Laser frequency stabilization to spectral hole burning frequency references in erbium-doped crystals :
material and device optimization
by Thomas Bottger
A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of
Philosophy in Physics
Montana State University
© Copyright by Thomas Bottger (2002)
Abstract:
Narrow spectral holes in the absorption lines of Er^3+ doped crystals have been explored as references
for frequency stabilizing external cavity diode lasers at the important 1.5 μm optical communication
wavelength. Allan deviations of the beat signal between two independent stabilized lasers as low as
200 Hz over 10 ms integration time have been achieved using regenerative spectral holes in
Er^3+Y2SiO5 and Er^3+:KTP, while drift was reduced to ~ 7 kHz/min by incorporating the
inhomogeneous absorption line as a fixed reference. During active stabilization, the transient spectral
hole was continuously regenerated as hole burning balanced relaxation. In contrast, persistent spectral
holes in Er^3+:D^-:CaF2, with lifetimes of several weeks, provided programmable and transportable
secondary frequency references that maintained sub-kilohertz stability over several seconds and
enabled 6 kHz stability over 1.6xl0^3s. The error signal was derived from the spectral hole
transmission using frequency modulation spectroscopy. A servo amplifier applied fast frequency
corrections to the injection current of the laser diode and slower adjustments to the piezo-driven
feedback prism plate.
These stabilized lasers provide ideal sources for spectral hole burning applications based on optical
coherent transients, where laser stability is required over the storage time of the material. Since the
lifetime of the frequency reference is exactly the material storage time, this requirement is
automatically met by using our technique. This was demonstrated in Er^3+:Y2SiO5 and successfully
transferred to high-bandwidth signal processing applications.
The material Er^3+Y2SiO5 was optimized for these applications. The 4I15/2 and 4I13/2 crystal field
levels were site-selectively determined by absorption and fluorescence spectroscopy. The excited state
lifetime was measured to be 11.4 ms for site 1 and 9.2 ms for site 2. Zeeman experiments and
two-pulse photon echo spectroscopy as a function of magnetic field orientation were used to determine
the anisotropic electronic g-values for both Er^3+ sites and established a preferred magnetic field
orientation for minimizing homogeneous line broadening by spectral diffusion. The spectral diffusion
was characterized by stimulated photon echo spectroscopy and successfully described with established
theories. In a 0.02 atomic percent Er^3+:Y2SiO5 crystal at B = 0.8 T and T = 1.6 K, line broadening
became significant after 10 μs, increasing the homogeneous linewidth from 7.5 kHz to 75 kHz after
120 μs. Spectral diffusion, primarily caused by direct phonon driven Er^3+ spin-flips in the ground
state, can be controlled to negligible levels with proper magnetic field strength and orientation,
temperature, and erbium concentration. In optimizing Er^3+:Y2SiO5, the narrowest optical resonance
in any solid-state material of 73 Hz was measured.
LASER FREQUENCY STABILIZATION TO SPECTRAL HOLE BURNING
FREQUENCY REFERENCES IN ERBIUM-DOPED CRYSTALS:
MATERIAL AND DEVICE OPTIMIZATION
by
Thomas Bottger
A dissertation submitted in partial fulfillment
of the requirements fdr the degree
of
Doctor of Philosophy
in
Physics
MONTANA STATE UNIVERSITY
Bozeman, Montana
April 2002
ii
APPROVAL
. of a dissertation submitted by
Thomas Btittger
This dissertation has been read by each member of the dissertation committee
and has been found to be satisfactory regarding content, English usage, format,
citations, bibliographic style, and consistency, and is ready for submission to the ,
College of Graduate Studies,
Rufus L. Cone, III -Md
(Date)
100 2.______
John C. Hermanson
Approved for the Department of Physics
I
(Date)
Approved for the College of Graduate Studies
Bruce R. McLeod ^ ~ ^ ^
(Signature) / ' _ (Date)
iii
STATEMENT OF PERMISSION TO USE
In presenting this dissertation, in partial fulfillment of the requirements for a
doctoral degree at Montana State University, I agree that the Library shall make it
available to borrowers under the rules of the Library. I further agree that copying of
this dissertation is allowable only for scholary purposes, consistent with “fair use” as
prescribed in the U.S. Copyright Law. Requests for extensive copying or
reproduction, of this dissertation should be referred to Bell & Howell Information
and Learning, 300 North Zeeb Road, Ann Harbor, Michigan 48106, to whom I have
granted “the exclusive right to reproduce and distribute my dissertation in and from
microform along with the non-exclusive right to reproduce and distribute my abstract
in any format in whole or in part.”
Signature. / / 4 f}'
Date ^ / eY — 2_eo 2_
iv
ACKNOWLEDGMENTS
I want to thank my advisor, Prof. Rufus L. Cone, for the opportunity to work on
this extraordinary project. I very much enjoyed the freedom and independence he
gave me. He has been a major influence in my development as a physicist.
I would like to thank Dr. Yongchen Sun for teaching me rare earth spectroscopy
and sharing his experimental expertise and Charles W. Thiel for assistance with the
spectral diffusion theory and computer data acquisition. I also would like to thank
Dr. Geoff J. Pryde who has been a coworker for two years On this project. My fellow
lab mates, graduate students Gregory Reinemer and Todd L. Harris and postdoctoral
researchers Nick Strickland, Flurin Kdnz, and Alain Brand all have contributed to
my knowledge of physics. Tm greatly indebted to G. Casey Dodge, Dustin Rich and
Norm Williams for their quality machining work, to Dr. Gregg W. Switzer and Prof.
John L. Carlsten for sharing the initial laser design, and to Drs. Peter B. Sellin and
Kevin S. Repasky for teaching me the basics of external cavity diode lasers and laser
frequency stabilization. I thank Steve Kelly and Kevin Rosen in the MSU electronic
shop for assistance in circuit board fabrication. Fm also grateful to Dr. C. Michael
Jefferson of IBM Almaden Research Center for sharing circuit ideas, noise reduction
techniques, as well as memorable days at Bridger Bowl.
A special thanks goes to my wife, Tini, who certainly has made the most
sacrifices caused by my frequent absence from home and our years of. separation.
Her love and encouragement made this possible.
TABLE OF CONTENTS
LIST OF TABLES............................................................................... viii
LIST OF FIGURES................................................................... ..........:....................ix
ABSTRACT........... ........................................................ .................. ................. ,.xvii
1. INTRODUCTION. ................ .............................................................................. I
Frequency Stabilized Lasers and Their Applications............ ......................... 4
Spectral Hole Burning and Optical Coherent Transients................... . ........ 7
Overview of the Dissertation......... ........... ...... ..................................... ......11
References........................................ ...13
2. MATERIALS - BACKGROUND AND THEORY.............. .......... .................18
Homogeneous and Inhomogeneous Broadening................... 20
Spectral Hole Burning......... .......................... ....................... ................... .23
Photon Echoes................................. 25
Two Pulse Photon Echo..................... ,...! .......,..26
Stimulated Photon Echo................................ ...,29
References...................................... ,...,.31
3. LASER FREQUENCY STABILIZATION - BACKGROUND
AND THEORY. .................................................................... ................... ..33
Frequency Modulation Spectroscopy........ ,........... 33
Maximizing the Slope of the Error Signal.............. .............................. ,41
Measuring and Characterizing Laser Frequency Stability.. . .____...........49
Measurement T echniques ................................ ............ ......... ............ ,50
Frequency Domain-The Spectral Noise Density.. ............................ ....53
Time Domain-The Allan Variance......................................... ........ ..,.56
The External Cavity Diode Laser................. ,..............59
Principle of Operation......... .............................. ................................... 61
ECDL Construction and Characterization...................... 63
ECDL Transducer Response. ........ ....... ............... ...............;...... ' 70
Spectral Noise Density of the free running ECDL................ .......... . .'.74
References........... ........ ........................... ........... ............ ..................... „.80
4. LASER FREQUENCY STABILIZATION TO SPECTRAL HOLES...............83
Laser frequency stabilization to regenerative spectral holes in
Er^YzSiOs....................................;.........:............................................ ......85
Methods and. Apparatus...................... .86
Results and Discussion...... ........................................ 92
Incorporating the Absorption Line as a Fixed Reference................ .......94
Improved Photon Echo Stability for Applications............................... .100
Frequency response of a regenerative spectral hole......... ......................... 103
Methods and Apparatus......................................... 104
Discussion................ ................ .............. ....................... ;....................107
Laser frequency stabilization to regenerative spectral holes in
Er3+:KTP............................. :....................................... ...................... ........109
Methods and Apparatus.............................................. .......... ...... . 110
Results and Discussion,............ ........................................... ................113
Conclusions.................................... 116
Laser frequency stabilization to persistent spectral holes in
Er3+ID-CaF2... . ............. 117
Methods and Apparatus.................. ..117
Results and Discussion............ .......... 122
Conclusion................................. 126
References....................................... ....... ......................................... ...... 128
5. SPECTROSCOPY AND, DYNAMICS OF Er3+IY2SiO5................ ................. 132
Introduction arid Motivation,....................................................................... 132
Conventional spectroscopy............................. .......... , , ......... . .......... ;....... 136
Methods and Apparatus........ ............................................................ 136
Results arid Discussion............... 140
Lifetiriie Measurements......................................... 145
Methods and Apparatus......................................... .....................;....... 146
. Results and Discussion................. 148
Zeeman experiments......... .................... 151
Methods and Apparatus....... .................................... 156
The Zeeman Effect............ ...................... 160
Zeeman Experiments as a Function of Field Orientation................ _.. 162
Results.......................................................... .......................... ....163
Discussion..................... 173
Nonlinear Spectroscopy.. , .■....................................................... 180
Methods and Apparatus.................. 180
Two-Pulse Photon Echo Spectroscopy as a Function of Field
Orientation................... 183
Results...................................................... 184
vi
Vii
. '
Discussion............................................ ............................ .............. . .187
Stimulated Photon Echo Spectroscopy and Spectral Diffusion........... .......191
Results................................ ....................................................... ...........193
Discussion pf Spectral Diffusion in 0.02% Er3+IYaSiOs as a
Function of Magnetic Field........................ ........ ..................................199
Discussion of Spectral Diffusion in 0.02% Er3+IY2SiOs as a
Function of Temperature................ .................................................... . 204
Discussion of Spectral Diffusion in Er3+IY2SiOs as a
Function of Erbium Concentration................ ............ ......................... 205
Stimulated Photon Echo T-Decays. .......... 206
Results and Discussion................. ...... ........... . . ................................. 208
Ultraslow Dephasing........................... ...... ........ ....................... ........,.,..211
Results and Discussion .......................... ........ ................... ........... 211
Operation of Er3+=Y2SiOs at elevated temperatures,............................ ......213
Results and Discussion......... ;....... ........ .................................. ......... 214
References....................................................... i........................................218
6. SUMMARY.................... 221
APPENDICES......................................... 227
APPENDIX A - ELECTRONIC FEEDBACK.... ....... .........................................228
APPENDIX B - REFERENCE CAVITY.......... .......... ........................................ 251
APPENDIX C - BAI-FAYER THEORY OF SPECTRAL
DIFFUSION............................... 258
References..... ...... ................ ................................................ ...............,263
viii
Table Page
1 Common noise types found in frequency standards and their
relation to the spectral noise density, Sv (f), and
Allan deviation, ay{f)................................................... .............................. 59
2 Crystal field levels of Er^iYzSiOg as determined from
absorption and site selective fluorescence experiments........................... . 145
3 Fitted g-tensor values for ground and excited state of site I and 2
with respective orientations I and 2 in the three optical
planes.................. ............................................................. ....................... 172
LIST OF TABLES
ix
Figure Page
■2.1 Line broadening and spectral hole burning .................. ....:.....22
2.2 A typical two pulse photon echo sequence. ................. ....................27
2.3 A typical stimulated photon echo pulse sequence................ ............. 27
3.1 Block diagram of the experimental setup for frequency
modulation (FM) spectroscopy based on phase sensitive
detection........... .........•........ ....... ....... ........... ................................... 34
3.2 Spectrum of the electric field after passing the phase
modulator......................................................... .35
3.3 Calculated FM-signals for F = I 5M= I , and com- 10.......................40
3.4 Calculated FM-signals for modulation index, M= I, as a
function of modulation frequency, CQm.................................... 43
3.5 Calculated FM-signals for modulation frequency CQm = 5 as a
function of frequency, the modulation index, M,
varies between subplots................................................... 45
3.6 Calculated FM-signal slopes at line center as a' function
of modulation frequency, GOm............................................ 47
3.7 Calculated FM-signal slopes at line center as a function
of modulation index, M.................................................... 48
3.8 Experimental setup to characterize the frequency stability
of a laser using error signal analysis................................. ............... 51
3.9 Experimental setup to characterize the frequency stability
of a laser using beat note analysis.................... .................................. 51
3.10 Illustration of the power law model for the spectral noise .
density as a function of frequency (a) and the Allan
LIST OF FIGURES
deviation as a function of integration time (b)....................................55
3.11 Schematic of the external cavity diode laser in the
Littman-Metcalf configuration..................................... ..................... 64
3.12 ECDL-spectmfn near 1535 nm showing a side-mode
suppression ratio of ~ 51 dB............................................................. 66
3.13 Optical output power versus injection current (PI curve)
for the ECDL “Max”............ :....................................... ................... 67
3.14 Continuous tuning curve for ECDL “Max”, Continuous
tuning is demonstrated over ~ 47 GHz, center wavelength
~ 1535 nm, 1 = 60 mA............ ....... ....................................................69
3.15 Experimental setup used to measure the transducer response
of the ECDL with a low finesse cavity as a
frequency discriminator................... 71
3.16 Transducer response of the tuning elements of the
ECDL............................................................................................ .....72
3.17 Experimental setup for measuring the spectral noise
density of the free running laser.......................................................... 75
3.18 Spectral noise density of the free running ECDL as a function
of noise frequency; . note double logarithmic
scales.............................. 78
4.1 Transmission spectrum of 0.005% Er3+IY2SiO5 scanned by a
diode laser showing the entire inhomogeneOusly broadened
absorption profile at zero applied field
(B = OT)............................................. 88
4.2. Experimental apparatus. for laser frequency locking to
spectral holes and beat frequency measurement of laser
stability..................... 90
4.3 (a) Transmission spectrum, as probed by a phase-modulated
laser with sidebands, of a single spectral hole burned in the
inhomogeneously broadened absorption profile by an
unmodulated second laser, in an applied field of B = 0.2 T.
(b) Error signals derived from the spectral hole and the
inhomogeneous line using different
phase delay settings 91
xi
4.4 Allan deviation values for the beat between two lasers: (a)
lasers free-running (triangles), (b) locked to spectral holes in
different crystals using straight quadrature detection of the
error signal at applied field B = 0.5 T (squares), and (c)
locked to spectral holes using the strategy of intermediate
phase detection of the combined error signal from the
spectral hole and inhomogeneous line at applied field
B = 0.2 T (circles)...............................................................................94
4.5 Change in heterodyne beat signal between (a) free-running
and (b) independently locked lasers to separate spectral holes
and. inhomogeneous lines in different
crystals at field B = 0.2 T .....,.......... ..98
4.6 Stimulated photon echo decay on the 4I15y2 (I) —> 4I13y2 (I)
transition in Er3+: Y2SiO5................................................................... 102
4.7 Experimental setup used to measure the frequency response
of a regenerative spectral hole ' in 0.001 %
Er3+IY2SiO5..................................... 105
4.8 Frequency response of the error signal generated from a
regenerative spectral hole frequency reference in 0.001 %
Er3+=Y2SiO5...................................................................................... 108
4.9 (a) Transmission spectrum of 0.004 % Er3+=KTP at 1537 nm
showing the entire inhomogeneously broadened 4I15y2 —» 4I13y2
optical absorption scanned by a diode probe laser. The arrow
indicates a spectral hole, which has been burned by a second
laser, (b) Transmission of a phase-modulated probe laser
through a single spectral hole created by a second laser, using
an applied magnetic field of B=0.25T. (c) Demodulated FM-
error signal derived from the spectral hole in (b) ..............................I l l
4.10 Allan deviation for the heterodyne beat frequency between
two lasers: (a) lasers free-running, (b) independently locked
to transient spectral holes in the
4Ii5/2—>4Ii3/2 transition in Er3+=KTP at 1537 nm............................... 114
4.11 (a) Transmission spectrum of Er3+=D =CaF2 at 1523 nm. A
number of spectral holes have been burned into the
xii
inhomogeneously broadened 4Ii5z2 —» 4Ii3z2 optical absorption
for demonstrating the programmability of the material.
Spectral hole burning is not limited to the center of the line.
The arrow indicates a spectral hole, which is enlarged in (b).
(c) Demodulated FM-error signal derived
from the spectral hole in (b)..................... .......... . ............................120
4.12 Root Allan variance for the heterodyne beat frequency
between two lasers: (a) lasers free-running, (b) independently
locked to persistent spectral holes in the
4Ii5/2 —> 4Ii3/2 transition in Er3"1": D =CaF2 at 1523 nm.. .....................123
4.13 Subset of the change in heterodyne beat frequency between
(a) free running and (b) independently locked lasers to
persistent spectral holes in separate crystals over
a period of 10 minutes......... .................. ..........................................125
5.1 Experimental setup for broadband absorption and site-
selective fluorescence....................... 137
5.2 Polarized E H D2 lamp absorption spectrum of 2%
Er3+=Y2SiO5 at T= 1.95 K...................... 141
5.3 Site selective fluorescence spectra of 0.001 % Er3+:Y2SiO5 at
T = 10 K; linecenters are given in
wavenumbers........................................ 143
5.4 Crystal field levels of 4Ii5z2 and 4Ii3z2 multiplets of
Er3+=Y2SiO5 as determined from absorption and site
selective fluorescence excitation for site I and site 2.......................144
5.5 Experimental setup to measure the 4Ii3z2 fluorescence
lifetime for E r3+=Y2SiO5............................................ ................ . 147
5.6 Fluorescence lifetime decay for 0.001% Er3+=Y2SiO5
4Zi3z2(Fi) —> 4Z1572 (Z1) transition at T = 10 K .................... ............149
5.7 Transition labelling scheme in Zeeman laser absorption
experiments................ ..................................................................... 153
5.8 Projections of the magnetic field B onto the primary g-
.tensor axes for the case of B lying in the plane ....................... 155
xiii
5.9 Experimental setup for laser Zeeman absorption; ECDL 2
serves to calibrate the optical frequency...........................................157
5.10 Experimental configuration for full rotational Zeeman
measurements illustrated for theDi-D2 plane..................... ......... ...158
5.11 Laser absorption Zeeman spectra for 0.001 % Er3+IY2SiO5 as
a function of magnetic field for BHD1,
k l lb a tT= IOK.......................................................... ................... 161
5.12 Angle dependent Zeeman laser absorption scans, with
the B-field in 6-D2 plane.................................................................. 164
5.13 (a) Orientational-dependent g values of magnetically
inequivalent orientations of site 2 in the 6-Z)2 plane
determined from data of (b)........................................ .....................166
5.14 (a) Orientational-dependent g values of ■ magnetically
inequivalent orientations of site I in the 6-D2 plane
determined from data of (b)..................................... .......... ............. 167
5.15 (a) Orientational-dependent g values of magnetically
inequivalent orientations of site I in the B-D1 plane -
determined from data of (b).................. ........ ................................ 168
5.16 (a) Orientational-dependent g values of magnetically
inequivalent orientations of site 2 in the b-Di plane
determined from data of (b)............................... .............. ...............169
5.17 (a) Full orientational dependent g values of site I in the
D2 D2 plane determined from data of (b)............... ............. .170
5.18 (a) Orientational-dependent g values of site 2 in the D2-
D2 plane determined from data of (b)..... ...... .......... ............... .......171
5.19 Schematic of the spin flip broadening mechanism in
Er3+: Y2SiO5............................ ........................................................174
5.20 Schematic of the direct phonon process in the Zeeman
split Er3+ ground state......... ...... ........... ........................................177
5.21 Experimental setup to measure two pulse photon echoes,
stimulated echoes and optical nutation with a spectral
hole stabilized laser......... ........................................ .............. ....... .181
xiv
5.22 (a) Orientation-dependent g values of site I and site 2 in the
Di-D2 plane, glg denotes the g value for the ground state of
site I, g2e the g value for the excited state of site 2, etc.; solid,
lines are fits to the data, (b) Correlation with the
homogeneous linewidth of site I measured with two pulse
echoes in the D2-D2 plane as a function of magnetic
field orientation............... .................................................. .........;. 185
5.23 (a) Orientational dependent g values of site I and 2 sub-sites
in the B-D2 plane, gig denotes the g value for the ground state -
of orientation I, g2e the g value for the excited state of
orientation 2, etc.; solid lines are fits to the data.
(b) Correlation with the homogeneous linewidth of site I
orientation 2 measured with two-pulse photon echoes in the
B-D2 plane as a function of magnetic field
orientation................................ ....................... ........... ......... ....... ..186
5.24 Energy level structure of Er3+IY2SiO5 at a magnetic field
of B = 3 T with angle
, B I/ Di. Solid lines are fits to
expression (2.5) to extract the homogeneous linewidth; each
case shows excellent agreement, (b) Homogeneous
line widths as a function of temperature obtained from (a), the
solid line serves to guide the eye.....................................................215
5.34 Evolution of the linewidth of site I in 0.005% Er3+=Y2SiO5 at
T = 4.2 K and B = 3 T as the waiting time, T, between pulse
two and three is varied in a stimulated
photon echo measurement.............................. 216
AT Conceptual block diagram of the laser frequency
stabilization system....................................................................... ...230
A.2 Block diagram of the diode laser frequency stabilization
Apparatus............................. ......233
xvi
A.3 Electronic schematic of the diode laser injection current
servo (I MHz bandwidth).......................................... ................ ,..,235
A.4 Electronic schematics of the PZT servo (low bandwidth)............... 238
A.5 Electronic schematic of the bridged-T notch filter....... ............... ..240
A.6 Electronic schematic of the resonant EOM-tank circuit..... ........... .242
A.7 Electronic schematic of the post-mixer amplifer............................. 244
A. 8 Electronic schematic of the RF phaseshifter (first
part)......................, .......... . .........................................:......... 246
A.9 Electronic schematic of the RF phase shifter (second part).......... ... 247
A. 10 Electronic schematic of the low noise diode laser
driver; arrows indicate the signal flow............................................. 249
B. l Electronic schematic for cavity-ringdown measurements.,..........................................................................255
B. 2 (a) Cavity-ringdown (lifetime) measurement, straight line is a
least square fit exponential fit to the data, (b) Transmission
spectrum of the TEMqo mode matched
cavity showing one full free spectral range (FSR).......................... ,257
C. l Two-level system with energy splitting, = and
populations in the upper (lower) state, /?++(/?__)
employed in the Bai-Fayer theory................................................... 259
xvii
ABSTRACT
Narrow spectral holes in the absorption lines of Er3"1" doped crystals have been
explored as references for frequency stabilizing external cavity diode lasers at the
important 1.5 pm optical communication wavelength. Allan deviations of the beat
signal between two independent stabilized lasers as low as 200 Hz over IOms
integration time have been achieved using regenerative spectral holes in Er3^ Y2SiO5
and Er3+IKTP, while drift was reduced to ~ 7 kHz/min by incorporating the
inhomogeneous absorption line as a fixed reference. During active stabilization, the
transient spectral hole was continuously regenerated as hole burning balanced
relaxation. In contrast, persistent spectral holes in Er3+:D":CaF2, with lifetimes of
several weeks, provided programmable and transportable secondary frequency
references that maintained sub-kilohertz stability over several seconds and enabled
6 kHz stability over LbxlO3S. The error signal was derived from the spectral hole
transmission using frequency modulation spectroscopy. A servo amplifier applied
fast frequency corrections to the injection current of the laser diode and slower
adjustments to the piezo-driven feedback prism plate.
These stabilized lasers provide ideal sources for spectral hole burning
applications based on optical coherent transients, where laser stability is required
over the storage time of the material. Since the lifetime of the frequency reference is
exactly the material storage time, this requirement is automatically met by using our
technique. This was demonstrated in Er3+IY2SiO5 and successfully transferred to
high-bandwidth signal processing applications.
The material Er3+IY2SiO5 was optimized for these applications. The 4Ii5z2 and
4Ii3z2 crystal field levels were site-selectively determined by absorption and
fluorescence spectroscopy. The excited state lifetime was measured to be 1L4 ms for
site. I and 9.2 ms for site 2. Zeeman experiments and two-pulse photon echo
spectroscopy as a function of magnetic field orientation were used to determine the
anisotropic electronic g-values for both Er3+ sites and established a preferred
magnetic field orientation for minimizing homogeneous line broadening by spectral
diffusion. The spectral diffusion was characterized by stimulated photon echo
spectroscopy and successfully described with established theories. In a 0.02 atomic
percent Er3+IY2SiO5 crystal at B = 0.8 T and T = 1.6 K, line broadening became
significant after 10 p,s, increasing the homogeneous linewidth from 7.5 kHz to
75 kHz after 120 jas. Spectral diffusion, primarily caused by direct phonon driven
Er3+ spin-flips in the ground state, can be controlled to negligible levels with proper
magnetic field strength and orientation, temperature, and erbium concentration. In
optimizing Er3+=Y2SiO5, the narrowest optical resonance in any solid-state material .
of 73 Hz was measured.
ICHAPTER I
INTRODUCTION
Many classes of solids have inhomogeneously broadened optical absorption
lines. When a narrow spectral region in such a material is saturated by a laser,
notches, called spectral holes, are “burned” into the line shape, either transiently or
persistently modifying the optical properties of the medium. At low temperatures,
these holes have useful lifetimes ranging from fractions of a second to weeks or
longer, and these narrow features have been widely used for spectroscopy at
Montana State University, IBM, and elsewhere. The transient or persistent
modification of optical properties that arises from spectral hole burning (SHB) also
offers opportunities to build powerful and interesting devices for signal processing or
data storage. Many of these devices rely on a combination of spectral hole burning
and extensions of the concepts of holography to the time domain, leading to
holography in “four dimensions”.
In this research, the ultra-narrow resonances provided by spectral holes have
been exploited as references for frequency stabilization of external cavity diode
lasers (ECDL). In particular, Er3+ doped materials that exhibit SHB in the important
optical communication band at 1.5 pm were investigated. Relatively short-lived
(transient) spectral holes with lifetimes of Ti ~ 10 ms and kilohertz hole linewidths
supplied frequency references that provided stabilities over timescales that are
2 ''i
especially appropriate for coherent nonlinear spectroscopy and SHB device
applications. [1,2] Continuously regenerated, lifetime-limited (regenerative) spectral
hole frequency references in Er3+-Y2SiO5 produced laser frequency stabilities (Allan
Deviation) of 500 Hz over 2 ms. [3] Further system development allowed
improvement to 200 Hz over 5 ms using regenerative spectral hole frequency
references in Er3+=KTP. [4, 5] Persistent spectral holes in Er3+=CaF2, with lifetimes of
several weeks, permitted extension of laser frequency stability to longer integration
times, achieving stabilities of 6 kHz over 1600 s while maintaining sub-kilohertz
stabilities over integration times of several seconds. [6]
For SHB applications using materials optimized for correlators or memories,
stable laser sources are required at wavelengths and timescales specific to each
individual material. We have demonstrated that using a spectral hole in a separate
spatial region or a separate piece of the SHB correlator or memory material
automatically provides frequency references meeting these requirements. Together
with previously developed 793nm laser systems [7, 8], these 1.5 pm lasers stabilized
to spectral holes have been successfully transferred to SHB correlator devices at the
Montana State University Spectrum Lab, Cone lab, and Babbitt lab. [9, 10]
The material Er3+=Y2SiO5 plays an important role in the arena of SHB devices
covering the telecommunication band at 1.5 pm. [11] It has been used for SHB
praof-of-principle demonstrations at Montana State University such as real-time
address header decoding for optical data routing [12] and spatial-spectral
holographic correlation. [13] Very recently, the improved material and laser
3frequency stabilization that is reported in this thesis allowed demonstration of much
higher bandwidth (500 MHz) analog signal processing at temperatures of 4.2 K .
[10] Other research groups around the world, including the University of Colorado,
the Laboratoire Aime Cotton, Orsay, France, and the Australian National University,
'
are already using this material to develop SHB applications that will include
massively parallel computing, radio frequency spectrum analysis [14], electro-
magnetically induced transparency, and quantum information demonstrations [15]
based on our development.
The importance of Er3+IYiSiOs for SHB applications provided the motivation for
fundamental research to further explore the parameters that influence materials
critical to these applications. Spectroscopic investigations using conventional and
coherent nonlinear methods allowed the characterization and optimization of the
spectral hole burning in the Er3+IY2SiO5 material [16] and have led to the
measurement of a 73 Hz linewidth, the narrowest optical resonance, to the best of
our knowledge, in any solid-state material. Furthermore, the methods developed
during the material optimization are directly applicable to other Er3"1" doped
compounds. These studies have advanced our fundamental understanding of these
materials.
4Frequency Stabilized Lasers and Their Applications
Many applications require laser frequency stability that is impossible to achieve
with a free-running laser, Fortunately, dramatic stabilization is possible with an
external frequency reference and feedback control of the laser cavity. Such lasers
find widespread and important uses. Laser frequency stabilization is important in
long-baseline interferometry for gravitational wave detection [17], ultrahigh-
resolution spectroscopy of solids, molecules, and atoms [18], optical communication
systems [19], for solid state optoelectronic devices based on spectral hole burning
(SHB) [20] such as GHz-scale time-domain optical signal processing [21, 22, 13, 9]
and network packet switching [23, 12], for precision laser ranging, for spatial
coordination of satellite arrays, and for optical communication using coherent light
detection. These stabilized lasers are also suitable for sensitive vibration monitoring
devices and a variety of other optical and fiber optical sensors.
One of the most prominent applications for which stable laser sources are being
developed is optical frequency standards based on trapped ions [24, 25,, 26, 27, 28]
and neutral atoms [29, 30], particles which are nearly at rest in the laboratory frame,
eliminating Doppler effects while producing very narrow, several-Hz-wide
resonance lines. Recent developments suggest that an optical frequency clock can
offer significant improvement over the resolution of the current Cesium microwave
T-
atomic clock. In the tera-hertz domain, optical frequencies provide four orders of
magnitude higher operating frequencies than microwave references, while the higher
5quality factors of optical reference transitions allow determination of center
frequencies to greater precision. The recent advent of air-silica microstructure optical
fiber, which broadens the frequency comb of a femtosecond laser to span the optical
octave from 532 nm to 1064 nm, furthers the trend toward an all-optical frequency
standard. For the first time, optical frequencies in the IO15 Hz range can be directly
measured and compared to microwave standards [31, 32, 33, 34] without the use of
elaborate frequency chains. These new techniques Will likely lead to new, super-
accurate clocks based on optical frequencies, with projected performance
approaching IO"18 accuracy. In addition, a stable frequency in the radio frequency
(RF) domain can now be derived with improved accuracy using an optical standard.
Recent reviews of the field can be found in Ref. [35, 36]. The basic ideas how to
build an optical frequency standard apply to any laser source to be stabilized.
In order to achieve an optical frequency standard, the laser system must include
the optical frequency reference, must maintain a single spatial and frequency mode,
and have an actuator that allows frequency tuning with sufficient bandwidth to
suppress the intrinsic laser noise so that the laser’s short-term linewidth can be
narrowed to allow interrogation of the optical reference transition. The lasers’ long
term stability derives from the optical frequency reference.
Pre-stabilizing a laser to a reflection mode Of a Fabry-Perot cavity often achieves
this goal. These laser clocks, frequently called flywheel oscillators, provide a cavity-
derived reference with sufficient short-term stability to interrogate the atomic or ion
optical frequency standard of interest. While our research goal did not encompass the
6development of absolute frequency standards, our lasers stabilized to spectral hole
frequency references have already been considered to become flywheel oscillators of
an optical frequency standard [37].
During the 19.80’s, a number of technical developments made Fabry-Perot
cavities an attractive choice to achieve an ultra-stable laser. The Pound-Drever-Hall
technique [38], a detection scheme capable of exploiting the narrow linewidth of a
high finesse cavity, proved to be the most effective method of locking a laser to a
cavity. The lowest relative instability for a laser that is frequency locked to a cavity
reached SxKT17, measured by locking two lasers to adjacent modes of a single
Fabry-Perot. [39] The main problem associated with Fabry-Perot cavities is
vibration-induced and thermally-induced length changes. The thermal instability was
reduced with introduction of Ultra Low Expansion (ULE) glass [40] or Zerddur [41]
as cavity spacer materials, providing nearly zero thermal expansion, with expansion
coefficient of a ~ 4xl0"8 / K at room temperature. Even more extreme measures
have been undertaken in the work of Mlynek and Schiller et al. [42] by operating
optical reference cavities at cryogenic temperatures to minimize thermal expansion,
which is described by a ^ T3 for T —>0, giving a value of a -SxlQ r13T31K in
sapphire. Next, the development of very high finesse dielectric mirrors led to
extremely narrow cavity linewidths, providing a very steep discriminator slope for a
tight laser servo lock.
It was also appreciated that achieving a tight servo lock is necessary but not
sufficient for an ultra-stable laser. The stability of the laser frequency can never
7exceed that of the frequency reference. To assess real stability requires a second,
independent reference cavity stabilized laser. The best absolute Stability for cavity-
stabilized lasers has been 3xlCT16 [43], which is half an order of magnitude less than
the SxlO-17 value mentioned previously.
Reduction of perturbations to the reference cavity, mainly due to temperature
fluctuations and vibrations exciting mechanical modes or causing deformations of
the cavity, has been the focus of much of the work in the past 15 years, Applying
these considerations to practical systems should take into account cost, size,
portability, and degree of complexity.
Spectral Hole Burning and Optical Coherent Transients'
Spectral holes provide alternative references that are complimentary to precision
atomic resonances or reflection modes of Fabry-Perot cavities. The relative
immunity of the spectral hole reference to vibrational disturbances greatly simplifies
the experimental setup. The entire system developed here fits on a 3’ by
4’breadboard including optics, two sets of lasers, and feedback electronics, cryostat,
and beat measurement for characterization. Further system development could lead
to compact, transportable, stable laser sources based on SHB technology,
miniaturized external cavity diode lasers, and mechanical closed-cycle cryo-coolers.
Other state-of-the-art stable laser systems based on Fabry-Perot cavities and atomic
resonances occupy several optical tables and require extreme isolation from the
8environment to stabilize the local oscillator. [43] Note also that the best performance
with traditional Fabry-Perot cavities, atomic or ion resonances requires cryogenic
temperatures. [24, 42]
Spectral hole burning is a property found in certain materials, such as inorganic
and organic solids, and Doppler broadened atomic vapors. Of particular interest for
this research are rare earth doped solids, which exhibit an inhomogeneously
broadened absorption at low temperatures. In particular, an individual homogeneous
packet of rare earth ions in a crystalline or glassy solid manifests an absorption
linewidth, known as the homogeneous linewidth, Fh. At optical frequencies and
cryogenic temperatures, this width can be as narrow as 73 Hz, as reported in this
work. These rare earth ion subgroups comprise part of the broader distribution of the
strain-broadened absorption line described by the inhomogeneous linewidth, Finh.
Inhomogeneous linewidths in rare earth doped crystals range from sub-GHz values
up to hundreds of GHz.
When ions are temporarily or permanently removed from the inhomogeneously
broadened absorption, a spectral hole is produced. This selective bleaching of a
particular subgroup of ions may occur when the ions are exposed to a narrow band
laser. The process of producing a narrow spectral hole in the absorption line is called
spectral hole burning. Spectral holes may be as narrow as 2 Th, which can approach
100 Hz, so they provide resonances whose widths are competitive with those of
super-cavities and isolated single atoms Or ions.
9The large difference between the magnitudes of inhomogeneous and
homogeneous linewidths can be exploited in optical memories. In the frequency
domain, the inhomogeneous line is subdivided into “frequency bins”, with each bin
having a potential minimum frequency width equal to the homogeneous linewidth.
Binary information can be spectrally addressed while stored in the combination of
either the presence (binary I) or absence (binary 0) of a spectral hole. The.
approximate number of spectral holes burned in a single spatial location depends on
the ratio of inhomogeneous to homogeneous linewidths, which has been measured to
be as high as IO8 in some materials, [44]
Spectral hole burning in the time domain leads to coherent phenomena, called
optical coherent transients. The same rare earth materials can be used to store
temporally structured, optical pulse patterns, with duration limited by the material
coherence time, T .^ Stored pulse patterns can be recalled temporally, leading to
optical memory [45] with a storage duration determined by the material’s hole
burning mechanisms. Storage times can vary between milliseconds, in cases of two-
level saturation or population bottlenecks formed by a metastable intermediate state,
to several weeks, for optical pumping of hyperfine-split sublevels of the ground state
or optically induced local ion site distortion. Stored pulse patterns can also be used
for correlation with an incident optical data stream, which leads to optical
processing. [46] Most applications of coherent transients use photon echoes, treated
in detail in chapter 2. A review of memory, processing, and routing applications can
be found in Ref. 47.
10
Stable laser technology accommodates frequency and time domain spectral hole
burning applications as well as spectroscopy. Frequency domain applications reach
their limit at the laser linewidth, often significantly larger than the associated
homogeneous linewidth of the particular material. In memory applications, this
causes considerable reduction of the possible number of holes burned into the
inhomogeneous line, thus limiting the achievable storage capacity. For optimal
exploitation of the photon echo and stimulated photon echo, which are the basis for
time-domain spectroscopy and the associated range of proposed optical devices [47],
laser frequency stability must exceed the spectral resolution needed to store the
spectrum of the excitation pulse sequences. The entire pulse sequence length is
limited by the storage duration for the material.
Lasers stabilized to spectral hole frequency references offer distinct advantages
in frequency as well as time-domain spectral hole burning applications and
Spectroscopy. In stabilizing the laser to a spectral hole in a second.piece of the same
signal processing material, automatic frequency compatibility between the signal
processing material and the stabilized laser source is provided. There is a natural
correspondence between the timescales governing optical processing and laser
frequency stabilization. The relative vibrational immunity of the spectral holes
provides an important simplification in system design and performance, especially
when both the frequency reference and spectroscopic sample or SHB device are
mounted on the same platform.
11
Overview of the Dissertation
After this chapter of introduction, Chapter 2 discusses general rare earth material
properties, spectral hole burning, and optical coherent transient phenomena.
Chapter 3 introduces the basic concepts of laser frequency stabilization. Special
emphasis is placed on frequency modulation spectroscopy as well as methods of
measurement and characterization of laser frequency stability. The external cavity
diode laser system, built for all the research conducted, is described and
characterized in this chapter.
Chapter 4 reports our results on laser frequency stabilization to regenerative
spectral holes in Er3^ YaSiOs and Er3+:KTP, as well as: persistent spectral holes in
Er3+: CaF2; The reliability of stimulated photon echoes produced by a laser stabilized
to a spectral hole, spatially separated in the same crystal, is demonstrated.
' Regenerative spectral holes, because of their limited lifetime, differ fundamentally
from traditional frequency references. They are dynamic references. The interplay of
the laser field with the dynamic hole reference was experimentally investigated.
Chapter 5 presents a study of the spectroscopic and dynamic properties of
Er3+--Y2SiO5 and explores optimization strategies for better performance in SHB
applications. Use of broadband' absorption and site-selective fluorescence
spectroscopy allowed mapping the relevant energy levels for operation at 1.5pm as .
well as measuring the fluorescence lifetime of the excited state. Paramagnetic g
12
values of the ground and excited states were characterized as a function of magnetic
field orientation with Zeeman spectroscopy. The g value, is one of the key variables
in controlling the Er3+ spin dynamics of the material,, which leads to a Iinewidth
broadening referred to as spectral diffusion. Stimulated photon echo spectroscopy
was used to characterize spectral diffusion as a function of magnetic field,
temperature, and erbium ion concentration. Experimental results were successfully
described by established theories and advanced our fundamental understanding of
these materials. [48] Conventional and nonlinear spectroscopic methods, when
utilized together, enabled material optimization for SHB applications and the
measurement of the narrowest optical resonance in any solid-state material.
Additional information is summarized in the appendices. A complete description
of the electronic feedback system is included in appendix A. Appendix B describes
the reference cavity. Appendix C presents a detailed derivation of Bai-Fayer theory
relevant to describe spectral diffusion in Er3+IY2SiOs.
13
REFERENCES
I. T. Bdttger, G. I. Pryde, N. M. Strickland, P. B. Sellin, and R. L. Cone, Optics-&-
Photonics-News 12, no.12; Dec. 2001; p.23.
2. R. L. Cone, T. Bdttger, G. I. Pryde, N. M. Strickland, Y. Sun, P. B. Sellin, and R.
L. Carlsten, Semiconductor lasers stabilized to spectral holes in rare earth
Crystals, Proceeding of SPIE 4283, 33:5 (2001). '
3. P. B. Sellin, N. M. Strickland, T. Bdttger, J. L. Carlsten, and R. L. Cone, Phys.
Rev. B 63 155111 (2001).
4. G. J. Pryde, T. Bdttger, R. L. Cone, and R. C. C. Ward, Semiconductor Lasers
Stabilized to Spectral Holes in Rare Earth Crystals to a Part in IO13and their
Application to Devices and Spectroscopy, J. oh Luminescence, to appear (2002).
5. G. J. Pryde, T. Bdttger, and R. L Cone, J. Lumin. 94-95,76 (2001)
6. T. Bdttger, G. I. Pryde, R. L. Cone, and G. D. Jones, Programmable sub-kHz laser
frequency stabilization at 1523 nm using persistent spectral hole burning,
submitted to Opt. Lett. 2002.
7. P. B. Sellin, N. M. Strickland, J. L. Carlsten, and R. L. Cone, Opt. Lett. 24, 1038
(1999).
8. N. M. Strickland, P. B; Sellin, Y. Sun, J. L. Carlsten, and R. L. Cone, Phys. Rev.
B 62,1473 (2000).
9. K. D. Merkel, R. W. Peters, P. B. Sellin, K. S. Repasky, and W. R. Babbitt, Opt.
Lett. 25, 1627 (2000).
14
10. Z. Cole, T. Bottger, R. Krishna Mohan, R. Reibel, W. R. Babbitt, R. L. Cone and
K. D. Merkel to be published. Accepted for presentation at Conference of Lasers
and Electro Optics (CLEO), Long Beach, CA, May 2002.
11. R. M. Macfarlane, T. L. Harris, Y. Sun, R. L. Cone, and R. W. Equall, Opt. Lett.
22, 871 (1997).
12. T. L. Harris, Y. Sun, R. L. Cone, R. M. Macfarlane, and R. W. Equall, Opt. Lett,
23, 636 (1998).
13. T. L. Harris, Y. Sun, W. R. Babbitt, R. L. Cone, J. A. Ritcey, and R. W. Equall,
Opt. Lett. 25, 85 (2000)
14. L. Menager, I. Lorgere, I. L. Le-Gouet, D. Dolfi, and J. P. Huignard, Opt. Lett.
26, 1245 (2001).
15. N. Ohlsson, R. Krishna Mohan, and Stefan Kroll, Opt. Comm, 201, 71 (2002).
16. T. Bottger, Y. Sun, G.l. Pryde, G. Reinemer, and R.L. Cone J. Lumin. 94-95, 55
(2001).
17. A. AbramoVici, W. E. Althouse, R. W. P. Drever, Y. Giirsel, S. Kawamura, F. J.
Raub, D. Shoemaker, L. Sievers, R. E. Spero, K. S. Thome, R, E. Vogt, R.
Weiss, S. E. Whitcomb, and M. E..Zucker, Science 256, 325 (1992).
!
18. R. G. DeVoe and R. G. Brewer, Phys, Rev. Lett. 50, 1269 (1983); R. M.
Macfarlane and R. M. Shelby in Spectroscopy of Crystals Containing Rare-Earth
Ions, edited by A. A. Kaplyanskii and R. M. Macfarlane (North-Holland,
Amsterdam, 1987); M. J, Sellars, R. S. Meltzer, P. T. H. Fisk, and N. BvManson,
J. Opt. Soc. Am. B 11, 1468 (1994).
15
19. S. L. Gilbert, W. C. Swann, and T. Dennis in Laser Frequency Stabilization,
Standards, Measurement and Applications, Proc. SPIE 4269, 184 (2001) and
references therein.
20. T. W. Mossberg, Opt. Lett. 7, 77 (1982); A. Rebane, R. Kaarli, P. Saari, A.
Anijalg, and K. Timpmann, Opt. Commun. 47, 173 (1983); M. Mitsunaga, R.
Yano, and N. Uesugi, Opt, Lett. 16, 1890 (1991); M. Mitsunaga, N. Uesugi, H.
Sasaki, and K. Karaki, Opt. Lett. 19, 752 (1994); H. Lin, T.. Wang, and T. W.
Mossberg, Opt. Lett. 20, 1658 (1995).
21. N. W. Carlson, L. J. Rothberg, A. G. Yodh, W. R. Babbitt, and T. W. Mossberg,
Opt. Lett. 8, 483 (1983); M. Zhu, W. R. Babbitt, and C. M. Jefferson, Opt. Lett.
20, 2514 (1995); K. D. Merkel and W. R. Babbitt, Opt. Lett. 24, 172 (1999).
22. T. R. Dyke, M. J. Sellars, G. J, Pryde, N. B. Manson, U. Elman, and S. Kroll, J.
Opt. Sci. Am. B 16, 805 (1999); G. J. Pryde, M. J. Sellars, and N. B. Manson,
Phys. Rev. Lett. 84, 1152 (2000); M. J. Sellars, G. J. Pryde, N. B. Manson, and
E. R. Krausz, J. of Luminescence 87, 833 (2000).
[23]. X. A. Shen and Ri Kachru, Opt. Lett. 20, 2508 (1995); T. Wang, H. Lin, and
T. W. Mossberg, Opt. Lett. 20, 2541 (1995).
24. D. J. Berkeland, J. D. Miller, J. C. Bergquist, W. M. Itano, and D. I. Wineland,
Phys. Rev. Lett. 80, 2089 (1998).
25. R. J. Rafac, B. C. Young, J. A. Beall, W. M. Itano, D. J. Wineland, and J. C.
Bergquist, Phys. Rev. Lett. 85, 2462 (2000).
26. M. Roberts, P. Taylor, G. P. BarwOod, P. Gill, H.A. Klein, and W. R. C. Rowley
Phys. Rev. Lett. 78, 1876 (1997).
27. J. E. Bernard, A. A. Made], L. Marmet, B. G. Whitford, K. M. Siemsen, and S.
Cundy, Phys. Rev. Lett. 82, 3228 (1999).
16
28. E. Peik, G. Hollemann, and H.Walther, Phys. Rev. A 49, 402 (1994).
29. H. Schnatz, B. Lippart, J. Helmcke, F. Riehle, and G. Zinner, Phys. Rev. Lett.
76, 18 (1996).
30. C. W. Oates, E. A. Curtis, and L. Hollberg, Opt. Lett. 25, 1603 (2000).
31. S. A. Diddams, D. J. Jones, J. Ye, S. T Cundiff, J. Hall, L R. Ranka, R. S.
Windeler, R. Holzwarth, T. Udem, and T. W. Hansch, Phys. Rev. Lett. 84, 5102
(2000).
32. D. J. Jones, S. A. Diddams, J. K. Ranka, A. Stehtz, R. S. Windeler, J. L. Hall,
and S. T. Cundiff, Science 288, 635 (2000).
33. S. A. Diddams, D. J. Jones, L.-S. Ma, S. T. Cundiffj and J. Hall, Opt. Lett. 25,
186 (2000).
34, J. Reichert, M. Niering, R. Holzwarth, M. Weitz, T, Udem, and T. W Hansch,
Phys. Rev. Lett. 84, 3232 (2000).
35. J. L. Hall and Jun Ye, Optics & Photonics News 44, February 2001,
36. J. L. Hall, IEEE Journal on Selected Topics in Quantum Electronics 6, 1136
(2000).
37. Jun Ye, private communication.
38. R.W. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A, J. Munley,
and H. Ward, Appl, Phys, B 31, 97 (1983).
39. D; Hils and J. L. Hall in Proceedings o f the 4th Symposium of Frequency
Standards and Metrology, Ed. A De Marchi, Springer Verlag Berlin 1989;
17
C. Salomon, D. Hils, and J. L. Hall, J. Opt. Soc. Am. B 5, 1576 (1988).
40. ULE, ultra low expansion glass, trademark of Coming Glass Works, Coming,
NY 14831.
41. Zerodur is a trademark of Schott Glaswerk, Mainz, Germany.
42. S. Seel, R. Storz, G. Ruoso, J. Mlynek, and S. Schiller, Rhys. Rev. Lett. 78, 4741
(1997).
43. B.C.Young, F.C.Cruz, W.M.Itano, and J.C. Bergquist, Phys, Rev. Lett. 82, 3799
(1999).
44. R. W. Equall, Y. Sun, R. L. Cone, and R. M. Macfarlane, Phys. Rev. Lett. 72,
2179 (1994).
45. T. W. Mossberg, Opt. Lett. 7, 77 (1982).
46. Y. S. Bai, W. R. Babbitt, N. W. Carlson, and T. W. Mossberg, Appl Phys. Lett.
45, 714 (1984).
47. W. R. Babbitt, Advanced Optical Memories and Interfaces to Computer Storage
Proc. SPIE 3468, 304 (1998).
48. Y. S. Bai and M. D. Payer, Phys. Rev. B 39, 11066 (1989).
18
CHAPTER 2
MATERIALS-BACKGROUND AND THEORY
The rare earth ions, or lanthanides, form a special group of elements in the
periodic table. Triply-ionized rare earth ions have a partially filled 4f shell, shielded
from the environment by the outer lying filled 5s2 and 5p6 electron shells. These
partially filled shells give rise to narrow spectral lines due to inner shell 4f-4f
transitions that span the spectrum from the far infrared to the vacuum ultraviolet. [I J
Even when doped into a crystal host material, the shielding of the 4f levels is so
efficient, that the “crystal field” acts only as a Weak perturbation to the free ion
levels. The levels obtained from observed spectra closely resemble those of free
ions.
Here the discussion is restricted to the Er3+ doped materials that were
investigated in this research. The Er3+ energy levels arise from the odd-numbered
4fn electron configuration. For the free ion, the angular momenta / and Mj are good
quantum numbers, and the energy levels are 2/+1 fold degenerate and give rise to
multiplets that are labeled by J. When the ion is incorporated into the crystal, the
reduced crystal field symmetry causes partial lifting of the 27+1 fold Mj degeneracy.
For ions with an odd number of electrons, Kramers theorem [2] dictates that all
levels have electronic degeneracy that can only be lifted by a magnetic field.
Because of Kramers degeneracy, the crystal field can lift the degeneracy only to a
maximum of !+1A crystal field levels depending on the ions’ site symmetry. As a
result, the 4I15y2 ground multiple! can be Split into 8 Kramers doublets, the 4Ii3y2
multiplet into 7. For the laser spectroscopy experiments, a simple two level system
model is used, where the lowest crystal field-split levels of the 4I15y2 and 4I13y2 serve
respectively as the ground and excited state. Applying an external magnetic field lifts
the remaining Kramers degeneracy. This electronic degeneracy and the associated
electronic magnetic moments lead to strong Zeeman interactions.
Transitions between the 4I15y2 —> 4I13y2 multiplets in Er3+ doped compounds are in
the 1.5 pm spectral region, where optical fiber transmission losses are a minimum,
This situation makes Er3+ materials attractive for technological applications and
explains why so much effort in recent years went into developing Erbium materials
for all-optical signal processing applications based on spectral hole burning. [3,4, 5],
Conversely, the well developed infrastructure for 1.5 pm telecommunication devices
including diode Iasersj with Er-doped fiber amplifiers, high bandwidth detectors and
modulators, fiber beamsplitters etc., can be exploited for spectroscopy and
demonstrations.of SHB devices. This interest continued in this work by exploiting,
optimizing and demonstrating the potential of Er3+Y2SiO5 for applications in laser
frequency stabilization and high bandwidth, all-optical correlation.
19
20
Homogeneous and Inhomogeneous Broadening
In rare earth doped materials, two major broadening mechanisms influence the
observed optical spectra - homogeneous broadening and inhomogeneous
broadening. Homogeneous broadening is experienced equally by individual ions in
the crystal and is governed by dynamical processes acting as perturbations on the
ion’s transition frequency or phase. For Er3+ doped compounds, the homogeneous
linewidth can be expressed as the sum of several contributions
^hom ^ p o p Er^Er ^ Phonon ^E r-H o s l • (2 .1)
The contribution Tpop corresponds to the fundamental linewidth associated with the
excited state population lifetime, 7/; Tpop relates to the population lifetime, T1, as
Tpop ^TfTl (2.2),
The T1 lifetimes can be extremely long for these rare .earth ion levels and have been
measured in excess of 10 ms for the lowest 4Ii372 level. Other erbium ions undergo
phonon-induced electronic spin flip transitions in their ground state and therefore
modulate the energy levels of the optical transition causing the contribution, Tsr-Er-
The TEr-Er contributions have been measured, modeled, and optimized in this work
by choosing a proper Er3+-ion concentration, operating temperature, and magnetic
field strength and direction. The TPhonon contribution includes dephasing from
temperature-dependent phonon scattering. Working at cryogenic temperatures
minimizes higher order phonon contributions. Only direct phonon processes are
21
important. Nuclear and electronic spins of the host lattice contribute TEr-Host- Using
a host such as YzSiO ^ that exhibits low or zero nuclear magnetic moment or low
isotopic abundance of magnetic nuclei, can control these contributions. Changes in
the local environment due to the optical excitation of neighboring ions, an effect
termed instantaneous spectral diffusion (ISD), contributes TISd and can be
minimized by using low optical excitation densities or low ion concentrations.
Under optimal experimental conditions, the homogeneous linewidth can be ultra
narrow (< I kHz). The values measured in this work approach the fundamental
lifetime limit. In fact, this thesis reports the narrowest optical transition, 73 Hz,
observed in any solid-state material.
Due to local strains in the crystal caused by crystal growth, impurities, or lattice
imperfections and dislocations, each individual optical center experiences a very
slightly different local environment in the host crystal. These strains and
imperfections cause the center of the homogeneous linewidth of individual optical
centers to subtly shift in frequency space, leading to a distribution of transition
frequencies. The combination of many homogeneously broadened lines, each with a
Lorentzian absorption profile centered at its own resonant frequency, results in a
much broader, often Gaussian, frequency distribution with a width called the
inhomogeneous linewidth, Tm -
Depending on rare earth dopant concentration and crystal composition, the
inhomogeneous linewidth can be as much as IO8 times broader than the individual
homogeneous linewidth in some material systems [3-8,11], and may be even
broader in rare earth doped glasses. As the crystal composition is changed, the
absorption frequency may be “tuned” over a considerable range. Figure 2.1 shows
inhomogeneous and homogeneous broadening; each homogeneous line or “packet”
represents a subgroup of .ions experiencing the same local environment in the
crystal. The erbium materials investigated in this work have inhomogeneous
linewidths between 150 MHz and 10 GHz.
Spectral
hole
frequency
Figure 2.1 Line broadening and spectral hole burning. Each homogeneous line
corresponds to a subgroup of ions experiencing the same local strain
environment in the crystal. The envelope over all homogeneous linewidths
defines the inhomogeneous line. A narrowband laser is used to selectively excite
a subgroup of ions underneath the inhomogeneous profile from the ground to the
excited state at the laser frequency ^ ser. A spectral hole is created at the laser
frequency ^ ser and can be seen as a reduction in absorption.
Spectral Hole Burning
Homogeneous and inhomogeneous broadening leads to a phenomenon called
spectral hole burning (SHB). In order to bum a spectral hole into the
inhomogeneously broadened transition, a narrowband laser selectively excites a
subset or packet of ions underneath the inhomogeneous profile, as shown in
Fig. 2.1. Ions resonant with the laser are pumped from the ground state to the
excited state, thereby bleaching the absorption and leaving behind a spectral hole,
which can be seen as a reduction of optical material absorption at the laser
frequency. This process of saturating a particular homogeneous packet in a material
is one mechanism for spectral hole burning. The hole lifetime is determined by the
lifetime of the population reservoir when SHB proceeds by this mechanism. In the
case of Er3+ materials such as Er3+IY2SiOs, Er3"1":KTP or Er3+IY2O3, hole burning
does take place by population storage in the excited state of the two-level system,
with the hole lifetime determined by the lifetime of the excited state, the lowest
level of the 4Ii3z2 multiplet. Since 4Ii3z2 lifetimes for the above materials are
nominally 10 ms, ions relax back into the ground state where the laser may excite
them again. Under continuous laser illumination, this process takes place until a
balance between spontaneous hole relaxation and hole burning occurs. Hole
burning, where the hole lifetime is limited by Ti of the excited state or an
intermediate bottleneck state [6], is called transient spectral hole burning.
23
24
When the hole lifetime surpasses Ti and extends over much larger timescales,
such SHB is called persistent. Longer spectral hole lifetimes have been measured at
1.5 pm in Er3+:D":CaF2, where a different photo-physical hole burning mechanism
changes the local environment of the optical center. This spectral hole burning
mechanism involves photo-induced D ion migration into nearby interstitial sites.
Spectral holes in Er3+:D":CaF2 with a full width at half maximum (FWHM) of
~40 MHz have been measured to be persistent without change for at least forty-
eight hours. [7] Based on temperature-dependent studies of this material, it is
reasonable to expect the persistence to be much greater.
In SHB materials covering the 1.5 pm wavelength region explored to date,
transient holes, such as those found in Er3LY2SiO5 and Er3LKTP, have proven to be
orders of magnitude narrower in frequency than persistent holes, such as those in
Er3LD": CaF2, which is the Only material known to date exhibiting persistent spectral
hole burning at 1.5 pm.
As mentioned previously, the ratio of inhomogeneous linewidth to
homogeneous linewidth can be extremely large. This becomes important, for
example, in the context of frequency domain optical memory. Binary information
can be stored using spectral holes, such as those found in Eu3LY2SiO5 [8] that are
being used in the development of massive optical memories at Montana State
University.
25
Photon Echoes
In 1981 T. Mossberg discussed the possibility of storing data as a temporal
stream in the time domain. [9] This method of data storage is equivalent to burning
the Fourier transform of the data stream into the inhomogeneous line. Coherent
phenomena in the time domain, called optical coherent transients, include photon
echoes of various types, optical nutation, and free induction decay. Optical coherent
transients can overcome the resolution limit imposed by inhomogeneous broadening
[11]. A large body of literature on coherent transients; exists and the reader is
referred to Ref. [10, 11, 12] and to references therein for a detailed theoretical
treatment. A brief summary relevant to the work presented in this thesis will be
given. '
Photon echoes are the optical analogue of spin echoes, long known in nuclear
magnetic resonance (NMR) [13], and Were first observed by Kumit et al [14] and
Abella et al. [15] in ruby. The photon echo technique, uses a sequence of short
pulses to indirectly measure the homogeneous line width without significant spectral
selection by the laser. The spectral width of the pulses may be large compared to
the homogeneous linewidth, thus relaxing the need for an ultra-stable laser source.
The effect of inhomogeneous broadening is removed by the pulse sequence itself.
26
Two Pulse Photon Echo
In a two-pulse photon echo experiment, two laser pulses, separated by a time
delay, excite the sample, as shown in Fig. 2.2. The first pulse creates a coherent
superposition of the ground and excited states. After the first pulse, this coherent
superposition state has a macroscopic oscillating dipole moment, whose re-radiation
is the free induction decay. [16] As time elapses, this dipole moment quickly
dephases as the ions accumulate phase according to their frequency offset from the
laser frequency within the inhomogeneous distribution. The second pulse, at time, r,
acts to exchange the amplitudes of ground and excited state in the coherent
superposition, which leads to a phase reversal for each individual ion; the ions
begin to rephase. After a waiting period, % following the second pulse, the net
phase shift cancels for each ion, leading to a rephasing of the coherence in the form
of a macroscopic oscillating dipole, detected as the photon echo. To optimize the
strength of the photon echo signal, the first pulse should be a 7t/2 pulse to excite the
sample to a coherent superposition state with equal amplitudes, meaning it should
be of area
©= \{finE {t)lh )d t= ^ (2.3)
with the transition dipole moment, fin , and the electric field strength, E, of the
pulse. The second pulse should be a Tt-pulse to exactly interchange amplitudes and
phase factors for the ground and excited states.
i
27
T i - pulse
tz/2 - pulse
photon echo
I
I "V
time
Figure 2.2 A typical two-pulse photon echo sequence. The first excitation pulse is
chosen to be of area tt/2 and the second pulse of area n follows pulse I after a time
delay of v, note the optical absorption of the pulses. The two pulse photon echo occurs
after a time delay % after the 7t-pulse. This figure shows an experimental trace; the
shape and heights of the transmitted excitation pulses are modified by coherent
absorption by the sample.
I 2 3
photon
echo
J L .....
stimulated
photon
echo
<---- -------------------->i
time
--------------
Figure 2.3 A typical stimulated photon echo pulse sequence. The first two excitation
pulses are separated by a time delay ti2 and pulse 2 and 3 are separated by the waiting
time T. The stimulated photon echo occurs after a time delay tiz after pulse 3, a two
pulse echo can be observed after a time delay tn after pulse 2; note that the excitation
pulse area is ~ tt/2 for all 3 pulses causing the stimulated photon echo to be stronger
than the two pulse photon echo.
28
As T is increased, a reduction in intensity of the echo reflects, the decay of
coherence during the time, Ir, as the result of stochastic processes in the crystal.
Measuring the photon echo intensity, as a function of the time delay, ^ between the
two pulses, yields for ideal two-level systems a single exponential decay as
exp(-4r / T2) whose time constant allows determining the dephasing time, T2. The
homogeneous linewidth can be determined from the dephasing time T2 by
r —hom
I
(2,4)
and provides a method to measure very narrow, sub-kilohertz linewidths with a
laser, whose linewidth can be several 100 kHz. In some material systems [8] the
Observed decays depend on the power of the pulses used due to instantaneous
spectral diffusion (ISD). The measured values of the linewidth must then be plotted
as a function of pulse power and extrapolated to zero power to obtain the true value
for T2\ for such cases, tt/2 and 7i-pulses are not optimum. The optical Bloch
equations and the optical Bloch vector model provide a visual picture of the
population dynamics and the coherence of an ensemble of ions in the interaction
with an excitation pulse sequence of the laser field. [10]
In the presence of spectral diffusion, where the homogenous linewidth evolves
on a timescale of the pulse sequence, observed echo decays are non-exponential and
can be described by the Mims [17] expression
/(t) = /0exp:
TV 1M j
(2.5)
29
first introduced in the context of electron spin echoes and later used in the analysis
of photon echoes. [18] The parameter, x, describes the deviation from a pure
exponential, while Tm corresponds to Th, in the specific case of x = I. An effective
homogeneous linewidth can be extracted from the phase memory time, Tm, as
I
h^om - ~ z r ■ (2.6)
Stimulated Photon Echo
Stimulated photon echoes conveniently allow studying spectral diffusion. The
stimulated photon echo requires a three 7t/2-pulse sequence and can be thought of as
a modified two-pulse echo, where the second 7t-pulse is broken up into two r e
pulses, separated by the waiting time, T. The pulse separation between pulses one
and two is ti2. Figure 2.3 shows a typical stimulated photon echo pulse sequence.
As in the two-pulse echo case, the first pulse creates a coherent superposition state.
After a period, Z72, the phases of the Superposition states evolve, according to the
frequency offset from the laser frequency within the excited packets. Rather than
completely reversing the phases with pulse two, the second pulse has an area of rt/2
and thus stores ions that have accumulated less than n in phase in the excited state
and ions that have accumulated more than n in phase in the ground state. As a
result, the total phase information is stored as a population grating between ground
and excited state with a frequency period given by M j2 ■ This grating decays due to
30
population decay with the lifetime of the excited state, T1, and gets smeared Out due
to frequency shifting interactions (i.e. spectral diffusion) during the waiting time T.
The grating can be probed with a short 7t/2-pulse after the waiting time, T, which
causes a rephasing of the stored coherence after a time delay, t12, and the stimulated
photon echo is emitted. The stimulated photon echo amplitude contains information
about dephasing during the two ^-delays and spectral diffusion and population
decay during the waiting time,-T. Systematic measurements of the stimulated echo
decay as a function of the waiting time allow mapping out the time evolution of the
homogeneous linewidth as excited ions undergo, spectral diffusion.. Results for
Er3+IYzSiOs are presented in chapter 5.
31
REFERENCES
E S . Hiifner, Optical Spectra of Transparent Rare Earth Compounds, Academic
Press, London (1978).
2. H, A. Kramers, Proc. Amsterdam Acad. 33, 959 (1930).
3. R. M. Macfarlane, T. L. Harris, Y. Sun, R. L. Cone, and R. W. Equall, Opt. Lett.
' 22, 871 (1997).
4. T. L. Hams, PhD thesis Erbium based Optical Coherent Transient Correlator for
the 1.5 Micron Communication Bands, Montana State University, Bozeman,
April 2001.
5. Y. Sun, C. W. Thiel, R. L. Cone, R. W. Equall, and R. W. Hutcheson, Recent
Progress in Developing New Rare Earth Materials for Hole Burning and
Coherent Transient Applications, to appear in I Lurndn. (2002).
6; R. M. Macfarlane, Opt. Lett. 18, 1958 (1993).
7. N. M. Strickland, R.L, Cone, and R.M. Macfarlane, submitted to Phys. Rev. B.
8. R. W. Equall, Y. Sun, R. L. Cone, and R. M. Macfarlane, Phys. Rev. Lett. 72,
2179 (1994).
9. T. W. Mossberg, Opt. Lett. 7, 77 (1982).
10. L. Allen and J. H. Eberly, Optical Resonance and two level atoms, Dover
Publications Inc,, New York, 1987.
32
11. R. M. Macfarlane and R. M. Shelby, Chapter 3 Coherent Transient and Hole
Burning Spectroscopy of Rare Earth Ions in Solids, in Spectroscopy of Solids
containing Rare Earth Ions, Eds. A.A. Kaplyanskii and R. M. Macfarlane,
Elsevier Science Publishers North Holland, Amsterdam, 1987.
12. M. D. Levenson, Introduction to Nonlinear Laser Spectroscdpyi Academic
Press, New York 1982.
13. E. L. Hahn, Phys. Rev. 80, 580 (1950).
14. N. A. Kumit, I. D. Abella, and S. R. Hartmann, Phys. Rev. Lett., 13, 567
(1964).
15.1. D. Abella, N. A. Kumitt, and S. R. Hartmann, Phys. Rev. 141, 391 (1966).
16. A. Z. Genack, R. M. Macfarlane, and R. G. Brewer, Phys. Rev. Lett. 37, 1078
(1976).
17. W. B. Mims, Phys. Rev. 168,370 (1968).
18. R. M. Macfarlane, R. Wannemacher, D. Boye, Y. P. Wang, and R. S. Meltzer, J.
of Lumin. 48-49, 313 (1991).
33
CHAPTER 3
LASER FREQUENCY STABILIZATION-BACKGROUND AND THEORY
Frequency Modulation Spectroscopy
The laser frequency stabilization technique reported uses ultra-narrow spectral
holes as a laser frequency reference. To detect the center of the spectral hole,
frequency modulation (FM) spectroscopy provided a very sensitive method. FM-
spectroscopy was described and refined about 20 years ago by Bjorklund [I, 2,] as a
means to sensitively detect weak absorption features, such as spectral holes, and
independently by Hall [3, 4] in the context of servo locking a tunable laser to a high
finesse cavity [5], Both techniques are closely related optical analogs of methods
developed in the microwave region by Pound in the 1940’s. [6] The following,
section presents the basics of general optical FM-spectroscopy theory following
references [7,'8, 9], The FM-Iineshapes were calculated, and the error signal slope
was determined as a function of modulation frequency and modulation index.
Figure 3.1 shows the basic setup for FM-spectroscopy [I]. The output of a
single-mode laser of carrier frequency, coc, passes though an electro-optic phase
modulator (EOM) driven sinusoidally by an applied electric field at the RF-
modulation frequency, a>m, with a modulation index, M. The modulation index, M,. in
34
radian units is related to the modulation amplitude, 2nSv, and modulation frequency.
nv> b y
, , InSv Sv
The optical field after passing through the EOM is given by
E(t) = E0.expz"(tocf+ M .sintiy),
which can be written using a Bessel series expansion as
(3.1)
(3.2)
E(i) = E0exp(zm (0| = E02 exp(-2tij[l + 2 cos Co jY j (£- J Z - I
J Z = O
+ 2 sin OJj Y J ,Jn+l (0-n-!
n=0
-4+1 + 5-n - S n)
- ( L + (L i-& ]
(3.8)
The signal consists of a DC part (constant) and two time Varying components,
modulated at CQm due to a beating of adjacent frequency sidebands. Terms coming
from higher order sidebands interfering with each other (~2com, ~3com, etc.) have
been omitted because the detection bandwidth is at com. Use of an AC-coupled
photodiode allows exclusive detection of the time-varying part of the signal. Phase
38
sensitive detection with a mixer and phase shifter allows isolation of both the
absorptive (~cos COmf) and dispersive(~sincomf) components. A mixer is a nonlinear
device whose output is the product of its inputs. Choosing the detected optical signal
as one input and the modulation signal of the local oscillator at com as the other input,
the mixer output will contain a signal at the sum and difference frequencies of the
input signals. The difference frequency signal, which will be at DC, can be further
isolated with a low pass filter. In practice, the phase shifter is important to
compensate for unequal delays between signals and to detect the pure absorptive or
dispersive FM-signal. Intermediate settings of the phase shifter produce mixed
Iineshapes and, as later described, allow implementing a hybrid locking technique to
improve the long-term laser frequency stability (see section: incorporating the
inhomogeneous absorption line as a fixed reference in Chapter 4).
In the case of FM-spectroscopy, the Bessel series expansion can be terminated
after the first set of side bands so that expression (3.8) simplifies to [1,7,8]
= K exp(-250)[l + M (&, - S1)coscoj + M ( ^ 1 + - l ^ s m c o j ] , (3.9)
In FM-spectroscopy, W m is chosen large compared to F (wm » F), so that the
spectral feature can be probed by an isolated sideband where either the carrier
frequency, wc, of the laser or alternatively the RF modulation frequency, wm, is
scanned across the optical transition. The coswmt component then directly displays
the absorption profile when the upper or lower sideband resonates with the optical
39
transition, whereas the sin&V component displays three dispersion curves, one for
each frequency component resonant with the spectral feature as shown in Fig. 3.3.
In the case of a Lorentzian line shape, a good representation for the
homogeneous linewidth given by a spectral hole, attenuation and optical phase shift
can be written as
S (co) — A
>icd) = A
F2 Y
( 4A / j tK f 2 ■ (3,16)
a=-2 J J
Five noise types can be classified by the exponent, a, on the Fourier frequency, Z for
the spectral density Sv(f) with C tr= -2 random walk frequency noise, C t r = - I flicker
frequency noise, a = 0 white frequency noise (random walk phase noise), a = I
flicker phase noise and oc = 2 white phase noise. These noise types are illustrated in
Fig. 3.10.
A
lla
n
de
vi
at
io
n
ay
(x
)
[H
z]
S
pe
ct
ra
l n
oi
se
d
en
si
ty
S
v(f
)
[H
z2
ZH
z]
55
Fourier frequency f [Hz]
Integration time t [s ]
Fig.3.10 Illustration of the power law model for the spectral noise density as a
function of frequency (a) and the Allan deviation as a function of integration
time (b); note double logarithmic scales. [20]
56
Elliot et al [13] have derived a relationship to calculate the laser linewidth from
the spectral noise density, assuming a laser with white noise spectral density, Sv,
only up to a bandwidth, B, and no noise above this frequency. If the rms frequency
fluctuations Svrms « B , then the laser line shape is Lorentzian with linewidth
Al/, =TtSv.. (3.17)
If Svrms » B , a Gaussian line shape is predicted with linewidth
Ay, = 2 . 3 5 ^ . (3,lg)
In practice, at low frequencies the spectral noise density of oscillators tends to show
random walk frequency noise and flicker frequency noise, invalidating the above
relations, and a full numerical integration of the rather complicated expressions in
Elliot’s paper [13] has to be done.
Time Domain - The Allan Variance
Because it was found that the classical variance diverged for some types of noise
commonly found in frequency standards, D. W. Allan [15] introduced what has
become known as the Allan variance, (T2y ( r ) , to characterize frequency stability of
oscillators in the time-domain. The Allan variance is the measure for instability
recommended by the Institute of Electrical and Electronics Engineers (IEEE) and the
International Telecommunication Union (ITU); references [11, 15, 16] give detailed
information.
57
The Allan variance assesses the stability of a frequency standard over a time
interval, % commonly referred to as the integration time. In an Allan variance
measurement, a frequency counter determines the average frequency, /„, over a
specific time interval, r, (Fig. 3.9) which is recorded by a computer for subsequent
statistical analysis. For each time interval, t, a series of IV measurements are made.
The Allan variance for a finite number of measurements is then estimated using the
formula
I /v-i
. (3 19)
Division by IV-1 normalizes O 2y to the number of entries in the sum, and division by
2 facilitates comparison to the classical variance in the case th en ’s are random and
uncorrelated (white noise). Only for white frequency noise (a= 0) does the classical
variance equal the Allan variance. The process is repeated for different values of r.
Fewer samples are typically available for large integration times, r. To avoid large
variations, a minimum of 3 samples is required, meaning that an Allan deviation for
a 10 minute integration time requires at least 30 minutes of data recording. The
longer the data set duration, the better the confidence in the estimate.
The Allan deviation or root Allan variance is given by taking the square root of
the Allan variance
(Ty (T) = ^ a 2y(T) . (3.20)
58
Plotting the Allan deviation versus the integration time, t, yields the Allan deviation
curve, which allows inference about the type and level of noise present in the
system.
Almost exclusively, Allan variance plots are actually Allan deviation plots in the
literature. To compare laser systems operating at different frequencies, dividing the
Allan deviation by the laser frequency, V0, produces the fractional Allan deviation,
(O/V
Similar to the case of spectral noise density, the nomenclature of frequency noise
distributions can be associated with a truncated power law model
The classification is shown for the Allan deviation in Figure 3.10(b). .
From frequency domain measurements, time domain predictions can be made
using the relation
where//, is the high frequency cutoff for the applicable measurement system, i.e. the
conversions from Allan variance to spectral noise density are impossible. This
relation is useful when only one laser system is available [17, 30, 26, 28], Since we
had the luxury of two independently stabilized lasers and wanted to avoid potentially
inaccurate conversions, we only used original experimental data for publications.
(3.21)
Sin4 (^-ZT)
(3,22)
measurement bandwidth. The conversion process loses some information. Generally,
59
Table I summarizes the most common noise types found in frequency standards
and their relation to the spectral noise density, Sv (f), and Allan deviation, (Ty(T). [18]
Table I Common noise types found in frequency standards and their relation to the
spectral noise density, Sv (/), and Allan deviation, o>(z). [18]
Type of noise a p O y (T )
Random walk frequency -2 1/2
Tl^ 3 " 2 ■
Flicker frequency -I 0 h - J - 1 21n 2
White frequency 0 -1/2
‘A
\F
\
to
I ^
Flickerphase I -I K f 1
White phase 2 -I
A h2A x i *'
The External Cavity Diode Laser
The semiconductor laser was invented by Basov et al. [19] in the early 1960’s
and its ease of operation has made it one of the most widely used laser systems
covering wavelength windows from ~ 635 nm (AJGaInP) to IO pm (Pb-salt). Diode
laser wavelength coverage is not continuous, and; performance characteristics vary
widely depending on the wavelength. Semiconductor lasers operating at the optical
transmission frequencies of optical fibers, at 229 THz (1.3 pm) and 193 THz
60
(1.55 (im), have become particularly important in the optical telecommunication
industry, because they can be mass-produced, have low power consumption, and are
reliable. For the commercially most important wavelengths, in addition to traditional
Fabry-Perot-type lasers, more complex resonator structures on the semiconductor
chip have been developed leading to distributed feedback (DFB) and distributed
Bragg reflector (DBR) lasers. For these lasers, the selective reflectivity of the
distributed grating determines the lasing oscillation wavelength. Recently, diodes
emitting blue light (GaN) have become commercially available, of particular
importance in optical data storage systems; because the storage density is inversely
proportional to the square of the wavelength of the light used, approximately four
times higher data storage can be achieved than with near-infrared laser diodes. [20]
High-power, wide-stripe laser diodes that run multimode have become important as
pump sources in solid-state lasers, fiber amplifiers, and tapered amplifier systems.
Many scientific applications, in particular high-resolution laser spectroscopy,
require the laser to operate narrowband, in a single spatial and frequency mode, over
the spectral range of interest. Linewidth narrowing, continuous tuning, and stable
operation can be accomplished by using dispersive optical feedback to the laser
diode.
The external cavity diode laser (ECDL) system used for all of our Stabilization
and spectroscopy experiments was developed at Montana State University. Gregg
W. Switzer and J. L. Carlsten first developed the local design for application in a
miniature water vapor lidar system suitable for measurement of water vapor in the
61
Martian atmosphere at a wavelength of - 935 nm. [21] G. C. Dodge further refined
and modified the mechanical design to allow easy pivot point adjustment and
operation at 1550 nm. Two external cavity diode lasers (named Max & Moritz) were
assembled, aligned, and characterized by the author. Performance characteristics of
the laser system will be discussed in subsequent sections, and information on
electronic drivers and feedback circuitry is presented in appendix A,
Principle of Operation
Free running solitary laser diodes have relatively large linewidths on the order of
10’s of MHz, very large relative to atomic transitions, due to a low optical cavity
quality factor, a result of the facet reflectivities of R ~ 30 % and cavity lengths of
~ 100 (am. Injecting a current between the n and p cladding layers in the active
region of the diode generates laser light. Current injection produces electron-hole
pairs that recombine and emit photons with high quantum efficiency. The
semiconductor material band gap, a function of temperature and carrier density,
determines the nominal emitted wavelength, and a broad gain curve typically leads
to multimode operation. Single wavelength operation requires careful selection of
operating temperature in conjunction with the diode laser injection current.
Wavelength tuning with temperature is often accompanied by sudden jumps in
amplitude and phase (mode-hops) as the laser oscillation jumps to the next cavity
mode.
62
External cavity diode lasers, in contrast, allow single mode operation and tuning
to any wavelength within the gain curve of the laser diode as long as tuning of
wavelength selective elements can be synchronized. Ideally, the laser diode itself
becomes a pure gain element; an external resonator acts as a mode selector, allowing
gain for only one mode of the external cavity. Apart from excellent tuning
characteristics, external cavity diode lasers have linewidths on the order of 10’s of
kHz over a few milliseconds, sufficient for many applications in spectroscopy and
important as a starting point for active frequency stabilization, since servo
requirements needed to suppress the inherent frequency noise become less
demanding.
The Littman-Metcalf optical feedback configuration, [22] first introduced for dye
laser oscillators, was chosen for the lasers constructed and used in this work because
it has distinct advantages: a) there is no output beam steering - important for the
alignment of subsequent optical elements, b) double passing the dispersion grating at
grazing incidence gives better spectral resolution, and c) using a knife-edge prism
instead of a plane mirror as the retro reflector makes the cavity configuration
mechanically more stable by making the tilt degree of freedom unimportant.
The Littrow configuration, in contrast, operates on only a single pass of the
dispersing grating and provides higher optical output power. Beam steering,
however, needs compensation, which makes the beam mechanically unstable and
causes difficulties in producing long continuous scans.
63
ECDL Construction and Characterization
Since the laser diode in an ECDL should act only as a gain element, longitudinal
facet modes caused by the miniature Fabry-Perot resonator are not desirable. Single
angled facet (SAP) diodes offer an alternative, and they were used instead of anti-
reflection coated diodes. [23] In SAF-diodes, the wave-guide region was grown so
that it intersects the cleavage plane at normal incidence on the back facet and at an
angle of ~ 7 degrees at the front facet. Photons reflecting off the front facet are
absorbed in the substrate material; this together with the low reflectivity of the
angled facet of R ~ 2XlCT5 removes the Fabry-Perot cavity for this diode structure.
The particular GaInAsP / InP diodes used were custom grown by Quantum
Photonics, Inc. and packaged in a standard 9 mm can with the output facet open to
potential outside contamination. Electrical pin outs were only provided for laser
anode and laser cathode, without a photodiode that monitors the output of the lasers’
back facet. The gain peak occurred at ~ 1535 nm, very close to the optical transitions
of interest, extending ~ 30 nm in either direction (measured @150 mA, T = 20° C).
The diodes were driven forward biased with a low noise laser driver constructed by
the author (see Appendix A).
The optical configuration of the laser is shown in Fig 3.11 (figure courtesy of G.
W. Switzer [21]). The highly divergent laser light emitted from the SAF-diode is
collimated using an aspheric collimating lens with numerical aperture NA = 0.55 and
focal length 3.1 mm (Thorlabs). An 8 mm wide region of a gold-coated holographic
64
grating with 1000 lines/mm obtained from American Holographic, Inc. disperses the
collimated light at a grazing incidence of 87 degrees. The grating efficiency was
measured to be ~35 % into the first diffraction order.
Prism
1st Order
Laser
Diode
Collimating
Lens
Diffraction
Grating
Figure 3.11 Schematic of the external cavity diode laser in the Littman-
Metcalf configuration (figure courtesy of G. W. Switzer). [22]
65
A knife-edge roof prism, custom made by Continental Optics, Inc. and anti
reflection coated by Optosigma, was used to retro-reflect a selected wavelength of
the first order back into the diode for oscillation. All other wavelengths exit the
cavity below lasing threshold. Using a knife-edge roof prism removes the
requirement of a critical tilt adjustment needed with a plane mirror. The 0th order of
the diffraction grating has about 50 % efficiency and serves as the optical output
coupler.
The ECDL’s constructed using SAF-diodes exhibit unparalleled performance
relative to those with anti-reflection coated diodes. The side-mode suppression ratio
(SMSR), defined as the ratio of optical power in the highest intensity mode to the
next highest intensity mode, was measured by coupling part of the laser light into an
optical spectrum analyzer Model HP7095OB. Figure 3.12 shows the ECDL
spectrum near 1535 nm, the center wavelength of the diode’s gain peak, while
operated at 100 mA injection current. As a result of the low reflectivity of the angled
facet, a side-mode suppression ratio of ~ 51 dB relative to the central mode was
achieved, a significant improvement over commercially available ECDL’s which
typically show SMSR of 40 - 45 dB.
The optical output power of the ECDL named “Max” measured as a function of
injection current (Pi-curve) is shown in Figure 3.13 along with the Pi-curve for the
plain diode without feedback. Threshold current is 43 mA with a slope efficiency Of
0.12 mW / mA. The optical output power reaches 4 mW at 75 mA. To extend the
lifetime and to avoid heating and damage resulting from high optical power densities
66
on the facets, both lasers were operated with 60 mA injection current, giving
~1.9 mW optical output power for applications. This performance was lower than
expected when the diodes were purchased. Amplifying the laser output with an
Erbium doped fiber amplifier (EDFAj readily produced higher output power.
-10
-20
-30
CO
S
Is -40
I
i
CL
-60
-70
1530 1532 1534 1536 1538 1540
Wavelength [nm]
Figure 3.12 ECDL-spectrum near 1535 nm showing a side-mode suppression
ratio of - 51 TB. The width of the peak was limited by the 0.5 nm resolution of
the optical spectrum analyzer.
-51 dB Side-Mode
Suppression Ratio @ 1535 nm, 100 mA
67
without optical feedback
with optical feedback
43 mAthreshold
I 100 1
Injection Current (mA)
Figure 3.13 Optical output power versus injection current (PI. curve) for the
ECDL “Max”. At a threshold current of 43 mA the ECDL “Max” begins
lasing.
As mentioned earlier, diode laser output characteristics are strongly temperature
, . . . ' ;
dependent. Hence, flawless diode laser operation requires strict temperature control.
This was achieved by placing the aluminum base plate of the laser cavity onto two
Melcor 15 X 30 mm thermo-electric coolers with feedback for a control circuit
provided by a calibrated 10 k£2 thermistor placed close to the coolers. A Wavelength
Electronics model LFI-3526 T- controller exhibiting excellent stability (< 2 mK over
I h, < 5 mK over 5 h) controlled the temperature and was tuned to stabilize the
coolers at just below ambient room temperature of ~ 19°C. Maintaining near
68
ambient room temperature of the laser cavity avoids condensation on the diode laser
facets. Due to the lethargic response (minutes) of the lasers’ large thermal mass
aluminum Components, temperature tuning was not used for active laser frequency
stabilization.
Frequency tuning the laser in single mode is important for spectroscopic
applications, and tuning transducers provide the leverage for frequency stabilizing
the ECDL. The output laser wavelength of the ECDL was tuned by changing the
angle of the feedback prism plate. The prism plate contains a fine pitch screw that
rests on the top of a 20 mm piezo-electric tuning (PZT) stack (Thorlabs) held in the
base plate. The screw provides coarse wavelength tuning, whereas applying a
voltage to the PZT stack provides continuous tuning. Mode-hop-free tuning of the
laser requires the physical length of the cavity and the wavelength selected by the
cavity to change simultaneously. The grating pass band must exactly synchronize
with the accompanying change in cavity length. This requires very careful
positioning of the prism plate pivot point using orthogonal dovetail slides to adjust
the pivot point’s vertical and horizontal position. [24] When adjusted correctly, all
mode hops were eliminated over the full range of the diode’s gain bandwidth. The
reliability of this technique eliminated retuning of the external cavity during
1.5 years of operation.
At an injection current of 70 mA, coarse tuning (Max) covered a range of 66 nm,
between 1493 nm and 1559 nm. Continuous tuning over ~ 47 GHz was possible by
applying a 150 V DC-voltage ramp to the PZT stack and is shown in Fig. 3.14.'
O
pt
ic
al
F
re
qu
en
cy
(G
Hz
)
Measuring the laser output frequency with a Burleigh WA 1500 wavemeter, gave a
PZT transducer transfer function of K = 0.3 GHz / V. The prism plate assembly , was
made as light as possible to reduce errors from inertia. Repetitive scans at up to
I kHz have been demonstrated, limited by mechanical resonances to be investigated
in the following sections. Since changing the feedback-prism plate angle Is a
mechanical process, the response on tuning is rather slow. In contrast, tuning the
laser by modulating the injection current to the laser diode is very fast, and a
transducer transfer function of K = 30 MHz/mA has been achieved. Knowledge of
the tuning limitations is crucial in the context of laser frequency stabilization.
69
1 9 4 0 1 0
1 9 4 0 0 0
1 9 3 9 9 0
1 9 3 9 8 0
1 9 3 9 7 0
1 9 3 9 6 0
1 9 3 9 5 0
Piezo - Voltage (V)
Figure 3.14 Continuous tuning curve for ECDL “Max”. Continuous tuning is
demonstrated over ~ 47 GHz, center wavelength ~ 1535 nm, I = 60 mA.
70
ECDL Transducer Response
Assessing the ECDL transducer response and the actual frequency noise present
on the free-running ECDL are important steps in designing a proper servo loop
necessary to suppress laser frequency noise. Figure 3.15 shows the experimental
setup used to measure the transducer frequency response. This setup was similar to
that for measuring the spectral noise density of the free running laser described
below. A low finesse Fabry-Perot cavity (Burleigh) Was used as a discriminator to
convert frequency modulation (FM) into amplitude modulation (AM). This cavity
had a length of 15cm, giving a free spectral range (FSR) of ~1 GHz. The cavity
finesse, F, of ~ 2.1 was determined by scanning the cavity across the laser line width
and calculating the ratio of the free spectral range to the cavity transmission band. In
order to measure the transducer response of the ECDL, the laser was manually tuned
to operate at the side of the transmission fringe, so that laser frequency fluctuations
could be translated into amplitude fluctuations, detected by a New Focus 1811
InGaAs photodiode. An HP network analyzer, Model HP 3589A, with its source
Output connected to the current modulation input port of the laser diode driver (see
appendix A) measured the injection current frequency response of the ECDL. The
HP network analyzer also measured the frequency response of the PZT driven
feedback prism plate through the Thorlabs MDT 691 PZT-amplifier. The detector
output was connected to the input port of the network analyzer. Logarithmic scans of
the network analyzer were performed, covering 10 Hz to 100 MIz for measuring the
71
PZT frequency response and 10 Hz to 100 MHz for measuring the injection current
frequency response. Results of these measurements are shown in Fig. 3,16 (a) and
(b) for the PZT amplitude and phase, and Fig. 3.16 (c) and (d) for the injection
current amplitude and phase, respectively. .
Low finesse cavity
Detector
ECDL Isolator
PZT
amplifier
Current
modulation
Source
Network
analyzer
Figure 3.15 Experimental setup used to measure the transducer response of the
ECDL with a low finesse cavity as a frequency discriminator.
72
PZT response
DQ 3 0
+- I I l M l 4— 1 I I l l lTo 180
% 135
Q- -1 8 0
IO1 10 c /LJ JO 3Frequency (Hz)
Current response
4 -H 4 4W 4-H+MM
oj-1 3 5
52 -2 7 0
-C -3 1 5
Frequency (Hz)
Figure 3.16 Transducer response of the tuning elements of the ECDL, relative
amplitude (a) and phase (b) of the piezo-driven (PZT) feedback prism plate
assembly, relative amplitude (c) and phase (d) of the diode current modulation.
73
The PZT displayed a fairly flat response up to about I kHz. Beyond I kHz, a
significant resonance occurs, which leads to a phase shift of 180 degrees. This
resonance was attributed to a mechanical resonance in the feedback prism plate
assembly of the ECDL, and similar resonances have been observed in other
laboratories. Independent spectral noise density measurements, reported in the
following section, further supported this conclusion. To properly design the feedback
servo circuit, the transducer tuning elements’ resonant frequencies must be known
and avoided. Positive feedback to the laser, caused by a phase shift of more than
180°, could lead to severe laser oscillations. The low bandwidth PZT loop was
designed to keep the high bandwidth loop of the laser diode’s injection current
frequency control in the center of its dynamic range. Therefore, the unity gain point
was confined Iq a value below I kHz (see appendix A).
The frequency and phase response to changes in the injection current showed a
fairly flat response up to about I MHz. At I MHz the amplitude of the response had
decreased by ~ 20 dB. Hence, with high bandwidth feedback electronics available,
intrinsic noise suppression up to these I MHz frequencies can be accomplished
without inordinate technical difficulty. Indeed, the injection current servo, detailed in
appendix A, provided bandwidths in the order of ^ I MHz.
74
Spectral Noise Density of the free running ECDL
Error signal analysis allowed a direct measurement of the spectral noise density
Sv( f ) of the free running laser using a static discriminator slope D [V/Hz], which
converted the frequency modulation [Hz] on the laser, caused by frequency
fluctuations, into amplitude modulation [V], a conveniently measurable quantity. To
obtain a large discriminator slope, D, a narrow optical transition in isotopically pure
(99.9 % 7Li) 170Er3"1*:YLiF4 was chosen as the frequency reference. [25] Roger M.
Macfarlane of IBM and Richard S. MeltzCr from the University of Georgia, Athens,
kindly provided this crystal, which was grown by Arlette Casanho. During the
measurement, the laser frequency was compared to the center frequency of the
transition, using an error signal proportional to the frequency difference between the
two frequencies. The experimental setup, shown in Fig. 3.17, was based on FM-
spectroscopy of a narrow inhomogeneously broadened ~ 150 MHz (FWHM)
Er:3"1":YLiF4 transition. This heterodyne technique overcame the problem of low
frequency AM of the laser, which could have been wrongly interpreted as FM noise,
by detecting at the high RF modulation frequency, where intrinsic laser intensity
noise was very small. The ECDL, optically isolated by a Faraday rotator (OFR
Model 10-4-1550-VLP), was externally phase modulated by a broadband New Focus
electro-optic phase modulator Model 4002 driven at 109.5 MHz by a PTS 500
frequency synthesizer through a Custom-made resonant tank built by the author (see
appendix A).
. 109,5 MHz
Cryostat
Isolator
W avemeter
PZT
amplifier
Spectrum
analyzer
Ramp
. generator
Adjustable phase
delay
Detector
Figure 3.17 Experimental setup for measuring the spectral noise density of the free running laser.
76
The sideband frequency was chosen to optimize the error signal slope within the
125 MHz bandwidth of the New Focus 1811 InGaAs photo detector. The laser
frequency was monitored with a Burleigh WA 1500 wavemeter, which allowed
calibration of the error signal to within the 0.1 ppm accuracy of the wavemeter. For
this purpose linear scans of the laser frequency were provided by applying a saw
tooth voltage ramp to the PZT driving the feedback prism plate of the laser. A
Stanford Research Systems SR 345 function generator amplified by a Thorlabs
model MDT 691 PZT amplifier generated the ramp. In order to optimize the signal, a
A/2-waveplate in front of the cryostat allowed for adjusting the polarization of the
beam. The detected transmitted light was heterodyned with the 109.5 MHz local
oscillator, and the phase shifter was adjusted to obtain the dispersive error signal.
Spectral analysis of the error signal above I kHz was performed with an HP-
spectrum analyzer Model HP E4411B while a Wavetek spectrum analyzer Model
5820 was used for frequencies below I kHz. The optical power of the laser beam
was adjusted using the X/2-waveplate/polarizing beam splitter (PBS) combination to
140 pW. The Er:3"1":YLiF4 crystal was placed in an Oxford SpectroMag cryostat at
T = 10 K to avoid any spectral hole burning. The crystal was oriented with its c-axis
parallel to the laser’s A:-vector and perpendicular to the external B-field. To separate
the narrow optical transitions contained in the multiplet located at 6534 cm'1, a Small
magnetic field of B = 0.2 T was applied. The error signal yielded a linear slope at
line center of D - 3.746x10"7 V /Hz, which was used to convert the spectrum
77
analyzer output [dBm] into spectral noise density, S1XX), of the laser frequency
fluctuations [Hz!4Hz ]. T o measure the spectral noise density of the free running
ECDL, the laser was manually tuned to the center of the Er:3+:LiYF4 transition using
the PZT. No active servo feedback was engaged for this measurement. The error
signal was spectrally analyzed using specific analyzer resolution bandwidths
appropriate to the different frequency measurement intervals; sufficient frequency
overlap between individual measurements was assured.
Figure 3.18 shows the result of this measurement on a log-log plot of the rms
spectral noise density versus frequency. Three noise types common to frequency
standards, indicated by asymptotes of distinct slopes, can be distinguished according
to the power law model relation (3.16) for the spectral noise density. For frequencies
up to tens of Hz, random walk frequency noise (a= -2, slope = -I) was prevalent,
followed by flicker frequency noise { a—-I, slope = -1/2) for frequencies up to tens
of kHz. The white noise. frequency floor ( G f = 0, slope = 0) was reached at
frequencies at approximately 100 kHz. Structural resonances. in the laser cavity
caused narrow-band frequency noise. A distinct bump was observed at ~ I kHz and
can be assigned to a mechanical resonance in the feedback prism plate assembly of
the external cavity excited by acoustical laboratory noise, prevalent in the kHz range,
coupling into the laser cavity. This issue was already evidenced in the PZT
transducer frequency response (see previous section). To alleviate this problem,
passive acoustical isolation of the laser was required. An acoustical isolation box
was constructed to cover the optical setup for locking experiments in Er3"1": KTP (see
section: Laser frequency stabilization to regenerative Spectral holes in Er3*: KTP in
Chapter 4). The sharp features at 60 Hz and 120 Hz are due to pickup from the
power supply.
. 78
Figure 3.18 Spectral noise density of the free running ECDL as a function of
noise frequency; note double logarithmic scales.
79
The substantial amount of noise present at low frequencies exhibiting a ~ 1/f and
~ 1/f2 frequency dependence, typical for external cavity diode lasers
[26,27,28,29,30], indicated the need for a servo loop of high gain at these
frequencies. However, a relatively low bandwidth (~1 MHz) electronic feedback
loop should be sufficient to suppress this laser noise compared to the feedback
bandwidths of tens of MHz needed to suppress the noise found in solitary diodes [31,
32]. The injection current servo is documented in Appendix A.
80
REFERENCES
1. G. C. Bjorklund: IBM Invention Disclosure SA 8790135 (March 1979).
2. G. C. Bjorklund, Optics Lett. 5, 15 (1980).
3. J. L. Hall, Science 202, 147 (1978).
4. J. L. Hall, L. Hollberg, T. Baer, and H. G. Robinson, Appl. Phys. Lett. 39, 680
(1981).
5. R.W. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A. J. Munley, and
H. Ward, Appl. Phys. B 31, 97 (1983).
6 . R. V. Pound, Rev. Sci. Instrum. 17,490 (1946).
7. J. M. Supplee, E. A. Whittaker, and W. Lenth, Appl. Optics 33, 6294 (1994).
8 . G C. Bjorklund and M. D. Levenson, Appl. Phys. B 32,145 (1983).
9. B. I. Bleaney and B. Bleaney, Electricity and Magnetism (volume 2), Oxford
University press, 1989, 3rd edition.
10. T. Day, E. K. Gustafson, and R. L. Byer, IEEE Journal of Quantum Electronics
28, 1106 (1992). 1
11. NIST Technical Note “Characterization of Clocks and Oscillators” (US
Government Printing Office), Vol. 1337 (1990).
81
12. M. Roberts, P. Taylor, P. Gill, NPL report CLM 8 , February 1999.
13. D.S. Elliot, R. Roy, and S.J. Smith, Phys. Rev. A 26,12 (1982).
14. B.C. Young, F.C. Cruz, W.M. Itano, and J.C. Bergquist, Phys. Rev. Lett. 82,
.3799(1999).
15. D.W. Allan, IEEE Transactions on Ultrasonics, FerroelectriCs, and Frequency
Control, UFFC 34, 647 (1987), ■
16. Hewlett Packard Application Note 1289, 56 (1997).
17. I Dirscherl, B. Neizert, T. Wegener, and H. Walther, Opt. Comm . 91, 131
(1992).
18. J. A. Barnes, A.R. Chi, L.S. Cutler, D J. Healey, D.B. Leeson, T.E. McGunigal,
LA. Mullen, Ir., W.L. Smith, R.L. Sydnor, R.F.C. Vessot, and G.M.R Wimkler ,
IEEE Trans, instrum. Meas. 20,105 (1971). "
19. N.G, Basov, O.N. Krokhin, and Y.M. Popov, JETV 40, 1320 (1961).
20. S. Nakamura , W. Kaenders, Laser Focus World 35 (4), 69 April (1999).
21. G. Switzer, PhD thesis, Montana State University 1998.
22. K. Liu and M.G. Littman, Opt. Lett. 6 , 117 (1981); P. McNicholl and M.J.
Metcalf, Appl. Opt. 24, 2757 (1985).
82
23. PJ.S. Heim, Z.F. Fan, S.-H. Cho, Keeyol Nam, M. Dagenais, F.G. Johnson, and
R. Leavitt, Electr. Lett. 33,1387 (1997).
24. The dovetail slides were added to the original external cavity design by
mechanical engineering student G. C. Dodge,
25. R.M. Maefarlane, R.S. Meltzer, and A. Cassanho, Phys. Rev. Lett. 69, 542
(1992).
26. T. Kurosu, J. Ishikawa, and N. Ito, Appl. Phys.B 63, 265 (1996).
27. J. Kawakami, M. Kourogi, and M. Ohtsu, Jpri. J. Appl, Phys. 33, 1623 (1994).
28. G. Bianchini, P. Cancio', F. Minardi, F.S. Pavone, F. Perrone, M. Prevedelli, M.
Inguscio, Appl. Phys. B 6 6 , 407 (1998).
29. A. Danielli, P. Ruslan, A. Aiie, M.H. Chou, arid M.M. Fejer, Opt. Lett. 25, 905
(2 0 0 0 ) and references therein.
30. C.-H.. Shin and M. Ohtsu, Opt. Lett. 15, 1455 (1990).
31. H.R. Telle and H. Li, Electr-Lett 26, 858 (1990).
32. R.W. Fox, L. Hollberg, and A.S. Zibrov, Experimental Methods in the Physical
Sciences 29C, 77, Academic Press (1997).
83
CHAPTER 4
LASER FREQUENCY STABILIZATION TO SPECTRAL HOLES
The availability of ultra-narrow SHB resonances down to 15 Hz in rare earth
doped crystals, the relative immunity of spectral holes to environmental disturbances
such as vibrations, and the portability and compactness of a stable laser system using
SHB references with a closed cycle cryo-cooler are important features that should
enable application in a variety of fields beyond those normally associated with
spectral hole burning. Stabilization of mode locked lasers [1] to spectral holes also
should be practical and will have applications in signal processing situations and in
other contexts that require short pulses, frequency combs, or optical clocks. The
SHB frequency references are well suited to applications where multiple frequencies
are required and where the programmability of SHB materials allows programmable
frequency differences up to the multi-GHz range or, if disordered solids are used, to
the THz range. With the development of suitable photon-gated (or two-photon) SHB
materials, the production of long term secondary frequency standards based on SHB
may become practical. [31]
The importance of stabilization in real time optical signal processing in SHB
materials is underscored by the observation that early moderate-speed
demonstrations have been limited by laser frequency jitter that led to a loss in signal
fidelity. [2,3,4,5,6 ,7] These problems can occur at several levels: a) uncontrolled
84
phase variations between programming pulses when repeated pulse sequences are
used for writing or refreshing spectral interference gratings, b) the more extreme
case where the jitter exceeds the Fourier width of the exciting pulses so that the
processed pulses fail to overlap spectrally with the programming pulses, and c) the
case where the jitter exceeds the Fourier width of the exciting pulses so that the
exciting pulses fail to overlap. These problems can be overcome by . the use of a
second piece of the same signal processing material as an SHB frequency reference
providing automatic frequency and phase compatibility between the signal
processing material and the stabilized laser source over the time scale of interest.
[12] The limits on device performance are set then by material parameters rather
than by instability of the laser. [12,14] The relative vibrational immunity of the
spectral holes provides an important simplification in system design and
performance for either spectroscopy or SHB devices [8,9]. This advantage is even
greater when both the frequency reference and spectroscopic sample or SHB device
are mounted on the same sample holder. This has been demonstrated here for
Er3+IY2SiO5. Lasers stabilized to spectral holes are already playing an important role
in proof-of-principle demonstrations of a variety of SHB devices [5]; the latest of
which, a high-bandwidth correlator, demonstrated in Er3+IY2SiO5 at 4.2 K. [10]
Laser frequency stabilization to persistent spectral holes [11] and regenerative
spectral holes [12] burned in the absorption lines of Tm3+-doped insulating crystals
at 798 nm and 793 nm provided excellent stability. In spectral hole burning materials
covering the 1.5 pm wavelength region explored to date, regenerative (transient)
'85
holes, such as those found in Er3+IY2SiO5 and Er3+:KTP, have proven to be orders of
magnitude narrower in frequency than persistent holes, such as in Er3+:D':CaF2, the
only material known to date exhibiting persistent spectral hole burning at 1 .5 pm.
Er3+-doped SHB crystals exhibiting regenerative SHB (Er3+=Y2SiO5, Er3+=KTP) and
persistent SHB (Er3+=D-=CaF2) have been used here to extend laser stabilization to
the important 1.5 pm telecommunications band where Er3+-doped crystals [13,14,29]
have the frequency selectivity required for optical storage, real-time address header
decoding for all-optical packet routing, [6,7] and all-optical correlation. [3,4,5]
Our demonstrations were carried out with crystals cooled to liquid helium
temperatures, but higher temperature operation with I kHz frequency stability at
4.2 K has been demonstrated using Er3+=Y2SiO5 and projection of operation up to
20 K can be made for the deuterated CaF2 [11,32], Alternatively, using a narrow
inhomogeneous absorption line as . a frequency reference, such as the ~ 150 MHz
wide line found in isotopically pure Er3+=LiYF4 or a part per million diluted
Er3+=Y2SiO5 already allowed operation at 15 K.
Laser frequency stabilization to regenerative spectral holes in Er3+=YoSiOs
The stabilized diode laser - Er3+=Y2Si0 5 system described here exploits the
regenerative SHB technique. [12] A transient spectral hole is continuously
regenerated by the stabilized laser and provides a frequency reference at an
arbitrarily chosen location in the inhomogeneous Er3+=Y2SiO5 absorption profile.
86
The stability of the laser will then be determined by the dynamical properties of the
SHB material together with the design of the locking system.
To substantially improve the long-term frequency stability, we have extended the
locking technique by using a combination of the error signal contributions from the
spectral hole and the inhomogeneous line. The reduction of longer-term drift to
7 kHz / min over several minutes obtained with the extended technique represents a
substantial improvement over the 600 kHz ,/min reported for Tm:YAG.[12] In the
old and new cases, anticipated refinements to the feedback system and frequency
modulators may be expected to provide further substantial improvement over both
the.long and short term stability reported here, which is already IO3 times better over
important integration time scales than that for commercial lasers. As shown in
Ref. 12, this provides new capabilities to probe dynamics in the neighborhood of the
active rare earth ions in SHB materials and to reveal small scale level structures and
dynamics out to the tens of milliseconds scale or even longer.
Methods and Apparatus
The SHB crystal chosen as a frequency reference was Er3"1": YaSiOs with an Er3*
concentration of 0.005 atomic percent. This material exhibits transient spectral hole
burning on the 4Ziszz (I) —»AI 13/2 (I) transitions at 1536.14 nm (site I) and
1538.57 nm (site 2) by population storage in the excited state of the optically active
ion. [13] The homogeneous linewidth for this crystal was determined from the
87
optical dephasing time T2 obtained from two-pulse photon echo experiments at 1.6 K
and B = 0.2 T, and the measured value Fh-S kHz is consistent with previously
published results. [13] The homogeneous linewidth at 0.5 T is about half that at
0.2 T. In principle, the minimum line width of a shallow spectral hole burned by a
narrow-band laser is 2 Th due to convolution Of burning and reading cycles; deeper
holes can be expected to become broader, since there is less saturation of material
absorption in the wings of a hole than at the center.
The frequency locking experiments reported here were performed using the
strongest absorption transition from the lowest Zeeman-split level for site I in
moderate magnetic fields of B = 0.2 to 0.5 T; the 0.2 T magnetic field value was
chosen to simulate field strengths that have been obtained using compact I cm
diameter Nd-Fe-B permanent magnets. [4,7] The inhomogeneous linewidth
Fmh = 500 MHz = 0.017 cm"1 is illustrated in Fig. 4.1, where a single spectral hole
has also been burned into the line.
Two frequency reference crystals were cut from the same crystal boule
(batch # 7-167) to give an inhomogeneously broadened absorption of ~ 50 % at line
center. Crystal dimensions were 5 mm (and 4 mm) along D1, 6 mm (and 5 mm)
along D2 and I mm (and 1.2 mm) along b. Each crystal was oriented with its Dj-axis
parallel to the magnetic field, the lasers’ k-vectors parallel to the 6 -axis, and the
lasers polarized with E along D2. Both crystals were immersed in superfluid helium
at 1.6 K in a single Oxford Instruments SpectroMag cryostat with a superconducting
magnet that provided for adjustment of the magnetic field. The laser beams were
Fr
ac
tio
na
l T
ra
ns
m
itt
ed
I
nt
en
si
ty
88
spatially separated, and the crystals were masked so that each crystal was exposed to
only one beam. The spectral holes were created with irradiances of 100 pW / cm2
using ~ 3 mm beam diameters. Beam irradiance was controlled using a A, / 2-plate
and a prism polarizer.
Laser Frequency Offset (GHz)
Figure 4.1 Transmission spectrum of 0.005% Er3^Y2SiO5 scanned by a diode laser
showing the entire inhomogeneously broadened absorption profile at zero applied
field (B = 0 T). The arrow indicates a spectral hole burned by a second laser.
\
89
The experimental apparatus shown in Fig. 4.2 was similar to that described in
Ref. 11 and Ref. 12. Two external cavity diode lasers in the Littman-Metcalf
configuration [15], as described in Chapter 3, were used for locking the laser
frequency to the spectral hole with the Pound-Drever-Hall technique. [16] The error
signal was derived from the spectral hole transmission using frequency modulation
(FM) spectroscopy as described in Chapter 3, [17] with the two lasers modulated by
New Focus 4002 broadband external electro-optic modulators (EOM) driven at
27 MHz and 30 MHz, respectively. These frequencies greatly exceeded the spectral
hole widths but were far less than the inhomogeneous absorption linewidth
Finh = 500 MHz. The primary laser side bands had a modulation index M = 0.4, and
secondary side bands were small but observable. The sharp resonance of the spectral
hole in the inhomogeneous absorption line creates a corresponding dispersion in the
refractive index. Figure 4.3(a) displays the transmission spectrum through a single
spectral hole in the inhomogeneously broadened absorption profile for a ,phase-
modulated laser (including side bands), and Fig. 4.3(b) plots the demodulated FM-
spectroscopy error signals obtained simultaneously for three values of relative phase,
that could be continuously adjusted using a custom built phase shifter (see appendix)
between the local oscillator and the EOM. In the stabilization system, each locking
laser bums a hole in its reference crystal, primarily at the carrier frequency but also
at the frequencies of the FM-sidebands.
Cryostat
Figure 4.2 Experimental apparatus for laser frequency locking to spectral holes and beat frequency measurement
of laser stability by combining the beam from the laser shown and that from an independent second laser system.
91
-o d 1.5
-p 0.5
(j) = 90
Laser frequency offset (MHz)
Figure 4.3 (a) Transmission spectrum, as probed by a phase-modulated laser
with sidebands, of a single spectral hole burned in the inhomogeneously
broadened absorption profile by an unmodulated second laser, in an applied field
of B = 0.2 T. (b) Error signals derived from the spectral hole and the
inhomogeneous line using different phase delay settings. The spectral hole is not
limited to line center as chosen for illustration in this particular figure but can
also occupy other positions in the inhomogeneous line.
The FM-error signal was processed by a servo loop that provided both fast
corrections to the injection current of the laser diode and reduced bandwidth signals
to the piezoelectric control of the laser’s external feedback prism plate. Control of
the prism plate angle keeps the current servo within its operating limits. A detailed
description of the servo electronics and locking apparatus constructed by the author
92
is given in the appendices. Due to laboratory constraints, the two lasers were on one
table, and the reference crystals, magnet dewar, and locking beam detectors were on
a separate table. Neither table was pneumatically floated, so the results reported here
demonstrate the immunity of this locking technique to vibrations. By contrast, most
other frequency references require extreme vibration isolation measures to reach this
short-term stability.
Evaluation of the frequency stability of a single laser at sub-MHz resolution is
difficult if one lacks a frequency standard at the appropriate wavelength for
comparison. For that reason, two independent lasers were constructed and locked to
two separate SHB crystal references. The frequency stability was determined by
beating unmodulated portions of the two stabilized laser beams on a New Focus
1811 photodiode detector, recording the beat frequency measured by a Stanford
Research SRS 620 frequency counter, and carrying out subsequent statistical
analysis of the time dependence of the beats using a computer. Allan deviations for
integration times up to 50 ms were directly measured with the frequency counter,
whereas for longer integration times the Allan deviation was calculated from the
recorded heterodyne beat frequency data, which was recorded at 50 ms time
intervals. All data acquisition programs were written in Labview.
Results and Discussion
The stability over broad time scales is characterized by the Allan deviation [18]
of the heterodyne beat frequency as introduced in Chapter 3. Performance of the free
93
running lasers is shown in Fig. 4.4(a), while that of the lasers locked to regenerative
spectral holes using the method of Ref. 12 is shown in Fig. 4.4(b). Shbrt term
stabilization giving a 500 Hz Allan deviation for a 2 ms integration time was
achieved with the conventional adjustment, of the phase sensitive error signal, that is,
with the detector signal and local oscillator in phase at the mixer as illustrated by the
(j) = 0° signal of Fig 4.3(a). Under these conditions, the dominant contribution to the
error signal comes from the spectral hole. Comparison of the curves in Fig. 4.4(a)
and Fig. 4.4(b) shows that slow frequency drift (integration times longer than
1 0 0 ms) of the stabilized lasers in these early experiments approached that of the
unstabilized laser. Much of this drift is attributable to slowly-varying DC voltage
offsets in the feedback servo loop arising from residual amplitude modulation at the
optical phase modulators, and temperature drift in the feedback electronics [1 2 ] and
modulators. The servo offsets cause the laser to lock off-center to the spectral hole
and consequently cause the regenerated spectral hole to drift continuously until the
drift in the voltage offset undergoes a change of sign. Further development of the
feedback servo system should substantially reduce these slowly varying offsets and
thus reduce this long-term drift.
94
A A a a a a o ° o
< 1 0 I
Time (s)
Figure 4.4 Allan deviation values for the beat between two lasers: (a) lasers free-
running (triangles), (b) locked to spectral holes in different crystals using straight
quadrature detection of the error signal at applied field B = 0.5 T (squares), and (c)
locked to spectral holes using the strategy of intermediate phase detection of the
combined error signal from the spectral hole and inhomogeneous line at applied
field B = 0.2 T (circles). Filled symbols, 300-sample Allan deviations measured
directly by a frequency counter; open symbols, values computed directly from beat
frequency data [cf. Fig. 4.5].
Incorporating the Absorption Line as a Fixed Referennr
A new locking strategy was devised, however, to reduce drift even with the
existing servo system by simultaneously exploiting the high resolution short term
frequency reference of the spectral hole and the long term stability of the
95
significantly broader inhomogeneous absorption line. Figure 4.3(b) illustrates the
variation of the error signal as the relative phase between the detector signal and the
local oscillator is varied through 90 degrees at the mixer, with the multiple sharp
features due to the narrow spectral hole superimposed on the broad background of
the inhomogeneous line. The relative position of these two contributions on the
frequency axis depends on the arbitrarily chosen position of the spectral hole. Crystal
transmission is increased at the frequency of the spectral hole, so the transmission
property of the hole and inhomogeneous absorption line have opposite signs. Their
pure absorptive and dispersive FM contributions also have opposite signs in simple
limiting cases. The shape of the FM signals, however, also depends on the
relationship between the modulation frequency vm and the width of the relevant
spectral feature. There is broad latitude for making this choice, since the hole width
~ 2 Fh can be from IO5 to IO8 times narrower than the inhomogeneous line width
Finh- In the present case, the line widths of the hole contribution and inhomogeneous
line contribution lie at opposite extremes relative to the modulation frequency:
Fh « vm « Finh. The two contributions to the error signal at line center are then
strongly phase dependent and are maximized for different quadratures - the
dispersive case for the spectral hole and absorptive case for the inhomogeneous
profile. These phases are represented as 0° and 90° respectively in Fig. 4.5(b) and
correspond to the S2 and SI phases.of Ref. 17. For a phase angle of (|) = O0 the
inhomogeneous line signal is negligible and the locking signal is derived primarily
from the spectral hole and for a phase angle of ([) = 90° the signal from the
96
inhomogeneous line is maximized, but the contribution from the spectral hole has
vanished at the line center of the hole. By choosing an intermediate phase
0°< itI13a (I) site (I) transition of Er3+W2SiO5 using a stabilized
1536 nm laser, an Er-doped fiber amplifier, and the echo apparatus to be described in
Chapter 5 and in Ref. 12. Approximately 5 mW of unmodulated continuous-wave
laser power was available for producing echo excitation pulses after continuous
101
wave amplification of the laser by the Er-doped fiber amplifier, which was located
outside the servo loop for laser stabilization, A portion of the un-amplified laser
output was used to frequency-lock the laser to a regenerative transient spectral hole
in the same transition as described above. Echo excitation pulses were produced
using two acousto-optic modulators in series to improve the on/off contrast ratio and
to cancel any net shift in the laser frequency, since the Er3+--Y2SiO5 spectral lines are
narrow. The resulting photon echo signal was gated from the transmitted beam by a
third acousto-optic modulator to discriminate against the exciting pulses. The echo
was detected with a fast New Focus .1811 InGaAs-photbdiode. To generate
stimulated photon echoes, three 2 p,s excitation pulses were incident on the crystal,'
with the delay ti2 between the first and second pulses fixed at 19 ps. The strength of
the stimulated echo was measured as a function of the delay t23 between the second
and third pulses.
With the laser frequency locked to a transient spectral hole, stimulated photon
echoes could be measured consistently for t23 delay times of several hundreds of
microseconds, giving the data in Fig. 4,6(b). The limiting factor for measuring
echoes with longer t23 delay times was the detector signal-to-noise ratio, rather than
laser frequency jitter, even though additional jitter may have been introduced by the
Er-doped fiber amplifier. After 800 |is total delay time the stimulated echo signal
was buried in the noise. In contrast, when the stimulated echo decay was measured
with the laser free running, the reproducibility of the stimulated, echo became
unreliable after only 200 p.s, as shown in Fig. 4.6(a).
St
im
. E
ch
o
St
re
ng
th
(
a.
u.
)
St
im
. E
ch
o
St
re
ng
th
(
a.
u.
)
102
Figure 4.6 Stimulated photon echo decay on the 4Ii5/2 (I) —> 4Ii3/2 (I) transition in
Er IY2SiO5. Each point represents a single shot, (a) Measured with a free running
laser, (b) Measured with a laser stabilized to a spectral hole and inhomogeneous
line at field B = 0.2 T.
103
All the data points of Fig. 4.6 were single-shot acquisitions of the stimulated
photon echo without thresholding to reject low-intensity echoes. Figure 4.6(a)
demonstrates that frequency jitter was the cause of the echo, signal amplitude
fluctuations when a free-running laser was used, since occasionally an optimum echo
was produced when the laser frequency of the third pulse happened to match that of
the first two. An envelope of “good” echoes can be seen, but most points fall well
below this. Clearly, averaging the data in Fig. 4.6(a) over multiple shots would lead
to a much different and erroneous echo decay rate.
Frequency response of a regenerative spectral hole
Regenerative SHB materials provide a fundamentally different type of frequency
reference from Fabry-Perot cavities or isolated atomic references, because the
incident laser probe field can modify the spectral hole reference. This is a
fundamental difference between ungated SHB frequency references and static
traditional frequency references such as the familiar Fabry-Perot cavity or atomic
references.
The interplay of the dynamics of materials and the stabilization process has been
investigated numerically and experimentally, and these results have been used to
optimize the system performance. [21,22] The spectral hole width and hole lifetime
are important parameters for the use of spectral hole burning as a means for laser
frequency stabilization. The hole width dictates the short-term performance of the
system, whereas over longer time scales the hole lifetime is important. When the
104
hole lifetime is increased for a specific hole width, the memory time of the spectral
hole is increased - it remains a valid frequency reference over longer periods.
Regenerative spectral holes are expected to exhibit a different frequency response
over the low frequency range up to the spectral hole width compared to static
frequency references, such as a Fabry-Perot cavity. This is because regenerative
spectral holes will lose their memory for times longer than the hole lifetime leading
to a roll off in gain relative to the cavity, which has a flat frequency response.
Methods and Apparatus
Experimentally the frequency response was measured in a two beam experiment
by probing a regenerative spectral hole that was burned with a laser stabilized to a
' i
spectral hole in the same transition in a different location of the same crystal
(Fig. 4.7). To generate a probe beam, part of the stabilized laser beam was split off
with a X/2-waveplate and polarizing beam splitter after the EOM, which was
operated at 109.5 MHz.
109.5 MHz
Mixer
Locking
detector
Cryostat
Crystal
Probe
detector
Noise sidebands
Mixer
Spectrum analyzer
ECDL Isolator
Lock in analyzer
Adjustable
phase delay
Low p a ss filter
Servo Amp.
Fast (current)
SIow(PZT)'
Figure 4.7 Experimental setup used to measure the frequency response
Er3+IY2SiO5.
or a regenerative spectral hole m
106
Both beams were adjusted to have the same spatial profile and optical power of
50 (iW therefore generating identical error signals by FM spectroscopy. The first
beam was used to lock to a regenerative spectral hole. The second beam was used as
a probe beam and modulated with a second EOM putting on noise sidebands with a
modulation index of 7t/42 radians; the optical power in the noise sidebands was
estimated at 220 nW per sideband. This second phase modulation was varied in
frequency to linearly probe the error signal response over the range of 100 Hz to
1.6 MHz. Both beams went through the same 0.001 at. percent Er3+IY2SiO5 crystal
(batch #1-544) spatially separated with a beam spot size radius of 1.07 mm
otherwise experiencing identical conditions. The crystal was oriented with its 6-axis
(2 mm) parallel to the lasers’ ^-vector, and its Braxis (12 mm) parallel to the
external magnetic field of B = 0.3 T. The laser light polarization was adjusted in both
beams using aX/ 2 wave plate to be parallel to the crystals 15 mm longD2 direction.
The crystal was immersed in a liquid Helium bath operated at a temperature of
T = 1.7 K. Generation of the noise EOM drive frequency and capture of the signal at
this frequency was accomplished by using two different techniques depending on the
frequency range being examined. This was required by the limited frequency ranges
of the different instruments. In the range DC-SOO kHz, the EOM was driven by the
Stanford Research Systems DS 345 function generator (24 dBm output into 50 Q.,
10 Vpp amplitude) and the signal monitored using an EG & G Princeton Applied
Research lock-in analyzer Model 5204. In the range 10 kHz -1.6 MHz, the Hewlett
Packard spectrum analyzerZ-tracking generator E4411 B (OdBm output) drove the
107
EOM with amplification provided by a Minicircuits ZHL-6A amplifier. This also
yielded a 10 Vpp amplitude signal at the EOM. The same spectrum analyzer
monitored the error signal. Sweeping the phase modulated noise sidebands over the
spectral range of interest allowed measurement of the phase-modulation spectrum of
the error signal. The frequency response of the error signal was obtained after
scaling the phase modulation spectrum by a factor of lnlf. Since the phase
modulation signal goes to zero at DC, the signal to noise ratio imposes a lower limit
on the measurable error signal response. The low frequency response of the hole can
be inferred beyond this limit by taking the ratio of the spectral noise density
measured relative to the regenerative spectral hole and that measured relative to a
static reference, in this case, a 150 MHz FWHM inhomogeneous line in isotopically
purified Er3+IliYF4. This assumes that the same noise is present in both
measurements. This measurement is the source of the experimental data in Fig. 4.8 at
frequencies below 300 Hz. The narrow features in the low frequency error signal
response are artifacts of this technique. The feature at -700 kHz can be attributed to
a piezo-electric resonance of the EOM.
Discussion
By employing the time domain model, which describes the material in terms of
the optical Bloch equations (developed by Dr. G. I. Pryde and C. W. Thiel and to be
described in detail elsewhere [22, 23]), it was possible to obtain the frequency
dependent response of the error signal, shown as open circles in Fig. 4.8. The
108
calculated AC-response of a Fabry-Perot cavity [24,25,26] is given by a dashed
curve for comparison; it has the shape of a low-pass filter. The cavity transmission
band of half width at half maximum (HWHM) - 16 kHz for this calculation was
chosen to have the same width as that of the spectral hole. Over the range of the
measurement, the model and experiment agree, and they are both in coincidence
with the response of the Fabry-Perot cavity.
. . .. . .. . .. . .. . .. . I 1 . . . . . . . . . .. . . . . . . . 1 1 r 1 1 1 1 1 1 I I T-TTTTTT]-- -- -- - 1 T T T T n r i ----- - 1— I I I l l l l | - - - - - - - -T
1 0 1 1 0 2 1 0 3 1 0 4 I O 5 1 0 6
modulation frequency (Hz)
Figure 4.8 Frequency response of the error signal generated from a regenerative
spectral hole frequency reference in 0.001 % Er3^ Y2SiO5; experimental
measurement (solid line) and using the model (open circles), compared with the
calculated response of a Fabry-Perot cavity (dotted line) of the same linewidth.
109
It is clear that the high frequency roll off is close to 20 dB / decade, which is ideal
for servo design since the accumulated phase shift from the reference is limited to
90°. At lower frequencies, the predicted roll-off of the gain is observed; consistent
with the idea that the system is losing its memory as a frequency reference because
of the limited hole lifetime. In principle, servo gain can partially compensate the low
frequency roll-off [27], but not low-frequency noise on the reference itself. Passive
reduction of this noise becomes essential for good long-term performance of
regenerative spectral hole burning references with short hole lifetimes. Of course,
linear drift can be removed after the fact, as with lasers stabilized to traditional
references [28]. A new technique has been devised that incorporates the
inhomogeneous absorption line itself as a fixed reference, dramatically improving
the long-term stability. However, this technique is only applicable to materials with
narrow < I GHz wide inhomogeneous absorption lines.
Laser frequency stabilization to regenerative spectral holes in Er3+:KTP
Potassium titanyl phosphate, KTiOPCL (KTP), is an important nonlinear electro
optic material primarily used as a frequency doubling crystal and for optical
waveguides. Here we report the achievement of 200 Hz laser stabilization utilizing
Er3"1": KTP with an Er3+ concentration of 0.004 at. percent as a spectral hole burning
frequency reference.
no
Methods and Apparatus
The Er3+:KTP crystal was kindly provided by Roger C. C. Ward of the
Clarendon Laboratory, University of Oxford, England. It was oriented with its a-axis
parallel to an external magnetic field and its b-axis parallel to the laser k-'vector. The
lowest energy 4Ii5Z2 —> 4Ii3/2 Er3+ transition has been observed to exhibit six distinct
sites near 1537 nm. [29] The site with the strongest absorption, located at
1536.87 nm (6506.69 cm"1), exhibits transient spectral hole burning by population
storage in the excited state of the optically active Er3+ ion; laser stabilization
experiments were performed using the absorption transition between the lowest
Zeeman split levels. Fig. 4.9(a) shows the transmission through the ~ 2 GHz
(FWHM) wide inhomogeneously broadened line for a magnetic field of B = 0.25 T
and temperature T = 1.9 K. The origin of the shoulder appearing at higher frequency
is most likely due to absorption from a spectrally similar site, as indicated by
fluorescence decay experiments. A narrow spectral hole, indicated by an arrow, has
been prepared by a second laser and can be placed anywhere within the
inhomogeneously broadened line. The homogeneous linewidth has been
characterized using two pulse photon echoes measured as a function of delay time
between the two excitation pulses. The measured dephasing time corresponds to a
homogenous linewidth of 2 -3 kHz for small magnetic fields below B = 0.4 T.
I l l
195025 195026 195027 195028 195029 195030 195031
Frequency (GHz)
P 100
O 75
Laser Frequency Offset (MHz)
Figure 4.9 (a) Transmission spectrum of 0.004 % Er3LKTP at 1537 nm showing
the entire, inhomogeneously broadened 4Ii5z2 -> 4Ii3z2 optical absorption scanned
by a diode probe laser. The arrow indicates a spectral hole, which has been
burned by a second laser, (b) Transmission of a phase-modulated probe laser
through a single spectral hole created by a second laser, using an applied
magnetic field of B=0.25T. (c) Demodulated FM-error signal derived from the
spectral hole in (b).
112
The experimental setup and techniques are similar to the previous demonstration
using Er3+: Y2SiO5. The lasers were modulated with modulation index M ~ 0.4 at
27 MHz and 30 MHz, respectively, values greatly exceeding the spectral hole width.
Fig. 4.9(b) shows the transmission through a spectral hole burned by a second laser
and Fig. 4.9(c) the corresponding demodulated FM error signal probed with a phase-
modulated laser; detector signal and local oscillator are in phase. To obtain a good
signal to noise ratio in the detection under low laser locking iiradiance, it is
beneficial to choose a large frequency reference crystal. Crystal dimensions were
12.87 mm along a, 14.89 mm along c, with a 5.18 mm optical path along b.\ both
lasers were locked to the same crystal. To avoid spatial overlap or interaction of the
two independently locked laser beams, a mask with separated 4 mm apertures was
placed over the crystal. A small magnetic field of B ~ 0.25 T has been conveniently
applied by sandwiching the sample holder between two permanent Nd-Fe-B magnet
disks of diameter 5 cm greatly simplifying the experimental apparatus. The mounted
crystal was immersed in liquid helium at T= 1.9 K. The entire two-laser optical
setup, including the cryostat, fit on a 3’ by 4’ optical breadboard, which was placed
on vibration damping foam on a commercially available pneumatically floated
optical table. An acoustical isolation enclosure box [30] made of medium density
fiberboard lined with W1 thick polyester embossed sound control mat covered the
experiment to provide passive acoustical isolation from ambient acoustic noise.
The laser iiradiance of ~ 100 pW/cm2 at the crystal was split from the laser beam
using a 1/2-plate / prism-polarizer combination, leaving most of the laser output
113
power to be used for experiments requiring a stabilized source or to saturate an
Erbium doped fiber amplifier for higher power applications. Using higher irradiance
at the locking crystal leads to a deeper spectral hole, which in turn becomes broader
due to stronger material absorption in the wings of the hole than at the hole center.
During active stabilization, each laser bums a spatially and spectrally separated
transient spectral hole into the inhomogeneously broadened absorption profile. Error
signal feedback to each laser leads to a continuous regeneration of the transient
spectral hole until a balance between spontaneous hole relaxation and hole burning
occurs.
The relative frequency stability of the two stabilized lasers was characterized by
the statistical Allan deviation [18] of the optical beat frequency since no absolute
frequency reference was available.
Results and Discussion
The Allan deviation for (a) the free running lasers and (h) actively stabilized
lasers is shown in Fig. 4.10. With the laser locked to transient spectral holes in
Er3"1": KTP an improvement in the Allan deviation over the free running lasers to
250 Hz has been achieved for integration times between I ms and 100 ms,
demonstrating the potential of Er3"1": KTP at the current stage of system development.
During quiet periods, Allan deviations of 200 Hzat 10 ms integration time have been
measured. The structure of the Allan deviation curve is reproducible.
114
Q 10
Integration Time (s)
Figure 4.10 Allan deviation for the heterodyne beat frequency between two
lasers: (a) lasers free-running, (b) independently locked to transient spectral
holes in the Ii5z2- A 3y2 transition in Er3+:KTP at 1537 nm.
Long-term drift of the laser, evidenced by an upturn of the Allan deviation at longer
integration times, has been greatly reduced by choosing an intermediate phase
setting at the mixer between detector signal and local oscillator. This technique
combines the excellent short-term stability of the spectral hole with the good long
term stability of the inhomogeneous line as described above; long-term drift of the
beat frequency between the two lasers has been suppressed to ~ 10 kHz per minute.
The demonstrated short-term stability is showing a clear improvement to the
115
previously reported results using transient spectral holes in Er3^ Y2SiO5 and is the
best result achieved in Er3+-doped compounds. As in the case of Er3+--Y2SiO5,
spectral diffusion plays, a major role in the achieved performance.[13,19] Even
though the transient spectral hole provides a narrow ~ kHz reference, the narrow
hole Width could not be fully exploited over longer periods because of the T%-limited
hole lifetime. In addition, the strong Er3"1" ion magnetic moments in the ground state
are perturbed by local field fluctuations due to electron spin flip-flops between
nearby Er3+ ions in the ground state. These perturbations lead to a broadening of the
homogeneous linewidth over time. The small magnetic field of B = 0.25 T applied
slows this spectral diffusion process by thermally depopulating the upper Zeeman
component of the Er3+-ion ground state but does not eliminate it completely.
Technical limitations are similar to the ones described above for Ef3+:Y2SiO5.
Environmental disturbances due to vibration and acoustical noise sources have been
partially addressed by floating the optical table and covering the experiment by the
acoustical enclosure described in the section; Methods and Apparatus; improvements
to the servo system and laser diode drivers have been made as well. Performance
without these modifications is indicated by the minimum 500 Hz Allan deviation
observed for Er3+Y2SiO5 in Fig. 4.4. [19]
116
Conclusions
The concept of laser frequency stabilization using regenerative transient spectral
hole burning has been extended to the technologically important 1.5 |am wavelength
region. Stable laser sources based on this method improve both spectroscopic
capability and the performance of SHB devices such as all-optical network routers
and address header decoders. Moreover, by stabilizing the laser source to the Same
SHB material that is already employed in the SHB device, system complexity is
significantly reduced. A hybrid method to control long-term frequency drift has been
demonstrated, using an intermediate phase delay in the phase sensitive detection of
the frequency locking error signal that exploits contributions from the narrow
spectral hole and from the inhomogeneous absorption profile. This stabilization
method is particularly well suited for spectroscopy and for optical data processing
devices based on time-domain spectral hole burning. Substantial improvement in
stimulated photon-echo reproducibility was demonstrated, showing the impact of
this technique on spectroscopy of rare earth materials.
As materials having gated persistent spectral holes with very long lifetimes are
developed, [31] the locking techniques reported here for regenerative SHB can be
applied to produce sources with long term stability and perhaps provide highly
portable secondary frequency standards.
117
Laser frequency stabilization to persistent spectral, holes in Er3+:D":CaFo
Here we report stabilization of external cavity diode lasers to persistent spectral
holes at 1523 nm in Er3+:D":CaF2 which we believe is the first demonstration of a
programmable frequency reference in the important 1.5 pm telecommunication band
based on persistent spectral holes. The results presented here extend the excellent
performance of spectral hole stabilized lasers to integration times of tens to hundreds
of seconds as indicated by a 5.7 kHz Allan deviation over 1600 s integration time,
opening up new regimes for spectroscopy and interferometry applications. Persistent
spectral holes in Er3+:D":CaF2 are qualitatively, different from persistent spectral
holes found in other materials due to their extremely long lifetime; no measurable
degradation of the persistent holes in Er3+:D":CaF2 has been detected over 48 hours
in the absence of continuous laser radiation, which suggests that the lifetime could
be indefinitely long. A reference frequency programmed as a persistent spectral hole
in Er3LD :CaF2 could potentially be used as a long-term Secondary frequency
standard in a scheme where a flywheel oscillator infrequently probes the persistent
spectral hole to maintain its long-term stability.
Methods and Apparatus
Er3+:D":CaF2, to our knowledge, is the only material known to exhibit persistent
spectral hole burning [32] in the spectral region at 1.5 pm. The crystals used in these
118
experiments have a 0.05 atomic percent Er3"1" concentration and were prepared by
Glynn D. Jones at the University of Canterbury, Christchurch, New Zealand. The
introduction of deuterium (D ) ions into Er34" doped CaF2 crystals leads to a
substitution for F" ions on interstitial or lattice sites in the vicinity of the Er3"1" centers,
giving rise to additional absorption lines. Among these, the R center, one of the
strongest multiple D'-ion centers, has been shown to exhibit persistent hole burning
on the 4Ii 5/2 —> 4Ii 3/2 Er3"1" transition located at 1523 nm and is used here for this laser
fl
frequency stabilization demonstration. [32] The spectral hole burning mechanism is
known to involve excited D ion migration into nearby interstitial sites [33]. Spectral
holes of ~ 40 MHz (FWHM) have been measured to be fully persistent for at least
forty-eight hours. That together with the high activation energy of the hole burning
process implies that the hole lifetime could be indefinitely long if the sample is held
at liquid helium temperature. Thermal cycling of the material leads to a full recovery
of the original hole shape depending on the temperature reached and time held;
shallow holes burned at 1.7 K were largely recovered after thermal cycling up to
70 K for 10 minutes. These properties make Er3+:D":CaF2 an interesting candidate
for a secondary programmable transportable frequency reference since the
absorption profile can be modified in a controlled way. For instance laser beat
frequencies or information can be recorded in the crystal with arbitrary frequency
separations in the form of persistent spectral holes burned anywhere within the
absorption profile and read out at a later time or location as long as the sample
temperature and pressure is maintained. Fig. 4.11 (a) shows a transmission spectrum
119
of Er3+:D":CaF2 and illustrates a number of spectral holes that have been stored into
the inhomogeneous absorption line of the R center. The use of this material as a
frequency reference is feasible up to a temperature of 30 K where the hole width
reaches 320 MHz (FWHM). Commercial closed-cycle cryocoolers can readily cool
below 10 K, supporting the practical operation of materials like Er3+:D":CaF2 without
cryogenic fluids.
It should also be pointed out that the R-site hole burning mechanism is a property
of the host, rather than being dopant-specific, providing greater wavelength
versatility by replacing the Er3+ ions with other suitable rare earth ions, such as
Tm3+, Pr3+ [33],
Laser frequency stabilization was implemented as described above. The lasers
i
were externally modulated with electro-optic phase modulators. Custom-built
resonant tanks were constructed by the author (see Appendix A), which allowed
tailoring the modulation frequency and modulation index to the specific material
parameters. The primary sidebands chosen had a modulation index of M = 0.4, and
the modulation frequencies of 93 MHz and 109.5 MHz for the two lasers were
chosen to lie outside the ~ 40 MHz FWHM of the spectral holes used as frequency
references. Fig. 4.11(b) shows a transmission spectrum through a single spectral hole
burned into the R absorption line and the corresponding demodulated FM error
signal (Fig. 4.11(c)).
Er
ro
r S
ig
na
l (
a.
u.
)
In
te
ns
ity
(a
.u
.)
In
te
ns
ity
(a
.u
.)
120
Laser Frequency Offset (GHz)
Laser Frequency Offset (MHz)
Figure 4.11 (a) Transmission spectrum of Er3+:D":CaF2 at 1523 nm. A number
of spectral holes have been burned into the inhomogeneously broadened
L5/2 —> Ii3/2 optical absorption for demonstrating the programmability of the
material. Spectral hole burning is not limited to the center of the line. The
arrow indicates a spectral hole, which is enlarged in (b). (c) Demodulated FM-
error signal derived from the spectral hole in (b).
121
To demonstrate high performance under simple conditions, the entire setup was
implemented on a 3? by 4’ tabletop breadboard placed on an unfloated optical table
without any acoustical isolation.
The two lasers were independently stabilized to spectral holes in separate
reference crystals; each crystal was masked to avoid any coupling by stray light from
the other beam and was immersed in a single liquid Helium Oxford Optistat bath
cryostat held at T= 1.9 K.
Since width and depth of the programmed persistent spectral hole determine the
slope of the error signal for active laser frequency stabilization, careful preparation
of the initial persistent spectral hole is important. An increase in hole width as a
function of bum time was observed due to hole burning centers having only a partial
frequency overlap of their homogeneous linewidth with the laser, which leads to a
lower effective transition probability and therefore hole, burning on a longer time
scale. [34] Also, the hole depth increases and eventually saturates as a function of
bum time due to the finite number of centers at the laser frequency; as the persistent
spectral hole bums deeper its width increases due to earlier saturation in the center
than in the wings of the hole.
A tradeoff between hole depth and hole width has to be made in preparing the
initial persistent spectral hole. Good results were achieved in using persistent
spectral holes of FWHM ~ 40 MHz, prepared by illuminating the sample for
20 seconds with incident light intensities of ~ 300 pW/cm2. Since the short-term
laser linewidth. of the free running laser is much less than the narrowest spectral hole
122
width in this material no active stabilization was engaged in preparing the initial
spectral hole. The laser irradiance was controlled using a polarizing beam
splitter/half-waveplate combination.
The persistent spectral hole in each crystal was recorded at a similar frequency to
allow the beat frequency between the two lasers to lie well within the 125 MHz
bandwidth of the beat signal detector; otherwise the beat frequency can be chosen
arbitrarily by the relative frequency between the two spectral holes. Laser irradiance
enters into the locking stability in two opposing ways. High irradiance leads to good
signal to noise ratios on the locking detectors but modifies the spectral hole shapes
through continued hole burning; this degrades the reference by making the holes
broader and deeper especially when the locking is inadvertently off center relative to
the hole. To alleviate this trade off, large samples (~ 9 mm diameter) were chosen,
allowing the use of a 4.8 mm beam diameter to preserve the signal to noise ratio
needed for the servo feedback loop and to minimize continuous modification of the
spectral hole. For active stabilization, the laser illumination was reduced by one
order of magnitude to ~ 30 pW/cm2 to minimize hole burning during locking. The
laser frequency stability was characterized as described before.
Results and Discussion
Results for the achieved laser frequency stability are shown in Fig 4.12, with the
Allan deviation for the beat between (a) the free running lasers contrasted (b) with
that for the spectral hole stabilized lasers. The free running lasers show frequency
123
stability comparable or even better than similar commercially available systems and
are already sufficient for many applications in spectroscopy. With the laser locked to
~ 40 MHz (FWHM) wide persistent spectral holes, an improvement in the Allan
deviation over the free running lasers of at least one order of magnitude has been
achieved for integration times longer than 2 ms; for integration times longer than
300 s the improvement reaches more than three orders of magnitude.
(a) Z
Integration Time (s)
Figure 4.12. Allan deviation for the heterodyne beat frequency between two
lasers: (a) lasers free-running, (b) independently locked to persistent spectral
holes in the 4Ii5/2 —> 4Ii3/2 transition in Er3+: D":CaF2 at 1523 nm
124
Sub-kilohertz Allan deviations down to 680 Hz have readily been reproduced over a
wide range of integration times without requiring vibration isolation of the laser or
the crystal frequency reference. This demonstrates laser frequency stabilization to a
persistent spectral hole to better than 6 parts in IO5 of the ~ 40 MHz hole width or
better than 3 parts in IO12 of the optical frequency. The measured Allan deviation
curve is reproducible.
Characterization and optimization of spectral hole burning references by
experiment and simulation [21,22] show that the Allan deviation over short
integration times is determined by the width of the spectral hole leading to a high
signal to noise ratio and consequently good short-term stability, whereas stabilization
over long integration times is determined by the lifetime of the spectral hole. The
Er3+:D":CaF2 material used for this demonstration is not ideal, since the width of the
spectral holes (~ 40 MHz) is much wider than found in many other rare earth doped
material systems and is therefore limiting the short-term stability. The long lifetime
of the persistent Er3+:D":CaF2 spectral holes leads, however, to a good long-term
stability demonstrated in Fig. 4.13, where a subset of the beat frequency change over
10 minutes between the (a) free running and (b) actively stabilized lasers is shown.
Laser frequency drift has been reduced to less than I kHz per minute. The achieved
stability demonstrates the relative immunity of the spectral hole frequency reference
to environmental disturbances like vibrations and acoustical noise.
It should be noted that the initial spectral hole becomes broader and bums deeper
over time during continued laser locking illumination. This results in a reduced
125
precision of the frequency reference over long stabilization times. To overcome this
problem a hybrid locking technique could be considered where the short-term
stability is given by a flywheel oscillator, stabilized to a regenerated transient
spectral hole or Fabry-Perot, and long term drift is controlled by comparing its
frequency to the frequency of the programmed spectral hole.
Figure 4.13 Subset of the change in heterodyne beat frequency between (a) free
running and (b) independently locked lasers to persistent spectral holes in
separate crystals over a period of 10 minutes; the lowest trace is an expanded
view.
126
Technical limitations in our demonstrated laser frequency stability are similar to the
ones described in the case of Er3+IYaSiOs locking. Already significant improvements
over longer time scales have been made since locking to persistent spectral holes in a
similar material, Tm3+ID-1ICaF2. [11] has been reported.
Conclusion
In conclusion, we have extended the use and performance of persistent spectral
holes as programmable laser frequency references to the important 1.5 pm optical
communication window, achieving sub-kilohertz laser frequency stability over broad
time scales. A compact laser frequency stabilization that utilizes rather inexpensive,
low maintenance external cavity diode lasers has been demonstrated. The system is
not limited to low optical power applications, since Erbium doped fiber amplifiers or
injection locked high power laser diodes can be used to boost the output power.
Other sources like DFB-Iasers can be stabilized using this technique.
Suitable spectral hole burning materials for laser frequency stabilization and
spectral hole burning based optical processing are under constant. development
spanning a wide wavelength range, facilitating the stabilization of other sources and
proving the versatility of the technique. Materials combining very narrow few-
hundred-Hertz-wide holes [35] with long lifetimes of several weeks and minimal
spectral diffusion will provide good stability over a wide range of integration times.
127
Permanent hole burning references with gated spectral holes under development will
provide portable compact secondary frequency standards.
128
REFERENCES
I. R. J. Jones, J.-C. Diehls, I. Jasapara, and W. Rudolf, Opt. Comm. 175, 409 (2000);
S. A. Diddams, D. J. Jones, J. Ye, S. T. Cundiff, J. L; Hall, J. K. Ranka, R. S.
Windeler, R. Holzwarth, T. Udem, and T. W. Hansch, Phys. Rev. Lett. 84, 5102
(2000).
2. N. W. Carlson, L. J. Rothberg, A. G. Yodh, W. R. Babbitt, and T. W. Mossberg,
Opt. Lett. 8, 483 (1983); M. Zhu, W. R. Babbitt, and C. M. Jefferson, Opt. Lett.
20, 2514 (1995); K. D. Merkel and W. R. Babbitt, Opt. Lett. 24, 172 (1999)..
3. T. R. Dyke, M. J. Sellars, G. J. Pryde, N. B. Manson, U. Elman, and S. Kroll, J.
Opt. Sci. Am. B 16, 805 (1999); G. J. Pryde, M. J. Sellars, and N. B. Manson,
Phys. Rev. Lett. 84, 1152 (2000); M. J. Sellars, G. J. Pryde, N. B. Manson, and
E. R. Krausz, J. Lumin. 87, 833 (2000).
4. T. L. Harris, Y. Sun, W. R. Babbitt, R. L. Cone, J. A. Ritcey, and R. W. Equal!,
Opt. Lett. 25, 85 (2000).
5. K. D. Merkel, R.W. Petersj P. B. Sellin, K. S. Repasky, and W. R. Babbitt, Opt.
Lett. 25,1627 (2000).
. 6. X. A. Shen and R. Kachru, Opt. Lett. 20, 2508 (1995); T. Wang, H. Lin, and
T. W. Mossberg, Opt. Lett. 20, 2541 (1995).
7. T. L. Harris, Y. Sun, R. L. Cone, R. M. Macfarlane, and R. W. Equall, Opt. Lett.
23,636 (1998).
8. R. G. DeVoe and R. G. Brewer, Phys. Rev. Lett. 50, 1269 (1983); R. M.
Macfarlane and R. M. Shelby in Spectroscopy of Crystals Containing Rare-Earth
Ions, edited by A. A. Kaplyanskii and R. M. Macfarlane (North-Holland,
129
Amsterdam, 1987); M. J. Sellars, R. S. Meltzer, P. T. H. Fisk, and N. B. Manson,
J. Opt Soc. Am. B 11, 1468 (1994).
9. T. W. Mossberg, Opt. Lett. 7, 77 (1982); A. Rebane, R. Kaarli, P. Saari, A.
Anijalg, and K. Timpmann, Opt. Commun. 47, 173 (1983); M. Mitsunaga, R.
Yano, and N. Uesugi, Opt. Lett. 16, 1890 (1991); M. Mitsunaga, N. Uesugi, H.
Sasaki, and K. Karaki, Opt. Lett. 19, 752 (1994); H. Lin, T. Wang, and T. W.
Mossberg, Opt. Lett. 20, 1658 (1995).
10. Z. Cole, T. Bbttger, R. Krishna Mohan, R. Reibel, W. R. Babbitt, R. L. Cone,
and K. D. Merkel to be published. Accepted for presentation at Conference of
Lasers and Electro Optics (CLEO), Long Beach, CA, May 2002.
11. P. B. Sellin, N. M. Strickland, I. L. Carlsten, and R. L. Cone, Opt. Lett. 24, 1038
(1999).
12. N. M. Strickland, P. B. Sellin, Y. Sun, J. L. Carlsten, and R. L. Cone, Phys. Rev.
B 62, 1473 (2000).
13. R. M. Macfarlane, T. L. Harris, Y. Sun, R. L. Cone, and R. W. Equall, Opt. Lett.
22,871 (1997).
14. Y. Sun, R. L. Cone, T. L. Harris, R. M. Macfarlane, and R. W. Equall, Photon
Echo Study of Optical Dephasing in Er3+-doped Oxide Crystals at 1.5 pm. Bull.
Am. Phys. Soc. 43, 526 (1998) and to be published.
15. K. Liu and M. G. Littman, Opt. Lett. 6, 117 (1981); P. McNichol and H. J.
Metcalf, AppL Opt. 24, 2757 (1985), K. C. Harvey and C. J. Myatt, Opt. Lett. 16,
910 (1991).
16. R. W. P. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A. J. Munley,
and H. Ward, Appl. Phys. B 31, 97 (1983); M. Zhu and J. L. Hall, J. Opt. Sci.
Am. B 10, 802 (1993).
130
17. G. C. Bjorklund, M. D. Levenson, W. Lenth, and C. Ortiz, Appl. Phys. B 32, 145
(1983); J. M Supplce, E. A. Whittaker, and W. Lenth, Appl. Opt. 33, 6294
(1994).
18. D. W. Allan, IEEE Trans. UFFC-34, 647 (1987).
19. T. Bottger, Y. Sun, G. J. Pryde, G. Reinemer, and R. L. Cone,. J. Lumin. 94-95,
565 (2001).
20. N. C. Wong and J. L. Hall, J. Opt. Sci. Am. B 2, 1527 (1985); F L. Walls, IEEE
Trans. UFFC-34, 592 (1987); R. W. Fox, L. D'Evelyn, H. G. Robinson, C. S.
Weimer, and L. Hollberg, Prop. SPEE Conference 2378, 58 (1995).
21. G. J. Pryde, T. Bottger, and R. L. Cone, J. Lumin. 94-95, 587 (2001).
22. G. J. Pryde, T. Bottger, and R. L. Cone, J, Lumin., accepted for publication
(2002).
23. G. J. Pryde, T. Bottger, C. W. Thiel, and R. L. Cone, and M. J. Sellars, to be
published
24. M. Zhu and J. L. Hall, J. Opt. Soc. Am. B 10, 802 (1993).
25. M. J. Sellars, PhD thesis, Australian National University, Brisbane, Australia
(1995)
26. O. Mbr and A. Aiie, IEEE J. Quant. Elect. 33, 532 (1997).
27. J. L. Hall, M. S. Taubman, and Jun Ye, Handbook of Optics, 2nd edition,
McGraw-Hill, New York 2001, Vol. IV, Chapter 27.
131
28. B. C. Young, F. C. Cruz, W. M. itano, and J. C. Bergquist, Phys. Rev. Lett. 82,
3799(1999).
29. G. Reinemer, Y. Sun, R. L. Cone, and R. C. C. Ward, submitted to Phys. Rev. B.
30. D. Hils, J. E. Faller, and J. L. Hall, Rev. Sci. Intrum. 57, 10 (1986)
31. C. W. Thiel, H. Cruguel, H. Wu, Y. Sun, G. J. Lapeyre, R. L. Cone, R. W.
Equall, and R. M. Macfarlane, Phys. Rev. B 64, 085107 (2001), C. W. Thiel, H.
Cruguel, H. Wu, Y. Sun, G. J. Lapeyre, R. L. Cone, R. W. Equall, and R. M.
Macfarlane, Optics-&-Photonics-News 12, no. 12; Dec. 2001; p.64.
32. N. M. Strickland, R. L.Cone, and R. M. Macfarlane “Persistent spectral
holebuming in deuterated CaF2: Er3+”, Phys. Rev. B, submitted for publication.
33. R. M. Macfarlane, R. J. Reeves, and G. D. Jones, Opt. Lett. 12, 660 (1987); R.J.
Reeves and R.M. Macfarlane, J. Opt. Soc. Am. B 9, 763 (1992); N.M Strickland,
R. L. Cone, and R. M. Macfarlane, Phys. Rev. B 59,14328 (2000).
34. N. M. Strickland, R. L. Cone, and R. M. Macfarlane, Phys. Rev. B 59, 14328
(1999).
35. R. W. Equall, Y. Sun, R. L. Cone, and R. M. Macfarlane, Phys. Rev. Lett. 72,
2179 (1994).
132
CHAPTER 5
SPECTROSCOPY AND DYNAMICS OF Er3+--Y2SiO5
Introduction and Motivation
The optical material Er3+IY2SiO5 has recently received considerable attention for
spectral hole burning applications in the important 1.5 pm optical communication
band. This material has also been investigated for solid-state laser applications [1]
because of its good chemical and thermo-mechanical properties as well as the
potential for high rare-earth doping. In 1997, Macfarlane et al. [2] measured the first
two-pulse photon echoes and observed dephasing times up to 580 ps, triggering a
number of important SHB proof-of-principle device demonstrations. These
demonstrations were motivated by the desire to develop SHB technologies that
would enable all-optical memory, switching, and processing at communication
wavelengths [3, 4], These potential SHB devices would store information encoded in
optical pulse sequences by employing stimulated photon echoes. With this approach,
the maximum temporal length of a recorded data stream is limited by the material
coherence lifetime, T2, requiring long coherence lifetimes for some practical device
applications. Even with continuous programming and accumulation [5, 10], T2
remains an important parameter. The importance of T2 for SHB applications
provides a motivation for fundamental research to fully explore the parameter space
that influences the coherence lifetime so that optimized materials may be developed.
The material YzSiOs is well known as an excellent host for achieving ultraslow
optical dephasing at low temperatures. [6] These extremely long T2 values are
expected because the constituent elements in Y2SiO5 have small magnetic moments
(- 0.137 pN for 89Y) or small natural abundance of magnetic isotopes (4.7 % with -
0.554 Pn for 29Si, 0.04 % with - 1.89 Pn for 17O). [2] By minimizing the magnetic
moments in the host lattice, dephasing due to nuclear and electronic spin fluctuations
can be dramatically reduced.
In any Er3+ material, the primary sources for homogeneous broadening are
magnetic interactions between the large electronic magnetic moments of the erbium
ions. Lowering the Er3"1" concentration can minimize these.
The complex magnetic dipole interactions between both electronic and nuclear
spins have been the subject of research over the last six decades in the context of
nuclear magnetic resonance (NMR) and electron paramagnetic resonance (EPR) as
well as optical experiments. [7, 8, 9j. In the literature, ions, and their corresponding
spins, are commonly divided into two groups, where the ions directly being probed
in the experiment are referred to as the A-ions, and the remaining ions present in the
environment, but not directly probed, are referred to as the B-ions. To achieve long
dephasing times, it is necessary to suppress the spin-flip broadening caused by the
environment B-ions and the optically active A-ions. Motivated by the need to
minimize spin-flip broadening, we have characterized the spin-flip induced spectral
133
134
diffusion. The approach taken in this work is to characterize the effects of erbium
ion concentration, magnetic field strength and orientation, and temperature on the
spectral diffusion to develop a systematic strategy to optimizing the SHB material.
This chapter presents the conventional and nonlinear spectroscopy of
Er3+:Y2SiC>5 relevant to SHB applications. Early sections cover the site-selective
spectroscopy to determine the crystal field level structure and to identify spectral
hole burning transitions at 1.5 pm. Crystal field levels were mapped out for the 4I15z2
ground state and the 4I15z2 excited state. Time-resolved fluorescence spectroscopy
was used to measure the excited state lifetime, which determines the maximum
achievable coherence times in the material. Full rotational Zeeman experiments Were
performed in the three separate optical planes of the crystal to determine the ground
state and excited state g values for both sites, revealing a preferred direction for
applying the external magnetic field.
Coherent nonlinear spectroscopy was used to determine material properties not
accessible by conventional methods, such as the coherence lifetime, T2, and its
spectral domain counterpart, the homogeneous linewidth, Thom. Two-pulse photon
echoes were measured as a function of magnetic field orientation in the optical
extinction planes and supplemented the Zeeman experiments. The time evolution of
the homogeneous linewidth due to spectral diffusion was studied with stimulated
photon echo spectroscopy as a function of magnetic field, erbium concentration,
temperature, and crystal orientation. The spectral diffusion observed in Er3+:Y2SiO5
can be successfully described in the. framework of established theories [18] with the
135
explicit inclusion of a direct phonon process driving the Er3+ spin-flips, as presented
in detail in appendix C.
The spectroscopic studies described in this chapter led to significant material
optimization for SHB applications while improving our fundamental understanding
of these materials. This has enabled (to the best of our knowledge) the measurement
of the narrowest optical resonance in a solid with a homogeneous linewidth of 1
Thom = 73 Hz. Photon echoes are reported at elevated temperatures of up to 5 K,
which is particularly important to facilitate operation with mechanical cryo-coolers
to eliminate the requirement for liquid Helium. The SHB device potential of
Er3+IY2SiO5 was demonstrated for a high bandwidth 0.5 GHz correlator at T = 4.2 K
[10].
The crystals were grown by Scientific Materials Inc. of Bozeman, Montana using
the Czochralski method. The Er3+- concentrations varied between 0.001 and 2 atomic
percent. The crystal Y2SiO5 belongs to the space group C^h with eight formula units
per monoclinic cell. The Y3+ ions -occupy two crystallographically inequivalent sites
of Ci symmetry [11] and the Er3+ ions substitute for Y3+ host ions without charge
compensation. All crystals were transparent and appeared colorless. The material
Er3+=Y2SiO5 has 3 mutually perpendicular optical extinction axes, the 6 -axis
equivalent with the (0 1 0 ) direction, and the Dj and D2 axes correspond to optical
extinction directions when the sample is viewed along (0 1 0 ) between crossed
polarizers. [1] All crystals were oriented, cut and optically polished perpendicular to
the three optical extinction axes.
136
Conventional spectroscopy
Spectroscopy on Er3+^ S iO s has been previously reported by Li et. al. [1] on
highly doped (up to 10 at. percent Er3+ concentration) powder and single crystal
samples for temperatures between T= IOE arid 300 K in the context of solid-state
laser material development. Absorption and emission measurements yielded 16
crystal field components of the 4Ii5z2 ground state for the two crystallographical sites
and 14 components of the 4Ii5z2 excited state. Crystallographic site identifications
reported by Li et al. are uncertain. Their assignments were made according to line
strength and assuming that both sites have similar energy level structure slightly
shifted with respect to each other without the use of site-selective spectroscopy.
For that reason, broadband absorption experiments and site-selective
fluorescence experiments were used to resolve this issue. Absorption measurements
located the excited-state crystal field levels, and site-selective fluorescetice
confidently assigned the excited and ground state level structure to each
crystallographic Er3+ site. A number of the previous assortments have been
superceded.by our more detailed and precise measurements.
Methods and Apparatus
Figure 5.1 shows the experimental setup used for absorption and site-selective
fluorescence measurements.
JL I X / l l ------
Halogen v >
Lamp
Detector
Cryostat
ECDL
Wavemeter
Computer
Spectrometer
Oscilloscope
Figure 5.1 Experimental setup for broadband absorption and site-selective fluorescence.
138
For absorption, a 2 % Er3+IY2SiO5 crystal (batch#0-84) was immersed in a
liquid Helium bath held at T= 1.95 K using an Oxford Instruments SpectroMag
cryostat with four windows providing optical access. The chosen Erbium
concentration proved ideal to avoid over-absorption. Sharp absorption lines were
detected. The crystal was aligned with its 6 -axis (2 mm) parallel to the light ^ -vector.
Other crystal dimensions were 3 mm along D1 and 4 mm along D2. For absorption
measurements, an f = 150 mm lens (not shown) focused an image of a 55 W
tungsten filament inside the crystal. A color filter (not shown) was placed in the
beam path between the filament and cryostat with band-pass between 950 nm and
2900 nm to eliminate visible light that could be transmitted at second order by the
spectrometer. A Glan-Thompson polarizer (not shown) selected the light
polarization. An f = 150 mm lens (not shown) collimated the light transmitted
through the crystal. The light was then focused with an f = 2 0 0 mm lens (not shown)
on the entrance slit of a SPEX model 1000M spectrometer. The spectrometer
grating had 600 lines / mm blazed at 1.5 pm, giving 1.6 nm / mm dispersion at the
exit slit. Entrance and exit slit width were 2 0 pm, giving a bandpass of 0.32 A or
0.14 cm'1. The exit slit of the spectrometer was imaged onto either a liquid Nitrogen
cooled Hamamatsu NIR photomultiplier tube (PMT) Model R 5509-72 or a liquid
Nitrogen cooled Advanced Detector Corporation Gemanium detector Model 403 L
(not shown). The detector signal was further amplified using the vertical plug-in of
an analog Tektronix 7904 oscilloscope, digitized by a 12-bit National Instruments
Model AT-MIO-16F-5 analog to digital converter (not shown) and recorded with a
139
data acquisition computer that also Controlled the spectrometer grating position.
Signal averaging was accomplished by reading the analog to digital converter for a
fixed number of times, typically 1Q0 0 , at each setting of the spectrometer.
Interference fringes caused by the Germanium detector window as well as a rich
absorption spectrum of water lines partially obscured the recorded absorption
spectra. Water molecules in the beam path absorbed light in the spectral region of
interest resulting in sharp absorption lines superimposed on the Er3"1" absorption lines.
The water lines conveniently allowed calibrating the measured absorption spectrum
using the water line atlas. [12] The interference fringe background was reduced
using a wedge-shaped YAG window in place of the original and was removed using
a software routine. [13] Final calibrations of the spectra were verified using the
narrowband ECDL as a marker in the spectrum whose wavelength was determined
with the Burleigh WA 1500 wavemeter to very high precision..
Operating at T = 1.95 K ensured that only the lowest crystal field level Zi of the
4I i 5/2 J-multiplet was initially populated. Absorption measurements mapped the
crystal field levels of the 4Ii3/2 J-multiplet. The o3.bserved absorption spectra contain
two interspersed sets of absorption lines from both Er3"1" sites, with each site
experiencing slightly different crystal field splittings in the Y2SiO5 host.
In order to unambiguously assign the observed absorption lines to a specific
crystallographic site, and to site-selectively map out the energy levels of the 4I1572 J-
multiplet, site-selective fluorescence experiments were performed. In these
experiments, a narrowband laser was used to individually excite each absorption
140
line. The fluorescence from the 4Ii372 to the 4Ii572 level carries a distinct signature
depending on whether it was excited from an absorption line belonging to site I or
site 2. Two distinct fluorescence spectra were observed when individually pumping
all absorption lines with the narrowband ECDL, and this allowed each crystal field
level of the 4Ii572 ground state and 4Ii372 to be assigned to site I or site 2. For these
experiments, a 0.001 % Er3+: Y2SiO5 crystal (batch # I -544-Top) was aligned with its
Di axis parallel to the laser k vector. The fluorescence was collected from the b-D2
plane at an angle of 90° with respect to the laser beam. Crystal dimensions were
3 mm, 5 mm, and 4 mm along b, Di , and D2. Monitoring the ECDL'wavelength with
a Burleigh WA 1500 wavemeter ensured site-selective pumping. Laser absorption of
the individual 4Ii372 levels was also verified by scanning the ECDL over the
respective absorption line and detecting the transmitted intensity with a New Focus
Model 1811 InGaAs PIN-photodiode. The fluorescence spectra were obtained by
scanning the spectrometer and recording the detector signal with the data acquisition
computer.
Results and Discussion
Figure 5.2 shows a polarized E HD2 absorption spectrum at T= 1.95 K for the
2 % Er3^ Y2SiO5 crystal. Water lines and interference fringes have been removed
from the spectrum for clarity. The spectrum contains very sharp well-separated
absorption lines; 13 out of the possible 14 lines of the 4I1372 J-niultiplet were
observed. Line centers were determined by fitting the absorption lines.
20
6599.14
™ 2% Er^YS iO , ,T = 1.95K
(I) (2) 2 5
6549.82
E//D2,k/ /b
6800.036726.97
6498.68
6752.56 6871.68
Wavenumber (cm^1)
Figure 5.2 Polarized E H D2 lamp absorption spectrum of 2% Er3+IY2SiO5 at T= 1.95 K. Water lines and interference
fringes have been removed for better visibility. Absorption line centers are given in wavenumbers and numbers in
parenthesis show the crystallographic site assignments from site-selective fluorescence experiments (see text). Laser
site-selective excitation was not possible above 6700 cm'1.
142
Energies in wavenumbers are indicated above each line and site assignments are
given in parentheses. The levels above 6700 cm"1 could not be selectively assigned
due to limits of the site-selective fluorescence experiment, described below.
Figure 5.3(a) shows the fluorescence spectra excited from the 4Ii3z2JYi level of
site I at a temperature of T = 10 K. Figure 5.3(b) shows that all of the eight possible
lines for each site were clearly identified. Fluorescence line centers were obtained
from a Gaussian fitting routine and are labeled in the graph. Due to the limited
tuning range of the ECDL, it was not possible to pump levels above 4In^Y 4, so
levels above 6700cm"1 could not be assigned to a site using this method.
The crystal field level structure of Er3+IY2SiOs determined from absorption and
site-selective fluorescence experiments is shown in Fig. 5.4. The level structure
obtained by Li et al. [1] is given for comparison. The ground state 4I15z2 crystal field
levels have been labeled Z1-Zg and the 4I12z2 excited state crystal field levels with Y1-
Y4; higher lying Yn assignments are undetermined. The large discrepancies of the
levels Li et al. determined in their simpler experiment are evident. Differences can
be ascribed to the superior spectral resolution, experimental conditions, calibration
techniques, and sample quality used in our experiments. Spectral lines in absorption
and fluorescence are sharp, well-separated and clearly resolved, while site-selective
fluorescence experiments allowed unambiguous site assignment. Table 2
summarizes the crystal field level structure of the Er3+IY2SiO5 4I15z2 and 4I13z2
manifolds obtained in this work.
R
el
at
iv
e
In
te
ns
ity
(a
.u
.)
R
el
at
iv
e
In
te
ns
ity
(a
.u
.)
143
Selective excitation o f fluorescence for site I
0.001 %Er3+:Y2SiO5, T = I O K
laser @ site I: 6508.39 cm'1
6508.39
6469.02
6424.09
6028.77
6338.67' 67 6405.93
A__JU5998.81 6088.82
6100 6300
Wavenumber (cm"1)
6471.54Selective excitation o f fluorescence for site 2
0.001% Er3+:Y2SiO5 , T = 10 K
laser @ site 2: 6498.14 cm"1
6498.14
6435.296184.98
6372.38
6151.53
X 3.5 6330.496086.15
6100 6200 6300 6400 6500
Wavenumber (cm"1)
Figure 5.3 Site selective fluorescence spectra of 0.001 % Er3+IY2SiO5 at
T=IOK; linecenters are given in wavenumbers, (a) Laser excitation from site I
4Ii3/2’Yi, (b) Laser excitation from site 2 4Ii3/2:Yi.
en
er
gy
(c
m
1)
144
6900-
6800-
6700-
Er +: Y2SiO5 energy levels
this work Li et al. this work Li et at.
- f - ..........-cl
. ^
-------- flj.____ , 7
-...I..Y -CS 6
■M
6600-
6500-
500 -
400 -
300 -
2 0 0 -
100 -
0 -
Y„
I
Y.
Y„
Y„ Y„
- Z,
I 13/2
15/2
Site I Site 2
Figure 5.4 Crystal field levels of 4Ii5z2 and 4I1372 raultiplets of Er3+IY2SiO5 as
determined from absorption and site selective fluorescence excitation for site I
and site 2. The right columns show values from Li et al. for comparison.
145
Table 2 Crystal field levels of Er3+^ S iO s as determined from absorption and site
selective fluorescence experiments.
Label Site I Site 2 unassigned
4Iisza energies (cm"1)
Y4 6625.08 6622.51
6871.68
6852.95
6800.03
6752.56
6726.97
Y3 6599.14 6594.50
Y2 6549.82 6565.65
Yi 6508.39 6498.14
4Iisza energies (cm'1) Z8 509.58 411.99
Z7 - ■ 479.62 346.61
Z6 419.57 313.16
Z5 169.72 167.65
Z4 102.46 125.76
Z3 84.3 62.85
Z2 39.37 26.6
Zl ; 0 0
Lifetime Measurements
This section describes investigations of the fluorescence dynamics of the
metastable 4Iisza: Yi excited state for Er3"1" in both crystallographic sites. Knowledge
of the excited state lifetime, Ti, is important since it establishes an upper bound for
the coherence lifetime, T2, given by T2 < IT1. In the experiment, a laser pulse was
used to pump the 4IiszaiZi to 4Iisy2=Y2 transition of a single site. Rapid nonradiative
relaxation from Y2 to Yi within the 4Iisz2 J-multiplet of that site results in
fluorescence occuring from 4IiszaiYi to 4Iisy2=Zi-Zg, This fluorescence intensity was.
146
recorded as a function of time with the PMT and spectrometer that was tuned to
individual ground state levels. The 4Ii5/2.' Zi to 4Iisza-' Y2 transition was pumped to
allow the laser and fluorescence signal to be spectrally , separated.
Methods and Apparatus
Figure 5.5 shows a schematic of the experimental setup used to investigate the
fluorescence lifetime of the metastable 4Ii3z2 state. The same Er3+I Y2SiO5 crystal and
orientation was used as described earlier for fluorescence experiments. The low
0.001 % Er3+ concentration minimized radiative energy trapping effects that can
artificially lengthen fluorescence decay times.
The ECDL light source provided 1.8 mW of single-frequency light that saturated
the output of an ILX Model FOA-8100 Er fiber amplifier (EDFA) at 35 mW. A
Crystal Technology Model 3165-1 acousto optic modulator (AOM) gated I ms
duration pulses at IOHz repetition rate from the amplified laser beam to excite
fluorescence. A X/2-plate (not shown) in front of the cryostat defined the orientation
of the linear polarization. A PTS-500 RF synthesizer (RF-source) amplified by a
40 dB Minicircuits ZHL-5W-1 amplifier drove the AOM at 165 MHz. Laser pulses
were obtained by switching the RF-power to the AOM with a Watkins-Johnson SI
RF-switch controlled by a Hewlett Packard model 8013B pulse generator at the RF
switching threshold. The required pulse sequence was provided by the HP pulse
generator, triggered by a Stanford Research model DG 535 delay generator.
Cryostat
Crystal
Detector
AOMEDFAECDL
RF-switch
Wavemeter
RF-source RF amplifier
generator
Pulse
generator
Computer
Spectrometer
Oscilloscope
' Figure 5.5 Experimental setup to measure the 4I i3z2 fluorescence lifetime for Err3+IY2SiO5.
147
148
Fluorescence from the crystal was collected at 90° to the laser beam geometry
and focused onto the spectrometer entrance slit with the spectrometer tuned to the
respective 4Iisz2: Zn transition. The spectrally resolved light leaving the spectrometer
exit slit was detected by the PMT in a time-resolved fashion. The signal from the
PMT was captured on a Tektronix TDS 520 D digitizing oscilloscope triggered by
the DG 535 delay generator. A Burleigh WA 1500 wavemeter continuously
monitored the laser wavelength. The transmitted intensity through the crystal was
monitored with a New Focus 1811 InGaAs photo detector to verify absorption lines
for pulsed excitation. For this purpose, the ECDL was scanned and the AOM
operated in continuous wave mode.
Results and Discussion
Figure 5.6(a) shows the measured 4ImrYi -> 4Ii5z2IZi fluorescence decay for
site I and Figure 5.6(b) for site 2, both at a temperature of T = 10 K. The observed
fluorescence decays were exponential over several decades and exponential least
squares fits to the data are shown as solid lines. A fluorescence lifetime of
Ti = (11.44 ± 0.01) ms was obtained, for site I, whereas Ti = (9.20 ± 0.01) ms was
obtained for site 2 from the exponential fit. Exciting the Zj-Y2 inhomogeneous line
on the low energy side yielded a 0.2 ms shorter lifetime. A decrease of 0.5 ms in
lifetime was observed by increasing the sample temperature to T = 40 K.
149
0.001 % Er3+IY2SiO5, T = 1 0 K
w 0.01
SitelQ) 1E-3
T1 = (11.44 ± 0.01) ms
OC 1E-4
S 0.01
Site 2> 1E-3
(9.2 ±0.01) ms
2.0x10' 4.0x10 6.0x10 8.0x10 2 1.0x10
Time (s)
Figure 5.6 Fluorescence lifetime decay for 0.001% Er3+IY2SiO5
iInn(Xx) —> T 15/2 (Z1) transition at T=IOK; (a) site I, (b) site 2. Straight
lines correspond to exponential least square fits to the data.
150
The lifetimes were unaffected by varying the spatial location of excitation (edge
versus center).
In contrast, pronounced variations in the fluorescence lifetimes of a 10 %
Er3+:Y2SiO5 crystal were observed by Li et but without distinguishing
between crystallographic sites. Li et al. observed a very strong geometrical effect by
moving a slit in front of the crystal along the excitation direction and found a linear
increase of the lifetimes from 10 ms to 17 ms with increasing excitation path length.
Measuring the fluorescence temperature dependent lifetime in 1% Er3+: Y2SiO5, they
obtained variations between 13 ms at T= 12 K and 8 ms at T = 300 K. Li et al.
attributed these variations to a radiative energy trapping process, during which the
emitted photon is reabsorbed (trapped) and then reemitted, slowing down the overall
observed fluorescence. Increasing the crystal temperature reduces the absorption
coefficient, and therefore reduces the probability of trapping but it also changes the
population of the individual levels in the 4Ii3/2-multiplet. Increasing the excitation
path length increases the number of ions participating in the reabsorption process,
which enhances the probability of trapping. The large spread in Li et al. results does
not allow a direct comparison to our measurements.
The ultra-low Erbium concentration of 0.001% used in our experiments
minimized effects associated with trapping as the small variation in fluorescence
lifetime with temperature and excitation energy suggest. Hence, the values of
11.4 ms for site I and 9.4 ms for site 2 provide an accurate result and establish an
upper bound for the fluorescence lifetime of the 4Ii5/2:Yi level in Er3+:Y2SiO5.
151
Stimulated photon echo T-decay experiments discussed in the nonlinear
spectroscopy section set a lower limit for the fluorescence, lifetime by measuring a
Ti of 9.8 ms for site I.
Zeeman experiments
Practical operation of Er3^ YzSiO5 in SHB devices demands the application of an
external magnetic field in order to obtain long dephasing times. A magnetic field
lifts the Kramers degeneracy by splitting each doubly-degenerate crystal field level.
The level splitting, characterized by the g value, is a key variable influencing the
microscopic dynamics in Er3^ Y2SiO5. Maximizing the level splitting relative to the
thermal energy, kT , can “freeze out” the thermal population in the upper Zeeman
level and thus reduce dephasing due to electron spin fluctuations of neighboring Er3+
ions. Zeeman spectroscopy is used to determine the relevant ground and excited state
g values that influence the optical dephasing. Controlling the level splitting is an
important part of the material optimization strategy.
Implementing such a strategy was made difficult by the extraordinary complexity
encountered in Er3^ Y2SiO5, which has two distinct crystallographic sites that both
exhibit low Ci site symmetry. This complexity results from the anisotropic level
splitting of individual Kramers doublets and magnetic inequivalence for each
crystallographic site. Hence, Zeeman experiments were carried out as a function of
magnetic field orientation to find a direction that simultaneously maximizes g for all
152
levels important for optical dephasing. The results of these measurements led to the
identification of a preferred magnetic field orientation that minimized dephasing,
and thus optimized Er3+IY2SiO5 for practical SHB-device applications.
As we discussed in the previous section, the crystal field partially lifts the 2/+T
degeneracy of the free ion, yielding up to /+1/2 crystal field levels. Because of
Kramers degeneracy, all crystal field levels determined in the previous section are
doubly degenerate and have first order magnetic moments. The 4I1572 ground J-
multiplet consists of 8 Kramer’s doublets and the 4I1572 excited state J-muliplet
consists of 7 Kramer’s doublets. Application of an external magnetic field lifts the
remaining degeneracy, and each crystal field level splits into two.Zeeman sub-levels.
Figure 5.7 schematically shows the Zeeman splitting for the ground (g) 4I1572: Z1 and
excited state (e) 4I1572: Y1 levels, where g and e label the ground and excited state
and + and - the upper and lower Zeeman components. The letters a, b, C, and d
indicate the four different optical transitions that are possible between the levels. The
g-factors for the ground gg and excited state ge can be found from the transition
energies using
Se =
(Ea-E h)+ (En- E d)
2/4,# (5.1)
and
£g 2/j.bB (5.2)
Transition energies, a, usually given in wavenumbers, can then be obtained from
153
(5.3)
(5.4)
where g is the particular g value for the site of interest, a 0 is the zero field transition
energy between the 4Ii5^ Z i -> 4Inh=Y] crystal field levels, jU B is the Bohr magneton
and B is the magnetic field strength.
c rM “ c rO ± ' ~ ( ^ + S e x c ' t B
aOtC = crO ± - ^ ( g s ,
4 113/2 - Yi
e 6+ ^
— < l
< ■>V
9 SeVnB
a
9 ^
C
>
d
' t
g+ i
4 115/2 ^ Zi g' SgP1BB
T
Figure 5.7 Transition labelling scheme for Zeeman laser absorption
experiments.
154
Application of a magnetic field splits each Kramer’s doublet differently,
characterized by an anisotropic g-tensor unique to each doublet. For small magnetic
fields, the splitting scales linearly with the applied magnetic field strength. This can
be described by an effective spin Hamiltonian of the form
particular levels and crystal site symmetry. The g-tensor is a symmetric tensor of
rank 2 that has six independent components in general. [14] In the principal axis
representation, the components correspond to the lengths of the three major axes and
the three angles that determine the orientation of the axes with respect to the
crystallographic site. For high symmetry sites, crystal symmetry dictates the
orientation of the g-tensor and the number of independent components can be
dramatically reduced. The Er3+IY2SiO5 system, however, is of Ci crystal site
symmetry and no simplification is possible. Each crystal field level has distinct
values of gx, gy and gz that have to be specified by 3 angles relative to the local axes
of the crystallographic site. The Zeeman Hamiltonian of equation (5.5), rewritten in
components of these principal g-tensor values and components of the external
magnetic field, becomes
g 2 ), (5.5)
with S as the effective spin (S = ±’/2), and g the g-tensor determined by the
(5.6)
155
It is customary to write the projections of the applied magnetic field B on the
principal g-tensor axes. In the case of the two crystallographic sites of E r3+Y2SiO5,
none of the principal g-tensor axes lies within a crystallographic plane or optical
polarization extinction plane; such as D1-D2, B-D1 and b-D2 Of the crystal. Hence, a
magnetic field constrained to lie in one of these planes selects the projection of the
principal g-tensor axes onto this plane. Figure 5.8 illustrates the projections for the
case of the magnetic field B lying in the b-Dj plane with gx and gy the principal axes
of the ellipse.
D 1
....
.....
Figure 5.8 Projections of the magnetic field B onto the primary g - tensor axes
for the case of B lying in the B-D1 plane. The angle (Preferences the magnetic
field B to the 6-axis and the angle .areferences the gx -axis to the 6-axis.
156
The angle O references the magnetic field B orientation' to the 6 -axis, and is
measured in the experiment. The angle or references the g^-axis to the 6 -axis and is
unknown. The projection of B onto gx and gy follows from trigonometry as
and allows writing g in terms of these projections. From Zeeman experiments, g is
obtained as a function of e+
J-
6509.00
0.5 1.0 1.5 2.0
Magnetic Field (T)
Figure 5.11 Laser absorption Zeeman spectra for 0.001 % Er3"1": YaSiOs as a
function of magnetic field for B H D1 and & // £ at T = 10 K. (a) Sample Laser
absorption scans across site I as the magnetic field is varied, subplots have been
shifted for better visibility, (b) Zeeman transitions for site I with linear fits to the
data between B = OT and B = 2 T. (c) Zeeman transitions for site 2 with linear
fits to the data between B = OT and B = 2 T.
162
Zeeman Experiments as a Function of Field Orientation
Because the Zeeman level splitting is a key variable in achieving slow
dephasing, we needed to determine the g values for both crystallographic sites as a
function of applied magnetic field direction. The goal was to find a magnetic field
direction where the g values of both crystallographic sites are simultaneously
maximized in the ground as well as excited state. A splitting, large compared to the
thermal energy, kT, reduces thermal populations in the upper Zeeman level and
suppresses spectral diffusion by Er3"1" spin-flips.
As stressed earlier, the situation in Er3+-iY2SiO5 is complex due to two
crystallographic sites with both exhibiting a low Q site symmetry. Each
crystallographic Er3"1" site has multiple orientations in the unit cell. When a
distinction between individual site orientations is impossible without an applied
magnetic field, the Crystallographic site orientations are equivalent. However,
applying an external magnetic field to the crystal along an arbitrary direction
removes the orientational equivalence because the magnetic field makes different
angles with the two different sets of local site axes. Even if the field is parallel to a
local axis of symmetry for one ion, in general, it will not be parallel for the others
and the crystallographic site breaks up into two magnetically inequivalent
orientations. Each of these orientations allows observation of four transitions
according to Fig. 5.7. Magnetic inequivalency was observed when the magnetic field
was oriented in the b-Di or A-D2 plane. Eight lines for each crystallographic site
were observed corresponding to transitions within two magnetically inequivalent
163
orientations. When the magnetic field was applied in the optical Di-D2 plane, both
crystallographic sites were magnetically equivalent, leading to 4 transitions for each
crystallographic site. Identifying and tracking up to 16 individual transitions as the
magnetic field direction (angle
=O0 and
2
'i '
•2 line
- > -
- o -
- O -
- O -
-A— d, - line
• b„ - line
Angle (degrees)
Figure 5.13 (a) Orientational-dependent g values of magnetically inequivalent orientations of site 2 in the B-D2 plane
determined from data of (b), ggi denotes the g value for the ground state of orientation I, ge 2 the g value for the excited
state of orientation 2 etc. Solid lines are fits to the data; deviations are due, to a misalignment of the sample (see text),
(b) Transition frequencies for all possible orientations of site 2 in the B-D2 plane.
0.005 % E r+:Y2SiO5, site I, k // D1, B = 0.5 T, T = 5 K
12
10
O 8
o
a 6
o) ^
2
0
■ 195200
X 195160
CD
>, 195120
c .
CD=S 195080
CT
2
U- 195040
14
■ I ' I ' I I I
D ^2g
V
-p— C2 4 line
b1 - line
-a— a2 - line
- v — Cl1 - line
-o— c. - line
60 80 100 120 140 160 180
Angle 0 (degrees)
Figure 5.14 (a) Orientational-dependent g values of magnetically inequivalent orientations of site I in the B-D2 plane
determined from data of (b), gig denotes the g value for the ground state of orientation I, g2e the g value for the excited
state of orientation 2, etc.; solid lines are fits to the data, (b) Transition frequencies for all possible orientations of site I in
the B-D2 plane.
S10
8
6
4
2
12
0
195200
195160
1
CD
c .
195080
CT
2
LJL 195040
0.005 % Er :Y2Si05, site I, k // D2, B = 0.5 T, T = 5 K
195120
-4—I—i—I—i— h —I—H —i—I—i—I—I— b
D 3Ze
0 9Zs
^ %=
v SfI3
- D ^ a 1 - line
—°— a2 - line
—4— b1 - line
—v— c - line
60 80 100 120 140 160 180
Angle 0 (degrees)
Figure 5.15 (a) Orientational-dependent g values of magnetically inequivalent orientations of site I in the b -D j plane
determined from data of (b), gig denotes the g value for the ground state of orientation I, g# the g value for the excited
state of orientation 2, etc.; solid lines are fits to the data, (b) Transition frequencies for all possible orientations of site I in
the b -D i plane.
D ffl.
° 5f1e
A %
— D—
- O —
- A —
-V —
a2 - line
S1 - line
C1 - line
20 40 60 80 100 120 140 .160 180
Angle 0 (degrees)
Figure 5.16 (a) Orientational-dependent g values of magnetically inequivalent orientations of site 2 in the A-D7 plane
determined from data of (b), gig denotes the g value for the ground state of orientation I, g%. the g value for the excited
state of orientation 2 etc.; solid lines are fits to the data, (b) Transition frequencies for all possible orientations of site 2 in
the b -D i plane.
0.005 % E r+:Y2SiO5, site I, k //b , B = 0.5 T, T = 5 K
° SfIe
D Sfl9
OO
- A ^ a - line
—v—c - line
—D— b - line
— d - line
0 20 40 60 80 100 120 140 160 180
Angle
t 195120
c 4Iisz2: Yi crystal field levels. The optical Er3Non center is
surrounded by neighboring Er3+ environment ions in the ground state. Mutual
spin flip-flop transitions and even more severe single spin-flip transitions of
environment ions will cause a shift of the crystal field levels at the location of
the Er3Non optical center resulting in dephasing of the optical center ion.
To investigate the likelihood of the mutual spin flip-flop interaction, it is
worthwhile to estimate the average Er3+-JEr3+ ion distance, the corresponding
magnetic dipole field strength, and the resulting frequency shift experienced at the
optical center when a neighboring ion flips. The magnetic field magnitude in
spherical coordinates from a magnetic dipole is given by [15]
5 = - ^ y ("2cos© r+ s in 0 © ,
Anr L -I
(5.8)
175
with the magnetic dipole moment, m, the permeability of the material, //, and the
distance to the origin, r. The maximum field strength is along r when 0 = 0 with a
value
_ /nm
~ In rz ‘ (5.9)
The average magnetic field, 5 , for a fixed distance, r, can be calculated by taking
the absolute value and averaging over the angle, 0 , from 0 to 7t/2. This calculation
involves a complete elliptical integral that is tabulated. The result is
|B| = 0.771x£max. (5.10)
Using m = — gpB, with the g-value for the Er3"1"
2
ion in the ground state, g, and the
Bohr mageton, equation (5.10) becomes
LB| = 0 .357x4lT -A ], (5.11)
r
where r is in units of Angstroms [A] and the field, Bmax, is in units of Tesla [7],
Assuming g = 6.1 (as measured for BHD1), the magnetic field at a distance of
r = I A from an Er3+-ion becomes B = 2.18 T. However, due to the low Er3"1"
concentration used in our crystals, neighboring Er3+ ions will be much further apart.
To estimate the average distance between neighboring Erbium ions in a dilute 0.005
atomic percent Er3+: Y2SiOs crystal, consider the monoclinic unit cell of Y2SiOg with
dimension a = 10.419. A, b = 6.726 A, c = 12.495 A, and (3 = 102.63° [11], where (3
defines the angle between a and c, giving a unit cell volume of V = 854.44 A3. The
176
unit cell contains 16 Y3+ ions giving a Y3+ ion density of 1.87x 1022cm"3 . Assuming
each Y3+ ion occupies the center of a cube with side-length, /, the Y3+ ion-ion
distance will be given by the dimension, I, of the cube. In general, the average
distance can be estimated using
\
(5.12)I =
y
^NvcJ
with Nv, the number of sites in the volume, V, and c, the percentage of occupation of
these sites. The unit cell contains 16 Y3+ ions, so that the average Y3+ ion-ion
distance will be approximately/ = (854A3 /16)1/3 =3.11 k . For a 0.005 atomic
percent Er3^ Y2SiO5 crystal, the Y3+ site is only 0.005 % occupied by Er3+ions. The
Er3+ ions substitute for Y3+ ions with equal site occupation into two crystallographic
Er3"1" sites. Assuming that only ions belonging to the same crystallographic site
16contribute to dephasing gives Nv = — = 8. This assumption is justified, as we will
2
see later, for the preferred magnetic field direction and the average distances
between neighboring Er3+Tons according to (5.12) will be 129A. Hence using
relation (5.11), a neighboring Er3"1" ion causes a magnetic field of B -LOxlO-6T at
the location of the optical center, and a single environment ion spin-flip in the
ground state induces a frequency shift, k v = pi-kB /h, at the optical center of
Ay s 43 kHz . Indeed, stimulated photon echo spectroscopy at small magnetic fields
in the long waiting time limit, described later, measures homogeneous linewidths of
this magnitude. Because neighboring ions are -129 A apart and the magnetic dipole-
177
dipole interaction falls off with I / r6, mutual spin flip-flops are not expected to be
significant for this concentration. Even for the higher erbium concentrations used in
our experiments, such as 0.02 % and 0.1 %, the spin flip-flop process is expected to
. be weak.
Figure 5.20 shows the “direct” phonon scattering process. The direct process
gives rise to single phonon emission or absorption with phonon energies equal to the
ground state Zeeman splitting. [16] Absorption of a phonon promotes a spin from the
lower Zeeman level into the upper Zeeman level, and spontaneous emission of a
phonon causes a spin in the upper level to relax to the lower Zeeman level. Due to
the low cryogenic temperatures, the direct phonon process is expected to be the
dominant interaction that can cause spin-flips of Er3+ environmerit ions. The two-
phonon processes involve higher-energy phonons. Stimulated photon echo
spectroscopy, described later in the chapter, supported this argument.
Optical Er3+- ion
Figure 5.20 Schematic of the direct phonon process in the Zeeman split Er3"1"
ground state, (a) Phonon absorption with energy of the Zeeman splitting raises a
spin to the upper Zeeman level, (b) Spontaneous phonon emission relaxes a spin
to the lower Zeeman level.
178
The g values measured describe the magnetic field strength and directional
dependence of ground state splitting, AE5,, and excited state splitting, AE„. In our
optimization strategy to suppress dephasing’ we required the energy splittings, AE1, to
be as large as possible with respect to the thermal energy, kT, available to the ions.
In addition to maximizing the level splittings, a magnetic field orientation that
ensures magnetic equivalence of Er3+ ions is important. In general, all Er3+ ions play
a role in dephasing optically excited Er3+ ions. However, for a device application
only those ions that are resonant with the laser are useful. In the best case, all ions
are in resonance with the laser, which makes them magnetically equivalent.
Magnetic equivalence minimizes the overall doping concentration required for
optical absorption and increases the overall Er3+-Er3+ inter-ion distance, thereby
reducing interactions. In addition, a reduced number of site orientations simplifies
the anisotropic g value-patterns providing an easier guide for a “good” magnetic
field orientation. Of course, in a real material system, these optimization strategies
are often mutually incompatible.
Following these optimization strategies, the magnetic field should be applied in
the D1-D2 plane of the crystal because both crystallographic sites remain
magnetically equivalent upon rotation of the magnetic field in that plane. Site I is
the preferred site in this plane because it exhibits higher optical absorption (see
Fig. 5.2) and longer excited state lifetime (see Fig. 5.6 (a)). From the orientationally
dependent g-values of the ground and excited states shown for site I in Fig. 5.17 (a),
it is clear that maximum splitting occurs for both ground and excited state when the
179
magnetic field is at an angle (B jDj) = QSpi k l l b
6520-
L ,■ Y 9.44 cm"6515-
6510-
6505-
6500- L-: Y 0.66 cm"
resonant
energy
transfer
14.31 cm"L ,: Z
L - : Z. 3.58 cm"
Sitel Site 2
Figure 5.24 Energy level structure of Er3^ Y2SiO5 at a magnetic field of B = 3 T
with angle (P= 95° between B and Di. Laser excitation of the b-line at
6510.82 cm'1 of site I will cause a resonant energy transfer between site l ’s c-line
at 6496.52 cm"1 and site 2’s c-line at 6496.68 cm'1 and d-line at 6496.02 cm"1.
190
Two-pulse photon echo decay measurements were also used to determine the
homogeneous linewidth for orientation 2 of site I in the b-D2 plane summarized in
Fig. 5.23. The observed echo strengths for the field in this plane were considerably
weaker and dephasing times shorter due to the higher Er3+-Concentration of 0.005%
along with the magnetic inequivalency of the ions and the consequent lack of a
distinct magnetic field orientation to optimize dephasing. The structured linewidth
versus magnetic field direction behavior, compared to the D1-D2 plane, was due to
the different angular g-value dependences of the magnetically inequivalent ions.
However, even in this complex picture, a clear correlation could be verified between
g-values and homogeneous linewidth. The narrowest homogeneous linewidth of
4.4 kHz was measured for B along b where g-factors of all orientations were
relatively large. In contrast, no echo signal was detectable when the g 2g and g 26-
values of orientation 2 of site I were at their minima and the site 2 g-values were
also small. Site 2 ions also contributed to dephasing as the local maximum of
linewidth at
0.1
S-t—»
CO
O
JZO
LU
. i o .o i4—'
CO
5 10 15 20 25 30 35 40 45 50
t12-Delay (|4S)
Figure 5.25 Stimulated photon echo decays in 0.02 % Er3^ Y2SiOs at a
magnetic field of B = 1.75 T at 1.6 K as the waiting time T is varied between
0 jjs (equivalent to a two-pulse photon echo decay) and 5000 (is. Echo
decays for short waiting times are non-exponential indicating the existence of
spectral diffusion,. longer waiting times yield exponential decays and
observed decay constants don’t change anymore. Solid lines are “Mims” fits
using relation (2.5) and show excellent agreement with the data;
homogeneous linewidths were extracted from the decay constants using
relation (2.6).
0.02 % Er3+:Y2S i0 5
T = 1.6 K, B = 1.75 T
T = 0 [is
T = 10 ] is
T = 20 its
T = 50 JLtS
T = I 00 (js
T = 350 (js
T = 5000 |os
195
* (
0.02 % Er IY2SiO,
T = 1.6 K
o B = 0.8 I
A B = 1.5 T
v B = 1.75 I
o B = 2 T
< B = 2.25 T
> B = 3 TZr 20
B =2.25 T
V B = ST
100 1000 10000
I (.US)
Figure 5.28 Evolution of the homogeneous linewidth of site I in 0.1 %
E r+IYaSiOs at 1.6 K as the waiting time, T, between pulses two and three is
varied in a stimulated photon echo measurement. The magnetic field has
been changed from B = 0.8 T to B = 3 T between subplots. Increasing the
magnetic field strength suppresses spectral diffusion. Solid lines are least
square fits to the data using relation (5.13); each case shows good agreement.
198
0.001 % Er : Y SiO
O B = 0.8 T
A B = IT
o B = 1.5 T
> B = 1.75 T
□ B = 2 T
v B = 3 T
1000 10000
T (gs)
Figure 5,29 Evolution of the homogeneous linewidth of site I in 0.001 %
E r+:Y2SiOs at 1.6 K as the waiting time, T, between pulses two and three is
varied in a stimulated photon echo measurement. The magnetic field has been
changed from B =0.8 T to B = 3 T between subplots. Increasing the magnetic
field strength suppresses spectral diffusion. Solid lines are least square fits to
the data using relation (5.13); each case shows excellent agreement.
199
Figure 5.27 shows spectral diffusion in the same 0.02% sample at a fixed
magnetic field as the temperature was varied between T = 1.6 K and T = 2.1 K. Solid
lines in both figures are least squares fits to the data using relation (5.10). Each case
shows excellent agreement.
Measurements were also carried out in 0.1% and 0.001% Er3+^ Y2SiO5 crystals
and are summarized in Fig. 5.28 and 5.29. Solid lines in both figures are least
squares fits to the data using relation (5.10). Each case shows excellent agreement.
Discussion Cf Spectral Diffusion in 0.02%
Er3+: Y2SiO5 as a Function of Magnetic Field
Significant broadening of the homogeneous linewidth, as the waiting time, T,
J
was lengthened beyond 10 jis, as shown in Fig. 5.26, is an indication of spectral
diffusion. Clearly, a larger magnetic field suppressed homogeneous linewidth
broadening and a plateau was reached after more than 100 Lis, where the
homogeneous linewidth saturated. Solid lines in Fig. 5.26 represent least squares fits
to a variation of the model proposed by Bai and Payer [18] with the functional form
F(F) = F0+F1 [l-exp (-l? r)] . (5.13)
The parameter, Fq, is the homogeneous linewidth as measured by two-pulse photon
echoes. Fitted parameters include Fi, the saturation value of the linewidth for large
waiting times, T, and R, the rate of perturbations experienced by the ions due to the
environment; each case shows good agreement. Expression (5.13) was derived under
200
the simplifying assumption that dephasing only occurs during the waiting time, T.
The data of Fig. 5.26 justify this assumption. A detailed step-by-step derivation of
the model relevant to our particular system is given in appendix C.
The Bai-Fayer model is based on dephasing induced by localized two-level
perturbers, which can undergo, sudden jumps between the levels. This model was
directly applicable to our situation. Neighboring Er3+ ions in the ground state
represent the localized perturbers and sudden jumps between the levels correspond
to spin-flip transitions. The sudden jumps were treated with coupled two-level rate
equations, whose solution provided an expression for the population deviation from
thermal equilibrium. In thermal equilibrium, the number of spin-up transitions equals
the number of spin-down transitions and, hence, does not affect the energy levels of
the optical center. In the model, it is the deviation of the population from thermal
equilibrium that causes dephasing. If the spin population deviates from thermal
equilibrium, there will be an excess of spins that produce a deviation from the
equilibrium field experienced by the Er3"1" ions.
In the present model, the population between the two levels at thermal
equilibrium was treated with the Boltzmann distribution according to relations
(C.13) and (C.14) in. appendix C. The model predicts saturation of the linewidth
when all spins deviating from thermal equilibrium have flipped. The saturated
linewidth value, F1(B1T ), for long waiting times, T, was derived in equation (C.19)
of appendix C and is related to the size of the average deviation of the population
from equilibrium (variance) as
201
F1CBT) = FmaxSech2
IkT 1’ (5.14)
where g is the g value of the ground state, the Bohr magneton, B the magnetic
field, k the Boltzmann constant, T the temperature, and Fmax an empirical constant
that relates spin, flips to average frequency shifts, The ratio between the ground state
energy level splitting, BJEg - g/uBB , and the thermal energy available to the ions,
kT , determines the saturation level of the homogeneous linewidth, F1. If the
splitting is large compared to the thermal energy, deviations of the spin population
from the thermal equilibrium will be small and therefore only contribute to a small
spin-flip background at the optical center.
The rate, R, in (5.13), is the sum of the spin-up and spin-down transition rates
and describes how fast the spin population returns to thermal equilibrium. No
assumptions about the rate, R, were made in the original Bai-Fayer model This
information had to be explicitly incorporated into the present model taking into
account the mechanism driving the spin-flip transitions. From the previous
discussion, the one-phorion (direct) process was a likely candidate. The relaxation
rate for a direct process applied to Kramers ions as a function of magnetic field and
temperature is given by [16]
R (B ,T ) = R (gjUB)3B 5 coth
I t r )
(5.15)
202
This rate increases rapidly with increasing applied magnetic field due to an increase
of the accessible density of phonon states degenerate with the splitting of the
Zeeman levels.
Field dependent linewidth saturation values, Fi(B), and relaxation rates, B(B),
could be obtained from the fits shown in Fig 5.26 and have been plotted with the
respective error bars in Fig. 5.30. Figure 5.30(a) depicts the rate, B, as a function of
magnetic field. The solid line is a least squares fit to the magnetic field dependent
rate, B(B), using equation (5.15) with an added offset, Bo; it shows excellent
agreement. Fitting parameters are given in the figure. It should be noted that the
exact measured g-value of g = 6.1 for B H D1 was used in the fit. The constant offset
rate, Ro = 2 kHz, physically corresponds to an additional less significant mechanism
driving spin flips, independent of the applied magnetic field strength. Several
interpretations are possible. So far, the nuclear spin fluctuations due to the Y3+ host
ions have not been taken into consideration. It can be verified that a nuclear spin flip
of a neighboring Y3+ ion produces a frequency shift of the optical center ion
comparable with an Er3+Spin flip for these concentrations. The Y3+ nuclear spin-flip
rates, however, are on the order of 10-20 Hz [19], values too slow to be measured
over the time scale of our stimulated photon echo experiment, and they cannot
account for the observed offset, Rq. The Y3+ spin-flips are expected to make
significant contributions only at times longer than our longest waiting time, T. Two-
phonon processes that depend only weakly on the magnetic field, such as the Raman
and Orbach processes, could contribute to the offset.
203
0.02 % Er :Y SiO
R (B) = R0 + R(g(iB)3B5 coth(gnBB/2kT)
Two-Parameter Fit
R0= (2.35 ± 0.44) kHz
g = 6.1 (fixed)
Magnetic Field B (T)
0.02 % Er : Y SiO
r I (B) = r maxSeCh2(9ftB/2kT)
One-Parameter Fit5 150 Fma= (704 ±60) kHz
g = 6.1 (fixed)^ 1 0 0
Magnetic Field B (T)
Figure. 5.30 (a) Relaxation rate, R, as a function of magnetic field as measured in
0.02 % Er3+:¥ 2 8 1 0 5 at 1.6 K. Solid lines are least square fits using the expression
(5.15) and show excellent agreement; fitting parameters are given in the figure,
(b) Saturated spectral diffusion linewidth, F1, for long waiting times, T, as a
function of magnetic field in 0.02 % Er3"1":Y2SiOs at 1.6 K. Data points Were
fitted to expression (5.14) and show good agreement. Fitting parameters are
given in the figure.
204
The temperature dependence of these processes predicts that they should be very
weak at T = 1.6 K. [20, 21, 22] Since only a few data points are available at low
fields, additional measurements giving supplemental data points would be desirable,
along with independent measurements of the spin-lattice relaxation rate, for example
by electron paramagnetic resonance, to resolve this issue.
Figure 5.30(b) shows the corresponding saturation value, Ti, as a function of
magnetic field strength, B. The solid line is a least squares fit to the data using
relation (5.14) showing good agreement. Error bars on the data points for small
magnetic fields are large because the onset of saturation was not clearly observed for
those fields. Note that the exact measured g-value for this magnetic field orientation
was used in the fit. Clearly, as discussed above, a Zeeman, splitting large compared
to the thermal energy “freezes-out” thermal populations in the upper Zeeman level of
the ground state; deviations of the spin population from the thermal equilibrium
become small and therefore only contribute to a small spin-flip background at the
optical center.
Discussion of Spectral Diffusion in 0.02%
Er3*: Y7SiOs as a Function of Temperature
'
The dependence of spectral diffusion on temperature was investigated in 0.02 %
Er3+IY2SiOs at a fixed magnetic field strength of B = 2.25 T, with the results shown
in Fig. 5.27. Solid lines are fits to equation (5.13), showing excellent agreement with
the data. Fitting parameters are shown in the figure. Raising the temperature clearly
205
counteracts the magnetic field effect on the saturated spectral diffusion according to
the ratio ~
kT
in equation (5.14). The larger thermal energy available to the ions
causes the upper Zeeman level of the ground state to populate, which leads to larger
deviations of the spin population from thermal equilibrium and consequently to line
broadening. The rate, R, for the spin population to return to thermal equilibrium was
not measurably affected by the small temperature range, AT-0 .5 K , and is
consistent with a direct phonon process, which according to equation (5.15) is linear
in temperature at these low temperatures.
Discussion of Spectral Diffusion in Er3+: Y?SiOs as
a Function of Erbium Concentration
The erbium concentration dependence on spectral diffusion was investigated
using distinctly higher (0.1%) and lower (0.001%) concentration Crystals under
otherwise identical experimental conditions. The results of. these measurements are
summarized in Fig. 5.28 and Fig. 5.29. Solid lines in the figures correspond to least
squares fits to expression (5.13). In both cases, larger magnetic fields clearly
suppressed homogeneous IineWidth broadening. The data was successfully described
with the Bai-Fayer model as the excellent fits demonstrate. In particular, choosing a
lower Erbium concentration emphasizes the leading role of Er3+ spin-flips in spectral
diffusion. Whereas at a magnetic field of B = 2.25 T spectral diffusion in the 0.1%
and 0.02% sample is quite significant, it is almost negligible in 0.001% Er3+IY2SiOs.
206
In conclusion^ stimulated photon echo spectroscopy was performed to study the
effect of spectral diffusion on the homogeneous linewidth in dilute Er3"1": Y2SiOs. The
origin of spectral diffusion was attributed to direct one-phonon induced Er3"1"
environment ion spin-flips perturbing the energy levels of the optical Center. The
experimental results were successfully described in the framework of Bai-Fayer
theory by including the direct one-phonon process. It was shown that spectral
diffusion can be well-controlled with increasing magnetic field strength, lower
operating temperature, and lower .erbium doping concentration. Of course, trade offs
in the choice of these parameters must be made since they are often mutually
incompatible in a real system. Together, these are important results showing how to
optimize the Er3^ Y2SiOs material for laser frequency references and SHB-based
signal processing applications.
Stimulated Photon Echo T-Decavs
The rationale behind these measurements was to explore the material storage and
integration time for practical SHB applications, given by the lifetime of the transient
spectral hole. An upper limit for achievable storage times has been set by
fluorescence lifetime measurements, yielding 11.4 ms and 9.4 ms for site I and 2
respectively. Using the stimulated photon echo technique, the storage time can be
measured under practical conditions, i.e. in the presence of Spectral diffusion.
207
For an ideal two-level system, without the presence of spectral diffusion the echo
amplitude depends on the separation between the first two pulses and the waiting
time, T, the separation between the second and third pulse, as [23]
Thus, for an ideal system stimulated photon echo 7-decays allow the relaxation time,
Ti, to be determined by varying the waiting time, 7, and keeping T2 fixed.
diffusion over a timescale of 10’s to 100’s of (is alters T2 as discussed in the
previous sections. Hence, stimulated echo 7-decays in the presence of spectral
diffusion are expected to be non-exponential and relation (5.16) does not provide an
adequate description of the behavior. However, stimulated echo 7-decays will
indicate the “practical” storage time that can be exploited in a SHB device. In the
limit of completely quenched.spectral diffusion, these decays represent a lower limit
for the excited state population relaxation time, 7/.
The experimental apparatus and conditions were identical to the ones described
in the previous section. Measurements were carried out in 0.001% Er3+^ S iO 5 at
T = 1.6 K with B HDi and k l l b as a function of time delay, 7, with >12 fixed to
f T l f (5.16)
v -zIy v 12 J
In Er3"1": YaSiOs the situation is more complicated because the presence of spectral
50 (is.
208
Results and Discussion
Figure 5.31 shows the stimulated echo T-decays. The magnetic field strength was
varied between subplots from B = 0.8 T to B = 7 T. Individual decay curves are
normalized to I for a time delay, T = 0. Solid lines represent single exponential fits
to the exponential part of the decay with time constants given in the figure. Each
case shows good agreement with the data.
At low magnetic fields, the stimulated echo decays displayed a strong deviation
from exponential behavior. A rapid decay dominated the initial I ms part of the
curve, and then the decay became single-exponential. Spectral diffusion, being faster
than population decay, dominated the rapid initial part. The population grating stored
was being “smeared out” by frequency shifting interactions caused by direct phonon-
induced environment Er3"1" ion spin-flips. As we learned from the previous section,
these interactions take place on a timescale of 10’s to 100’s of (is with a rate,
determined by the direct phonon relaxation rate. After all environment ion spins had
flipped, the stored population grating could not smear out any further. At this time,
the homogeneous linewidth broadening reached the spectral diffusion saturation
value, Ti, and the observed decays became single exponential. Increasing the
magnetic field clearly suppressed the fast initial decay, that is caused by spectral
diffusion, and time constants for the exponential part increased from 2.9 ms at
B = 1.75 T to 4.9 ms at B = 7 T. At a magnetic field of B = 7 T, stimulated echo T-
decays were single exponential, consistent with the complete suppression of the
209
spectral diffusion seen earlier in Fig. 5.29. The observed time constant at B = 7 t
approached the theoretical limit set by population decay 2r < ZJ =11 ms very closely
with 2r= 9.8 ms according to equation (5.16). The value of 9.8 ms sets a lower limit
on the population decay time. Hence, under appropriate conditions, the storage time
for an optical memory based on Er3"1": YaSiOs can reach the lifetime limit.
T = (4.87 ± 0.02) ms
a B = 2 T, T = (3.99 ± 0.07) ms
o B = 1.75 T, T =(2.89 ±0.19) ms
< B = 0.8 T
0.001 % Er IY2SiO5, T = 1.6 K
B H Dj i IcIIb , T12 = 50 ps,
12.5
T - Delay (ms)
Figure 5.31 Stimulated echo T-decays in 0.001% Er3+IYaSiOs at T = 1.6 K,
B H D ii fc // 6 as a function of time delay, Ti between pulse two and three. The
time delay, tn, between pulse one and two was kept fixed at tn = 50 (is. The
magnetic field strength was varied from B = 0.8 T to B =7 T between subplots.
Solid lines represent single exponential fits with time constants given in the
figure.
210
The single exponential behavior of the decays after the rapid initial part also
indicates that no electronic Zeeman sublevel storage occurs in the ground state for
Er3+-iY2SiOs. In electronic Zeeman sublevel storage, the population grating persists
in the ground state, due to a long spin lattice relaxation time, often much longer than
the population lifetime of the excited state. For the material Nd3+=LaF2, for example,
electronic Zeeman sublevel storage was reported with storage times of 100 ms [24],
and in Tb3+=LiYF4 the sublevel storage lasted for minutes [25]. The absence of a
grating component with a different decay rate indicates that rapid spin-lattice
relaxation prevents Zeeman sublevel storage in Er3+=Y2SiOs. In fact, this supports the
earlier determination that the direct one-phonon process induces Er3"1" spin flips in
spectral diffusion. The measured spin relaxation rates (see Fig. 5.30) in the 10’s of
kHz range indicate a rapid spin-lattice relaxation preventing Zeeman sublevel
storage on the time scale of Ti in this material.
In conclusion, stimulated echo T-decays were used to measure the material
storage time. Spectral diffusion, taking place on a faster timescale than population
decay, dominated the decays for short T-delays and small magnetic fields. Higher
magnetic fields suppressed spectral diffusion and stimulated echo T-decay times
became population lifetime limited.
211
Ultraslow Dephasing
The stimulated echo T-decay experiments showed that the material storage time
can become lifetime limited. Utilizing the material optimization strategy, it seemed
worthwhile to investigate the material coherence time, T2 (or homogenous linewidth,
Th), under similar operating conditions. ■
For this purpose, the most diluted crystal was chosen with 0.001 atomic percent
Er3+ concentration, Dephasing of site I was investigated using two-pulse photon
echo measurements. The experimental apparatus was identical to the one described
earlier. In order to. approach the homogeneous linewidth limit and to reduce
contributions from Tsr-Er and Tphonon, a magnetic field of B = 7 T was applied in the
D1-D2 plane. The field was oriented along the preferred direction, where B takes on
an angle of 0 - 140° with respect to the Di crystal axis. This angle gave the longest
dephasing time (narrowest homogeneous linewidth) in two pulse echo experiments
as a function of field orientation. In order to “freeze out” the thermal population in
the upper Zeeman level of the ground state, the sample was immersed in a liquid
helium bath held at T = 1.5 K with the laser k vector along b.
Results and Discussion ,
Figure 5.32 displays a two-pulse photon echo decay in 0.001% Er3+IY2SiO5 at
B = 7 T, and T = 1.5 K. The observed decay was fitted to a single exponential.
212
represented by the solid line, and shows good agreement. From the fit, a dephasing
time of T2 = 4.38 ms was determined, corresponding to a homogeneous linewidth of
73 Hz. This dephasing time is the longest dephasing time ever measured in any
solid-state material, with the corresponding linewidth being the narrowest optical
resonance. Previous measurements of Equall et al. [6] in Eu3+--Y2SiOs yielded
T2 = 2.6 ms corresponding to Fhom = 122 Hz as . the narrowest known optical
resonance thus far in a solid.
*=L
(Z)
C
CD
C
I
LU
0.001 % Er -Y2SiO5 1-544-Top
Angle 0 ( B 3 D1) = 140'
exponential fit:
T2 = (4.38 ± 0.08) ms
= (72.6 ± 1,3) Hz
Figure 5.32 Two-pulse photon echo decay in 0.001% Er3+:Y2SiOs at B = 7 T,
T = 1.5 K, the lasers k vector is along b and B is in the D1-D2 plane at an angle
of 0 - 140° to the Di axis. The solid line is an exponential least square fit to the
data yielding a dephasing time T2 = 4.38 ms and a homogeneous linewidth of
73 Hz.
213
Major contributions to this narrow linewidth are population decay, Fpop,
instantaneous spectral diffusion, Fbo, and contributions from the 89Y nuclear spin
fluctuations, Fgr-Host- The lifetime of Ti = 11.4 ms contributes Fpop = 14 Hz to the
homogeneous linewidth. Further measurements are required to separate the other
individual contributions. Fluctuations of the earth’s magnetic field as well as stray
electro-magnetic fields in the laboratory may influence the linewidth as well, as the
results by Equall et. al. [26] suggest. Equall et. al. shielded the sample from
alternating magnetic fields to obtain the longest decays.
In conclusion, the intrinsic homogeneous 4Ii5z2: Zi(I) —> 4Ii3/2:Yi(l) linewidth
limit for site I in Er3+: Y2SiO5 has been closely approached by eliminating the effects
of Er3+ spin flips, Fgr-Er, and phonon contributions, Tphonon. The measured
homogeneous linewidth of Fhom = 73 Hz contains contributions from population
decay, Fpop ~ 14 Hz, instantaneous spectral diffusion, Fz5s, and spin fluctuations of
the 89Y nucleus, Fgr-Host- To the best of our knowledge, it is the narrowest optical
resonance measured in any solid.
Operation of Er3+: Y2SiOs at Elevated Temperatures
Motivated by the encouraging results achieved using the optimized material,
two-pulse photon echoes at B = 2 T were measured as a function of temperature to
quantify practical operational temperatures for SHB applications. In addition,
stimulated photon echo spectroscopy characterized the spectral diffusion at elevated
214
temperature and revealed important material information. The experimental setup
was identical to the one described earlier.
Results and Discussion
The 0.005 % Er3^ YaSiOs crystal was oriented With Dj along the magnetic field
B and the 6-axis along the laser ^-vector. Figure 5..33(a) shows two-pulse photon
echo decays for a fixed magnetic field strength of B = 2 T as the temperature was
varied from T =1.5 K to T = 5 K. Solid lines in Fig. 5.33(a) are least squares fits
using expression (2.5) that allowed extracting the homogeneous linewidth using
relation (2.6); each case shows excellent agreement. Figure 5.33(b) plots the
homogeneous linewidth as a function of temperature. The homogeneous linewidth
was observed to increase from ~ 3 kHz at T = 1.5 K to ~ 160 kHz at T = 5 K. At a .
temperature of T = 4.2 K, a homogeneous linewidth of 38 KHz was obtained. This is
the first report of photon echoes in Er3+-doped materials at 1.5 pm at a temperature
ofT = 4.2 K and an important result for practical SHB applications since mechanical
closed-cycle cryo-coolers can readily cool to these temperatures.
Li
ne
w
id
th
(k
Hz
)
Ec
ho
S
tre
ng
th
(a
rb
. u
ni
ts
)
215
o T= 1.5 K
D T = 1.8 K
a T = 2.0 K
v T = 2.4 K
O T = 2.8 K
< T = 3.2 K
> T = 3,8 K
o T = 4.4 K
o T = 5.0 K
0.005 % E r:Y 2SiO,
B = 2 T , site 1
45.0
t12 - Delay (ns)
0.005 0ZoEr=Y2SiO,
B = 2 1 , site 1
5.0 5.5
Temperature (K)
Figure 5.33(a) Two-pulse photon echo decays measured in 0.005% Er3+^ Y2SiOs
as a function of temperature for a fixed magnetic field strength of B = 2 T, // 6,
B H Dj. Solid lines are fits to expression (2.5) to extract the homogeneous
linewidth; each case shows excellent agreement, (b) Homogeneous line widths as
a function of temperature obtained from (a), the solid line serves to guide the
' eye.
216
Next, stimulated photon echo spectroscopy was used to investigate the time
evolution of the homogeneous line width. by spectral diffusion in the 0.005%
Er3+IY2SiO5 crystal at a temperature of T = 4,2 K and magnetic field of B = 3 T.
Stimulated photon echoes were measured as the time delay, t12, between pulse I and
2 was stepped for fixed waiting times, T. The observed decays were fitted to
expression (2.5) that allowed extracting the homogeneous linewidth using equation
(2.6). Figure 5.34 shows the result of this analysis.
120
100
£
80
£
%
' i
o 60
40
20
0.1 1 10 100 1000 10000
T (its)
Figure 5.34 Evolution of the linewidth of site 1 in 0.005% Er3+=Y2SiO5 at
T = 4.2 K and B = 3 T as the waiting time, T, between pulses two and three is
varied in a stimulated photon echo measurement.
OOOCtiy OO O o1
d 50. A note of caution is in order. BNC connectors and cables
have stray capacitance that can modify the resonant frequency and the EOM crystal
has piezo resonances that the user should avoid. To verify proper construction, the
reflected RF power from the resonant tank when connected to the EOM should be
measured with a directional coupler and network analyzer.
■Impedance matching
Transformer L
off the board
Figure A.6. Electronic schematic of the resonant EOM-tank circuit.
243
Post-Mixer Amplifier
The Post-Mixer Amplifier’s purpose in the feedback loop was threefold, to
amplify the error signal amplitude, to provide monitoring of the error signal for
locking quality diagnostics, and to allow to adjust the error signal baseline. Figure
A.7 shows the schematic. The first stage is a non-inverting amplifier providing a
gain of 30; the voltage-offset compensation adjusted the baseline of the error signal.
A voltage follower op-amp sampled the amplified error signal, without disturbing
the circuit operation, for display on an Oscilloscope; a 50 Q impedance provided the
output to the current servo. The post-mixer amplifier was assembled into a shielded
metal box (Pomona Electronics) with BNC connectors for input and output.
Phase Shifter
Chapter 2 (FM spectroscopy) and chapter 4 (Incorporating the inhomogeneous
line as a fixed reference) provide a rationale for the presence of the phase shifter.
The phase shifter design exploited the trigonometric identity
sin cot sin f
Figure C.l Two-level system with energy splitting, AE = (J.B, and populations in the
upper (lower) state, yO++(/?__) employed in the Bai-Fayer theory.
The rate equations describing the jumps can be written as
^ = -A A -+ A -R h. , .■
f ( C l )
at
where /?++(/?__) is the population in the upper (lower) state, and 2?_+ (i?+_)is the up
(down) transition rate. The coupled equations can be solved as
Ap{t) = Ap(eq) + [AyO(O) - Ap(eq)] exp(-ifr) (C.2)
with
R = K - + R.+ , (C.3)
AyO = yO++ - p__, (C.4)
P++ + P— ~ I (C.5)
260
where t±p{eq) is the population difference at equilibrium and the term in square
brackets in (C.2) describes the deviation from thermal equilibrium. The rate, R, is the
relaxation rate of the two levels toward equilibrium. From (C.4) and (C.5) it follows
that
AyO = 1 -2 /7 ,. (C 6)
Making the assumption that dephasing occurs only during the waiting time, T, of the
echo sequence, the probability of a spin-flip from up to down occurring during this
time can be obtained from
4 - = F ( ^ ( 0 ) = l ) f ( y O _ ( r ) = 1 1 y O + + (0 ) = 1 ) ( C . 7 )
with P+_ the probability for a spin flip from up to down, which is given, by the
conditional probability of finding the perturber in its lower state after r = 7, given
that it was initially in the upper state. In thermal equilibrium, the probability of a
spin-flip up is equal to the probability of a splin-flip down, so that we only need to
investigate one case.
The probability for finding the perturber initially in the upper state is given by the
percentage of the population in the upper state at thermal equilibrium. Hence,
P (/Th. (0) = l) = /9++ (eq) (C.8)
and the conditional probability of a spin flip to occur during the waiting time, 7,
given that the spin was initially up state is given by the percentage of population in
the lower state after the time, t - T
f (/?__(;) = 1 1 /7^(0) = 1 ) = / 7 _ ( f ) ( C . 9 )
261
Hence (C.7) becomes
P+-= P++(eq)p__(t) ■
Usingp++ (O)=I, p__(t) can be found from (C.2) using (C.6) to be
P - (0 = p - (eq) (I - exp(-to)).
Plugging (C .ll) into (C.10) yields
P+_ = p++(eq)p__ (eq) [I - exp(-Pt)].
(C.10)
(C.ll)
(C.12)
From the Boltzmann distribution, the populations in the upper level, p++ (eq) , and
. ' ' ■ '
lower level, p__(eq) , at thermal equilibrium can be found as
exp
p^{eq) = -
pB
2kT
exp
IkT
+ exp
\2A:Ty
(C.13)
exp
p__ ( e q )= -
exp z p B ''
/ 2 ^ r y
/ ,,n X
+ exp pB
(C.14)
Inserting (C.13) and (C.14) into (C.12) yields
P+_ =—sec Zi2r pB } [ l—exp(-Pt)];. (C.15)
In thermal equilibrium, the probability of a spin-flip up is equal to the probability of
a splin-flip' down. Hence,
P+_= P + . (C.16)
I262
Each up or down spin-flip of a perturber will cause an average frequency shift,
2Afavg , at the optical center. The total number of spin flips (up or down) occurring,
Nfiip, is given by
Ww = (P ._ + ^ ) iV , (C.17)
where N is the total number of ions. The total linewidth, F ro (T) , of the optical center
due to spectral diffusion (SD) after the waiting time, T, is given by the total number
of flips times the average frequency shift of each spin flip (assuming Lorentzian
broadening), so that
T^(T) = 2AL,A%. (C.18)
Plugging (C.15)-(C.17) into (C.18) yields the final result
Z , , D x
Tgg(T) = NAf sech JL lB
^2&T/ [l-exp(-RO]-
(C.19)
Expression .(C. 19) was used to fit the experimental data of the linewidth as a function
of waiting time, T, where a constant offset, To, has been included to account for the
initial linewidth as measured with a two-pulse photon echo. The saturated linewidth
after long waiting times was fit to
r i (5 ) = TmaxSech2 SMbb
2kT
(C.20)
with M = SMb -
263
REFERENCES
I. J. L. Hall, “Stabilizing Lasers for Applications in Quantum Optics”, in Quantum
Optics IV, J. D. Harvey and D. F. Walls, Eds. Springer Verlag: Proceedings of the
Fourth International Conference, Hamilton, New Zealand, (1986).
2. T. Day, E. K. Gustafson, and R. L. Byer, IEEE Journal of Quantum Electronics
28,1106(1992).
3. J. Helmcke, “Optical Wavelength Standards”, M, Zhu and J. L. Hall “Frequency
Stabilization of Tunable Lasers” in Experimental Methods in the Physical Sciences
Vol. 29C, F.B. Dunning and R. G. Hulet, Eds. Academic Press, San Diego (1997).
4. C. Salomon, D. Hils, and J. L. Hall, JOSA B 5,1576 (1988).
5. A. B. Williams, F. J. Taylor, Electronic Filter Design Handbook, 3rd edition, Mc
Graw Hill, New York 1995, p. 6.18.
6. New Focus Application Note 2 “Practical Uses and Applications of Electro-Optic
Modulators”; T. Day, Laser Focus World May 1992, 183.
7. K. G. Libbrecht, J. L. Hall, Rev. Sci. Instrum. 64, 2133 (1993).
8. M. Zhu and J. L. Hall in Frequency Stabilization of Tunable Lasers, Experimental
Methods in the Physical Sciences 29C, Academic Press San Diego 1997, 103
(1997).
9. Dana Z. Anderson, Josef C. Frisch, and Carl S. Masser, Appl. Optics 23, 1238
(1984).
I F f f i N A STATE UNIVERSITY - BOZEMAN
3 1762 10358944 4