day 12. An additional 12 standard
159
measurements were made in one or the other well on alternating days to improve the
signal to noise ratio in the data collected and to confirm that 4 measurements was
adequate to describe the system.
LF Probe HF Probe
Expt 1 Expt 2 Expt 1 Expt 2
Echo time, tE (ms) 1.5 1.5 1.3 1.3
Repetition time, Tr (ms) 1500 5000 1500 5000
Acquisition time (ms) 50 500 50 500
No. of echoes 34 334 39 385
No. of averages 600 150 600 150
Table 6.2. NMR Experimental Parameters
Results and Discussion
Both the LF and HF NMR logging probes recorded changing distributions of T2
relaxation times during the experiment with reductions of 62% and 43%, respectively, in
the mean log T2 during the biofilm growth phase (Figure 6.2). In the LF well, the T2
relaxation time was 29 ms on Day 1 and fell to an average of 11 ms between Days 5-12.
Over the same period, the HF well T2 relaxation time fell from 50 ms to an average of 28
ms. Data from both wells initially showed a T2 distribution with two peaks that
transitioned to a single peak distribution by Day 6 (Figure 6.3 (a) and (b)).
160
Figure 6.2. Mean log T2. Mean log T2 relaxation times decreased 62% (LF) and 43% (HF)
indicating that protons became more rotationally constrained as biofilm was cultivated in
the soil. (LF well data is shown with square markers, HF well data with triangle markers).
The first measurements were performed on Day -1 after bleaching and flushing both wells.
Inoculation occurred on Day 1. Substrate was injected daily Days 2-10. Days 11-14, the
bacteria were starved. On Day 14, T2 relaxation was measured, then the wells were flushed
with high flows of groundwater from the test cell and T2 was measured again. Then a
bleach solution was injected to oxidize remaining organics. Day 15 data was collected
after flushing the bleach solution from the wells.
161
Figure 6.3. T2 distribution for LF well (a) and HF well (b). The curves show the transition
over time of the distribution of T2 relaxation times in each well, beginning at inoculation
(short dash line). As biofilm grew in the NMR probe’s sensitive region, the T2 relaxation
times shifted to a single peak distribution centered about a shorter mean log T2 time (solid
lines). After flushing and bleaching each well, the T2 relaxation distributions closely
resembled the initial distributions (dash-dot lines).
In the Day 1 data, the relaxation time distribution is bimodal. Given that there
was only one fluid type at this time, the relaxation distribution peaks represent the
relative pore size distribution. Water in the smaller pores relaxes more quickly due to
the higher S/V ratio and the more frequent interactions of the diffusing fluid molecules
with the grain surfaces [26]. Water in the larger pores experiences fewer interactions on
the measurement timescale and yields slower signal decay.
162
As bacteria grow in the soil pores, the creation of gel phase EPS and the resulting
exchange of hydrogen in the polymerized structure produces a secondary relaxation
mechanism [21]. We attribute the collapsing the bimodal distribution to this mechanism,
where the otherwise long relaxation time components shift to shorter relaxation times. In
both wells, T2 distributions measured Day 6-12 show a single peak, on the order of 101
ms. The transition to shorter relaxation times is an indication of biofilm accumulation
[19, 27, 28]. Data collected on Day 15 after flushing and bleaching each well shows a
return to the bimodal T2 distribution that typified the system prior to biofilm growth.
On Day 12, following two days of biofilm starvation, the two probes were raised
from the well bottom to record T2 relaxation over 3.0-3.5 m of saturated well depth in 0.5
m increments. In the HF well data set (Figure 6.2, triangles), the two data points
collected on Day 12 show the difference in the measurements due to raising and replacing
the probe in its original position. In the LF well dataset, the Day 12 data point (Figure
6.2, squares) was measured prior to raising the probe. The measurement made after re-
positioning the probe at the bottom of the LF well was made on Day 14. For both data
sets, the open marker represents the data collected after the probe was repositioned at the
bottom of the well. The variance exists both because of inherent noise in the data and
because the probe was likely exciting a slightly different shell within the soil where
heterogeneities exist in both the soil and biofilm. The difference is 17% and 14%,
respectively, in the LF and HF wells and is significantly less than the change in signal
response measured during the biofilm growth phase. The lower-value data points
collected on Day 14 show that the biofilm surrounding both wells remained intact and
163
relatively robust despite four days of starvation, compared to earlier measurements during
the biofilm growth phase. This is consistent with the durability of the biobarrier
originally constructed in the test cell where measured conductivity remained 2 orders of
magnitude lower than the initial conductivity after approximately 6 months of
starvation[7].
The higher-value data points from each well on Day 14 were measured after a
high-flow flush with groundwater from the test cell. The final data point, on Day 15, was
collected after the wells were each treated with the same sodium dichlor solution used in
the site preparation step to oxidize organics (Supporting Information). It was expected
that the high shear stress and bleach solution in the soil pores would detach and remove
most of the biofilm previously formed, causing the T2 distribution to return to longer
relaxation times.
When the test cell was constructed, coarser soil from an off-site location was used
in the region where the HF well was located. The finer textured soil surrounding LF well
was excavated on-site, sieved, and replaced. This variation in the soil texture between
the two wells explains the differing magnitudes of the two relaxation peaks seen in the
Day 1 and 15 data (Figure 6.3 (a) and (b)), with relatively more of the pore space around
the HF well composed of larger pores and a more even distribution of pores sizes near the
LF well. It is likely that the soil mineralogy also differs between the two wells, though
examination of such parameters was beyond the scope of this experiment.
More signal was recovered following removal in the HF well than in the LF well,
likely due to the coarser soil and larger pores in the vicinity of the HF well. These larger
164
pores would allow for bleach solution and detached EPS to be transported more easily
through the excitation shell. In the soil around the LF well, bleach solution penetration
may have occurred along preferential flow paths and biofilm sloughed from one pore
may have been trapped in another smaller pore where it may have still contributed to the
measured signal. Furthermore, the initial bleach injection was performed prior to
installing the NMR probes, so the bleach was free to migrate up the casing and may not
have penetrated the target area at design strength. The final bleach injection, which
occurred with the probes in place and below the diversion disks, would have more
effectively isolated the biofilm growth area. Some soil pores may have experienced the
second bleach pulse but not the first. The measurements collected Days 14 and 15
confirm that the mechanism responsible for the change in T2 relaxation could be reversed
with flushing and bleaching the wells and is consistent with biofilm growth.
T2 measurements were recorded with respect to depth in each well prior to
inoculation and again on Day 12 (Figure 6.4). These well logs show notable differences.
In both well logs, there was a change in T2 distribution over the entire measured depth
after 10 days of substrate injection at the bottom of the well where the NMR logging tool
was deployed. On Day -1, the two well logs show broad T2 distributions marked by two
peaks at most depth levels. After the biofilm growth phase, the Day 12 data show a
single peak distribution of T2 relaxation times centered about a shorter mean relaxation
time.
165
Figure 6.4. LF and HF well profiles measured Day -1 (pre-inoculation) and Day 12 of
the biofilm growth phase. T2 distribution as a function of depth shifted to faster decay
times over the entire measured depth of each well when biofilm was cultivated at the well
bottom.
The observed changes in the well log are consistent with biofilm cultivation at the
well bottom. The changes in the distributions are most pronounced at the bottom of the
well, where the substrate was injected. The diversion disk attached to the top of each
probe was located approximately in the middle of the distributions shown, meaning that
substrate was directly accessible to the bottom half of each well log. As biofilm
accumulated in the soil pores, the substrate would have encountered increased resistance
to flow through the soil [10] and the gap between the casing and the diversion disk
(~1mm) would have become a preferential flow pathway. Substrate would diffuse into
the soil above the probe, in addition to the substrate diffusing through the soil itself.
Again the difference between the LF and HF well is instructive, as the upper levels of the
166
HF well show less intensity of effect than the upper levels of the LF well. The HF well is
surrounded by coarser soil, making it more difficult for the biofilm to clog the soil pores.
Conversely, the LF well region had smaller soil pores initially and experienced more
abundant biofilm accumulation as shown by the larger reduction in mean log T2
relaxation times.
Microbiological Data and Water Chemical Analysis
Heterotrophic plate counts (HPC) in both wells were approximately 1x103 colony-
forming units (cfu)/mL after bleaching and prior to inoculation, representing the
culturable ‘native’ heterotrophic population of the test cell (Figure 5(a)). After
inoculation and substrate injection, the HPC increased in both wells to approximately
1x105 cfu/mL by Day 4. Prior to inoculation and following the injection of the sodium
dichlor solution 2-3 days previously, pH measurements were approximately pH 8 (Figure
6.5 (b)). The pH in both wells was stable through Day 3 following inoculation. On Day
4, the pH measurements diverged with the HF well pH increasing to pH 9, while the LF
well pH decreased to approximately pH 7. Also on Day 4, the NMR signal in both wells
changed significantly, dropping to near the minimum for mean log T2 relaxation time in
the LF well where the effect was more pronounced (Figure 6.3). Taken together, the
higher HPC values, optimum pH conditions, and faster NMR signal decay indicate that a
measureable accumulation of biofilm had grown within the pore spaces of the logging
tool’s sensitive zone by Day 4 of the experiment. Since the HPC reflect bacterial cells
which are necessarily not attached in biofilms, the NMR logging tool indicates itself as a
167
valuable addition to the bioremediation toolbox through its sensitivity to the biofilm EPS,
which is not measurable with plate counts.
Figure 6.5. Heterotrophic Plate Counts (HPC) (a) and pH data (b). Both HPC and pH
measurements decreased on Day 5 in the two wells compared to earlier values, with
increased variability in the measurements noted during the starvation phase when no
substrate was injected (LF well data is shown with square markers, HF well data with
triangle markers). This water chemical variability was not reflected in the NMR signal
response which remained consistent over the same period.
Water samples from days 5-10 of the experiment show a decrease in HPC to
approximately 1x104 cfu/mL and a decrease in pH to approximately 4.5-5 in both wells.
Lower pH promotes bacterial attachment to surfaces [29], while not significantly
adversely affecting the biofilm matrix over the relatively short time frame measured [30].
This increased adhesion of bacteria may have reduced the number of suspended cells
within the bulk fluid for capture during sampling or caused preferential sampling of the
well casing fluid, rather than pore water, due to biofilm clogging the soil pores. It is also
possible that the acidic conditions selected for other native bacterial strains that are not
168
culturable on agar plates, like autolithotrophs participating in oxidation and reduction of
chemical species in the soil and water.
To explore potential causes for the observed drop in pH during testing, benchtop
tests were performed using isolates of the two most populous cell morphologies from the
agar plates for Days 5-10. These bacteria were cultivated in 100 mL of molasses
substrate in Erlenmeyer flasks. The pH of the broth decreased by more than an order of
magnitude, from pH 6.2 on average to pH 5.0 on average by Day 4 of the 8-day benchtop
experiment. These results are consistent with results from the field study and indicate
that metabolic processes, including production of nucleic acids and CO2, contributed to
the observed decrease in pH observed in the field study.
Another possible explanation of the observed pH change is related to the redox
chemistry of sulfur and iron cycling [31, 32]. The groundwater in the lined test cell was
static for more than 10 years before the current field study was initiated, creating
reducing conditions at the bottom of the test cell. The distinctive odor of hydrogen
sulfide (H2S) was readily perceptible in the water pumped from the cell. When exposed
to the atmosphere, the reduced Fe(II) in the groundwater oxidized to form a rust-like
layer of Fe(III) compounds, likely ferrihydrite, in the large-capacity water storage tanks.
The injection of the sodium dichlor solution followed by substrate containing organic
carbon, nitrate, and dissolved oxygen would have promoted the oxidation of reduced
species in the subsurface. Heterotrophic bacteria, like Pseudomonas fluorescens, use the
electron acceptors oxygen and nitrate to oxidize the organic carbon in the media,
producing new cells, EPS, and waste products. Oxidation of sulfide leads to
169
acidification, as in the case of acid mine drainage [33], and may account for some of the
observed decrease in pH after the initial biofilm growth between Days 1-4.
Laboratory studies have shown that NMR relaxation measurements are sensitive
to changing soil redox conditions [34] as well as the mineralogic form of iron species [35,
36]. To conclude that the observed changes in relaxation response are due to biofilm
growth, it is important, therefore, to consider the possible impact of geochemical
changes. First, the timing of the changes is indicative of biofilm growth. Were
oxidation-driven geochemical changes strongly influencing T2 during the approximately
two days between the initial bleach and inoculation, we would expect to see significant
changes in the T2 relaxation behavior of the two well systems Days -1 and 0. Instead we
see relatively small changes—first a slight decrease in T2 which may, in fact, be due to
abiotic Fe(III) precipitation, then an increase during inoculum injection (Figure 6.2). The
increase is most likely due to the make-up of the inoculum broth which was made with
DI water rather than water from the test cell. The tight temporal relationship at the end of
the experiment between the final flush and bleach and the observed recovery of long T2
times provides further strong evidence that the oxidizing conditions themselves do not
drive the shortening of the T2 response.
A second reason we expect limited influence from mineralogical transformations
pertains to competition for the electron acceptors in the subsurface. The 4 log reduction
between the inoculum colony count and the Day 2 HPC (Figure 6.5a) indicates
approximately 104 colony forming units (cfu)/mL attached to soil surfaces in the well-
bore, creating strong competition for oxygen and nitrate. Oxidation of the H2S would
170
have been more thermodynamically favorable compared to oxidation of iron compounds
and would have occurred preferentially [37] where the electron acceptors available
exceeded bacterial demand. We therefore expect the contribution of iron precipitation to
be moderated and would not expect the iron-driven T2 changes to be any larger than was
observed in the small decrease between Days -1 and 0 before competing heterotrophic
bacteria were present in the system.
Finally, we have also considered the magnitude of the observed T2 shortening to
assess whether biofilm growth or moderate changes in iron geochemistry are more likely
mechanisms. The T2 shortening observed during the experiment generally shows long
components, T2 > >100ms, transitioning to shorter relaxation time components,
T2<<100ms. Thus, the dominant relaxation mechanism(s) must be ones that can result in
T2 relaxation times much shorter than 100ms.
With the exception of magnetite and hematite, the literature reports elevated, but
moderate surface relaxivity for most Fe(III) minerals at a measurement frequency of 2
MHz [35]. Large reductions in relaxation times resulting from the change of Fe(II) to
Fe(III) have been shown at 90 MHz [34], but this measurement frequency is more than
100 times higher than the current downhole measurements. Relaxation times exhibit
significant and complex frequency dependence over this wide range. Generally, the
geochemical changes are expected to show a reduced influence on T2 at lower field
strengths [38] since the amplitude of internal gradients associated with diffusion
relaxation are reduced at lower field.
171
With evidence suggesting moderate ferrihydrite formation is unlikely to cause the
absolute T2 shortening we observe, we consider evidence of T2 shortening due to the
mechanism of biofilm polymerization. Laboratory measurements at 275 kHz by
Sanderlin et al. (2012) showed growth of biofilm and polymer gels resulted in T2 values
as short as 50ms [21]. Furthermore, in previous laboratory work using the same LF
probe [20], moderate biofilm growth in 1mm quartz sand resulted in a decrease in the
mean log T2 of 45% which is similar in magnitude to the changes observed in this study.
Geochemical changes in the pore fluid would have occurred slowly over the two
week period, and the drop in T2 would be expected to be considerably less than was
observed. Given the timing and magnitude of the T2 changes, and the evidence from
previous lab work that biofilm has a large influence on NMR relaxation measurements,
the experimental results provide compelling evidence that the low field NMR logging
tools can detect and monitor biofilm growth in the subsurface. How biofilms interact
with the geochemical environment will vary from field site to field site, as well as over
space and time. Future experiments and implementation of these methods, therefore,
should include an informed monitoring of the chemistry of extracted pore fluid and the
changing NMR relaxation.
We have shown the NMR logging tools detected significant and sustained change
in signal response during the biofilm growth phase, measured by changing T2 relaxation
distributions. Both the HF and LF probes were sensitive to the changes, and the
differences in soil geology between the probe locations likely resulted in larger
differences in the signal response than the variation between the two probes. Mean log T2
172
relaxation times decreased from 29 ms to an average of 11 ms in the LF well, and from
50 ms to an average of 28 ms in the HF well while biofilm was cultivated in the
surrounding soil. We have further shown that high shear flow and oxidative stress
resulting from the bleach solution, applied with the intent to denature and remove biofilm
from the tool’s sensitive zone, produced a return of signal response similar to initial
conditions. The timescale of these changes is consistent with biofilm formation and
subsequent removal. These results provide an important demonstration of the advantages
of incorporating an NMR measurement into future bioremediation toolkits. NMR T2
relaxation measurements provide unique complementary data that, together with other
monitoring techniques such as hydrological conductivity measurements, can improve our
ability to draw the correct conclusion about the subsurface environment with regard to
biofilm growth.
Supporting Information
Site Preparation
A bleach solution consisting of 2 lb (907 g) sodium dichloro-s-triazinetrione
dihydrate (C3H4Cl2N3NaO5) (Spa Guard Chlorinating Concentrate, Bio-Lab Inc.,
Lawrenceville, GA) was pumped into each well to oxidize any pre-existing biofilm. The
solution was mixed in a 55 gal (208 L) drum using enough water to dissolve the granules,
approximately 20-30 gallons (75-115 L). The solution was pumped to the target region
through a standard garden hose using a 1/3 hp centrifugal pump (AMT Pump Model
3680-975-97 by Gorman Rupp, Royersford, PA) operating at approximately 12 gpm
173
(45.5 L/min). The bleach solution was allowed to react in the wells overnight, and was
followed by a 30 min high-flow (12 gpm (45.5 L/min)) flush of groundwater from the test
cell to detach EPS and dead cells and purge any remaining bleach solution. Then, the
probes were lowered to the bottom of the wells, approximately 18 ft (5.5 m) deep and 15
ft (4.5 m) below the water table in the test cell.
Bacterial Culture
The target organism, Pseudomonas fluorescens CPC211A, is the environmental
isolate used in the original experiments in the test cell and, therefore, was known to grow
successfully in that environment.[7] The bacteria was cultured from cryo-stock using a
nutrient broth consisting of 10 g/L molasses (Aunt Patty’s Blackstrap Molasses, Eugene,
OR), 3 g/L sodium nitrate (SQM industrial grade prills, 98%, SQM North America,
Atlanta, GA), 1 g/L yeast extract (Acros Organics, Geel, Belgium), 0.12 g/L potassium
phosphate dibasic, and 0.04 g/L potassium phosphate monobasic (Thermo-Fisher
Scientific, Waltham, MA). The inoculum was cultured in successively larger volumes at
ambient temperature without mixing over seven days to produce a final inoculum volume
of 60 L with a viable cell count of 8.5x107 colony forming units (cfu)/mL.
The active biofilm growth region of the well-bore environment was designed to
be approximately 18 gal (67 L) of pore space, corresponding to the height of the NMR
logging device (4.5 ft), a 9 in (22.9 cm) radius and an estimated porosity of 0.3 [7]. This
active region of soil in each well was conditioned with 15 gal (56.7 L) of substrate prior
to injection of the inoculum broth. The substrate was pumped at a rate of 1.2 L/min with
a peristaltic pump (Masterflex L/S Model 900-1255, Cole-Parmer, Vernon Hills, IL)
174
down each well casing to the sensitive zone of the NMR probe using four lines of 1/8-
inch (3.2mm) ID vinyl tubing (Clearflex, McMaster-Carr, Santa Fe Springs, CA). Then
30 L (7.9 gal) of inoculum was injected into each well, followed by 5 gal (18.9L) of fresh
substrate to push the bacteria into the soil from the well casing. Finally, water from the
test cell was injected for approximately 1-2 minutes to rinse the injection tubing. No
attempt was made to maintain a monoculture.
During the biofilm growth phase of the experiment, Days 1-10, the molasses
substrate was prepared just prior to use in 50-55 gal plastic drums using groundwater
from the lined test cell. Groundwater was pumped from another of the test cell’s eleven 4
inch (10.2 cm) wells located 36 ft (11 m) from the nearest experimental well and stored
prior to use in a large capacity water storage tank (Ace Roto-Mold, Den Hartog
Industries, Hospers, IA). Each well received approximately 100-110 gal (379-416 L) per
day of substrate over a 5-6 hour period. This translates to a Darcy velocity at the logging
tool’s excitation shell of approximately 4.2 cm/hr and an interstitial velocity of 14.4
cm/hr in the LF well during substrate injection, and 7.2 cm/hr and 23.4 cm/hr,
respectively, in the HF well. NMR measurements were collected during the pumping of
substrate. There was no pumping or flow in the wells for the remainder of each day
following substrate injection. Continuous flow of substrate was not feasible given
experimental constraints.
175
NMR Measurements
T2 relaxation measurements were conducted using a Carr-Purcell-Meiboom-Gill
(CPMG) pulse sequence [24, 25], consisting of a 90° excitation pulse followed by a
series of 180° refocusing pulses separated by the echo spacing, tE. Frequencies and echo
spacings for each probe were fixed. The repetition time, Tr, which is the time between
90° rf excitation pulses, was either 1500 ms or 5000 ms for each measurement (Table
6.2). Faster repetition of the excitation pulse, with a Tr of 1500 ms, allowed for the rapid
collection of data from the fastest decaying components of the NMR signal; recording
600 averages improved resolution of the early part of the decay curve [14]. A longer
repetition time of 5000 ms, recorded with 150 averages, allowed signal to be collected
from the slower decaying components, such as water in larger pores.
Microbiological and Water Chemical Analysis
Two 10-15 mL samples of water were collected daily in sterile 15 mL Falcon
tubes (Becton, Dickinson and Co., Sparks, MD) from each well prior to injection of fresh
substrate. Flow on the substrate injection pumps was reversed and allowed to flow for 2-
3 minutes in order to collect a water sample from each well’s biofilm growth region. The
samples were placed on ice in the field and then frozen for later analysis at the conclusion
of the field test. The samples were analyzed for pH (VWR sympHony benchtop SB70P
pH meter) and by drop plate for heterotrophic plate count (HPC) [23].
Samples were drop plated on Difco tryptic soy agar (Becton, Dickinson and Co.,
Sparks, MD) plates in duplicate and cultured at room temperature on the benchtop and in
anaerobic pouches (BD GasPak EZPouch, Becton, Dickinson and Co., Sparks, MD). The
176
aerobically grown plates produced higher and more consistent numbers of culturable
heterotrophic cells.
Results and Discussion
Another measure of signal attenuation is given by the Square of Echoes (SOE)
method which is used to improve the low signal to noise ratio that is typical for a low-
field NMR device in natural geologic material. The SOE is the squared value of the
mean echo in the signal decay curve. A reduction in the SOE value over the course of the
experiment qualitatively indicates that the log mean value of T2 relaxation is also
decreasing. When normalized with respect to water content squared, data from the field
test shows an SOE decrease of 91% in the LF well and 70% in the HF well during the
biofilm growth phase of the experiment. The percent reduction is calculated using the
SOE at inoculation (Day 1) and the average of SOE values Day 5-12 when the values
were stable (Figure 6.6).
177
Figure 6.6. Square of Echoes (SOE) data from the HF (triangles) and LF (squares) wells
shows a decrease of 70% and 90% respectively.
178
References
1. LaGrega, M.D., P.L. Buckingham, and J.C. Evans, Hazardous Waste
Management. 2nd ed. 2001, New York: McGraw-Hill.
2. Kao, C.M., et al., Biobarrier system for remediation of TCE-contaminated
aquifers. Bulletin of Environmental Contamination and Toxicology, 2004. 72(1):
p. 87-93.
3. Kao, C.M., et al., Enhanced PCE dechlorination by biobarrier systems under
different redox conditions. Water Research, 2003. 37(20): p. 4885-4894.
4. Kim, G., S. Lee, and Y. Kim, Subsurface biobarrier formation by microorganism
injection for contaminant plume control. Journal of Bioscience and
Bioengineering, 2006. 101(2): p. 142-148.
5. Pal, A. and A.K. Paul, Microbial extracellular polymeric substances: central
elements in heavy metal bioremediation. Indian Journal of Microbiology, 2008.
48(1): p. 49-64.
6. Taylor, S.W. and P.R. Jaffe, Enhanced in-situ biodegradation and aquifer
permeability reduction. Journal of Environmental Engineering-ASCE, 1991.
117(1): p. 25-46.
7. Cunningham, A.B., et al., Subsurface biofilm barriers for the containment and
remediation of contaminated groundwater. Bioremediation Journal, 2003. 7(3-4):
p. 151-164.
8. Careghini, A., S. Saponaro, and E. Sezenna, Biobarriers for groundwater
treatment: a review. Water Science and Technology, 2013. 67(3): p. 453-468.
9. Callaghan, P.T., Translational Dynamics & Magnetic Resonance: Principles of
Pulsed Gradient Spin Echo NMR. 2011, New York: Oxford University Press.
10. Cunningham, A.B., et al., Influence of biofilm accumulation on porous-media
hydrodynamics. Environmental Science & Technology, 1991. 25(7): p. 1305-
1311.
11. Taylor, S.W. and P.R. Jaffe, Biofilm growth and the related changes in the
physical properties of a porous medium. 1. Experimental investigation. Water
Resources Research, 1990. 26(9): p. 2153-2159.
12. Vandevivere, P. and P. Baveye, Effect of bacterial extracellular polymers on the
saturated hydraulic conductivity of sand columns. Applied and Environmental
Microbiology, 1992. 58(5): p. 1690-1698.
179
13. Johns, M., et al., Mobile NMR and MRI: Developments and Applications. New
Developments in NMR. 2015: Royal Society of Chemistry.
14. Walsh, D., et al., A small-diameter NMR logging tool for groundwater
investigations. Groundwater, 2013. 51(6): p. 914-926.
15. Kleinberg, R.L., Kenyon, W.E., Mitra, P.P., Mechanism of NMR relaxation of
fluids in rock. Journal of Magnetic Resonance Series A, 1994. 108: p. 206-214.
16. Kleinberg, R.L. and M.A. Horsfield, Transverse relaxation processes in porous
sedimentary rock. Journal of Magnetic Resonance, 1990. 88(1): p. 9-19.
17. Timur, A., Pulsed nuclear magnetic resonance studies of porosity, movable fluid,
and permeability of sandstones. Journal of Petroleum Technology, 1969.
21(JUN): p. 775-&.
18. Godefroy, S., et al., Surface nuclear magnetic relaxation and dynamics of water
and oil in macroporous media. Physical Review E, 2001. 64(2): p. 021605-1 -
021605-13.
19. Codd, S.L., et al., NMR relaxation measurements of biofouling in model and
geological porous media. Organic Geochemistry, 2011. 42(8): p. 965-971.
20. Kirkland, C.M., et al., Biofilm Detection in a Model Well-Bore Environment
Using Low-Field NMR. Groundwater Monitoring and Remediation, 2015: p.
DOI:10.1111/gwmr.12117.
21. Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media
using a low-field magnetic resonance system. Environmental Science &
Technology, 2013. 47(2): p. 987-992.
22. Vogt, S.J., et al., Permeability of a growing biofilm in a porous media fluid flow
analyzed by magnetic resonance displacement-relaxation correlations.
Biotechnology and Bioengineering, 2013. 110(5): p. 1366-1375.
23. Herigstad, B., M. Hamilton, and J. Heersink, How to optimize the drop plate
method for enumerating bacteria. Journal of Microbiological Methods, 2001.
44(2): p. 121-129.
24. Carr, H.Y. and E.M. Purcell, Effects of diffusion on free precession in nuclear
magnetic resonance experiments. Physical Review, 1954. 94(3): p. 630-638.
25. Meiboom, S. and D. Gill, Modified spin-echo method for measuring nuclear
relaxation times Review of Scientific Instruments, 1958. 29(8): p. 688-691.
180
26. Korb, J.P., S. Godefroy, and M. Fleury, Surface nuclear magnetic relaxation and
dynamics of water and oil in granular packings and rocks. Magnetic Resonance
Imaging, 2003. 21(3-4): p. 193-199.
27. Bayer, J.V., F. Jaeger, and G.E. Schaumann, Proton Nuclear Magnetic Resonance
(NMR) Relaxometry in Soil Science Applications. Open Magnetic Resonance
Journal, 2010. 3: p. 15-26.
28. Jaeger, F., E. Grohmann, and G.E. Schaumann, 1H NMR relaxometry in natural
humous soil samples: insights in microbial effects on relaxation time
distributions. Plant and Soil, 2006. 280(1-2): p. 209-222.
29. Kristian Stevik, T., et al., Retention and removal of pathogenic bacteria in
wastewater percolating through porous media: a review. Water Research, 2004.
38(6): p. 1355-1367.
30. Chen, M.J., Z. Zhang, and T.R. Bott, Effects of operating conditions on the
adhesive strength of Pseudomonas fluorescens biofilms in tubes. Colloids and
Surfaces B: Biointerfaces, 2005. 43(2): p. 61-71.
31. Juncher Jørgensen, C., et al., Microbial Oxidation of Pyrite Coupled to Nitrate
Reduction in Anoxic Groundwater Sediment. Environmental Science &
Technology, 2009. 43(13): p. 4851-4857.
32. Melton, E.D., et al., The interplay of microbially mediated and abiotic reactions
in the biogeochemical Fe cycle. Nat Rev Micro, 2014. 12(12): p. 797-808.
33. Evangelou, V.P. and Y.L. Zhang, A review: Pyrite oxidation mechanisms and
acid mine drainage prevention. Critical Reviews in Environmental Science and
Technology, 1995. 25(2): p. 141-199.
34. Bryar, T.R. and R.J. Knight, Sensitivity of nuclear magnetic resonance relaxation
measurements to changing soil redox conditions. Geophysical Research Letters,
2002. 29(24).
35. Keating, K. and R. Knight, A laboratory study to determine the effect of iron
oxides on proton NMR measurements. Geophysics, 2007. 72(1): p. E27-E32.
36. Keating, K. and R. Knight, A laboratory study of the effect of Fe(II)-bearing
minerals on nuclear magnetic resonance (NMR) relaxation measurements.
Geophysics, 2010. 75(3): p. F71-F82.
37. Benjamin, M.M., Water Chemistry. Series in Water Resources and Environmental
Engineering. 2002, New York, NY: McGraw-Hill.
181
38. Koenig, S.H. and K.E. Kellar, Theory of 1/T-1 and 1/T-2 NMRD profiles of
solutions of magnetic nanoparticles Magnetic Resonance in Medicine, 1995.
34(2): p. 227-233.
182
CHAPTER SEVEN
DETECTING MICROBIALLY-INDUCED CALCITE PRECIPITATION
(MICP) IN A MODEL WELL-BORE USING
DOWNHOLE LOW-FIELD NMR
Contribution of Authors and Co-Authors
Manuscript in Chapter 7
Author: Catherine M. Kirkland
Contributions: Helped conceive and implement study design. Collected and analyzed
data. Wrote manuscript.
Co-Author: Sam Zanetti
Contributions: Helped implement study design. Collected and analyzed data.
Co-Author: Elliot Grunewald
Contributions: Helped conceive and implement study design. Provided supervision and
oversight on data collection and analysis. Provided feedback and comments on the
manuscript.
Co-Author: David O. Walsh
Contributions: Provided feedback and comments on the manuscript.
Co-Author: Sarah L. Codd
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
Co-Author: Adrienne J. Phillips
Contributions: Helped conceive and implement study design. Provided feedback and
comments on the manuscript.
183
Manuscript Information Page
Catherine M. Kirkland, Sam Zanetti, Elliot Grunewald, David O. Walsh, Sarah L. Codd,
Adrienne J. Phillips
Environmental Science and Technology
Status of Manuscript:
____ Prepared for submission to a peer-reviewed journal
____ Officially submitted to a peer-review journal
____ Accepted by a peer-reviewed journal
_X___ Published in a peer-reviewed journal
American Chemical Society
Vol. 51, No. 3, Pages 1537-1543, February 2017
184
DETECTING MICROBIALLY-INDUCED CALCITE PRECIPITATION
(MICP) IN A MODEL WELL-BORE USING
DOWNHOLE LOW-FIELD NMR
Abstract
Microbially-induced calcite precipitation (MICP) has been widely researched
recently due to its relevance for subsurface engineering applications including sealing
leakage pathways and permeability modification. These applications of MICP are
inherently difficult to monitor non-destructively in time and space. Nuclear magnetic
resonance (NMR) can characterize the pore size distributions, porosity, and permeability
of subsurface formations. This investigation used a low-field NMR well-logging probe to
monitor MICP in a sand-filled bioreactor, measuring NMR signal amplitude and T2
relaxation over an 8-day experimental period. Following inoculation with the ureolytic
bacteria, Sporosarcina pasteurii, and pulsed injections of urea and calcium substrate, the
NMR measured water content in the reactor decreased to 76% of its initial value. T2
relaxation distributions bifurcated from a single mode centered about approximately 650
ms into a fast decaying population (T2 less than 10 ms) and a larger population with T2
greater than 1000 ms. The combination of changes in pore volume and surface
minerology accounts for the changes in the T2 distributions. Destructive sampling
confirmed final porosity was approximately 88% of the original value. These results
indicate the low-field NMR well-logging probe is sensitive to the physical and chemical
changes caused by MICP in a laboratory bioreactor.
185
Introduction
Biofilms form when bacteria secrete a matrix of extracellular polymeric substance
(EPS), attaching themselves to solid surfaces in colonies akin to multicellular organisms
and buffering their micro-scale environment[1]. Bacterial biofilms are known to induce
metal corrosion [2], cause persistent infections [3], treat wastewater [4], or remediate
contaminated groundwater [5]. When composed of ureolytic microbes, biofilms can also
induce calcite precipitation [6], a process referred to as biomineralization or microbially-
induced calcite precipitation (MICP). Many strains of bacteria found naturally in soil and
groundwater are ureolytic, meaning they can hydrolyze urea for energy and a source of
nitrogen [7]. Catalyzed by the microbially produced urease enzyme, water cleaves urea
to form ammonia and carbon dioxide which, due to a pH increase, shifts the carbonate
equilibrium toward bicarbonate and carbonate. Then, with sufficient calcium and
carbonate activity, calcium carbonate (CaCO3 or calcite) precipitates (Eqn. 7.1).
(NH2)2CO + 2H2O + Ca2+ ⇒2NH4+ + CaCO3(s) (7.1)
Sporosarcina pasteurii, the ureolytic bacteria used in this experiment, forms a
thin biofilm in porous media [8] where the EPS matrix, a 3-dimensional diffusion-limited
hydrogel, can either facilitate or inhibit MICP over microscales. The organic molecules
comprising the EPS matrix restrict mass transfer, creating localized chemical gradients
within the hydrogel structure [6]. Ca2+ ions are not used in metabolic processes and
accumulate near cell surfaces where ureolysis produces an alkaline environment. Thus,
the microbial biofilm matrix provides nucleation sites for calcite precipitation [9]. In
186
porous media, the precipitated calcite binds together media grains and fills pore spaces
[10].
MICP has engineering applications [11] that include soil stabilization [10, 12] and
subsurface barriers [13], sealing of cap rocks and well-bore regions for carbon dioxide
sequestration [14-16], and limestone and concrete remediation [9]. Many of these
beneficial applications of MICP occur in the subsurface, raising the question of how the
process can best be monitored spatio-temporally. Nuclear magnetic resonance (NMR) is
commonly used non-destructively and non-invasively to characterize the pore size
distributions, porosity, and permeability of subsurface geologic formations [17]. These
are the same physical properties affected by MICP, indicating that NMR well-logging
tools may have potential for monitoring subsurface engineering applications of MICP.
This study used a low-field NMR well-logging tool designed for subsurface
hydrogeologic investigations [18] to detect changes in NMR signal response indicative of
MICP in the pore spaces of sand-filled radial-flow bioreactor.
Background
There are limited examples in the scientific literature where NMR methods have
been applied to the study of biomineralization in porous media relevant for engineering
applications [8, 19]. These previous studies have used high field strength magnetic
resonance imaging (MRI) along with other NMR methods to probe hydrodynamic
properties of biomineralization in model porous media systems.
Fridjonsson et al. [8] used high-field NMR to measure changes in hydrodynamic
dispersion resulting from MICP in model porous media to compare flow dynamics
187
between systems influenced by either solid precipitates or a biofilm matrix. The authors
used a combination of NMR displacement measurements, relaxation mapping, MRI, and
microscopy methods.
Sham et al. [19] used MRI and NMR flow measurements on both a model bead
pack and a Bentheimer sandstone rock core to examine structure and transport properties
of each system following MICP. The authors report a reduction of 3.7% in absolute
porosity in the bead pack, which correlated to a 98% reduction in permeability. In the
sandstone, a 7.2% reduction in absolute porosity yielded a 96.5% reduction in
permeability. In both systems, preferential fouling of the inlet region of the column was
observed.
The low-field NMR well-logging tool used in this study (Javelin JP350, Vista
Clara, Inc., Mukilteo, WA) is sensitive to biofilm growth in the pore spaces of a sand-
filled bioreactor [20] and in the subsurface soil of an engineered field testing site [21]. In
both of these studies, biofilm growth caused enhanced relaxation with T2 relaxation times
decreasing by approximately 40 – 60%.
These previous studies show 1) NMR methods are useful for analyzing changes
resulting from MICP in porous media and 2) the well-logging tool is sensitive to small
changes over time in the micro-scale pore environment. To our knowledge, field scale
low-field NMR instruments have not been applied to the measurement or monitoring of
MICP. In the current study, CaCO3 precipitation was expected to change the NMR signal
response by reducing the liquid fraction from which the signal is obtained, causing a
decrease in signal amplitude over time as the pores accumulate calcite. MICP will also
188
change the pore sizes and mineral surface of the porous media, thereby influencing the
signal relaxation response. A correlation between the signal response and reduction of
porosity due to MICP may indicate the use of a NMR well-logging tool as a sensor for
biomineralization in field applications where optical or destructive monitoring methods
are not possible. This study represents a first step toward that end by demonstrating that
a NMR well-logging tool is sensitive to MICP.
NMR Theory
The NMR well-logging tool is sensitive to the hydrogen protons in water, called
‘spins,’ such that the behavior of the NMR signal over time is related to the various
micro-scale water environments in the surrounding formation. The tool measures 1.37 m
long and 8.9 cm in diameter and is designed to be lowered into small-diameter cased or
uncased borehole wells (Figure 7.1) [18]. The dual frequency probe used in this
experiment operates at approximately 250 and 300 kHz, and is composed of an array of
permanent magnets and radio-frequency (RF) induction coils [18]. The permanent
magnets establish a static magnetic field, B0, along the direction of the borehole, where
the field strength depends on the radial distance from the tool. The RF pulses produce
two mm-scale cylindrical excitation shells at radial distances of 17-19 cm from the probe
center and in the middle of the reactor’s sand annulus. The excited shells are 50 cm in
height. Only spins in these two excitation shells contribute to the measured NMR signal
response, which is averaged over all the spins in each shell.
The initial amplitude of the NMR signal is proportional to the amount of water in
the excitation shell and reflects the volumetric water content, or porosity, of the porous
189
media. The NMR signal amplitude decreases when water is displaced by mineral
formation in the pores.
The observed decay rate reflects spin-spin, or T2, relaxation, which occurs as
protons interact with each other in the transverse plane. These interactions cause a
dephasing of spin coherence and signal attenuation. In geologic materials, the observed
T2 relaxation rate comprises the bulk relaxation rate of the pore fluid, 1
𝑇𝑇2𝐵𝐵
, the surface
relaxation rate, 1
𝑇𝑇2𝑆𝑆
, related to interactions between the fluid and the pore walls, and the
diffusion relaxation rate, 1
𝑇𝑇2𝐷𝐷
, related to diffusion of fluids within pores due to
inhomogeneities in the local magnetic field (Equation 7.2) [22, 23].
1
𝑇𝑇2
= 1
𝑇𝑇2𝐵𝐵
+ 1
𝑇𝑇2𝑆𝑆
+ 1
𝑇𝑇2𝐷𝐷
(7.2)
At the low magnetic field strength used in this study, the experimental parameter
of the echo spacing, tE, can be selected to make the influence of diffusion relaxation, T2D,
sufficiently small to be neglected [20, 23]. For the current study, changes in the fluid
properties of the pore liquid, such as viscosity, are not expected be a significant factor in
the overall change of the system T2 relaxation time [8]. The influence of changes in T2B
can therefore also be neglected. Changes in surface relaxation, 𝑇𝑇2𝑆𝑆, are expected to
dominate changes in the observed T2 of this experimental system.
The low-field NMR signal response in most saturated natural geologic media is
dominated by surface relaxation [22, 24]. Surface relaxation occurs as excited spins
approach and interact with the pore walls. Thus, the rate of surface relaxation is most
190
strongly related to pore size and the mineral surface of the solid matrix. Surface
relaxation occurs faster in small pores with a high surface-area-to-volume-ratio because
the diffusing water molecules are more likely to interact with the grain surface. The
surface relaxation rate also depends on the propensity of the surface for inducing
relaxation, a characteristic referred to as surface relaxivity, 𝜌𝜌. Greater concentrations of
paramagnetic ions like Fe3+ and Mn2+ produce higher magnitudes of 𝜌𝜌 and faster
relaxation rates [25, 26]. In heterogeneous materials with a range of pore sizes or
variable 𝜌𝜌, there may be a distribution of relaxation rates making up the bulk response.
Thus, an Inverse Laplace Transform yields a decay-time distribution that can be
interpreted as a distribution of pore environments.
In our experiments, we expect MICP to have several combined influences on the
NMR response. First, we expect that growth of calcite within the pore space will reduce
the total porosity and water content. We also expect the growth of CaCO3 to influence the
observed relaxation rate due to changes in mineralogy and pore size [27-29]. The quartz
sand used in this study is coarse-grained and contains small percentages of paramagnetic
species including iron oxide (Fe2O3) at a mean weight percent of 0.04 (2095 Granusil®
silica sand, Unimin Corp., Ottawa, MN). We expect that CaCO3 forming on the quartz
grain surfaces will decrease the total macro-pore dimension which could drive faster
relaxation rates. On the other hand, CaCO3 precipitating on the grain surface may shield
water from the paramagnetic ions on the sand, thus decreasing the average 𝜌𝜌 of the grain
surface. A lower average 𝜌𝜌 would tend to decrease the surface relaxation rate in the
macro-pores, resulting in longer overall T2. Further, the CaCO3 may form
191
microcrystalline structures that incorporate significant micro-porosity of nanometer scale.
We expect water in the very small geometry of these micro-pores to exhibit very short
relaxation times. Thus, we anticipate these changes in the pore structure concurrent with
MICP will manifest themselves as multiple changes to the NMR T2 relaxation time
distribution. These observed changes are expected to indicate which mechanism
dominates in the bioreactor where there exists a particular initial pore size distribution
and surface minerology.
Materials and Methods
Bioreactor
The radial flow bioreactor is designed to model the near well-bore environment
and consists of four concentric polyvinyl chloride (PVC) pipe sections sealed with
grooved top and bottom plates (Figure 7.1). The reactor is the same as was used in a
previous study to detect biofilm growth in sand using the same NMR logging tool [20].
In the current experiment, the height of the reactor was 50 cm. The inner and outer pipes
are solid while the two inner pipes are slotted to allow radial flow through the sand
annulus between them. The inner and outer annuli are the influent and effluent
reservoirs, respectively. The sand annulus measures 7.6 cm wide and was filled with 1
mm nominal quartz sand (2095 Granusil® silica sand, Unimin Corp., Ottawa, MN). The
liquid volume of the reactor is approximately 30 L, including the sand pore volume and
influent and effluent reservoirs.
192
Figure 7.1. The radial flow bioreactor and NMR logging tool were housed in a Faraday
cage to reduce detection of electromagnetic noise from the laboratory.
Media and Injection Strategy
Two kinds of substrate media were used in this study, a bacterial growth medium
(growth medium) and a calcite mineralization-promoting medium (calcium medium).
Both were urea- and yeast extract-based (1 g/L yeast extract (Arcos Organics, Gheel,
Belgium), 20 g/L urea, 1 g/L NH4Cl, and 24 g/L NaCl). The calcium medium contained
an added 49 g/L CaCl2∙2H2O. Commercial-grade chemicals were used for urea (Urea
193
Fertilizer, Espoma, Millville, NJ), calcium chloride (various brands of commercial ice
melt), and sodium chloride (Morton Table Salt, Chicago, IL). Media were mixed just
prior to use in a non-sterile manner using tap water.
A pulsed-flow injection strategy promoted an even distribution of CaCO3
precipitation by balancing reaction and transport rates [30]. Each 30 L pulse of substrate
was pumped at a flow rate of 1 L/min, producing a pore velocity of approximately 0.4
cm/min and ensuring that the fresh substrate would penetrate the full width of the
sandpack. Calcium medium was injected four times per day during the biomineralization
phase Days 4 – 7. A 2-hour batch reaction period (no flow) followed each injection of
calcium medium. One pulse of growth medium was injected each evening to stimulate
the bacteria for the following day’s calcium medium injections. A 10 L brine rinse (24
g/L NaCl) was injected into the reactor first each morning to reduce mixing of the two
substrate media in the influent reservoir and minimize clogging of the slotted pipe.
Bacterial Culture
The bacteria used in this experiment, Sporosarcina pasteurii (ATCC 11859),
formerly known as Bacillus pasteurii, is widely used in laboratory experiments related to
urea hydrolysis and biomineralization [11]. S. pasteurii is a non-pathenogenic natural
soil organism capable of producing relatively large amounts of the urease enzyme needed
to catalyze urea hydrolysis [11]. For the inoculum, one mL of frozen stock of S. pasteurii
was cultured in 100 mL of growth medium on a shaker table at 150 rpm for 24 hours.
The 100 mL culture was then added to 10 L fresh growth medium and mixed on a stir
plate at 1150 rpm for 24 hours. Finally, the 10 L culture was added to 20 L of fresh
194
growth medium and mixed as before to produce a final inoculum volume of 30 L. No
attempt was made to maintain a monoculture in the inoculum or in the reactor.
The reactor was inoculated by first injecting 5 L of fresh growth medium to
condition the reactor at a flowrate of 1 L/min, followed by the 30L inoculum. An
additional 5 L of fresh growth medium was injected last. Bacteria were allowed to attach
to the sand for approximately 15 hours with no flow before the first injection of calcium
medium. There was no calcium present in the reactor during the 3-day control period or
during inoculation. The initial period was used as the control. Note that previous
experiments have shown no permeability reduction was achieved when urea and calcium
containing solutions were injected into glass bead filled columns that were not inoculated
with ureolytic microbes [31].
NMR Measurements
Low-field NMR measurements typically consist of repeated scans which are
stacked and averaged to reduce noise in the data. In this study, two experiments were
conducted sequentially and together constitute one CPMG scan for measurement of T2
relaxation. Experiment 1 collects T1-weighted fast-decaying signal (tE=1.3ms, Tr=800ms,
54 echoes, 360 averages). Experiment 2, on the other hand, collects the signal from spins
with longer relaxation times (tE=1.3ms, Tr=5000ms, 334 echoes, 60 averages). All NMR
measurements were collected under no flow conditions. Measurements during the
control period, Days 1 – 3, consisted of 24 CPMG scans. Three (3) CPMG scans were
stacked and averaged for each daily measurement during the biomineralization phase,
195
Days 4 – 8, because of the timing of repeated substrate injections on a 2-hour cycle. Data
presented here was collected with a noise level of approximately 1.4%.
As only one tool was available on loan for a limited period, it was not possible to
run replicate experiments. However, previous work with this tool [20, 21] has allowed
multiple experimental runs whilst monitoring biofouling in both a sand pack and in the
subsurface. The tool’s performance has been consistent and repeatable.
Sampling
Influent and effluent samples were collected for each injection of brine, calcium
medium, and growth medium. The sample pH was measured shortly after collection,
then the sample was filtered (0.2 µm membrane, VWR International, Radnor, PA) and
refrigerated for later analysis with the Jung Assay [32] to evaluate the sample urea
concentration. After the final measurement on Day 8, the reactor was drained and
destructively sampled. The outer pipe was cut away in sections, leaving the
biomineralized sand annulus exposed for sampling (Figure 7.2). Twenty-four (24) cores
were collected: 2 radial cores of approximately 1 inch diameter (2.5 cm) and 3 inch
length (7.5 cm) at each of 3 depths were sampled in 4 orthogonal directions. Each core
sample was divided into 3 subsamples which were then weighed and subjected to nitric
acid digestion to remove the solid precipitates. The liquid was extracted for calcium
content analysis by ICP-MS using an Agilent 7500ce (Santa Clara, CA) with a collision
cell (helium mode) and a certified environmental calibration standard from CPI
International (product number 4400-12 1116NCO2). Additionally, micrograph images
were acquired using a Zeiss Supra 55VP scanning electron microscope (Zeiss, USA).
196
Biomineralized sand samples from the reactor and control sand samples were sputter
coated with iridium and high-resolution images were taken at 1.0 kV at a working
distance of 3—4 mm.
Figure 7.2. The biomineralized sand annulus was destructively sampled to quantify
CaCO3 precipitation. a) The outer pipes of the bioreactor were cut away to expose the
biomineralized sand annulus. A saw was used to cut the annulus into quarters, producing
the large crack shown here. b) Six radial core samples were collected from each quarter.
Results and Discussion
The influence of CaCO3 precipitation on the NMR signal response is reflected in
the daily signal decay curves and resulting T2 distributions where significant changes
were observed over time. Representative data, collected on Days 2, 4, 6, and 8, are
presented in Figure 7.3; the top panel shows fits to recorded signal decay curves, and the
197
bottom panel presents the T2 distributions for those decay curves. First, we will address
the change in water content which corresponds to a drop in the porosity of the sandpack.
Then we will discuss the relaxation distributions, which give insight to changes in the
relaxation mechanism.
Figure 7.3. Signal decay curves (top) and the corresponding T2 distributions (bottom) are
shown with each curve representing a day. Day 2 occurred during the control period.
Inoculation occurred on Day 3 (not shown). The calcium media injections occurred
between Day 4 – 7. The Day 8 data was collected prior to flushing the reactor with brine
and destructively sampling. Both graphs show fits to the raw data.
198
Water Content and Porosity
Decreasing signal amplitude over time is an indication of CaCO3 precipitation,
since CaCO3 will displace water in the pore volume. During the control period the initial
porosity indicated by the NMR-measured total water content was approximately 30%
which is slightly less than the 35-39% expected from a sand pack with relatively uniform
grains. The observation of entrained air leaving the system after the first flow pulse
following inoculation, and the subsequent increase in the water content signal on Day 4,
leads us to conclude that the sand pack was not fully saturated during the control period.
This also explains why the measured water content value of ~30% is less than the
expected value of 35-39%. The NMR-measured total water content in the reactor
decreased to approximately 76% of its original value between the control period (Day 2
data) and the end of the biomineralization phase on Day 8 (Figure 7.3, top panel, and
Figure 7.4). This reduction in total water content indicates that the pore volume within
the reactor decreased significantly during the biomineralization phase. If we consider
Day 4 to represent full saturation, then the NMR-estimated porosity reduction is 70% of
the initial value, indicating the sensitivity of the NMR measurement to partial saturation.
199
Figure 7.4. The measured total water content in the radial flow reactor decreased from
approximately 29% during the control period Days 1-3 to approximately 22% by Day 8.
Note that the increase on Day 4 is real and well outside expected error bounds. The
increase follows the observation of entrained air leaving the bioreactor, indicating the
desired fully saturated state may not have been obtained until after the control period.
CaCO3 formation was confirmed by scanning electron microscopy (SEM). There
appeared to be a relatively uniform CaCO3 coating on the sand samples viewed with
SEM. Figure 7.5 shows an SEM micrograph showing the crystals formed on a grain of
sand from the reactor (a) and the surface of a control sand grain (b). The surface of the
CaCO3 -encrusted sand reveals micro-scale cavities and pores between crystals. No
bacteria were visible in the sand samples viewed with SEM; it is most likely that the cells
are entombed within the crystals.
200
Figure 7.5. SEM image of a) CaCO3 crystals attached to a grain of sand from the reactor
following 4 days of MICP and b) control sand without CaCO3. Scale bar is 20 µm. Note
that the CaCO3 crystals completely cover the sand surface and the sand is not visible in
(a), whereas in (b) the smooth sand surface is observed.
Several methods were applied to estimate the volume of CaCO3 formed in the
reactor in order to independently determine the reduction in pore volume achieved.
These methods include a mass balance on urea, ICP-MS detection of Ca2+, and
gravimetric methods. An initial porosity estimate of 37%, typical for the sand in the
reactor, was used in these calculations. Because the reactor was not fully saturated during
the control period, the total porosity is greater than the NMR water content. The results
of these three methods are in good agreement with each other and support the NMR data
showing a significant pore volume reduction due to calcite precipitation. Mass balance
on urea: Influent and effluent samples of each pulse of media were analyzed using the
Jung Assay [32] to quantify the urea content. A mass balance on urea showed that
approximately 4.2 kg of urea was consumed within the reactor, stoichiometrically
producing approximately 6.9 kg of calcite. This mass of calcite would occupy at
maximum approximately 15% of the pore space in the sand annulus. Since CaCO3 also
201
formed in the tubing and on the reactor walls, we consider the urea mass balance method
to provide an approximation of the upper bound of CaCO3volume. ICP-MS: ICP-MS
was used to measure the concentration of Ca2+ in the acid extraction liquid from 24
samples of biomineralized sand from the reactor. A mean value of 9.36 g/L Ca2+ was
obtained with a sample standard deviation of 1.89 g/L, which equates to an average total
mass of 6.3 kg CaCO3 within the sand. The ICP-MS data translates to an average pore
volume reduction of approximately 12% (+/- 2.4%). Gravimetric method: The 24 sand
samples were also weighed before and after the acid digestion removed the precipitate,
resulting in an average mass of calcite of 63.6 mg CaCO3/g sand with a sample standard
deviation of 13 mg/g. By this method, the average total mass within the sandpack was
5.5 kg calcite. The gravimetric method indicates that approximately 11% (+/- 2.2%) of
the pore space in the sand annulus was occupied by CaCO3at the end of the experiment.
Unlike the mass balance method, ICP-MS and gravimetry account only for CaCO3
attached to the sand. On the basis of these complementary and independent methods, we
estimate that CaCO3 occupied approximately 11 – 12% of the pore space in the sandpack
by Day 8 of the experiment.
This estimated porosity reduction is significantly higher than those previously
reported in other NMR/MICP studies [8, 19]. The pulsed flow injection strategy used
here promotes relatively uniform CaCO3 precipitation, as evidenced by the small standard
deviation of the samples collected from the reactor. The uniform calcite precipitation
implies spatially uniform porosity reduction. Consequently, only an insignificant
reduction in permeability was observed in this study.
202
Compared to the methods described above which found final porosity to be
approximately 88% of the original value, NMR measurements of water content
overestimate the porosity reduction achieved. Final NMR water content was 76% of the
initial value, or 70% of the Day 4 value. The overestimation can be attributed to carbon
dioxide (CO2) gas production inside the reactor. The excess CO2 produced by microbial
oxidation of the yeast extract in the substrate can be trapped in the reactor pore spaces,
displacing water and reducing signal amplitude without changing the pore geometry. Gas
formation was also observed in previous NMR studies of MICP [8, 19]. Furthermore,
signal decaying in the interval before the first echo acquisition will underestimate the
water content and may explain in part the NMR overestimation of porosity reduction.
Relaxation
The tall initial peak (Day 2 data) in the bottom panel of Figure 7.3, centered about
approximately 600-700 ms and associated with water in large pores, first increases then
decreases in amplitude over time as the biomineralization phase proceeds. At the same
time, there is an increase in both the occurrence of very fast T2 relaxation times less than
10 ms, and an increase in the proportion of spins experiencing very long relaxation times,
greater than 1000 ms. At the left-hand limit of the T2 distribution (Figure 7.3, bottom),
the NMR logging tool cannot capture NMR signal that decays faster than the
measurement echo time (tE=1.3ms). We note that since the time of this study, the echo
time of the Javelin tool has been reduced to 0.7 ms. At the right of the distribution,
signals with T2 between 1-5 s are not tightly resolved on the T2-axis because the signal is
sampled only to 500 ms. However, the amplitude of these long signals is accurately
203
measured (Figure 7.3, top). By Day 8 of the experiment, the mean log T2 time of the
distribution had increased to greater than 1000 ms from approximately 650 ms during the
control period.
The data shows that T2 relaxation in the macro-pores of the sandpack is more
significantly affected by the reduction in 𝜌𝜌 than by the decrease in the macro-pore
dimension. As seen in the SEM images (Figure 7.5a), the CaCO3 crystals are on the
order of 101 µm thick. In a large pore on the order of 102 µm in diameter, there is a
relatively minor change in pore dimension due to calcite precipitation. On the other
hand, the relatively thin and uniform coating of CaCO3 crystals is sufficient to minimize
molecular interactions between the pore fluid and paramagnetic species on the sand,
making the surface much less likely to induce relaxation. The combination of a large
change to 𝜌𝜌 and a small change to the pore size explains the lengthening of the overall
mean log T2 relaxation time. At the same time, CaCO3 precipitation also creates micro-
pores between and within the crystals. In these pores, the pore size effect dominates and
T2 relaxation occurs rapidly for the small population of spins within the crystals.
Previous NMR/MICP studies reporting the opposite [19] or no relaxation effect
[8] are not at odds with this interpretation of the data. Both previous studies used smaller
diameter (~100 – 250 µm) model porous media (borosilicate or polystyrene beads,
respectively) with a low initial 𝜌𝜌 and small initial pore size, where we would expect more
potential influence from a change in the pore geometry than from a reduction in ρ. We
expect if CaCO3 precipitation had continued to progress in the current experimental
system, the reduction in the macro-pore dimension would eventually become the
204
dominate influence, driving relaxation times to decrease. Thus, the potential complexity
of the relaxation response leaves open the possibility of different relaxation signatures in
other porous materials where pore sizes or surface properties are more heterogeneous.
Our results show that changes in NMR signal response due to MICP include 1) a
decrease in signal amplitude over time, indicating a reduction in porosity, and 2) a
lengthening of the overall T2 relaxation time in the quartz sand of the bioreactor. NMR
measured water content in the reactor decreased to approximately 76% of the initial
value, which corresponds well to the measured reduction in porosity to approximately
88% of the typical initial value. The extent of the decrease in porosity, and the
corresponding minimal change in permeability, is related to the pulsed-flow injection
strategy employed to achieve the MICP. T2 relaxation distributions bifurcated from a
single mode centered about approximately 650 ms during the control period into a very
fast decaying population (T2 less than 10 ms), associated with water in the porous CaCO3,
and a larger population with relaxation times greater than 1000 ms, corresponding to the
bulk water in the large crystal-coated pores. Slower relaxation is caused by CaCO3
crystals on the mineral surface of the macro-pores shielding paramagnetic species from
the pore fluid, reducing 𝜌𝜌 of the pore. In the CaCO3 micro-pores, the pore size effect
dominated and enhanced relaxation. Future work will evaluate the NMR signal response
to MICP in natural soils and porous rock where surface relaxivity and pore sizes are more
heterogeneous. This study demonstrates that a NMR well-logging tool is sensitive to
MICP and has potential as a sensor for biomineralization in field applications where
optical or destructive monitoring methods are not possible.
205
References
1. Stoodley, P., et al., Biofilms as complex differentiated communities. Annual
Review of Microbiology, 2002. 56: p. 187-209.
2. Beech, I.B. and J. Sunner, Biocorrosion: towards understanding interactions
between biofilms and metals. Current opinion in biotechnology, 2004. 15(3): p.
181-6.
3. Costerton, J.W., P.S. Stewart, and E.P. Greenberg, Bacterial biofilms: A common
cause of persistent infections. Science, 1999. 284(5418): p. 1318-1322.
4. de Kreuk, M., J.J. Heijnen, and M.C.M. van Loosdrecht, Simultaneous COD,
nitrogen, and phosphate removal by aerobic granular sludge. Biotechnology and
Bioengineering, 2005. 90(6): p. 761-769.
5. Cunningham, A.B., et al., Subsurface biofilm barriers for the containment and
remediation of contaminated groundwater. Bioremediation Journal, 2003. 7(3-4):
p. 151-164.
6. Decho, A.W., Overview of biopolymer-induced mineralization: What goes on in
biofilms? Ecological Engineering, 2010. 36(2): p. 137-144.
7. Ferris, F.G., et al., Kinetics of calcite precipitation induced by ureolytic bacteria
at 10 to 20 degrees C in artificial groundwater. Geochimica Et Cosmochimica
Acta, 2003. 67(8): p. 1701-1722.
8. Fridjonsson, E.O., et al., NMR measurement of hydrodynamic dispersion in
porous media subject to biofilm mediated precipitation reactions. Journal of
Contaminant Hydrology, 2011. 120-21: p. 79-88.
9. De Muynck, W., N. De Belie, and W. Verstraete, Microbial carbonate
precipitation in construction materials: A review. Ecological Engineering, 2010.
36(2): p. 118-136.
10. DeJong, J.T., et al., Bio-mediated soil improvement. Ecological Engineering,
2010. 36(2): p. 197-210.
11. Phillips, A.J., et al., Engineered applications of ureolytic biomineralization: a
review. Biofouling, 2013. 29(6): p. 715-733.
12. van Paassen, L.A., et al., Quantifying Biomediated Ground Improvement by
Ureolysis: Large-Scale Biogrout Experiment. Journal of Geotechnical and
Geoenvironmental Engineering, 2010. 136(12): p. 1721-1728.
206
13. Mortensen, B.M., et al., Effects of environmental factors on microbial induced
calcium carbonate precipitation. Journal of Applied Microbiology, 2011. 111(2):
p. 338-349.
14. Mitchell, A.C., et al., Microbially Enhanced Carbon Capture and Storage by
Mineral-Trapping and Solubility-Trapping. Environmental Science and
Technology, 2010. 44(13): p. 5270-5276.
15. Phillips, A.J., et al., Fracture Sealing with Microbially-Induced Calcium
Carbonate Precipitation: A Field Study. Environmental Science & Technology,
2016. 50(7): p. 4111-4117.
16. Phillips, A.J., et al., Potential CO2 Leakage Reduction through Biofilm-Induced
Calcium Carbonate Precipitation. Environmental Science & Technology, 2013.
47(1): p. 142-149.
17. Behroozmand, A.A., K. Keating, and E. Auken, A Review of the Principles and
Applications of the NMR Technique for Near-Surface Characterization. Surveys
in Geophysics, 2015. 36(1): p. 27-85.
18. Walsh, D., et al., A small-diameter NMR logging tool for groundwater
investigations. Groundwater, 2013. 51(6): p. 914-926.
19. Sham, E., et al., Monitoring bacterially induced calcite precipitation in porous
media using magnetic resonance imaging and flow measurements. Journal of
Contaminant Hydrology, 2013. 152: p. 35-43.
20. Kirkland, C.M., et al., Biofilm Detection in a Model Well-Bore Environment
Using Low-Field NMR. Groundwater Monitoring and Remediation, 2015: p.
DOI:10.1111/gwmr.12117.
21. Kirkland, C.M., et al., In Situ Detection of Subsurface Biofilm Using Low-Field
NMR: A Field Study. Environmental Science & Technology, 2015. 49(18): p.
11045-11052.
22. Grunewald, E. and R. Knight, A laboratory study of NMR relaxation times in
unconsolidated heterogeneous sediments. Geophysics, 2011. 76(4): p. G73-G83.
23. Kleinberg, R.L. and M.A. Horsfield, Transverse relaxation processes in porous
sedimentary rock. Journal of Magnetic Resonance, 1990. 88(1): p. 9-19.
24. Kleinberg, R.L., Kenyon, W.E., Mitra, P.P., Mechanism of NMR relaxation of
fluids in rock. Journal of Magnetic Resonance Series A, 1994. 108: p. 206-214.
207
25. Bryar, T.R., C.J. Daughney, and R.J. Knight, Paramagnetic effects of iron(III)
species on nuclear magnetic relaxation of fluid protons in porous media. Journal
of Magnetic Resonance, 2000. 142(1): p. 74-85.
26. Foley, I., S.A. Farooqui, and R.L. Kleinberg, Effect of paramagnetic ions on NMR
relaxation of fluids at solid surfaces. Journal of Magnetic Resonance Series A,
1996. 123(1): p. 95-104.
27. Kenyon, W.E. and J.A. Kolleeny, NMR surface relaxivity of calcite with adsorbed
Mn2+. Journal of Colloid and Interface Science, 1995. 170(2): p. 502-514.
28. Coates, G.R., et al., A new characterization of bulk-volume irreducible using
magnetic resonance. Log Analyst, 1998. 39(1): p. 51-63.
29. Zhang, Q., et al. Some exceptions to default NMR rock and fluid properties. in
Transactions of the SPWLA Annual Logging Symposium (Society of Professional
Well Log Analysts). 1998.
30. Ebigbo, A., et al., Darcy-scale modeling of microbially induced carbonate
mineral precipitation in sand columns. Water Resources Research, 2012. 48.
31. Wheeler, L.A., Establishment of ureolytic biofilms and their influence on the
permeability of pulse-flow porous media column systems. 2009, Montana State
University, Bozeman, MT.
32. Jung, D., et al., New colorimetric reaction for endpoint, continuous-flow and
kinetic measurement of urea Clinical Chemistry, 1975. 21(8): p. 1136-1140.
208
CHAPTER EIGHT
NMR INVESTIGATION OF WATER DIFFUSION IN
DIFFERENT BIOFILM STRUCTURES
Contribution of Authors and Co-Authors
Manuscript in Chapter 8
Author: Maria P. Herrling
Contributions: Conceived the concept and designed the experiment. Conducted diffusion
measurements, characterized the biomass. Analyzed the data and wrote the manuscript.
Co-Author: Jessica Weisbrodt
Contributions: Conducted diffusion experiments. Collected and characterized the
biomass. Analyzed data.
Co-Author: Catherine M. Kirkland
Contributions: Assisted with data collection for D-T2 measurements. Assisted with data
analysis for D-T2 measurements. Provided feedback and comments on manuscript.
Co-Author: Nathan H. Williamson
Contributions: Assisted with data analysis using the Gamma distribution. Provided
feedback and comments on manuscript.
Co-Author: Susanne Lackner
Contributions: Discussed the data and Provided feedback and comments on manuscript.
Co-Author: Sarah L. Codd
Contributions: Provided feedback and comments on the manuscript.
209
Co-Author: Joseph D. Seymour
Contributions: Provided supervision and oversight on D-T2 data collection and analysis.
Provided feedback and comments on the manuscript.
Co-Author: Gisela Guthausen
Contributions: Conceived the concept and designed the experiment. Developed data
analysis tools, analyzed the data, and discussed the data. Provided feedback and
comments on the manuscript.
Co-Author: Harald Horn
Contributions: Designed the experiment and discussed the data. Provided feedback and
comments on the manuscript.
210
Manuscript Information Page
Maria P. Herrling, Jessica Weisbrodt, Catherine M. Kirkland, Nathan H. Williamson,
Susanne Lackner, Sarah L. Codd, Joseph D. Seymour, Gisela Guthausen, Harald Horn.
Biotechnology and Bioengineering
Status of Manuscript:
____ Prepared for submission to a peer-reviewed journal
__X_ Officially submitted to a peer-review journal
____ Accepted by a peer-reviewed journal
____ Published in a peer-reviewed journal
Wiley-Blackwell
Submitted February 10, 2017
211
NMR INVESTIGATION OF WATER DIFFUSION
IN DIFFERENT BIOFILM STRUCTURES
Abstract
Mass transfer in biofilms is determined by diffusion. Different mostly invasive
approaches have been used to measure diffusion coefficients in biofilms, however data on
heterogeneous biomass under realistic conditions is still missing. To non-invasively
elucidate fluid-structure-interactions in complex multispecies biofilms pulsed field
gradient-nuclear magnetic resonance (PFG-NMR) was applied to measure the water
diffusion in five different types of biomass aggregates: one type of sludge flocs, two
types of biofilm, and two types of granules. Data analysis is an important issue when
measuring heterogeneous systems and is shown to significantly influence the
interpretation and understanding of water diffusion. With respect to numerical
reproducibility and physico-chemical interpretation, different data processing methods
were explored: (bi)-exponential data analysis and the Γ distribution model. Furthermore,
the diffusion coefficient distribution in relation to relaxation was studied by D-T2 maps
obtained by 2D inverse Laplace transform (2D ILT). The results show that the effective
diffusion coefficients for all biofilm samples ranged from 0.36 to 0.96 relative to that of
water. NMR diffusion was linked to biofilm structure (e.g. biomass density, organic and
inorganic matter) as observed by magnetic resonance imaging and to traditional biofilm
parameters: Diffusion was most restricted in granules with compact structures, and fast
diffusion was found in heterotrophic biofilms with fluffy structures. The effective
212
diffusion coefficients in the biomass were found to be broadly distributed because of
internal biomass heterogeneities, such as gas bubbles, precipitates, and locally changing
biofilm densities. Thus, estimations based on biofilm bulk properties in multispecies
systems can be overestimated and mean diffusion coefficients might not be sufficiently
informative to describe mass transport in biofilms and the near bulk.
Introduction
Biofilms or more generally biomass aggregates are pervasive in natural aquatic
systems [1, 2] as well as in technical systems [3]. Their substrate conversion depends on
bulk and internal structures influencing the mass transfer into the matrix. Generally, this
mass transfer in biomass aggregates is driven by diffusion and plays a key role for
metabolic activity [4-6]. The investigation and understanding of mass transfer and
substrate consumption is essential for the development of strategies to improve design
and operation of biofilm-based technical applications as well as for modeling. In-depth
knowledge is required - especially for multispecies biofilms under technical and realistic
conditions.
To experimentally explore fluid-structure-interactions, several analytical methods
have been applied. Imaging techniques [7] and micro sensors [8, 9] have been used the
most. Furthermore, mathematical modeling leads to mechanistic understanding of mass
transfer phenomena in complex biofilm systems [10]. Biofilm modeling in combination
with imaging indicates that rough biofilms show higher mass transfer of substrates
compared to smooth biofilms under stagnation conditions. The larger interfacial surface
213
of a rough biofilm provides better contact to substrates [11]. The diffusion coefficient D
of substrates into the biofilm is assumed constant over the biofilm depth [12]. In transport
models, D is usually set to 20-30 % less than Dwater [12]. Diffusion of water and nutrients
strongly varies between different biofilm systems, geometries, and growth conditions.
Actual mass fluxes inside biofilms are unknown, and a generalization of mass transport is
not possible. Therefore, data on heterogeneous systems is needed to address measures for
mass transport in real systems. Pulsed-field gradient nuclear magnetic resonance (PFG-
NMR) allows the measurement of translational diffusion completely non-invasively and
non-destructively [13, 14], also in micro-porous systems and in restricted geometry.
In biofilm research, alginate is often used as model system due to the chemical
composition which is comparable to the biofilms’ extracellular polymeric substances
(EPS). Gel heterogeneities were detected in transverse relaxation T2 maps [15]. Filtration
processes have been imaged in hollow fiber membranes using alginate as model system
[16]. Basic knowledge gained from investigations of artificial biofilms can be transferred
to real biofilm systems. Monoculture and multispecies biofilms were investigated by
NMR flow and diffusion in porous media and flow cells to study water dynamics and
biofilm growth at different time and length scales [17-21]. Furthermore, combined
diffusion and Magnetic Resonance Imaging (MRI) linked mass transfer to biofilm
structure [7, 20]. Diffusion strongly depends on biofilm systems, growth conditions and
biofilm geometries. For example, in methanogenic granular sludge internal
heterogeneities significantly influenced the mass transfer [22]. Strong correlations
between the diffusion of substrates and biofilm parameters have been reported [4, 23, 24].
214
For example, Renslow et al. (2010) correlated the effective diffusion coefficient with
biofilm depth by means of PFG-NMR and imaging in Shewanella oneidensis biofilms.
The same was found for phototrophic Phormidium biofilms [25]. Other studies also
confirm that D is biofilm-specific and depth-dependent with a linear decrease of D with
biofilm depth [26, 27].
Mass transport and diffusion in biofilms are difficult to analyze and to model
because of the high complexity of these biomass aggregates. Apart from the influence of
specific compounds, PFG-NMR measurements and data interpretation deserve scrutiny.
Different data processing approaches are summarized in Röding, Bernin [28], [29]. The
most commonly used data processing approach for PFG-NMR data is the (bi)-
exponential fit. Diffusion coefficients are obtained by fitting a single exponential function
to the measured signal attenuation, leading to effective or apparent diffusion coefficients.
An alternative and recently introduced approach is the Γ distribution model which offers
mean diffusion coefficients and their distribution width. Different data processing
schemes are compared for biomass aggregates in this study to show their limitations and
relevance. For mathematical description we refer to [28], Röding, Williamson [29].
Water diffusion coefficients in multispecies biofilms with diverse geometries and data
processing schemes are compared in this paper. To identify fluid-structure interactions,
five different types of biofilms or biomass aggregates were investigated: sludge flocs,
fluffy and compact biofilms grown on carriers, and aerobic and (an)aerobic granules. The
goals of the study were to i) characterize the biomass by means of physico-chemical
parameters and imaging, ii) compare Dwater in different structures to identify correlations
215
between biomass properties and water diffusion, iii) compare PFG-NMR data processing
schemes, including (bi)-exponential fit, Γ distribution and 2D inverse Laplace transform.
Materials and Methods
Biofilm Sample Preparation
Activated sludge flocs, carrier based biofilms, and granules were chosen to cover
a broad range of physical morphologies which are technically relevant, e.g. for
wastewater treatment. The properties and functionalities are summarized in Table 8.1.
Sludge flocs (sludge) were collected from a full-scale wastewater treatment plant in
Weinheim (Germany). The sludge is a mixed culture of auto- and heterotrophic biomass
used for carbon (C), nitrogen (N) and phosphorus (P) removal from wastewater. The
sludge was rinsed and sieved to a size fraction < 200 µm prior to characterization. Two
different carrier based biofilms were investigated. Fluffy biofilms (biofilm_1) grown on
K1 carrier materials (plastic carrier, diameter: 10 mm, AnoxKaldnes, Sweden) were
cultivated in a laboratory-scale moving bed biofilm reactor fed with acetate [30].
Biofilm_1 comprised heterotrophic biomass and was mainly used for removal of easily
degradable carbon compounds. Compact biofilms (biofilm_2) on K3 (plastic carrier,
diameter: 25 mm, AnoxKaldnes, Sweden) were obtained from a full-scale wastewater
treatment plant (AnoxKaldnes' Biofarm) in Malmö, Sweden. Biofilm_2 comprised
multiple species, i.e. heterotrophs, nitrifiers, denitrifiers and anammox bacteria [31] and
was mainly used for N removal. Two different types of granules, i.e. granulated biofilms,
were also selected for the investigation. Aerobic granules (granules_1) were cultivated in
216
a laboratory-scale sequencing batch reactor (SBR) with acetate as the main substrate.
Granules_1 were used for N, P and easily degradable C removal. Large sized granules
with an approximate diameter of 10 mm were manually selected. Additionally,
(an)aerobic granules (granule_2) were collected from a side-stream SBR of a full-scale
wastewater treatment plant in Heidelberg (Germany). Granules_2 mainly contributed to
the removal of N and some slowly degradable C compounds. Granules_2 were rinsed and
sieved to a size fraction of 700-900 µm.
Table 8.1: Characterization of the biofilms. The abbreviation n.m. means not measured.
Data marked with asterisk (*) refer to diameter of the granules. “C” refers to carbon
removal, “N” to nitrogen removal and “P” to phosphorous removal. The density of water
at 20°C is 998 g/L.
Parameter Sludge Biofilm_1 Biofilm_2 Granules_1 Granules_2
microbes autotrophic heterotrophic autotrophic autotrophic autotrophic
removal C/N/P C N C/N/P N/C
TSS 6 g/L TSS 86 mg
TSS/carrier
270 mg
TSS/carrier
40 mg
TSS/granule
3 g/L TSS
VSS [%] 80 90 69 78 80
biofilm
thickness [µm]
< 200* < 200 < 1000 10 000* 400-700*
biofilm
density [g/L]
1033
±15
<1010
1047
±4
1029
±2
1046
±14
Ca [g/kg TSS] 37 n.m. 490 97 286
P [g/kg TSS] 53 n.m. 132 49 72
Fe [g/kg TSS] 37 n.m. 590 0.6 327
S [g/kg TSS] 10 n.m. 6 6.8 84
Si [g/kg TSS] 2 n.m. 65 0.2 37
217
All biofilms were rinsed prior to the experiment to remove particulate matter and
stored in tap water at 4°C. Biofilms were imaged by light microscope SMT4 (Mikroskop
Technik Rathenow) in combination with a DSLR camera (Canon EOS 600D) and
characterized by their total suspended solid concentration (TSS) and volatile suspended
solid concentration (VSS) according to DIN-EN-12880 (2001). The biomass density was
determined by pycnometer measurement (n=4, 10 ml, Blaubrand) according to DIN ISO
35079 and the solids density was calculated according to Loosdrecht, Nielsen [32].
Element analysis was conducted by atomic emission spectroscopy (ICP-OES, Varian
VistaPro, Agilent Technologies, l detection limit: 10 µg/L) after acid digestion. ICP-OES
data for biofilm_1 is not available due to low biofilm mass.
MRI and PFG-NMR
MRI. MRI experiments were performed on a 200 MHz MRI instrument (Bruker
Avance 200 SWB, Bruker BioSpin GmbH, Rheinstetten, Germany) with a magnetic flux
B0 of 4.7 T, 150 mm vertical bore and equipped with a Bruker gradient system micro2.5
and a 1H-NMR bird-cage (25 mm inner diameter). All measurements were performed
temperature controlled at 25°C using the Bruker software ParaVision. The often used
multi-slice multi-echo imaging sequence (MSME) [14, 33] was applied to acquire
predominantly proton density-weighted images due to the minimum echo time used in the
experiments. The biofilms were imaged using the same acquisition parameters as in a
previous study [34] with TR = 10 s, τE = 50 ms, number of averages = 4, pixel matrix
128x128, slice thickness 0.8 mm. To maintain the original structure, sludge and
granules_2 were filled into glass vials with 1 mL of tap water from Karlsruhe to avoid
218
osmotic stress. The Karlsruhe tap water (average values of 2015, Stadtwerke Karlsruhe)
had a pH of 7.2, total organic carbon concentration of 0.84 mg/L and an electrical
conductivity of 653 µS/cm (c(Ca2+) = 110 mg/L, c(Na+) = 11.1 mg/L, c(K+) = 1.7 mg/L,
c(Cl-) = 22.8 mg/L, c(HCO-) = 232 mg/L).
Wet biofilm_1, biofilm_2 and granules_1 were placed into plastic wrap (without
bulk water) and directly inserted into the bird-cage. The measured signal intensity is
encoded on a gray-scale and physically corresponds to the proton density, slightly
weighted by the transverse relaxation T2. The slightly T2-weighted proton density images
allow the discrimination of the main components within the biofilm system, being either
1H containing liquid (signal) or solid or gas (no signal intensity in MSME images).
Additionally, materials without 1H, i.e. minerals, do not show up in the 1H images.
Diffusion Measurements. Translational motion was measured by PFG-NMR, i.e.
the pulsed application of magnetic field gradients which encode and decode the position
across an ensemble of molecules at different times. The time between the encoding and
decoding pulses is the diffusion time Δ. The PFG-NMR acquisition parameters are
summarized in Table 8.2. Biomass samples were measured using the PFG stimulated
echo (PFG-STE) [13] on the above described NMR spectrometer equipped with the
Bruker software Topspin and a Diff30 probe, which allowed z-gradients g of up to 12
T/m. Acquisition parameters were chosen according to Table 8.2, measurement I. The
logarithmic, normalized signal attenuation ln(S/S0) was measured as a function of q,
where q = γδg, γ being the gyromagnetic ratio, δ the gradient pulse duration, and g the
gradient amplitude. Complementary measurements were performed on a 250 MHz MRI
219
tomograph (Bruker Avance 250 SWB, Bruker BioSpin GmbH, Rheinstetten, Germany)
also equipped with a Diff30 Probe (maximum gradient 17 T/m) employing acquisition
parameters of measurement II and III, (Table 8.2). Images taken at the beginning and end
of the experiments showed stable biofilm structure. All diffusion measurements were
performed as single measurements at 25°C.
Table 8.2: Acquisition parameters for PFG-NMR diffusion measurements. Measurement
II and III were used to characterize biofilm_1.
Acquisition
Parameter
Measurement I Measurement II Measurement III
PFG-STE PFG-STE D_T2_STE
gradient pulse
duration
3 ms 1 ms 1.136 ms
diffusion time 40, 100, 200, 400,
500, 800 ms
50, 100, 200 ms 100 ms
first rf pulse delay 4.26 ms 21.21 ms 21.21 ms
recycle delay 8 s 900 ms 12 s
diffusion gradient
amplitude
0.016 T/m to 0.5 T/m
linear in 32 steps
-0.32 T/m to +0.32
T/m linear in 64 steps
0.015 T/m to 0.2 T/m
linear in 64 steps
number of scans 4 16 32
number of dummy
scans
1 0 4
gradient direction z z z
220
Data Processing
(Bi)-exponential Model. Dwater was obtained by fitting a single exponential
function to the measured signal attenuation as it is expected for a homogeneous liquid of
small molecules [35]. In the biofilm matrix, the motion of water is partially restricted.
The samples are highly heterogeneous; therefore, a D-distribution is expected. However,
as little is known about the details and mechanisms, the question about an appropriate
model is essential for a confident data interpretation. As a first approach, the commonly
used (bi)-exponential function 𝑆𝑆
𝑆𝑆0
= 𝐴𝐴1 exp �−𝐷𝐷1𝑞𝑞2 �𝛥𝛥 − 𝛿𝛿3�� + 𝐴𝐴2 exp �−𝐷𝐷2𝑞𝑞2 �𝛥𝛥 −
𝛿𝛿
3
�� (eq. 1) with the gradient duration 𝛿𝛿. D1 and D2 in biofilms were obtained, meaning
that there is a fast (D1) and a slow (D2) diffusion fraction described by the relative
amplitudes Ai. D1 makes up more than 80 % of the signal amplitude and is the dominant
part. A small portion with signal amplitudes of 4-20% can be associated with D2 (in the
range of 10-10 to 10-11 m2/s), which is significantly lower than D1. Due to the dominant
relative signal amplitude, the discussion focuses on D1 in the following. The relative
effective diffusion coefficient is defined by fD_exp = D1/ Dwater (eq. 2), which is used to
compare the diffusion properties of the samples.
Data was processed using self-written scripts in Matlab® (version R2012a,
Matlab Works Inc.; Natick, Massachusetts, USA) which take special care of the small
amplitude of the second diffusion contribution. This (bi)-exponential rather than a mono-
exponential approach is essential for accurately describing the majority of the data to
avoid miss- or over interpretation due to numerical errors during data processing. Only
data points with signal above the noise level were used in the fits.
221
Gamma Distribution Model. For heterogeneous systems, the Stejskal-Tanner
approach for self-diffusion of homogeneous liquids composed of small molecules does
not apply necessarily. D-distributions are physically more meaningful than a discrete
number of D in these highly complex and heterogeneous systems. Apart from the
conventionally used (bi)-exponential function for modelling the signal attenuation, a
recently developed approach by Röding et al. was used to reveal the distribution of
diffusion coefficients [28, 29]. The Γ distribution function is a generalization of
distribution functions known in mathematical statistics. One of the advantages is that the
equation for the experimentally observed magnetization as a function of q2 becomes
rather simple and accurate, for details we refer to Röding et al. (2012). Data was
processed using self-written scripts (Matlab®) to obtain Dmean and the distribution width
σ. The relative effective diffusion coefficient within the Γ distribution is defined by fD_g =
Dmean/Dmean, water (eq. 3). The ratio between Dmean/D1 listed in Table 8.4 indicates the
difference between the data analysis approaches and underlines the importance of using a
physical and numerically appropriate model.
2D Inverse Laplace Transform. Another approach to model data deviating from a
strictly mono-exponential behavior is a purely numerical approach, the inverse Laplace
transform (ILT). Diffusion measurements can additionally be combined with relaxation
measurements, and the data can be Laplace transformed in two dimensions. In the present
case, D-T2 maps were obtained by performing a 2D inverse Laplace transform (2D ILT)
[19, 36, 37] to correlate diffusion and relaxation measurements for biofilm_1.
222
Results and Discussion
Biofilm and Biomass Characterization:
Comparison of Common Quantities and Images
Biomass is commonly characterized by i.e. thickness, density, elemental
composition, TSS, and VSS, which do not consider spatial and structural
inhomogeneities. On the other hand, mass transport and diffusion properties strongly
depend on these spatial inhomogeneities but are not unique in the sense of a one-to-one
correspondence of structure and diffusion coefficient. In a first approach, the microscopic
findings and the macroscopic chemical parameters were collected and compared to link
these physical parameters to water diffusion.
As evident in stereomicroscopic and MRI images (Table 8.3), sludge consisted of
a complex, but loose network of flocs with a small-scale heterogeneity. The measured
biomass density of 1033 g/L is within the expected biomass densities for active sludge
flocs with 1020-1060 g/L [38] and is higher than pure water density (998 g/L at 20°C).
The mineral content was relatively low with 37 g Ca per kg TSS. It is expected that the
water diffusion is hindered in the sludge flocs, and is therefore, slower than free water.
Compared to the heterogeneous structure of the sludge flocs, biofilm_1 comprised
a fluffy and homogenous structure (Table 8.3). The fluffiness of the biofilm is indicative
for the low biofilm density (not distinguishable from free water) and high VSS of 90%.
The pycnometer measurements for biofilm_1 were not reproducible due to low density
and low biomass concentration. Furthermore, spatial variation in biomass density, which
influences water dynamics in the biomass, cannot be captured by a simple pycnometer
223
measurement. Another approach is to get insight into the biomass structure and density
by MRI. MRI image contrast delivers qualitative information: biofilm_1 was visible in
the slightly T2-weighted images due to the reduced T2 in the biofilm compared to bulk
water. This effect is based mainly on the exchange of protons between EPS biopolymers
(mainly OH-groups) and water molecules [39] as well as by the different molecular
mobility of molecules in biofilm matrix and water. In proton-density weighted images,
biofilm_1 was hardly distinguishable from bulk water because of similar proton
concentration [34]. Based on this knowledge, the density was estimated to be ~1010 g/L.
Biofilm_2 was cultivated for ~12 months in a WWTP, TSS and mineral
accumulation were significantly higher than in all other biofilms, therefore, diffusion is
expected to be highly restricted. The iron content was especially high due to iron-
accumulating bacteria (anammox) [31]. The internal structure was heterogeneous: outer
biofilm layers were less dense than layers close to the carrier material reflected in the
gray values of corresponding MRI voxels. Solids (precipitates, possibly CaCO3) and gas
bubbles (air, CO2, N2 gas due to microbial activity) appear black with the usual phase
susceptibility artifacts. The local heterogeneities in biofilm structure can be related to
advanced biofilm age, density and growth conditions. Those heterogeneities suggest that
there are regions in the biofilm which for which water diffusion is restricted to a larger
extent than in other regions.
224
Table 8.3: Summary of stereomicroscopic images and MRI (slightly T2-weighted image)
of the investigated biofilms. The resolution for MRI images was approximately 100 µm x
100 µm.
Biomass Stereomicroscopic Image MRI
Sludge
Biofilm_1
Biofilm_2
Granules_1
Granules_2
225
Under certain cultivation conditions, biofilm formation is also possible without
carrier material, commonly known as granules. Due to high shear stress applied during
cultivation, the surface of granules_1 and granules_2 were compact and smooth.
However, in the large granules_1 strong internal heterogeneities became apparent in the
MRI images, such as structural layers of microorganisms and precipitates which might
lower diffusion in the biomass. The biomass density was within the conventional range
for aerobic granules (1005 g/L-1070 g/L), which depends among others on the cultivation
conditions and volume fraction of solid [40-42]. Granules with a high solids content of
precipitates can have biomass densities up to 1200 g/L [43]. Granules_2 originated from
a WWTP accumulated significantly more minerals (e.g. Ca, P, Fe, P) than granules_1,
which were cultivated in a lab scale reactor. Furthermore, biomass density of granules_2
was in the same range as biofilm_2. As density is known to be one of the most important
parameters for diffusion in biomass (Renslow et al. 2010) the correlation between usually
measured parameters and diffusion coefficients were explored.
Diffusion of Water in the Presence of Biomass
To relate biofilm geometry and composition to water diffusion, PFG-NMR
diffusion experiments were carried out on different biomasses. As mentioned, the
biofilm’s heterogeneity poses some challenges for modelling diffusion in these systems.
The logarithmic signal attenuation was measured as a function of q2, results of biofilm_2
and granules_2 are shown in Figure 8.1 together to provide examples of the data and the
numerical fits. Diffusion data from the biofilms were interpreted with both the (bi)-
exponential fit (indicated as (bi.)-exp.-func.) and the gamma distribution model
226
(indicated by Γ distr.). The levels for good description of the data are depicted in form of
2 % and 4 % of the maximum amplitude, respectively, corresponding to the goodness of
Γ distr. fit to the data. A significant difference in the signal attenuation between the
samples is evident (inset in Figure 8.1), which can be explained by the different biomass
properties (see Table 8.1) influencing the water dynamics significantly. The compact
morphology and low VSS lead to a reduced water diffusion in granules_2 compared to
biofilm_2. The (bi.)-exp.-func. leading to D1 and D2 does not describe the measured data
sufficiently well, especially the first data points (inset in Figure 8.1). The reason is found
in the discrete approach which is physically not really appropriate for a multicomponent
system. A discrepancy results between D1 (from the (bi.)-exp fit) compared to Dmean
(from the Γ distr. fit). The Γ distr. represents the measured data above the noise level
significantly better than the (bi.)-exp.-func., resulting in lower residuals leading to a
better data quality and higher accuracy. The results for D1 and Dmean differ by 15 %
(biofilm_2) and 5 % (granules_2) which supports the above argument. The findings are in
agreement with previously published results for other heterogeneous samples such as
motor oils, where a better description of measured diffusion data was achieved by Γ distr.
[44]. The importance of an appropriate data analysis is obvious as the diffusion
coefficients influence simulation for substrate conversion in biological systems.
227
Figure 8.1: The logarithmic signal attenuation of biofilm_2 (top) and granules_2 (bottom)
are depicted exemplarily for all biofilm samples as a function of q2. The mono- and bi-
exponential decay functions and the gamma distribution model were fitted to the data.
The levels for good description of the data are depicted in form of 2% and 4% of the
maximum amplitude, respectively, corresponding to the goodness of fit of the gamma
model to the data. Deviations of the mono-exponential model are evident.
228
Influence of Biomass Structure on Water Dynamics
D1 and Dmean were obtained by (bi.)-exp.-func. and Γ distr., respectively (Table
8.4). Dwater = 2.09 m2/s·10-9 was slightly lower than the reported values [19, 45] due to
different water purity and temperature. When comparing the two data processing
approaches, Dmean and D1 for water were similar within approximately 2%. Dmean/D1 is
close to 1 for free water as expected.
Table 8.4: Summary of diffusion coefficients (at Δ 200 ms) of the used biofilms gained
by two different data analysis approaches. Diffusion coefficients D1 and Dmean obtained
by (bi)-exponential fit (indicated by (bi.)-exp.-func.) and gamma distribution (indicated
by Γ distr.), respectively. Data indicated by asterisk (*) were fitted using mono-
exponential function. Dr is the relative effective diffusion coefficient for both data
processing approaches, see Eq. 2 and 3.
Sample (bi).-
exp.-
func.
(bi).-
exp.-
func.
Relative
Amplitude
of D1
Contribution
Γdistr. Γdistr. Γ distr.
width
Ratio
D1
[m2/s]
·10-9
Dr_exp
[-]
Arel
[-]
Dmean
[m2/s]
·10-9
Dr_g
[-]
σ
[m2/s]
·10-10
Dmean
/D1
[-]
Water 2.09* 1 - 2.04 1 3.19·10-3 0.98
Sludge
1.45 0.69 0.93 1.62 0.79 8.55 1.11
Biofilm_1 1.76 0.84 0.97 1.96 0.96 5.17 1.12
Biofilm_2 1.47 0.70 0.95 1.69 0.83 7.18 1.15
Granules_1
1.27 0.61 0.94 1.60 0.78 4.96 1.26
Granules_2 0.76 0.36 0.84 0.73 0.36 4.71 0.95
229
D1 and Dmean measured in the presence of biofilms deviated by 20-30 % (D1/Dmean
between 0.95-1.26). This indicates a good overall agreement for both data processing
approaches despite the better fitting of the measured data by Γ distribution. Dmean is
typically slightly larger than D1 (Table 8.4) as in the (bi)-exponential fit only D1 with Arel
> 84% was considered. The minor part of the signal is associated with D2, which is in the
range of 10-10 to 10-11 m2/s. D2 might be assigned to intracellular water [46]. Previous
studies have also demonstrated that water diffusion coefficients in biofilms are lower than
in pure water due to restricted diffusion in the biomass matrix, cells, and EPS, resulting in
a wide range of diffusion coefficients between 10−9 and 10−13 m2/s [46]. fD_exp and fD_g
vary strongly with the type of biofilm. The lowest fD_g was found for granules_2 (0.36)
and highest fD_g for the biofilm_1 (0.96). Diffusion was most restricted in granules_2
likely due to the compact structure and high content of inorganic matter (Table 8.1). In
comparison, the diffusion coefficients in heterotrophic biofilm_1 were closest to Dwater
because of the fluffy structure and high VSS of 90%. Results are within the typical range
of reported relative diffusion coefficients for biofilm from 0.2 to 0.8 [47-49] except
biofilm_1.
In Figure 8.2, D1 and Dmean are directly compared for the different biofilm
samples. The “error” bars for Dmean represent the width of the Γ distr. σ, listed in Table
8.4: σ for pure water was very small and not visible, whereas σ for all biofilms was large.
Sludge and biofilm_2 show the highest σ possibly due to their highly heterogeneous
internal structure (Table 8.4). The diffusion coefficients D1 and Dmean vary for all
biomass, with biofilm_1 showing fastest and granules_2 showing lowest diffusion
230
coefficients. The water diffusion in sludge and biofilm_2 are similar, although the
biomass structure parameters differ significantly. As visible in Table 8.3, these biomasses
comprise completely different morphologies being either an open network or a compact
biofilm. Furthermore, biomass structural parameters (e.g. VSS, biofilm thickness and
density) differ strongly and suggest that the diffusion in biofim_2 is more hindered than
in the sludge. However, this is not the case. This leads to the conclusion that the
assumptions regarding mass transfer in biofilms based on bulk physical properties are not
directly related to the classical parameters or optical appearance.
Figure 8.2: Diffusion coefficients D1 and Dmean obtained by (bi)-exponential fit and Γ
distribution for water and different biomasses measured with a diffusion time Δ= 200 ms.
“Error” bars around Dmean represent not statistical errors but the width of the Γ
distribution σ (Table 8.2).
wa
ter
bio
film
_1
bio
film
_2
slu
dg
e
gra
nu
les
_1
gra
nu
les
_2
0
5.0x10-10
1.0x10-9
1.5x10-9
2.0x10-9
2.5x10-9
D
[m
2 /s
]
Dmean D1
231
No clear correlations are found between certain integral biofilm properties and
diffusion coefficients: the biofilm thickness and geometry seemed to have a minor
influence on D in the present case. For example, biofilm_2 and granules_2 had similar
densities, but D differed by approximately 50% when using both data processing
methods. Liquid channels in biofilm_2 correspond to an enhanced water diffusion within
the biomass resulting in similar apparent diffusion values as highly porous sludge.
The broad range of relative diffusion coefficients highlights the fact that biofilm
structure determines the mass transfer and ultimately the performance of productive
biofilms, such as the mentioned wastewater biofilms or fungal bio pellets [50]. Whether
the measured water diffusion can be transferred to relevant substrates for biofilms
remains unclear. The substrates’ molecular weight, interaction with the surroundings as
well as diffusivity and penetration into the biofilm differ from water molecules and could
be explored by heteronuclear NMR diffusion measurements and theoretical models can
help to answer this question. Using a simple calculation, we estimated the difference
between COD (chemical oxygen demand) turnover for the highest (biofilm_1) and lowest
(granule_2) diffusion coefficient. The oxygen flux can roughly be estimated by: D/Δ z *
(concentration DO at biofilm surface – concentration DO at z) (eq. 4), where z is the
biofilm depth and DO the dissolved oxygen concentration. Dmean was used a diffusion
coefficient due to the best fit of the data. Keeping all biofilm related parameters constant
(only changing D) the lower diffusion results in 20 % less turnover based on flux
estimates, thus knowing the diffusion coefficient is highly relevant for biofilm modelling.
Further infestations are planned in the future.
232
Besides that, the results contribute to link transport processes to biofilm structure
which was shown for nanoparticle and metal transport in biofilms [27, 51]. The
comparison of classical integral quantities with optical and MRI images shows that the
approach of just using one measure for biofilm characterization is insufficient. A
comprehensive characterization on different time and length scales is needed when
aiming for an understanding of biofilms and the processes of mass transport and
especially diffusion. The link between structure, diffusion and biological activity
deserves further research.
Influence of Diffusion Time on Water Dynamics
In homogeneous liquids of small molecules, D is independent of diffusion time ∆.
In porous media, D decreases with increasing ∆ due to barriers like physical restriction
and adhesion of water molecules [14]. Biofilms can also be regarded as porous media
consisting of biofilm matrix and bulk water. Our results show, that D1 decreased with Δ
for all biofilms (Figure 8.3). D1 was selected as the trends were most pronounced. Dmean
showed a similar but less marked effect. Water diffusion was most restricted in
granules_2, indicated by the strong decline of D1 and low diffusion coefficient asymptote
of 6.5 m2/s·10-10 at 400 ms, followed by granules_1. A different composition of EPS and
microbial communities as well as the length scales of the heterogeneous internal
structures are already visible in MRI images (Table 3) and are possible explanations for
our observations. The mobility of water molecules may also be decreased due to
entrapment in EPS molecules [46, 52]. As already mentioned, sludge and biofilm_2 both
displayed fast diffusion, similar distribution width, and a small dependence on Δ.
233
Biofilm_1 did not show a strong diffusion barrier; only a slight decline of D1 relative to
Dwater was found which leads to the conclusion that mainly unrestricted bulk water is
observed. The calculated D also include the diffusion in the near bulk water which may
be influenced by the presence of the biofilm, as has been shown for emulsions [53].
Figure 8.3: Diffusion coefficients D1 of biofilms at different diffusion times Δ.
Classically, the dependence of the diffusion coefficients on diffusion time is a hint on
hindered or restricted diffusion and is most pronounced in the granules. Data points for
500 ms and 800 ms are only given for sludge.
The distributions of the effective diffusion coefficients obtained within Γ distr. are
shown in Figure 8.4. Granule_2 exhibits the narrowest distribution (σ of 4.71 m2/s·10-10)
together with the slowest Dmean. Faster diffusion was found for sludge and biofilm_2 with
a broader distribution (σ of 8.55 and 7.18 m2/s·10-10). Narrow distributions, but with
0 100 200 300 400 500 600 700 800 900
6.0x10-10
8.0x10-10
1.0x10-9
1.2x10-9
1.4x10-9
1.6x10-9
1.8x10-9
sludge
biofilm_1
biofilm_2
granules_1
granules_2
D
1
[m
2 /s
]
∆ [ms]
234
faster Dmean were found for biofilm_1 and granules_1. The sludge and biofilm_2 results
demonstrate that increased internal heterogeneity over small length scales in the biomass
tends to broader distributions. The distributions give unique indications for the estimation
of the overall diffusion properties of diverse biofilm systems that single parameters
determined by (bi.)-exp.-func. cannot express. However, the Γ distr. model also depicts
portions of 1H-NMR signals which are diffusing significantly faster than free water
(Dwater = 2.04·10-09 m2/s) highlighted in gray. This portion differed for all biofilms and
was especially pronounced for biofilm_1. Possible explanations for this phenomenon are
convection due to temperature or enhanced diffusion by concentration gradients. Larger
bulk water compartments, which were evident for biofilm_1 (Table 8.3), are more prone
to convective transport, resulting in larger influence on D. Additionally, it is known that
diffusion can increase in the presence of small concentrations of nonpolar compounds,
such as methanol or acetonitrile in water mixtures [54]. The convection could also be a
result of vibrations induced by the large field gradients as it is difficult to immobilize the
granules and biofilm carriers.
235
Figure 8.4: Distribution of the diffusion coefficients according to Γ distribution. D =
Dmean for free water is 2.04·10-9 m2/s. Gray region indicates diffusion values, which are
larger than free water.
Correlation of Diffusion and Transverse Relaxation
The effective D-T2 relaxation correlation was measured for biofilm_1 and
compared to free water (Figure 8.5) using a regularization parameter α = 1*108 and 64
steps during 2D-ILT. The peak in the upper edge of the maps is an artifact of 2D-ILT due
to insufficient sampling of higher gradients. To confirm the location of the main peak in
the 2D distribution, a single line of data was extracted in each dimension to produce a 1D
spectrum of diffusion and relaxation as the 1D ILT is numerically more stable than the
2D [55]. As expected, bulk water is characterized by Dwater = 2.4 x 10-9 m2/s with a
reasonably narrow distribution width and by an average T2 of 2.6 s.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
D [m2/s]
0
2
4
6
8
10
12
pr
ob
ab
ilit
y
[s
/m
2 ]
x10-9
x108
sludge
biofilm_2
biofilm_1
granules_2
granules_1
236
Figure 8.5: Effective diffusion spin-spin relaxation correlation maps for free water (left)
and biofilm_1 (right) at a diffusion time of 100 ms. Data was processed using 2D-ILT
and reveals the correlation between diffusion and transverse relaxation both being
influenced by the structural properties of the biomass.
Compared to free water, the presence of biofilm_1 produces a significant shift of
a portion of the signal to smaller D and lower T2. Typically, T2 and D in biofilms are less
than that of bulk water [19, 34]. These results are consistent with previous observations.
Another portion of signal attributed to the bulk water, about T2 of 3 s, shifts towards
larger D. This strongly points to convection as a possible explanation for the observation
of larger diffusion coefficients in 1D diffusion experiments as well as in the correlation
experiment. As mentioned earlier this could well be due to vibrations induced by the
gradients and enhanced by the biofilm as it is hard to immobilize the plastic carriers in
the test tubes. It should be noted that the effect is well known and occurs in low viscosity
liquids even at almost no temperature gradient if the solid structures in the sample cannot
be completely immobilized.
237
Conclusions
In this study, the water diffusion coefficients and their distributions were
determined using PFG-NMR for five different biomass samples. Diffusion data was
processed using conventional (bi)-exponential data analysis, the Γ distribution model and
2D ILT for data interpretation. Stereomicroscopic images are compared to MRI images
and insights of D-T2 correlation maps are introduced. The experimental results lead to the
following conclusions: Similar diffusion coefficients for both data processing approaches,
(bi)-exponential data analysis and Γ distribution model, were obtained with 5-26%
difference between Dmean and D1. A better representation of the data was achieved by the
Γ distribution function with respect to numerical reproducibility and physico-chemical
interpretation. No direct correlation between Dwater and typical biofilm properties (e.g.
compactness, mineral content, VSS) was observed. Dmean is not sufficient to describe the
mass transport in multispecies biomass. Additional information is provided by MRI
images and distributions of Dwater in the biomass obtained by Γ distribution model: some
areas in the biofilm seem to have exhibited more restricted transport than others,
associated to gas bubbles, precipitates, and changing biofilm densities. D-T2 correlation
maps confirmed these findings and proved a shift in T2. More research in this field is
needed, especially regarding the spatial distribution of diffusion coefficients in
multispecies biomass where NMR methods are a promising approach (e.g. diffusion-
weighted imaging).
238
References
1. Costerton, J.W., et al., Microbial Biofilms. Annual Review of Microbiology,
1995. 49: p. 711-745.
2. Sutherland, I.W., The biofilm matrix - an immobilized but dynamic microbial
environment. Trends in Microbiology, 2001. 9(5): p. 222-227.
3. Metcalf & Eddy, I., Wastewater engineering : treatment and reuse. Fourth edition
ed. 2003, Boston: McGraw-Hill, 2003.
4. Stewart, P.S., Diffusion in biofilms. Journal of Bacteriology, 2003. 185(5): p.
1485-1491.
5. Flemming, H.-C. and J. Wingender, The biofilm matrix. Nature Reviews
Microbiology, 2010. 8(9): p. 623-633.
6. Wäsche, S., H. Horn, and D.C. Hempel, Influence of growth conditions on biofilm
development and mass transfer at the bulk/biofilm interface. Water Research,
2002. 36: p. 4775-4784.
7. Neu, T.R., et al., Advanced imaging techniques for assessment of structure,
composition and function in biofilm systems. FEMS Microbiology Ecology, 2010.
72(1): p. 1-21.
8. Guimera, X., et al., Dynamic characterization of external and internal mass
transport in heterotrophic biofilms from microsensors measurements. Water
Research, 2016. 102: p. 551-560.
9. Billings, N., et al., Material properties of biofilms-a review of methods for
understanding permeability and mechanics. Reports on Progress in Physics, 2015.
78(3).
10. Horn, H. and S. Lackner, Modeling of Biofilm Systems: A Review, in Productive
Biofilms, K. Muffler and R. Ulber, Editors. 2014, Springer-Verlag Berlin: Berlin.
p. 53-76.
11. Li, C.Y., et al., Assessing the influence of biofilm surface roughness on mass
transfer by combining optical coherence tomography and two-dimensional
modeling. Biotechnology and Bioengineering, 2016. 113(5): p. 989-1000.
12. IWA, IWA Task Group on Biofilm Modeling: Mathematical Modeling of Biofilms.
2006: IWA Publishing.
239
13. Callaghan, P.T., Translational Dynamics and Magnetic Resonance: Principles of
Pulsed Gradient Spin Echo NMR. 2011, Oxford: Oxford Press.
14. Callaghan, P., Principles of Nuclear Magnetic Resonance Microscopy. 1991:
Oxford University Press. 516.
15. Fabich, H., et al., Microbial and algal alginate gelation characterized by
magnetic resonance. Journal of Biotechnology, 2012. 61(3): p. 320– 327.
16. Arndt, F., et al., New Insights into Sodium Alginate Fouling of Ceramic Hollow
Fiber Membranes by NMR Imaging. AIChE Journal, 2016. 62(7 ): p. 2459-2467.
17. Seymour, J.D., et al., Magnetic resonance microscopy of biofouling induced scale
dependent transport in porous media. Advances in Water Resources, 2007. 30(6-
7): p. 1408-1420.
18. Seymour, J.D., et al., Anomalous fluid transport in porous media induced by
biofilm growth. Physical Review Letters, 2004. 93(19).
19. Vogt, S.J., et al., Permeability of a growing biofilm in a porous media fluid flow
analyzed by magnetic resonance displacement-relaxation correlations.
Biotechnology and Bioengineering, 2013. 110(5): p. 1366-1375.
20. Manz, B., et al., Investigation of Biofilm Structure, Flow Patterns and
Detachment with Magnetic Resonance Imaging. Water Science and Technology,
2005. 52(7): p. 1-6.
21. Herrling, M.P., et al., Determining the Flow Regime in a Biofilm Carrier by
Means of Magnetic Resonance Imaging. Biotechnology and Bioengineering,
2015. 112(5): p. 1023-1032.
22. Lens, P.N.L., et al., Diffusional properties of methanogenic granular sludge: H-1
NMR characterization. Applied and Environmental Microbiology, 2003. 69(11):
p. 6644-6649.
23. Bishop, P.L., T.C. Zhang, and Y.C. Fu, Effects of biofilm structure, microbial
distributions and mass transport on biodegradation processes. Water Science and
Technology, 1995. 31(1): p. 143-152.
24. Debeer, D., et al., Effects of biofilm structures on oxygen distribution and mass
transport. Biotechnology and Bioengineering, 1994. 43(11): p. 1131-1138.
25. Ramanan, B., et al., Magnetic Resonance Imaging of Mass Transport and
Structure Inside a Phototrophic Biofilm. Current Microbiology, 2013. 66(5): p.
456-461.
240
26. McLean, J.S., O.N. Ona, and P.D. Majors, Correlated biofilm imaging, transport
and metabolism measurements via combined nuclear magnetic resonance and
confocal microscopy. Isme Journal, 2008. 2(2): p. 121-131.
27. Phoenix, V.R. and W.M. Holmes, Magnetic resonance imaging of structure,
diffusivity, and copper immobilization in a phototrophic biofilm. Applied and
Environmental Microbiology, 2008. 74(15): p. 4934-4943.
28. Röding, M., et al., The gamma distribution model for pulsed-field gradient NMR
studies of molecular-weight distributions of polymers. Journal of Magnetic
Resonance, 2012. 222: p. 105-111.
29. Röding, M., N. Williamson, and M. Nydén, Gamma convolution models for self-
diffusion coefficient distributions in PGSE NMR. Journal of Magnetic Resonance,
2015. 261: p. 6-10.
30. Herrling, M.P., et al., Short and long term biosorption of silica-coated iron oxide
nanoparticles in heterotrophic biofilms. Science of The Total Environment, 2016.
544: p. 722-729.
31. Gilbert, E.M., et al., Low Temperature Partial Nitritation/Anammox in a Moving
Bed Biofilm Reactor Treating Low Strength Wastewater. Environmental Science
& Technology, 2014. 48(15): p. 8784-8792.
32. Loosdrecht, M.C.M.v., et al., Experimental Methods in Wastewater Treatment.
2016: Intl Water Assn, IWA.
33. Kimmich, R., NMR - Tomography, Diffusometry, Relaxometry. Springer Verlag,
1997.
34. Ranzinger, F., et al., Direct surface visualization of biofilms with high spin
coordination clusters using Magnetic Resonance Imaging. Acta Biomaterialia,
2016. 31: p. 167-177.
35. Callaghan, P.T., Principles of Nuclear Magnetic Resonance Microscopy. 1991,
New York: Oxford University Press.
36. Lee, J.H., et al., 2-Dimensional Inverse Laplace Transform Nmr - Altered
Relaxation-Times Allow Detection of Exchange-Correlation. Journal of the
American Chemical Society, 1993. 115(17): p. 7761-7764.
37. Callaghan, P.T., S. Godefroy, and B.N. Ryland, Diffusion-Relaxation Correlation
in Simple Pore Structures. Journal of Magnetic Resonance, 2003. 162(2): p. 320-
327.
241
38. Dammel, E.E. and E.D. Schroeder, Density of activated sludge solids. Water
Research, 1991. 25(7): p. 841-846.
39. Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media
using a low-field magnetic resonance system. Environmental Science &
Technology, 2013. 47(2): p. 987-992.
40. Winkler, M.K.H., et al., Factors influencing the density of aerobic granular
sludge. Applied Microbiology and Biotechnology, 2013. 97(16): p. 7459-7468.
41. de Kreuk, M.K., M. Pronk, and M.C.M. van Loosdrecht, Formation of aerobic
granules and conversion processes in an aerobic granular sludge reactor at
moderate and low temperatures. Water Research, 2005. 39(18): p. 4476-4484.
42. Etterer, T. and P.A. Wilderer, Generation and properties of aerobic granular
sludge. Water Science and Technology, 2001. 43(3): p. 19-26.
43. Juang, Y.-C., et al., Stable aerobic granules for continuous-flow reactors:
Precipitating calcium and iron salts in granular interiors. Bioresource
Technology, 2010. 101(21): p. 8051-8057.
44. Foerster, E., H. Nirschl, and G. Guthausen, NMR Diffusion and Relaxation for
Monitoring of Degradation in Motor Oils. Applied Magnetic Resonance, 2017.
48(1): p. 51-65.
45. Renslow, R.S., et al., In Situ Effective Diffusion Coefficient Profiles in Live
Biofilms Using Pulsed-Field Gradient Nuclear Magnetic Resonance.
Biotechnology and Bioengineering, 2010. 106(6): p. 928-937.
46. Vogt, M., H.C. Flemming, and W.S. Veeman, Diffusion in Pseudomonas
aeruginosa biofilms: a pulsed field gradient NMR study. Journal of
Biotechnology, 2000. 77(1): p. 137-146.
47. Stewart, P.S., A review of experimental measurements of effective diffusive
permeabilities and effective diffusion coefficients in biofilms. Biotechnology and
Bioengineering, 1998. 59(3): p. 261-272.
48. Wood, B.D., M. Quintard, and S. Whitaker, Calculation of effective diffusivities
for biofilms and tissues. Biotechnology and Bioengineering, 2002. 77(5): p. 495-
516.
49. Horn, H., H. Reiff, and E. Morgenroth, Simulation of growth and detachment in
biofilm systems under defined hydrodynamic conditions. Biotechnology and
Bioengineering, 2003. 81(5): p. 607-617.
242
50. Hille, A., et al., Effective Diffusivities and Mass Fluxes in Fungal Biopellets.
Biotechnology and Bioengineering, 2009. 103(6): p. 1202-1213.
51. Peulen, T.-O. and K.J. Wilkinson, Diffusion of Nanoparticles in a Biofilm.
Environmental Science & Technology, 2011. 45(8): p. 3367-3373.
52. Belton, P.S., NMR and the mobility of water in polysaccharide gels. International
Journal of Biological Macromolecules, 1997. 21(1-2): p. 81-88.
53. Guan, X.Z., et al., PFG-NMR on W1/O/W2-emulsions: Evidence for molecular
exchange between water phases. European Journal of Lipid Science and
Technology, 2010. 112(8): p. 828-837.
54. Derlacki, Z.J., et al., Diffusion coefficients of methanol and water and the mutual
diffusion coefficient in methanol-water solutions at 278 and 298 K. Journal of
Physical Chemistry, 1985. 89(24): p. 5318-5322.
55. Kausik, R. and M.D. Hurlimann, Sensitivity and resolution of two-dimensional
NMR diffusion-relaxation measurements. Journal of Magnetic Resonance, 2016.
270: p. 12-23.
243
STRUCTURE AND DIFFUSION OF AEROBIC GRANULAR
SLUDGE USING MAGNETIC RESONANCE
Abstract
Magnetic resonance imaging (MRI) and nuclear magnetic resonance (NMR)
allow non-invasive measurements describing both internal structures and transport
properties of opaque, complex materials like biofilms [1-3]. High-field MRI was used to
image aerobic granules collected from full-scale wastewater treatment plants in the
Netherlands. T1 and T2 relaxation-weighted images reveal heterogeneous internal
structures that include high and low density regions and solid inclusions. Additionally,
pulsed field gradient (PFG) NMR methods and multi-dimensional correlation and
exchange experiments were used to measure diffusion and transport properties within
undisturbed granules. Our results show differences in rates of water diffusion within the
heterogeneous granule structure and suggest that models employing a single diffusion
coefficient may be insufficient to capture the complexity of transport behaviour within
the granules.
Introduction
Compared to conventional activated sludge systems where the biomass exists in
dispersed flocs, aerobic granular sludge offers numerous benefits for wastewater
treatment including compact design, lower energy costs, and excellent biomass retention
[4]. By their very nature as spherical biofilm aggregates, granules are composed of a
244
variety of microniches—aerobic, anoxic, anaerobic —where diverse bio-chemical
conversions can occur simultaneously within the same granule [5, 6] (Figure 9.1). In the
last two decades, research into the formation, structure, and metabolism of granular
sludge has flourished [7-12] and today full-scale reactors are in operation in several
countries, including the Netherlands [13].
Figure 9.1. A conceptual model of aerobic granular sludge shows the different redox zones
within the granule as concentric structural layers where distinct biochemical conversions
occur simultaneously.
While it is known that hydrodynamic shear, substrate type and loading rate,
oxygen concentration, microbial growth rate, and microbial strain are all important
parameters in the formation of aerobic biofilm granules, the governing parameter has yet
to be identified [14]. Under appropriate operating conditions, floccular sludge self-
assembles into millimetre-scale biofilm granules in which EPS provides a robust
framework. The biofilm matrix of the granule has a complex structure [15], consisting of
245
hydrated gel-like fibers allowing for phenotypic heterogeneity and differentiation of cells
within the colony analogous to a multi-cellular organism [16-18]. Early MRI
experiments on anaerobic granules provided evidence of a cluster structure [19] while
imaging of methanogenic granules showed that the EPS matrix is organized in concentric
layers and the granules appear to have ‘hollow’ centers [20].
This complex morphology influences rates of diffusion and substrate utilization in
a manner that is not entirely well-defined. Moreover, granular sludge reactors may take
months or even years to reach steady state conditions, making experimental studies using
parametric analysis on large scale reactors infeasible. Mathematical models have been
developed to simulate substrate removal and the distribution of microbial populations and
processes within the biofilm granule [21-24], but their utility is limited by a lack of
knowledge. Until these models are refined with robust experimental results related to the
formation and activity of granular sludge, the technology will not by fully exploited for
wastewater treatment.
The current research explores how the internal granular structure is related to
reactive mass transport in the granule, with the goal of understanding how wastewater
components move through the granule and are degraded or captured, and how granule
activity is related to the structure of the aggregate. Most granules used in laboratory
research have been grown on simple soluble substrates like acetate, while natural
wastewaters contain complex substrates, including slowly biodegradable particulates [25-
27]. These particulates must be hydrolyzed before they are available to bacteria for
conversion, which influences the granulation kinetics, the resulting granule morphology,
246
as well as reactor performance and the required operational conditions [26].
Specifically, the research will examine in the future how particulates and complex
contaminants commonly found in wastewater influence the morphology and mass
transport properties of aerobic granular sludge [13, 25, 27].
NMR and MRI can provide spatially-resolved data on granule structure,
composition, and reaction-diffusion properties non-invasively and non-destructively
under various hydrodynamic conditions. These capabilities represent a significant
advantage over conventional methods like microscopy and micro-electrode studies which
are necessarily invasive and destructive. These methods also tend to be limited in terms
of the range of length scales that can be probed. For example, micro-electrodes provide
point or 1D data. Optical methods can achieve a high spatial resolution, but over a
limited field of view. NMR and MRI can better bridge multiple length scales, such as in
the case of a 5 mm sample field of view with 50 µm2 pixel resolution. As the research
described in this chapter is in the early stages of a multi-year international collaborative
project, this chapter will focus on two main research questions, each explored with
different NMR and MRI methods.
The first current research question is related to the internal granule structure.
Does the internal structure of aerobic granules used in practice for wastewater treatment
conform to the conceptual model shown in Figure 9.1? Secondary questions are related
to how consistent the observed structure is between samples from the same treatment
plant, from different treatment plants, over time as the granules age, and between
laboratory granules and those treating real wastewater. Imaging experiments give insight
247
to these structural questions.
Secondly, the current research explores how diffusive transport varies within a
granule by measuring the effective water diffusion coefficient using PFG-NMR methods.
Specifically, how large is the range of Deff within the granules? Multi-dimensional
diffusion – relaxation correlation (D-T2), relaxation – relaxation correlation (T1-T2), and
relaxation exchange (T2-T2) experiments can also provide information about how the
diffusive behaviour of water is related to structural regions with different relaxation rates,
and the timescale of exchange of water between those different regions.
Material and Methods
Sample Collection and Preparation
Samples of aerobic granular sludge were collected during the aeration phase of
the treatment cycle from sequencing batch reactors at the Utrecht and Garmerwolde
wastewater treatment plants in the Netherlands (Figure 9.2). The samples, which ranged
in volume from several hundred millilitres to several litres, were stored in airtight plastic
containers in the refrigerator without substrate addition. Granules from other sources,
including anammox granules from the B-stage of the Dokhaven treatment plant,
anammox granules treating the rejection water at Sluisjesdijk treatment plant, laboratory
VFA-producing granules, granules produced from extracted and re-constituted EPS, and
control samples of alginate beads were also imaged using the same parameters for
comparison with the aerobic granules. For the MRI experiments, a single granule was
added to tap water in a 5 mm NMR sample tube. The granules were elevated within the
248
sample tube by either a plug of Teflon tape, a stack of alginate beads, or a 3-D printed
polymer frame (Figure 9.3).
Figure 9.2. (Left to right) Samples were collected from the Utrecht wastewater treatment
plant’s Nereda ™ sequencing batch reactor. The granules settle first with the floccular
sludge on top. A single granule was then placed in a 5 mm NMR sample tube for
imaging experiments.
Samples for the PFG-NMR measurements and multi-dimensional correlation and
exchange measurements were collected from Garmerwolde treatment plant as described
above and stored in tap water in the refrigerator until measurements were collected,
approximately 4 months after sampling. As these measurements are not spatially
resolved in 2D, a stack of granules in tap water was added to the 5 mm NMR sample tube
to maximize the signal obtained from the granules relative to bulk water signal.
249
Figure 9.3. The granule samples were placed in 5 mm NMR sample tubes, elevated with
either Teflon tape (left), a stack of alginate beads (center), or a 3-D printed frame (right).
The granules shown include an extracted EPS granule, a VFA-producing laboratory
grown granule, and an anammox granule from the B-stage of Dokhaven treatment plant.
NMR and MRI Measurements
1H NMR and MRI are sensitive to hydrogen protons, typically water in most
biological systems. NMR relaxation rates provide information on the physico-chemical
environments in which different water populations exist. Timing of the pulses which
make up the NMR measurement can be adjusted to accentuate contrast between these
different water populations and distinguish between intercellular water, rotationally
restricted water in the EPS matrix, and free water in the bulk phase. T1 relaxation- and T2
relaxation-weighted images provide such a contrast and were collected on MRI
tomographs operating at 7 T and 22 T using parameters found in Table 9.1. The MRI
images were collected on both fresh granules (several days after sampling) and on aged
samples (approximately 2 months after sampling). The granules were also imaged using
250
traditional microscopy techniques including scanning electron microscopy (SEM) and
transmission electron microscopy (TEM).
22 T 7 T
T1 weighted T2 weighted T2
map
Repetition time, Tr (s) 0.550 5 6
Echo time, tE, (ms) 5.3 5.3 5.2
Number of echoes 16 16 128
Spatial resolution (μm) 47 x 47 47 x 47 47 x 47
Slice thickness (μm) 100 100 100
Measurement time 9 min 23 s 10 min 40 s 13 h 39 min
Table 9.1. MRI measurement parameters. The 22 T MRI system at uNMR-NL, an NWO-
funded National Roadmap Large-Scale Facility of the Netherlands, located at Utrecht
University, produced high-resolution and high-contrast images with a minimal
measurement time. The 7 T MRI system at Wageningen University achieved similar
resolution with longer measurement times.
Additionally, pulsed field gradient (PFG) NMR methods were used to measure
diffusion coefficients in aged granules from Garmerwolde treatment plant. The samples
were approximately 4 months old at the time of measurement. Experiments were
performed on a 250 MHz (5T) Bruker Avance III superconducting magnet with using a
high-power probe, Micro 5 gradient set (with a 2.81 T/m maximum gradient) and a 5 mm
radio-frequency coil. Two-dimensional correlation and exchange experiments were also
conducted to relate diffusion to T2 relaxation and to examine exchange between different
T2 populations.
The pulsed-gradient stimulated-echo (PGStE) experiment was used to collect
diffusion data for these granules given the relatively fast signal decay rates observed for
251
the Garmerwolde granules during imaging. During the diffusion time, Δ, the stimulated
echo experiment allows for the net magnetization to be stored in the longitudinal plane
where it is not subject to T2 relaxation processes (Figure 3.2). Therefore, measurable
signal remains following extended diffusion times in the range of tens to hundreds of
milliseconds.
Data Analysis
T2 Maps. Multi-slice multi-echo (MSME) images produce a stack of 2-D images
showing the echo amplitude per pixel in each sample slice. An image is collected of each
echo in the signal decay, such that the stack of images for each slice shows the
attenuation of signal in each pixel. Fitting the echo attenuation in each pixel as an
exponential decay produces the effective T2 relaxation rate in each pixel. These rates, or
the relaxation times, can be displayed in a 2D image of the sample where the pixel
intensity corresponds to the effective relaxation rate, R2,eff, or the effective relaxation
time, T2,eff. T2 map data was analysed with both IDL software at WUR by Dr. Frank
Vergeldt and also at MSU by the author using Prospa v3.13.
Diffusion Images. PGStE data for the 1D image was analysed first in Prospa by
Fourier transforming the echo data in the read direction. The data was then exported to
Matlab as a magnitude image where a fitting function was applied in the q-direction to
calculate the effective diffusion coefficient at each point in the 1D image. Diffusion-
weighted images collected in Paravision 5.1 were analysed in Prospa v3.13. A Prospa
macro first created a matrix of q-space data for each image slice prior to using the
252
standard Prospa makeDiffMap macro to produce an image in which the pixel intensity
corresponds to the apparent diffusion coefficient (ADC) in the sample.
Multidimensional Correlation and Exchange Experiments. Data collected for
correlation and exchange experiments were analysed in Matlab using the 2D Inverse
Laplace Transform (ILT) which uses a non-negative least squares fitting function with a
regularization parameter to minimize the error in the solution.
Results and Discussion
MRI of Granule Internal Structure
The initial research goal of this project was to simply identify internal structural
features within the granules and determine how consistent those features are across
granule sources, types, and over time. T1 and T2 relaxation-weighted images of granules
from wastewater treatment plants reveal heterogeneous internal structures that include
high and low density regions and solid inclusions. Images of granules from both the
Utrecht and Garmerwolde treatment plants reflected similar heterogeneity, though it was
significantly more difficult to collect high quality images from the Garmerwolde granules
because of rapid signal decay. For that reason, most of the images included here are of
granules collected from the Utrecht wastewater treatment plant.
253
Figure 9.4. The third echo of a T1-weighted image of a Nereda® granule (Utrecht
wastewater treatment plant) obtained from the 22 T MRI shows high contrast between
areas of differing density within the granule. Spatial resolution is 47 x 47 x 100 μm3,
echo time = 16 ms, repetition time = 550 ms. Because of the long echo time, brighter
regions in the granule correspond to less dense regions where diffusion is less restricted.
Darker areas indicate either decreased proton density, in the case of solid inclusions, or
more rotationally restricted water as would be found in cell clusters. All granules were
imaged in a 5mm tube of tap water, which is the relatively brighter fluid surrounding the
granule.
Figure 9.4 shows a fresh granule from the Utrecht treatment plant with
characteristic dark, dense regions and brighter less dense regions. Relatively bright bulk
water is visible in the sample tube outside the granule. This image is T1-weighted with a
short repetition time of 550 ms to enhance signal collection from the more dense regions
of the granule. It is also, however, an image of the third echo in the decay train making
the image also T2-weighted. The combination of T1- and T2-weighting provides excellent
contrast between regions of differing density and water mobility.
254
Figure 9.5. The first echoes of T1-weighted images of a laboratory VFA-producing
granule (left) and a Nereda® granule (Garmerwolde wastewater treatment plant) obtained
from the 22 T MRI show the high structural variability observed between laboratory and
treatment plant granules. Spatial resolution is 47 x 47 x 100 μm3, echo time = 5.3 ms,
repetition time = 550 ms. In the left image, the bright regions correspond to biomass,
which is selectively highlighted with T1-weighting relative to the bulk water outside the
granule. The right image, collected under the same measurement parameters, shows
dense biomass in the dark regions where the signal has already decayed due to T2
relaxation during the 5.3 ms echo time.
Figure 9.5 compares a laboratory-grown volatile fatty acid (VFA) producing
granule with a fresh granule from the Garmerwolde treatment plant. The VFA granule
appears in this image to be hollow, as was later confirmed with traditional microscopy
after sectioning the granule. The Garmerwolde granule, on the other hand, exhibits a
similar heterogeneous structure as seen in Figure 9.4, though more dense overall. Since
the VFA and Garmerwolde granules contain different bacterial populations and grow
under different operating conditions – anaerobic for the VFA granules and aerobic for the
Garmerwolde granules – it is not possible with these data to draw any conclusions about
the internal structural differences between laboratory and treatment plant granules in
255
general. However, as a starting point, it shows how different the internal structure can
be.
T2 Maps. Effective T2 maps of granules from the Utrecht treatment plant for
different age granules (Figure 9.6) show the same heterogeneous internal structure
observed in Figures 9.4 and 9.5. The maps were produced from MSME images (32
echoes) made on the 22 T system in the national NMR lab in Utrecht. Spatial resolution
is 47 x 47 x 100 microns. The effective T2 maps also show that neither the structure nor
the T2 relaxation behaviour changes significantly over the timescale of approximately 2
months. The image on the left in Figure 9.6 shows a fresh granule from Utrecht, while
the right image shows a different granule from the same sample, imaged after 2 months
of storage in the refrigerator. In both T2, eff maps, the bulk water T2 relaxation time is
approximately 18 ms, which is approximately equal to the maximum internal T2 time in
the granule ‘voids.’ In the dense biomass regions of the granules, the minimum T2 times
are approximately 6 ms with the apparent transition between ‘voids’ and biomass
occurring around a T2 time of 12 ms. T2 maps of alginate beads collected in the same
system with the same measurement parameters produced the same bulk water relaxation
time and a relatively uniform relaxation time within the bead of approximately 13 ms (not
shown).
256
Figure 9.6. T2 maps of granules from Utrecht wastewater treatment plant. The image on
the left was taken of a fresh aerobic granule, while the image on the right was collected
from the same sample of granules after aging for 2 months. The T2 relaxation times
range between approximately 18 ms in the bulk water and in the less dense regions of the
granules to approximately 6 ms in the dense cell clusters of the granules. There is no
discernible difference between the fresh and aged granules in terms of T2 relaxation. The
spatial resolution is 47 x 47 x 100 microns.
Boundary Layer. A dark apparent boundary layer is visible on the surface of both
the Utrecht granule in Figure 9.4 and the Garmerwolde granule in Figure 9.5, but was not
observed on any of the granules produced in laboratory reactors. This layer is very sharp
and well-defined in the MRI images, particularly those from the 22 T system. Two
recent studies have examined the surfaces of granular sludge and may provide insight
into the nature of the material observed in the MRI images. Poot et al. recently found
evidence of stratification of ammonia-oxiding bacteria (AOB) and nitritre-oxidizing
bacteria (NOB) at the surface of aerobic granules grown in low-strength wastewater [28]
using fluorescent in-situ hybridization (FISH). The authors suggest that the position of
the AOB on the granule surface provides a competitive advantage to the AOB in terms of
Ef
fe
ct
iv
e
T 2
re
la
xa
tio
n t
im
e (
m
s)
257
oxygen competition with the NOB, and also protects the NOB colonies from detachment.
In order to achieve complete nitritation for subsequent anaerobic ammonium oxidation,
NOB populations must be repressed relative to AOB populations. Thus, the layer may be
a dense biomass region, composed primarily of nitrifiers. Manas et al. reported
formation of calcium and phosphorus precipitates on the surfaces of anaerobic granules,
observed using SEM-EDX imaging [29]. These authors report that ‘rings’ of the mineral
were visible in the largest granules imaged, suggesting successive periods of
mineralization and overgrowth. The Garmerwolde granule shown in Figure 9.5 appears
to show successive rings as described by Manas et al.
As part of the effort to identify the apparent boundary layer observed with MRI,
SEM-EDX imaging was applied to the outer surfaces of samples of the same type of
granules used in the MRI experiments. A single Garmerwolde granule was also cut with
a razor blade prior to imaging with SEM-EDX to compare the internal and exterior
physical properties. Results (not shown) indicate that while some mineral precipitates
had formed in the granules collected from treatment plants, there was no significant
precipitate layer on the surface of the granule. The sectioned granule appeared to show a
more dense organic region at the surface, but the quality of the section was not high
enough to make a conclusive determination.
Images were also made using transmission electron microscopy (TEM) to more
directly observe the surface layer of aerobic granules from Garmerwolde treatment plant.
TEM imaging involves embedding an osmium-treated sample in resin which is then
258
baked until solid. The samples are then sectioned with a diamond blade to thicknesses of
70-90 nm and imaged in the electron microscope.
Figure 9.7 shows dense cell clusters at the surface of a granule in the top of the
image, with the embedding resin outside of the granule at the bottom. The black areas on
the edges of the image are the copper frame which holds the thin sectioned sample.
There is no precipitate or solid particulate debris visible on the granule surface. Figure
9.8 shows the edge of an internal cell cluster where the lower left area of the image is an
internal ‘void’ space. No cells are visible in the void, though some EPS matrix material
is visible in this and other TEM images of the less dense regions as imaged by MRI.
When viewed at the same scale, the granule exterior exhibits greater cell density than the
interior cluster, which may account for the dark apparent boundary layer.
Figure 9.7. TEM image showing the outer surface of a Garmerwolde granule (top of half
of image) relative to the embedding resin at the bottom. The cells shown extend
approximately 20-30 microns into the granule. Scale bar is 20 µm. A (~ 5µm)2 box is
shown for reference to Figure 9.8.
259
Figure 9.8. TEM image of the edge of an internal cell cluster and corresponding ‘void’
region. Scale bar is 5 µm. A (~ 5µm)2 box is shown for reference to Figure 9.7.
PFG and Multidimensional NMR
In addition to imaging experiments, PFG-NMR diffusion measurements and 2D
correlation and exchange measurements were recently collected at MSU at 5T field
strength using a stack of 4-month-old granules collected from Garmerwolde treatment
plant in a 5mm sample tube. These measurements confirm the presence of multiple T2
populations, a distribution of T1 populations, and variable diffusion coefficients within
the granules.
The PGStE pulse sequence was used in conjunction with imaging in the read
direction to produce a 1D image of the diffusion coefficients through a stack of alginate
beads and one aerobic granule in tap water (Figure 9.9, top). At a diffusion time, Δ, of
100 ms, the average effective water diffusion coefficient within the granule was
approximately 1 x 10-9 m2/s while the measured bulk water diffusion coefficient was
260
approximately 1.8 x 10-9 m2/s. The alginate beads exhibited slightly faster diffusion than
the granule with an effective water diffusion coefficient of 1.5 x 10-9 m2/s on average.
All of the measured effective diffusion coefficients are lower than the typical free water
diffusion coefficient of approximately 2 x 10-9 m2/s.
Figure 9.9. A 1D image of the diffusion coefficients was collected over a stack of
alginate beads and a single aged Garmerwolde granule in DI water. The top 2D image
shows the sample. The bottom image shows the calculated 1D diffusion coefficient for
each point in the image.
Diffusion-weighted images were collected of aged anammox granules sampled
from the B-Stage of the Dokhaven treatment plant in the Netherlands (Figure 9.10). The
images were collected on the 5 T system at MSU in a 10 mm NMR sample tube with a
spatial resolution of 78 µm x 78 µm x 300 µm and 16 averages. Eight gradient values
were used, ranging from 0 – 1 T/m. The resulting map of apparent diffusion coefficients
(ADC) shows reduced diffusion rates in the biomass within the granules and near-free-
bulk wateralginate bead
granule
D
( *
1
0-
9
m
2 /s
)
1
2
261
water diffusion rates in the central ‘void’ regions. Unfortuately, the spin echo-based
pulse sequence and diffusion measurement parameters used were not able capture and
resolve apparent diffusion coefficients in the dense outer layer of the granules. Future
work will involve optimizing the diffusion-weighted imaging measurement using a
stimulated echo to overcome the rapid signal loss in the outer granule layer.
Figure 9.10. Two anammox granules from the B-stage of Dokhaven treatment plant were
imaged on the 5 T system at MSU using diffusion weighting (left). The resulting ADC
map (right) shows reduced apparent diffusion coefficients within the biofilm matrix, but
was not able to quantify diffusion coefficients in the dense exterior regions of the
granules due to rapid signal attenuation.
Diffusion – relaxation (D-T2) correlation measurements were also collected, both
on the 5 T system at MSU and on a 0.3 T system at Wageningen University (WUR) in
the Netherlands (not shown). The D-T2 correlation measured at MSU with Δ = 50 ms
used a stack of aged granules from Garmerwolde in DI water. The correlation (Figure
9.11) shows multiple T2 populations with the largest peak corresponding to free water.
The distribution of diffusion coefficients suggests that the biomass, with the shorter
2
4
262
relaxation times, has a range of water diffusion coefficients which extends to lower
values.
Figure 9.11. D-T2 correlation was measured for a stack of aged Garmerwolde granules in
tap water on a 5 T system at MSU.
The T2 portion of the D-T2 correlation is very similar to the 1D T2 distribution
measured on the stack of aged Garmerwolde granules on the 5 T system at MSU (Figure
9.12). Here, the large peak on the order of T2 = 1-2 seconds corresponds to bulk water
outside the granules. When compared to T2 maps measured on the 7 T system at WUR
(Figure 9.12, right), the small peak at approximately 100-200 ms corresponds to the water
in the ‘void’ regions of the granules. The relatively large population with T2 relaxation
times of approximately 70 – 20 ms corresponds to the water within the dense biomass
regions of the granule.
263
Figure 9.12. The 1D T2 distribution measured on the 5 T system at MSU on a stack of
aged (4 month old) Garmerwolde granules (left) corresponds well to the T2 map of a fresh
Utrecht granule measured on the 7 T system at WUR (right) and the T2 distribution
measured as part of the D- T2 correlation shown in Figure 9.11. The intensity of the free
water peak is diminished in the current figure due to the loading of the sample tube to
maximize sample coming from the granules and minimize the volume of bulk water.
In addition to the D-T2 correlation, T1 – T2 correlation was also measured on the 5
T system at MSU (Figure 9.13). This correlation again shows the expected T2
distribution as well as a broadening of the T1 distribution relative to bulk water which
indicates the presence of more solid-like material in the granule.
264
Figure 9.13. The T1-T2 correlation measured on the 5 T system at MSU on a stack of
aged (4 month old) Garmerwolde granules.
T2 – T2 exchange was measured on the 5 T system at MSU using mixing times,
τm, of 0 ms and 150 ms (Figure 9.14). The measurement recorded at τm = 150 ms does
not show the classic off-diagonal peaks that would definitively show exchange between
the different T2 populations. However, since there is a distribution of T1 relaxation
populations present as well, exchange may result in averaging. Further data analysis and
measurement optimization is necessary to determine the extent of exchange between T2
populations.
265
Figure 9.14. T2-T2 exchange measurements conducted on the 5T MSU system using a
stack of aged Garmerwolde granules in DI water do not exhibit off-diagonal exchange
peaks.
Measurements are ongoing at the time of thesis submission. Preliminary results
indicate that aerobic granules from municipal treatment plants in the Netherlands exhibit
a heterogeneous structure comprised of dense cell clusters and ‘voids’ containing diffuse
EPS matrix. The structures observed are closer to the cluster structure observed in some
anaerobic granules [19] than to the concentric ring schematic shown in Figure 9.1. The
structures were stable over a storage period of several months. The apparent boundary
layer visibile in the MRI images appears to be of organic origin, composed of dense cell
clusters.
The preliminary findings also suggest that a distribution of diffusion coefficients
exists within each individual granule in which diffusion is slower than bulk water within
the dense biomass regions and on the order of free water in the ‘voids.’ This finding
suggests that models employing a single diffusion coefficient, often based on invasive
-3 -2 -1 0 1
-3
-2
-1
0
1
-3 -2 -1 0 1
-3
-2
-1
0
1
Log T2 (s) Log T2 (s)
Lo
g
T 2
(s
)
τm = 0 ms τm = 150 ms
266
micro-electrode measurements on laboratory granules, may be insufficient to capture the
complexity of transport behaviour within granules treating wastewater. Further work is
necessary to optimize measurement parameters and pulse sequences in order to measure
spatially-resolved apparent diffusion coefficients in the dense biomass regions of the
granules. Rapid signal decay renders spin-echo methods ineffective, but also suggests
significant differences in diffusion behavior exist in these regions. Finally, this research
demonstrates that NMR and MRI methods can provide novel and reliable data regarding
the structure and transport properties of aerobic granular sludge.
Future Work
Future experiments are planned to repeat the PFG-NMR and 2D correlation and
exchange experiments described above on fresh granule samples to produce data more
relevant for inclusion in mathematical models. Furthermore, propagator measurements
will be collected using a flow cell with a packed bed of granules to investigate the
timescale for exchange between the bulk flowing water and the intergranular water which
can shed light on not only transport properties of the system but also reaction kinetics.
Tracer particles will also be added to the flow to explore the fate of particulates within
the granular sludge system.
267
Acknowledgements
Experiments at the 22T (950 MHz) NMR instrument were supported by uNMR-
NL, an NWO-funded National Roadmap Large-Scale Facility of the Netherlands (project
184.032.207)
268
References
1. Codd, S.L., et al., Magnetic Resonance Microscopy of Biofilm and Bioreactor
Transport, in NMR Imaging in Chemical Engineering, S. Stapf and S.-I. Han,
Editors. 2006, Wiley: USA.
2. Van As, H. and P. Lens, Use of H-1 NMR to study transport processes in porous
biosystems. Journal of Industrial Microbiology & Biotechnology, 2001. 26(1-2):
p. 43-52.
3. Windt, C.W., F.J. Vergeldt, and H. Van As, Correlated displacement-T-2 MRI by
means of a Pulsed Field Gradient-Multi Spin Echo method. Journal of Magnetic
Resonance, 2007. 185(2): p. 230-239.
4. Lin, Y., et al., Characterization of alginate-like exopolysaccharides isolated from
aerobic granular sludge in pilot-plant. Water Research, 2010. 44(11): p. 3355-
3364.
5. Da-Wen, G. and T. Yu, Versatility and application of anaerobic ammonium-
oxidizing bacteria. Applied Microbiology & Biotechnology, 2011. 91(4): p. 887-
894.
6. de Kreuk, M., J.J. Heijnen, and M.C.M. van Loosdrecht, Simultaneous COD,
nitrogen, and phosphate removal by aerobic granular sludge. Biotechnology and
Bioengineering, 2005. 90(6): p. 761-769.
7. Beun, J.J., M.C.M. van Loosdrecht, and J.J. Heijnen, Aerobic granulation in a
sequencing batch airlift reactor. Water Research, 2002. 36(3): p. 702-712.
8. de Kreuk, M.K. and M.C.M. van Loosdrecht, Selection of slow growing
organisms as a means for improving aerobic granular sludge stability. Water
Science and Technology, 2004. 49(11-12): p. 9-17.
9. de Kreuk, M.K. and M.C.M. van Loosdrecht, Formation of aerobic granules with
domestic sewage. Journal of Environmental Engineering-ASCE, 2006. 132(6): p.
694-697.
10. Liu, Y. and J.H. Tay, The essential role of hydrodynamic shear force in the
formation of biofilm and granular sludge. Water Research, 2002. 36(7): p. 1653-
1665.
11. Nicolella, C., M.C.M. van Loosdrecht, and J.J. Heijnen, Wastewater treatment
with particulate biofilm reactors. Journal of Biotechnology, 2000. 80(1): p. 1-33.
12. Tay, J.H., Q.S. Liu, and Y. Liu, The role of cellular polysaccharides in the
269
formation and stability of aerobic granules. Letters in Applied Microbiology,
2001. 33(3): p. 222-226.
13. Pronk, M., et al., Full scale performance of the aerobic granular sludge process
for sewage treatment. Water Research, 2015. 84: p. 207-217.
14. de Kreuk, M.K., N. Kishida, and M.C.M. van Loosdrecht, Aerobic granular
sludge - state of the art. Water Science and Technology, 2007. 55(8-9): p. 75-81.
15. Vogt, S.J., et al., Permeability of a growing biofilm in a porous media fluid flow
analyzed by magnetic resonance displacement-relaxation correlations.
Biotechnology and Bioengineering, 2013. 110(5): p. 1366-1375.
16. Stoodley, P., et al., Biofilms as complex differentiated communities. Annual
Review of Microbiology, 2002. 56: p. 187-209.
17. Sauer, K., et al., Pseudomonas aeruginosa displays multiple phenotypes during
development as a biofilm. Journal of Bacteriology, 2002. 184(4): p. 1140-1154.
18. Stewart, P.S. and M.J. Franklin, Physiological heterogeneity in biofilms. Nature
Reviews Microbiology, 2008. 6(3): p. 199-210.
19. Gonzalez-Gil, G., et al., Cluster structure of anaerobic aggregates of an expanded
granular sludge bed reactor. Applied and Environmental Microbiology, 2001.
67(8): p. 3683-3692.
20. Lens, P.N.L., et al., Diffusional properties of methanogenic granular sludge: H-1
NMR characterization. Applied and Environmental Microbiology, 2003. 69(11):
p. 6644-6649.
21. Eberl, H.J., et al., A three-dimensional numerical study on the correlation of
spatial structure, hydrodynamic conditions, and mass transfer and conversion in
biofilms. Chemical Engineering Science, 2000. 55(24): p. 6209-6222.
22. Kreft, J.U., et al., Individual-based modelling of biofilms. Microbiology-Sgm,
2001. 147: p. 2897-2912.
23. Picioreanu, C., J.U. Kreft, and M.C.M. van Loosdrecht, Particle-based
multidimensional multispecies Biofilm model. Applied and Environmental
Microbiology, 2004. 70(5): p. 3024-3040.
24. Picioreanu, C., M.C.M. van Loosdrecht, and J.J. Heijnen, Effect of diffusive and
convective substrate transport on biofilm structure formation: A two-dimensional
modeling study. Biotechnology and Bioengineering, 2000. 69(5): p. 504-515.
25. de Kreuk, M.K., et al., Behavior of polymeric substrates in an aerobic granular
270
sludge system. Water Research, 2010. 44(20): p. 5929-5938.
26. Wagner, J., et al., Effect of particulate organic substrate on aerobic granulation
and operating conditions of sequencing batch reactors. Water Research, 2015.
85: p. 158-166.
27. Pronk, M., et al., Effect and behaviour of different substrates in relation to the
formation of aerobic granular sludge. Applied Microbiology and Biotechnology,
2015. 99(12): p. 5257-5268.
28. Poot, V., et al., Effects of the residual ammonium concentration on NOB
repression during partial nitritation with granular sludge. Water Research, 2016.
106: p. 518-530.
29. Manas, A., et al., Location and chemical composition of microbially induced
phosphorus precipitates in anaerobic and aerobic granular sludge.
Environmental Technology, 2012. 33(19): p. 2195-2209.
271
CONCLUSIONS
The work comprising this thesis demonstrates the versatility of NMR as a method
to explore complex and heterogeneous biofilm – porous media systems. NMR
measurements were applied at field strengths ranging from 250 kHz to 950 MHz to study
opaque systems ranging from macro-scale subsurface soils to millimeter scale granular
sludge biofilms. In all cases, the NMR measurements provided data non-destructively.
The low-field NMR experiments conducted during this thesis included 1) a
laboratory study to detect biofilm accumulation in a laboratory model well-bore reactor,
2) a field demonstration to detect biofilm accumulation in natural subsurface soils, and 3)
a laboratory study to detect microbially-induced calcite precipitation in the same model
well-bore reactor. In each of these experiments, T2 relaxation distributions were
measured over time to compare the signal response at the initial, clean condition to the
final, biofouled condition. Identification of the typical signal response for biofilm
accumulation and calcite precipitation in porous media may lead to the application of
low-field borehole NMR tools as sensors for bioremediation projects involving a
biobarrier.
The laboratory and field-scale studies to detect biofilm accumulation measured a
shortening of the mean log T2 relaxation time (T2ML) due to biofilm growth. In the lab
study, T2ML decreased by approximately 50%, while in the field study the decrease was
approximately 60% in the LF well (~275 kHz) and approximately 40% in the HF well
(~400 kHz). Biofilm growth was confirmed with microscopy and microbiological
272
methods in the lab study and by oxidizing and flushing the biofilm from the sensitive
zone in the field study.
In the lab study using the low-field borehole NMR tool to detect calcite
precipitation, T2 relaxation distributions recorded during the biomineralization process
showed an increase in the mean log T2 time and a bifurcation of the initial single
relaxation mode into a small population with very fast relaxation and a larger population
with slower relaxation. This result indicates that, in the relatively large pores of the
sandpack, the effect of changing the mineral surface of the pore space had a more
significant impact on T2 relaxation than the reduction in the pore size. The possibility
remains that a shortening of T2ML could be observed if the pores were initially smaller
than those in the reactor sandpack. In addition to the change in the T2 distribution, the
NMR measured water content in the reactor decreased to 76% of the initial value as
calcite displaced water in the pores and as excess CO2 gas collected. Calcite precipitation
was confirmed with destructive sampling followed by SEM imaging and acid digestion to
quantify calcium present. These measurements indicate that approximately 12% of the
pore space was occupied by calcite at the end of the experiment. The results of this
experiment indicate that low-field NMR is sensitive to calcite precipitation, though
further research is needed to evaluate the influence of initial pore size and mineral surface
on the T2 relaxation response.
The studies conducted at high magnetic field strength include collaborative work
with Dr. Maria Pia Herrling, a summer visitor to the Montana State University Magnetic
Resonance Lab from Karlsruhe Institute of Technology in Karlsruhe, Germany and
273
collaborative work conducted in part in the Netherlands with Dr. Merle de Kreuk and Dr.
Henk Van As as part of an NSF-funded 4-month visit by the author. In both cases, the
system of interest was biofilm used in wastewater treatment.
As part of her research into water diffusion in various biofilm structures,
including floccular sludge, granular sludge, and biofilm grown on plastic carriers, Dr.
Herrling measured diffusion-relaxation correlations at MSU with the author. In this
study, we found no direct correlation between the water self-diffusion coefficient, Dwater,
and typical biofilm properties like biofilm density and concentration of volatile
suspended solids. Additionally, the mean self-diffusion coefficient, Dmean, is not
sufficient to describe diffusion behavior in multi-species biofilms. High resolution MRI
found some areas in the biofilm exhibited restricted diffusion due to the presence of
precipitates and differing biofilm densities. The D-T2 correlations confirmed these
findings and showed shorter T2 relaxation in the carrier with biofilm attached.
The final experimental work included in this thesis describes ongoing
experimental work on the structure and diffusion properties of aerobic granular sludge
biofilms sampled from full-scale wastewater treatment plants in the Netherlands. High-
field MRI was used to image the complex and heterogeneous internal structure of the
granules. Additional high field PFG NMR experiments related to diffusion and transport
within the granules are currently being conducted at MSU using samples from Dutch
treatment plants. Preliminary findings from this work show a heterogeneous, cluster-like
structure in the aerobic granules obtained from the Utrecht treatment reactor, and a more
dense, heterogeneous structure in granules collected from the Garmerwolde treatment
274
plant, where the dense biomass regions and less dense EPS regions have differing
diffusive properties. These results have implications for modeling to optimize reactor
performance since the presence of multiple diffusion coefficients is not currently
accounted for in the models. Furthermore, the presence of inactive EPS regions may
affect granule activity and performance, or serve some other useful function not yet
understood. This experimental work is ongoing.
Thus, this thesis demonstrates that low-field borehole NMR has potential for use
as an in-situ sensor of biofilm accumulation and calcite precipitation for field applications
related to bioremediation. Moreover, high field PFG NMR and MRI can provide
spatially resolved data related to the structure and transport properties of aerobic granular
sludge biofilms used for wastewater treatment—data that is extremely challenging, if not
impossible, to obtain with other methods.
275
REFERENCES CITED
Allison, D.G., The biofilm matrix. Biofouling, 2003. 19(2): p. 139-150.
Anand, V., Hirasaki, George J., Paramagnetic relaxation in sandstones: Distinguishing
T1 and T2 dependence on surface relaxation, internal gradients and dependence
on echo spacing. Journal of Magnetic Resonance, 2008. 190: p. 68-85.
Arndt, F., et al., New Insights into Sodium Alginate Fouling of Ceramic Hollow Fiber
Membranes by NMR Imaging. AIChE Journal, 2016. 62(7 ): p. 2459-2467.
Bayer, J.V., F. Jaeger, and G.E. Schaumann, Proton Nuclear Magnetic Resonance (NMR)
Relaxometry in Soil Science Applications. Open Magnetic Resonance Journal,
2010. 3: p. 15-26.
Beech, I.B. and J. Sunner, Biocorrosion: towards understanding interactions between
biofilms and metals. Current opinion in biotechnology, 2004. 15(3): p. 181-6.
Behroozmand, A.A., K. Keating, and E. Auken, A Review of the Principles and
Applications of the NMR Technique for Near-Surface Characterization. Surveys
in Geophysics, 2015. 36(1): p. 27-85.
Belton, P.S., NMR and the mobility of water in polysaccharide gels. International Journal
of Biological Macromolecules, 1997. 21(1-2): p. 81-88.
Benjamin, M.M., Water Chemistry. Series in Water Resources and Environmental
Engineering. 2002, New York, NY: McGraw-Hill.
Beun, J.J., M.C.M. van Loosdrecht, and J.J. Heijnen, Aerobic granulation in a
sequencing batch airlift reactor. Water Research, 2002. 36(3): p. 702-712.
Billings, N., et al., Material properties of biofilms-a review of methods for understanding
permeability and mechanics. Reports on Progress in Physics, 2015. 78(3).
Binley, A., et al., The emergence of hydrogeophysics for improved understanding of
subsurface processes over multiple scales. Water Resources Research, 2015.
51(6): p. 3837-3866.
Bishop, P.L., T.C. Zhang, and Y.C. Fu, Effects of biofilm structure, microbial
distributions and mass transport on biodegradation processes. Water Science and
Technology, 1995. 31(1): p. 143-152.
Blumich, B., et al., The NMR-mouse: Construction, excitation, and applications.
Magnetic Resonance Imaging, 1998. 16(5-6): p. 479-484.
276
Blümich, B., F. Casanova, and S. Appelt, NMR at low magnetic fields. Chemical Physics
Letters, 2009. 477(4–6): p. 231-240.
Blumich, B., J. Perlo, and F. Casanova, Mobile single-sided NMR. Progress in Nuclear
Magnetic Resonance Spectroscopy, 2008. 52(4): p. 197-269.
Brownstein, K.R. and C.E. Tarr, Importance of classical diffusion in NMR studies of
water in biological cells. Physical Review A, 1979. 19(6): p. 2446-2453.
Bryar, T.R., C.J. Daughney, and R.J. Knight, Paramagnetic effects of iron(III) species on
nuclear magnetic relaxation of fluid protons in porous media. Journal of
Magnetic Resonance, 2000. 142(1): p. 74-85.
Bryar, T.R. and R.J. Knight, Sensitivity of nuclear magnetic resonance relaxation
measurements to changing soil redox conditions. Geophysical Research Letters,
2002. 29(24).
Butler, J.J., Jr., et al., Characterizing hydraulic conductivity with the direct-push
permeameter. Ground Water, 2007. 45(4): p. 409-419.
Callaghan, P.T., Principles of Nuclear Magnetic Resonance Microscopy. 1991, New
York: Oxford University Press.
Callaghan, P.T., Translational Dynamics & Magnetic Resonance: Principles of Pulsed
Gradient Spin Echo NMR. 2011, New York: Oxford University Press.
Callaghan, P.T., S. Godefroy, and B.N. Ryland, Diffusion-Relaxation Correlation in
Simple Pore Structures. Journal of Magnetic Resonance, 2003. 162(2): p. 320-
327.
Camper, A.K., Organic matter, pipe materials, disinfectants and biofilms in distribution
systems. 2014, IWA Publishing. p. 73-94.
Careghini, A., S. Saponaro, and E. Sezenna, Biobarriers for groundwater treatment: a
review. Water Science and Technology, 2013. 67(3): p. 453-468.
Carr, H.Y. and E.M. Purcell, Effects of diffusion on free precession in nuclear magnetic
resonance experiments. Physical Review, 1954. 94(3): p. 630-638.
Characklis, W.G. and K.C. Marshall, Eds., Biofilms. 1990, New York: John Wiley &
Sons, Inc.
Chen, M.J., Z. Zhang, and T.R. Bott, Effects of operating conditions on the adhesive
strength of Pseudomonas fluorescens biofilms in tubes. Colloids and Surfaces B:
Biointerfaces, 2005. 43(2): p. 61-71.
277
Coates, G.R., et al., A new characterization of bulk-volume irreducible using magnetic
resonance. Log Analyst, 1998. 39(1): p. 51-63.
Coates, G.R., L. Xiao, and M.G. Prammer, NMR Logging: Principles and Applications.
1999, Halliburton Energy Services: Houston.
Codd, S.L., et al., Magnetic Resonance Microscopy of Biofilm and Bioreactor Transport,
in NMR Imaging in Chemical Engineering, S. Stapf and S.-I. Han, Editors. 2006,
Wiley: USA.
Codd, S.L., et al., NMR relaxation measurements of biofouling in model and geological
porous media. Organic Geochemistry, 2011. 42(8): p. 965-971.
Costerton, J.W., Introduction to biofilm. International Journal of Antimicrobial Agents,
1999. 11(3-4): p. 217-221.
Costerton, J.W., et al., Microbial Biofilms. Annual Review of Microbiology, 1995. 49: p.
711-745.
Costerton, J.W., P.S. Stewart, and E.P. Greenberg, Bacterial biofilms: A common cause
of persistent infections. Science, 1999. 284(5418): p. 1318-1322.
Cunningham, A.B., E.J. Bouwer, and W.G. Characklis, Biofilms in porous media, in
Biofilms, W.G. Characklis and K.C. Marshall, Editors. 1990, John Wiley & Sons,
Inc.: New York. p. 697-732.
Cunningham, A.B., et al., Influence of biofilm accumulation on porous-media
hydrodynamics. Environmental Science & Technology, 1991. 25(7): p. 1305-
1311.
Cunningham, A.B., et al., Abandoned well CO2 leakage mitigation using biologically
induced mineralization: current progress and future directions. Greenhouse
Gases-Science and Technology, 2013. 3(1): p. 40-49.
Cunningham, A.B., et al., Subsurface biofilm barriers for the containment and
remediation of contaminated groundwater. Bioremediation Journal, 2003. 7(3-4):
p. 151-164.
Dammel, E.E. and E.D. Schroeder, Density of activated sludge solids. Water Research,
1991. 25(7): p. 841-846.
Da-Wen, G. and T. Yu, Versatility and application of anaerobic ammonium-oxidizing
bacteria. Applied Microbiology & Biotechnology, 2011. 91(4): p. 887-894.
278
de Kreuk, M., J.J. Heijnen, and M.C.M. van Loosdrecht, Simultaneous COD, nitrogen,
and phosphate removal by aerobic granular sludge. Biotechnology and
Bioengineering, 2005. 90(6): p. 761-769.
de Kreuk, M.K., et al., Behavior of polymeric substrates in an aerobic granular sludge
system. Water Research, 2010. 44(20): p. 5929-5938.
de Kreuk, M.K., N. Kishida, and M.C.M. van Loosdrecht, Aerobic granular sludge -
state of the art. Water Science and Technology, 2007. 55(8-9): p. 75-81.
de Kreuk, M.K., M. Pronk, and M.C.M. van Loosdrecht, Formation of aerobic granules
and conversion processes in an aerobic granular sludge reactor at moderate and
low temperatures. Water Research, 2005. 39(18): p. 4476-4484.
de Kreuk, M.K. and M.C.M. van Loosdrecht, Selection of slow growing organisms as a
means for improving aerobic granular sludge stability. Water Science and
Technology, 2004. 49(11-12): p. 9-17.
de Kreuk, M.K. and M.C.M. van Loosdrecht, Formation of aerobic granules with
domestic sewage. Journal of Environmental Engineering-ASCE, 2006. 132(6): p.
694-697.
De Muynck, W., N. De Belie, and W. Verstraete, Microbial carbonate precipitation in
construction materials: A review. Ecological Engineering, 2010. 36(2): p. 118-
136.
Debeer, D., et al., Effects of biofilm structures on oxygen distribution and mass transport.
Biotechnology and Bioengineering, 1994. 43(11): p. 1131-1138.
Decho, A.W., Overview of biopolymer-induced mineralization: What goes on in
biofilms? Ecological Engineering, 2010. 36(2): p. 137-144.
DeJong, J.T., et al., Bio-mediated soil improvement. Ecological Engineering, 2010. 36(2):
p. 197-210.
Dejong, J.T., et al., Biogeochemical processes and geotechnical applications: progress,
opportunities and challenges. Geotechnique, 2013. 63(4): p. 287-301.
Derlacki, Z.J., et al., Diffusion coefficients of methanol and water and the mutual
diffusion coefficient in methanol-water solutions at 278 and 298 K. Journal of
Physical Chemistry, 1985. 89(24): p. 5318-5322.
Dlubac, K., et al., Use of NMR logging to obtain estimates of hydraulic conductivity in
the High Plains aquifer, Nebraska, USA. Water Resources Research, 2013. 49(4):
p. 1871-1886.
279
Drescher, K., et al., Biofilm streamers cause catastrophic disruption of flow with
consequences for environmental and medical systems. Proceedings of the
National Academy of Sciences of the United States of America, 2013. 110(11): p.
4345-4350.
Eberl, H.J., et al., A three-dimensional numerical study on the correlation of spatial
structure, hydrodynamic conditions, and mass transfer and conversion in
biofilms. Chemical Engineering Science, 2000. 55(24): p. 6209-6222.
Ebigbo, A., et al., Darcy-scale modeling of microbially induced carbonate mineral
precipitation in sand columns. Water Resources Research, 2012. 48.
Etterer, T. and P.A. Wilderer, Generation and properties of aerobic granular sludge.
Water Science and Technology, 2001. 43(3): p. 19-26.
Evangelou, V.P. and Y.L. Zhang, A review: Pyrite oxidation mechanisms and acid mine
drainage prevention. Critical Reviews in Environmental Science and Technology,
1995. 25(2): p. 141-199.
Fabich, H., et al., Microbial and algal alginate gelation characterized by magnetic
resonance. Journal of Biotechnology, 2012. 61(3): p. 320– 327.
Fay, E.L. and R.J. Knight, Detecting and quantifying organic contaminants in sediments
with nuclear magnetic resonance. Geophysics, 2016. 81(6): p. EN87.
Fay, E.L., R.J. Knight, and Y.Q. Song, Investigating internal magnetic field gradients in
aquifer sediments. Geophysics, 2015. 80(3): p. D281-D294.
Ferris, F.G., et al., Kinetics of calcite precipitation induced by ureolytic bacteria at 10 to
20 degrees C in artificial groundwater. Geochimica Et Cosmochimica Acta,
2003. 67(8): p. 1701-1722.
Flemming, H.-C. and J. Wingender, The biofilm matrix. Nature Reviews Microbiology,
2010. 8(9): p. 623-633.
Foerster, E., H. Nirschl, and G. Guthausen, NMR Diffusion and Relaxation for
Monitoring of Degradation in Motor Oils. Applied Magnetic Resonance, 2017.
48(1): p. 51-65.
Foley, I., S.A. Farooqui, and R.L. Kleinberg, Effect of paramagnetic ions on NMR
relaxation of fluids at solid surfaces. Journal of Magnetic Resonance Series A,
1996. 123(1): p. 95-104.
Freedman, R., Advances in NMR Logging. Journal of Petroleum Technology, 2006.
58(1): p. 60-66.
280
Fridjonsson, E.O., et al., NMR measurement of hydrodynamic dispersion in porous media
subject to biofilm mediated precipitation reactions. Journal of Contaminant
Hydrology, 2011. 120-21: p. 79-88.
Fukushima, E., Nuclear magnetic resonance as a tool to study flow. Annual Review of
Fluid Mechanics, 1999. 31(1): p. 95.
Gilbert, E.M., et al., Low Temperature Partial Nitritation/Anammox in a Moving Bed
Biofilm Reactor Treating Low Strength Wastewater. Environmental Science &
Technology, 2014. 48(15): p. 8784-8792.
Gjersing, E.L., et al., Magnetic resonance microscopy analysis of advective transport in a
biofilm reactor. Biotechnology and Bioengineering, 2005. 89(7): p. 822-834.
Godefroy, S., et al., Surface nuclear magnetic relaxation and dynamics of water and oil
in macroporous media. Physical Review E, 2001. 64(2): p. 021605-1 - 021605-
13.
Goga, N.O., et al., Mobile NMR: applications to materials and biomedicine. Journal of
Optoelectronics and Advanced Materials, 2006. 8(4): p. 1430-1434.
Gonzalez-Gil, G., et al., Cluster structure of anaerobic aggregates of an expanded
granular sludge bed reactor. Applied and Environmental Microbiology, 2001.
67(8): p. 3683-3692.
Grunewald, E. and R. Knight, A laboratory study of NMR relaxation times in
unconsolidated heterogeneous sediments. Geophysics, 2011. 76(4): p. G73-G83.
Guan, X.Z., et al., PFG-NMR on W1/O/W2-emulsions: Evidence for molecular exchange
between water phases. European Journal of Lipid Science and Technology, 2010.
112(8): p. 828-837.
Guimera, X., et al., Dynamic characterization of external and internal mass transport in
heterotrophic biofilms from microsensors measurements. Water Research, 2016.
102: p. 551-560.
Hahn, E.L., Spin Echoes. Physical Review, 1950. 80: p. 580-594.
Herigstad, B., M. Hamilton, and J. Heersink, How to optimize the drop plate method for
enumerating bacteria. Journal of Microbiological Methods, 2001. 44(2): p. 121-
129.
Herrling, M.P., et al., Determining the Flow Regime in a Biofilm Carrier by Means of
Magnetic Resonance Imaging. Biotechnology and Bioengineering, 2015. 112(5):
p. 1023-1032.
281
Herrling, M.P., et al., Short and long term biosorption of silica-coated iron oxide
nanoparticles in heterotrophic biofilms. Science of The Total Environment, 2016.
544: p. 722-729.
Hertrich, M., Imaging of groundwater with nuclear magnetic resonance. Progress in
Nuclear Magnetic Resonance Spectroscopy, 2008. 53(4): p. 227-248.
Hille, A., et al., Effective Diffusivities and Mass Fluxes in Fungal Biopellets.
Biotechnology and Bioengineering, 2009. 103(6): p. 1202-1213.
Hills, B.P., C. Cano, and P.S. Belton, Proton NMR relaxation studies of aqueous
polysaccharide systems. Macromolecules, 1991. 24(10): p. 2944-2950.
Horn, H. and S. Lackner, Modeling of Biofilm Systems: A Review, in Productive Biofilms,
K. Muffler and R. Ulber, Editors. 2014, Springer-Verlag Berlin: Berlin. p. 53-76.
Horn, H., H. Reiff, and E. Morgenroth, Simulation of growth and detachment in biofilm
systems under defined hydrodynamic conditions. Biotechnology and
Bioengineering, 2003. 81(5): p. 607-617.
Hornemann, J.A., et al., Biopolymer and water dynamics in microbial biofilm
extracellular polymeric substance. Biomacromolecules, 2008. 9(9): p. 2322-2328.
Hoskins, B.C., et al., Selective imaging of biofilms in porous media by NMR relaxation.
Journal of Magnetic Resonance, 1999. 139(1): p. 67-73.
IWA, IWA Task Group on Biofilm Modeling: Mathematical Modeling of Biofilms. 2006:
IWA Publishing.
Jaeger, F., et al., Evaluation of 1H NMR relaxometry for the assessment of pore-size
distribution in soil samples. European Journal of Soil Science, 2009. 60(6): p.
1052-1064.
Jaeger, F., E. Grohmann, and G.E. Schaumann, 1H NMR relaxometry in natural humous
soil samples: insights in microbial effects on relaxation time distributions. Plant
and Soil, 2006. 280(1-2): p. 209-222.
Johns, M., et al., Mobile NMR and MRI: Developments and Applications. New
Developments in NMR. 2015: Royal Society of Chemistry.
Juang, Y.-C., et al., Stable aerobic granules for continuous-flow reactors: Precipitating
calcium and iron salts in granular interiors. Bioresource Technology, 2010.
101(21): p. 8051-8057.
282
Juncher Jørgensen, C., et al., Microbial Oxidation of Pyrite Coupled to Nitrate Reduction
in Anoxic Groundwater Sediment. Environmental Science & Technology, 2009.
43(13): p. 4851-4857.
Jung, D., et al., New colorimetric reaction for endpoint, continuous-flow and kinetic
measurement of urea Clinical Chemistry, 1975. 21(8): p. 1136-1140.
Kao, C.M., et al., Biobarrier system for remediation of TCE-contaminated aquifers.
Bulletin of Environmental Contamination and Toxicology, 2004. 72(1): p. 87-93.
Kao, C.M., et al., Enhanced PCE dechlorination by biobarrier systems under different
redox conditions. Water Research, 2003. 37(20): p. 4885-4894.
Kapellos, G.E., T.S. Alexiou, and A.C. Payatakes, Hierarchical simulator of biofilm
growth and dynamics in granular porous materials. Advances in Water
Resources, 2007. 30(6–7): p. 1648-1667.
Kausik, R. and M.D. Hurlimann, Sensitivity and resolution of two-dimensional NMR
diffusion-relaxation measurements. Journal of Magnetic Resonance, 2016. 270: p.
12-23.
Keating, K. and R. Knight, A laboratory study to determine the effect of iron oxides on
proton NMR measurements. Geophysics, 2007. 72(1): p. E27-E32.
Keating, K. and R. Knight, A laboratory study of the effect of magnetite on NMR
relaxation rates. Journal of Applied Geophysics, 2008. 66(3-4): p. 188-196.
Keating, K. and R. Knight, A laboratory study of the effect of Fe(II)-bearing minerals on
nuclear magnetic resonance (NMR) relaxation measurements. Geophysics, 2010.
75(3): p. F71-F82.
Kenyon, W.E., Petrophysical principles of applications of NMR logging. The Log
Analyst, 1997. 38(2): p. 21-43.
Kenyon, W.E., et al., A three-part study of NMR longitudinal relaxation properties of
water-saturated sandstones. SPE Formation Evaluation, 1988. 3(3): p. 622-636.
Kenyon, W.E. and J.A. Kolleeny, NMR surface relaxivity of calcite with adsorbed Mn2+.
Journal of Colloid and Interface Science, 1995. 170(2): p. 502-514.
Kim, G., S. Lee, and Y. Kim, Subsurface biobarrier formation by microorganism
injection for contaminant plume control. Journal of Bioscience and
Bioengineering, 2006. 101(2): p. 142-148.
Kimmich, R., NMR - Tomography, Diffusometry, Relaxometry. Springer Verlag, 1997.
283
Kirkland, C.M., et al., In Situ Detection of Subsurface Biofilm Using Low-Field NMR: A
Field Study. Environmental Science & Technology, 2015. 49(18): p. 11045-
11052.
Kirkland, C.M., et al., Biofilm Detection in a Model Well-Bore Environment Using Low-
Field NMR. Ground Water Monitoring and Remediation, 2015. 35(4): p. 36-44.
Kirkland, C.M., et al., Detecting microbially-induced calcite precipitation (MICP) in a
model well-bore using downhole low-field NMR. Environmental Science &
Technology, 2016. 10.1021/acs.est.6b04833.
Kleinberg, R.L., Kenyon, W.E., Mitra, P.P., Mechanism of NMR relaxation of fluids in
rock. Journal of Magnetic Resonance Series A, 1994. 108: p. 206-214.
Kleinberg, R.L., S.A. Farooqui, and M.A. Horsfield, T(1)/T(2) ratio and frequency-
dependence of NMR relaxation in porous sedimentary rocks. Journal of Colloid
and Interface Science, 1993. 158(1): p. 195-198.
Kleinberg, R.L. and M.A. Horsfield, Transverse relaxation processes in porous
sedimentary rock. Journal of Magnetic Resonance, 1990. 88(1): p. 9-19.
Knight, R., et al., Field experiment provides ground truth for surface nuclear magnetic
resonance measurement. Geophysical Research Letters, 2012. 39.
Knight, R., et al., NMR Logging to Estimate Hydraulic Conductivity in Unconsolidated
Aquifers. Groundwater, 2016. 54(1): p. 104-114.
Koenig, S.H. and K.E. Kellar, Theory of 1/T-1 and 1/T-2 NMRD profiles of solutions of
magnetic nanoparticles Magnetic Resonance in Medicine, 1995. 34(2): p. 227-
233.
Komlos, J., et al., Biofilm barriers to contain and degrade dissolved trichloroethylene.
Environmental Progress, 2004. 23(1): p. 69-77.
Korb, J.P., S. Godefroy, and M. Fleury, Surface nuclear magnetic relaxation and
dynamics of water and oil in granular packings and rocks. Magnetic Resonance
Imaging, 2003. 21(3-4): p. 193-199.
Kreft, J.U., et al., Individual-based modelling of biofilms. Microbiology-Sgm, 2001. 147:
p. 2897-2912.
Kristian Stevik, T., et al., Retention and removal of pathogenic bacteria in wastewater
percolating through porous media: a review. Water Research, 2004. 38(6): p.
1355-1367.
284
LaGrega, M.D., P.L. Buckingham, and J.C. Evans, Hazardous Waste Management. 2nd
ed. 2001, New York: McGraw-Hill.
Lauterbur, P.C., Image formation by induced local interactions: examples employing
nuclear magnetic resonance. Nature, 1973. 242: p. 190-191.
Lazarova, V. and J. Manem, Biofilm characterization and activity analysis in water and
waste-water treatment. Water Research, 1995. 29(10): p. 2227-2245.
Lee, J.H., et al., 2-Dimensional Inverse Laplace Transform Nmr - Altered Relaxation-
Times Allow Detection of Exchange-Correlation. Journal of the American
Chemical Society, 1993. 115(17): p. 7761-7764.
Lens, P.N.L., et al., Diffusional properties of methanogenic granular sludge: H-1 NMR
characterization. Applied and Environmental Microbiology, 2003. 69(11): p.
6644-6649.
Lewandowski, Z., et al., NMR imaging of hydrodynamics near microbially colonized
surfaces. Water Science and Technology, 1992. 26(3-4): p. 577-584.
Li, C.Y., et al., Assessing the influence of biofilm surface roughness on mass transfer by
combining optical coherence tomography and two-dimensional modeling.
Biotechnology and Bioengineering, 2016. 113(5): p. 989-1000.
Lin, Y., et al., Characterization of alginate-like exopolysaccharides isolated from aerobic
granular sludge in pilot-plant. Water Research, 2010. 44(11): p. 3355-3364.
Liu, Y. and J.H. Tay, The essential role of hydrodynamic shear force in the formation of
biofilm and granular sludge. Water Research, 2002. 36(7): p. 1653-1665.
Loosdrecht, M.C.M.v., et al., Experimental Methods in Wastewater Treatment. 2016: Intl
Water Assn, IWA.
Lotti, T., et al., Pilot-scale evaluation of anammox-based mainstream nitrogen removal
from municipal wastewater. Environmental technology, 2015. 36(9): p. 1167-77.
Maliva, R.G., E.A. Clayton, and T.M. Missimer, Application of advanced borehole
geophysical logging to managed aquifer recharge investigations. Hydrogeology
Journal, 2009. 17(6): p. 1547-1556.
Manas, A., et al., Location and chemical composition of microbially induced phosphorus
precipitates in anaerobic and aerobic granular sludge. Environmental
Technology, 2012. 33(19): p. 2195-2209.
285
Manz, B., et al., Investigation of Biofilm Structure, Flow Patterns and Detachment with
Magnetic Resonance Imaging. Water Science and Technology, 2005. 52(7): p. 1-
6.
Marble, A.E., et al., A constant gradient unilateral magnet for near-surface MRI
profiling. Journal of Magnetic Resonance, 2006. 183(2): p. 228-234.
McLean, J.S., O.N. Ona, and P.D. Majors, Correlated biofilm imaging, transport and
metabolism measurements via combined nuclear magnetic resonance and
confocal microscopy. Isme Journal, 2008. 2(2): p. 121-131.
Meiboom, S. and D. Gill, Modified spin-echo method for measuring nuclear relaxation
times. Review of Scientific Instruments, 1958. 29(8): p. 688-691.
Melton, E.D., et al., The interplay of microbially mediated and abiotic reactions in the
biogeochemical Fe cycle. Nat Rev Micro, 2014. 12(12): p. 797-808.
Metcalf & Eddy, I., Wastewater engineering : treatment and reuse. Fourth edition ed.
2003, Boston: McGraw-Hill, 2003.
Minsley, B.J., et al., Evidence for nonuniform permafrost degradation after fire in boreal
landscapes. Journal of Geophysical Research: Earth Surface, 2016. 121(2): p.
320-335.
Mitchell, A.C., et al., Microbially Enhanced Carbon Capture and Storage by Mineral-
Trapping and Solubility-Trapping. Environmental Science and Technology, 2010.
44(13): p. 5270-5276.
Mitchell, A.C., et al., Resilience of planktonic and biofilm cultures to supercritical CO2.
Journal of Supercritical Fluids, 2008. 47(2): p. 318-325.
Mortensen, B.M., et al., Effects of environmental factors on microbial induced calcium
carbonate precipitation. Journal of Applied Microbiology, 2011. 111(2): p. 338-
349.
Mualem, Y., Hysteretical models for prediction of hydraulic conductivity of unsaturated
porous-media. Transactions-American Geophysical Union, 1976. 57(8): p. 602-
603.
Neu, T.R., et al., Advanced imaging techniques for assessment of structure, composition
and function in biofilm systems. FEMS Microbiology Ecology, 2010. 72(1): p. 1-
21.
Nicolella, C., M.C.M. van Loosdrecht, and J.J. Heijnen, Wastewater treatment with
particulate biofilm reactors. Journal of Biotechnology, 2000. 80(1): p. 1-33.
286
Packer, K.J. and C. Rees, Pulsed NMR studies of restricted diffusion .1. Droplet size
distributions in emulsions. Journal of Colloid and Interface Science, 1972. 40(2):
p. 206-218.
Pal, A. and A.K. Paul, Microbial extracellular polymeric substances: central elements in
heavy metal bioremediation. Indian Journal of Microbiology, 2008. 48(1): p. 49-
64.
Parsekian, A.D., et al., Bootstrap Calibration and Uncertainty Estimation of Downhole
NMR Hydraulic Conductivity Estimates in an Unconsolidated Aquifer.
Groundwater, 2015. 53(1): p. 111-121.
Perlo, J., et al., Optimized slim-line logging NMR tool to measure soil moisture in situ.
Journal of Magnetic Resonance, 2013. 233: p. 74-79.
Petrova, O.E. and K. Sauer, Sticky Situations: Key Components That Control Bacterial
Surface Attachment. Journal of Bacteriology, 2012. 194(10): p. 2413-2425.
Peulen, T.-O. and K.J. Wilkinson, Diffusion of Nanoparticles in a Biofilm. Environmental
Science & Technology, 2011. 45(8): p. 3367-3373.
Pfeuffer, J., et al., Restricted diffusion and exchange of intracellular water: Theoretical
modelling and diffusion time dependence of 1H NMR measurements on perfused
glial cells. NMR in Biomedicine, 1998. 11: p. 19-31.
Pfeuffer, J., U. Flögel, and D. Leibfritz, Monitoring of cell volume and water exchange
time in perfused cells by diffusion-weighted 1H NMR spectroscopy. NMR in
Biomedicine, 1998. 11: p. 11-18.
Phillips, A.J., et al., Fracture Sealing with Microbially-Induced Calcium Carbonate
Precipitation: A Field Study. Environmental Science & Technology, 2016. 50(7):
p. 4111-4117.
Phillips, A.J., et al., Engineered applications of ureolytic biomineralization: a review.
Biofouling, 2013. 29(6): p. 715-733.
Phillips, A.J., et al., Potential CO2 Leakage Reduction through Biofilm-Induced Calcium
Carbonate Precipitation. Environmental Science & Technology, 2013. 47(1): p.
142-149.
Phoenix, V.R. and W.M. Holmes, Magnetic resonance imaging of structure, diffusivity,
and copper immobilization in a phototrophic biofilm. Applied and Environmental
Microbiology, 2008. 74(15): p. 4934-4943.
287
Picioreanu, C., J.U. Kreft, and M.C.M. van Loosdrecht, Particle-based multidimensional
multispecies Biofilm model. Applied and Environmental Microbiology, 2004.
70(5): p. 3024-3040.
Picioreanu, C., M.C.M. van Loosdrecht, and J.J. Heijnen, Effect of diffusive and
convective substrate transport on biofilm structure formation: A two-dimensional
modeling study. Biotechnology and Bioengineering, 2000. 69(5): p. 504-515.
Pintelon, T.R.R., et al., The effect of biofilm permeability on bio-clogging of porous
media. Biotechnology and Bioengineering, 2012. 109(4): p. 1031-1042.
Poot, V., et al., Effects of the residual ammonium concentration on NOB repression
during partial nitritation with granular sludge. Water Research, 2016. 106: p.
518-530.
Potter, K., et al., Assay for bacteria in porous media by diffusion-weighted NMR. Journal
of Magnetic Resonance. Series B, 1996. 113: p. 9-15.
Prange, M. and Y.-Q. Song, Quantifying uncertainty in NMR T-2 spectra using Monte
Carlo inversion. Journal of Magnetic Resonance, 2009. 196(1): p. 54-60.
Pronk, M., et al., Effect and behaviour of different substrates in relation to the formation
of aerobic granular sludge. Applied Microbiology and Biotechnology, 2015.
99(12): p. 5257-5268.
Pronk, M., et al., Full scale performance of the aerobic granular sludge process for
sewage treatment. Water Research, 2015. 84: p. 207-217.
Provencher, S.W., A constrained regularization method for inverting data represented by
linear algebraic or integral equations Computer Physics Communications, 1982.
27(3): p. 213-227.
Ramanan, B., et al., Magnetic Resonance Imaging of Mass Transport and Structure
Inside a Phototrophic Biofilm. Current Microbiology, 2013. 66(5): p. 456-461.
Ranzinger, F., et al., Direct surface visualization of biofilms with high spin coordination
clusters using Magnetic Resonance Imaging. Acta Biomaterialia, 2016. 31: p.
167-177.
Renslow, R.S., et al., In Situ Effective Diffusion Coefficient Profiles in Live Biofilms
Using Pulsed-Field Gradient Nuclear Magnetic Resonance. Biotechnology and
Bioengineering, 2010. 106(6): p. 928-937.
Robinson, D.A., et al., Soil moisture measurement for ecological and hydrological
watershed-scale observatories: A review. Vadose Zone Journal, 2008. 7(1): p.
358-389.
288
Röding, M., et al., The gamma distribution model for pulsed-field gradient NMR studies
of molecular-weight distributions of polymers. Journal of Magnetic Resonance,
2012. 222: p. 105-111.
Röding, M., N. Williamson, and M. Nydén, Gamma convolution models for self-diffusion
coefficient distributions in PGSE NMR. Journal of Magnetic Resonance, 2015.
261: p. 6-10.
Sanderlin, A.B., et al., Biofilm detection in natural unconsolidated porous media using a
low-field magnetic resonance system. Environmental Science & Technology,
2013. 47(2): p. 987-992.
Sauer, K., et al., Pseudomonas aeruginosa displays multiple phenotypes during
development as a biofilm. Journal of Bacteriology, 2002. 184(4): p. 1140-1154.
Seymour, J.D., et al., Magnetic resonance microscopy of biofilm structure and impact on
transport in a capillary bioreactor. Journal of Magnetic Resonance, 2004.
167(322-327).
Seymour, J.D., et al., Anomalous fluid transport in porous media induced by biofilm
growth. Physical Review Letters, 2004. 93(19).
Seymour, J.D., et al., Magnetic resonance microscopy of biofouling induced scale
dependent transport in porous media. Advances in Water Resources, 2007. 30(6-
7): p. 1408-1420.
Sham, E., et al., Monitoring bacterially induced calcite precipitation in porous media
using magnetic resonance imaging and flow measurements. Journal of
Contaminant Hydrology, 2013. 152: p. 35-43.
Song, Y.Q., Recent Progress of Nuclear Magnetic Resonance Applications in Sandstones
and Carbonate Rocks. Vadose Zone Journal, 2010. 9(4): p. 828-834.
Song, Y.Q., et al., T-1-T-2 correlation spectra obtained using a fast two-dimensional
Laplace inversion. Journal of Magnetic Resonance, 2002. 154(2): p. 261-268.
Staudt, C., et al., Volumetric measurements of bacterial cells and extracellular polymeric
substance glycoconjugates in biofilms. Biotechnology and Bioengineering, 2004.
88(5): p. 585-592.
Stejskal, E.O. and J.E. Tanner, Spin diffusion measurements: Spin echoes in the presence
of a time-dependent field gradient. Journal of Chemical Physics, 1965. 42: p. 288.
Stewart, P.S., A review of experimental measurements of effective diffusive permeabilities
and effective diffusion coefficients in biofilms. Biotechnology and Bioengineering,
1998. 59(3): p. 261-272.
289
Stewart, P.S., Diffusion in biofilms. Journal of Bacteriology, 2003. 185(5): p. 1485-1491.
Stewart, P.S. and M.J. Franklin, Physiological heterogeneity in biofilms. Nature Reviews
Microbiology, 2008. 6(3): p. 199-210.
Stewart, T.L. and H.S. Fogler, Biomass plug development and propagation in porous
media. Biotechnology and Bioengineering, 2001. 72(3): p. 353-363.
Stoodley, P., D. Debeer, and Z. Lewandowski, Liquid flow in biofilm systems. Applied
and Environmental Microbiology, 1994. 60(8): p. 2711-2716.
Stoodley, P., et al., Biofilms as complex differentiated communities. Annual Review of
Microbiology, 2002. 56: p. 187-209.
Straley, C., et al., Core analysis by low-field NMR. The Log Analyst, 1997. 38(2): p. 84-
94.
Sucre, O., et al., Low-field NMR logging sensor for measuring hydraulic parameters of
model soils. Journal of Hydrology, 2011. 406(1–2): p. 30-38.
Sutherland, I.W., Biofilm exopolysaccharides: a strong and sticky framework.
Microbiology, 2001. 147: p. 3-9.
Sutherland, I.W., The biofilm matrix - an immobilized but dynamic microbial
environment. Trends in Microbiology, 2001. 9(5): p. 222-227.
Tay, J.H., Q.S. Liu, and Y. Liu, The role of cellular polysaccharides in the formation and
stability of aerobic granules. Letters in Applied Microbiology, 2001. 33(3): p.
222-226.
Taylor, S.W. and P.R. Jaffe, Biofilm growth and the related changes in the physical
properties of a porous medium. 1. Experimental investigation. Water Resources
Research, 1990. 26(9): p. 2153-2159.
Taylor, S.W. and P.R. Jaffe, Enhanced in-situ biodegradation and aquifer permeability
reduction. Journal of Environmental Engineering-ASCE, 1991. 117(1): p. 25-46.
Thullner, M. and P. Baveye, Computational pore network modeling of the influence of
biofilm permeability on bioclogging in porous media. Biotechnology and
Bioengineering, 2008. 99(6): p. 1337-1351.
Timur, A., Pulsed nuclear magnetic resonance studies of porosity, movable fluid, and
permeability of sandstones. Journal of Petroleum Technology, 1969. 21(JUN): p.
775-&.
Torrey, H.C., Bloch Equations with Diffusion Terms. Physical Review, 1956. 104(3): p.
563-565.
290
Tychonoff, A.N. and V.Y. Arsenin, Solution of ill-posed problems. 1977, Washington:
Winston and Sons.
Van As, H. and P. Lens, Use of H-1 NMR to study transport processes in porous
biosystems. Journal of Industrial Microbiology & Biotechnology, 2001. 26(1-2):
p. 43-52.
Van Genuchten, M.T., A closed form equation for predicting the hydraulic conductivity
of unsaturated soils. Soil Science Society of America Journal, 1980. 44(5): p.
892-898.
van Paassen, L.A., et al., Quantifying Biomediated Ground Improvement by Ureolysis:
Large-Scale Biogrout Experiment. Journal of Geotechnical and Geoenvironmental
Engineering, 2010. 136(12): p. 1721-1728.
Vandevivere, P. and P. Baveye, Effect of bacterial extracellular polymers on the
saturated hydraulic conductivity of sand columns. Applied and Environmental
Microbiology, 1992. 58(5): p. 1690-1698.
Venkataramanan, L., Y.Q. Song, and M.D. Hurlimann, Solving Fredholm integrals of the
first kind with tensor product structure in 2 and 2.5 dimensions. Ieee Transactions
on Signal Processing, 2002. 50(5): p. 1017-1026.
Vogt, M., H.C. Flemming, and W.S. Veeman, Diffusion in Pseudomonas aeruginosa
biofilms: a pulsed field gradient NMR study. Journal of Biotechnology, 2000.
77(1): p. 137-146.
Vogt, S.J., et al., Permeability of a growing biofilm in a porous media fluid flow analyzed
by magnetic resonance displacement-relaxation correlations. Biotechnology and
Bioengineering, 2013. 110(5): p. 1366-1375.
von der Schulenburg, D.A.G., et al., Spatially resolved quantification of metal ion
concentration in a biofilm-mediated ion exchanger. Biotechnology and
Bioengineering, 2008. 99(4): p. 821-829.
von der Schulenburg, D.A.G., et al., Three-Dimensional Simulations of Biofilm Growth in
Porous Media. AIChE Journal, 2009. 55(2): p. 494-504.
Wagner, J., et al., Effect of particulate organic substrate on aerobic granulation and
operating conditions of sequencing batch reactors. Water Research, 2015. 85: p.
158-166.
Wagner, M., et al., Online assessment of biofilm development, sloughing and forced
detachment in tube reactor by means of magnetic resonance microscopy.
Biotechnology and Bioengineering, 2010. 107(1): p. 172-181.
291
Walsh, D., Environmental Geophysics: Portable NMR tools for measuring and
monitoring soil moisture, in Preview. 2015, CSIRO Publishing. p. 41-42.
Walsh, D., et al., A small-diameter NMR logging tool for groundwater investigations.
Groundwater, 2013. 51(6): p. 914-926.
Walsh, D.O., et al., Javelin: A slim-hole and microhole NMR logging tool. Fast Times,
2010. 15(3): p. 67-72.
Wäsche, S., H. Horn, and D.C. Hempel, Influence of growth conditions on biofilm
development and mass transfer at the bulk/biofilm interface. Water Research,
2002. 36: p. 4775-4784.
Wheeler, L.A., Establishment of ureolytic biofilms and their influence on the
permeability of pulse-flow porous media column systems. 2009, Montana State
University, Bozeman, MT.
Windt, C.W., F.J. Vergeldt, and H. Van As, Correlated displacement-T-2 MRI by means
of a Pulsed Field Gradient-Multi Spin Echo method. Journal of Magnetic
Resonance, 2007. 185(2): p. 230-239.
Winkler, M.K.H., et al., Factors influencing the density of aerobic granular sludge.
Applied Microbiology and Biotechnology, 2013. 97(16): p. 7459-7468.
Wood, B.D., M. Quintard, and S. Whitaker, Calculation of effective diffusivities for
biofilms and tissues. Biotechnology and Bioengineering, 2002. 77(5): p. 495-516.
Xavier, J.B., et al., Biofilm-control strategies based on enzymic disruption of the
extracellular polymeric substance matrix - a modelling study. Microbiology-Sgm,
2005. 151: p. 3817-3832.
Zhang, Q., et al. Some exceptions to default NMR rock and fluid properties. in
Transactions of the SPWLA Annual Logging Symposium (Society of Professional
Well Log Analysts). 1998.