PHYSICAL REVIEW B VOLUME 53, NUMBER 20 15 MAY 1996-II

Electronic structure of the N-V center in diamond: Experiments

A. Lenef, S. W. Brown, D. A. Redman, and S. C. Rand
Division of Applied Physics, 1049 Randall Laboratory, University of Michigan, Ann Arbor, Michigan 48109-1120

J. Shigley and E. Fritsch

Gemological Institute of America, 1630 Stewart Street, Santa Monica, California 90404-4088
~Received 4 August 1995!
Quantum-beat spectroscopy has been used to observe excited states of the N-V center in diamond. For the
1.945-eV optical transition, direct evidence is presented for the existence of GHz-scale fine structure, together
with a much larger 46-cm21 level splitting in the E state. An interference effect observed in transient four-
wave-mixing response is explained with a polarization selection rule involving Zeeman coherence among
magnetic sublevels. Also, detailed dephasing measurements versus temperature and wavelength have identified
the decay mechanisms operative among the various states. A comparison of these results with ab initio
calculations of excited electronic structure and interactions based on several multielectron models supports the
conclusion that the N-V center is a neutral, two-electron center governed by a strong Jahn-Teller effect and
weak spin-spin interactions.

I. INTRODUCTION very direct manner, these measurements permit a determina-

tion of electronic excited states and dynamics of the N-V
The center responsible for pink coloration in diamond, a center in the vicinity of the 637-nm zero-phonon transition.
rare tint in natural diamond,1 but one easily induced in syn- The experimental term diagram can then be compared with
thetic diamond containing substitutional nitrogen atoms,2 is calculations for centers containing two, four, or six active
the N-V center. It consists of a nitrogen atom and a vacancy electrons. Splitting patterns for these three models have been
on neighboring sites in the carbon lattice and may be formed predicted14 after considering the following electron interac-
through a simple process of irradiation and annealing. While tions: spin-orbit, spin-spin, strain, and the Jahn-Teller effect.
its optical3 and spin-resonance4 signatures, as well as its C 3 v Consistency between theory and experiment is obtained in
symmetry,5 have been known for a long time, the ground- the case of two electrons experiencing a Jahn-Teller effect
state electronic structure for the center was reassigned only strong enough to quench the spin-orbit interaction,15 with
recently, following the observation of unexpected satellite finer splittings arising from spin-spin interactions.
peaks in the persistent hole-burning spectrum.6 An important finding is that only two active electrons as-
Recent hole-burning7–9 and electron-paramagnetic- sociated with the nitrogen atom are necessary for understand-
resonance ~EPR! experiments10 have provided firm evidence ing the origin of the observed optical transitions of the cen-
that the ground state is a spin triplet, rather than a spin sin- ter. In particular, the dangling bonds on carbons adjacent to
glet as previously thought. The number of active electrons in the vacancy do not appear to contribute to optical interac-
the center must, therefore, necessarily be even. However, if tions. We are thus led to conclude that the N-V center is a
all unsatisfied bonding electrons in the center are counted neutral rather than a negatively charged nitrogen-vacancy
~two from nitrogen and three from carbons adjacent to the center, with properties determined by the two active nitrogen
vacancy!, the total is five, which is not even. The existence electrons.
of triplet levels originally prompted Loubser and van Wyk to
suggest4 that the center captures an extra electron to form a II. THEORETICAL
six-electron center, which is then negatively charged. How-
ever, this conclusion runs counter to the stability of the N-V The theory of a two- and three-pulse ~stimulated! photon
center of temperatures that permit even the heavy nitrogen echoes has been discussed previously by many authors.16 In
impurity atoms to become mobile.11 Studies of excess elec- this section, we merely outline the main results and provide
tron centers in alkali halide crystals12 indicate that they ion- extensions of earlier transient four-wave-mixing calculations
ize thermally at temperatures well below those necessary to necessary to understand and analyze our experimental results
initiate atomic diffusion. A six-electron model would seem in a multilevel system with magnetic degeneracy. Of main
unlikely on this basis, and the charge state and physical and interest are expressions for photon echo signal intensities
electronic structure of the N-V center are called into ques- versus interpulse delay times for ultrashort pulses, which
tion. generate quantum beats and Zeeman coherence described in
We believe this puzzling situation is resolved in the Sec. II A. When oscillations in the coherence decay are ob-
present work. Here, we present direct measurements of served as a function of delay between the forward beams
excited-state energy-level splittings and dynamics, obtained ~quantum and polarization beats!, relatively fine energy-level
using photon echo spectroscopy, to complement earlier in- splittings may be deduced for ground and excited states of
formation on magnetic interactions in the ground state.13 In a the center. When the center wavelength of the incident laser

0163-1829/96/53~20!/13427~14!/$10.00 53 13 427 © 1996 The American Physical Society

13 428 A. LENEF et al. 53

H 0 is the unperturbed Hamiltonian and V is the light-atom

interaction described by

V52eE~ R,t ! • (i ri . ~2!

The electric field E is evaluated at the position R of the

luminescent center at time t, and ri gives the coordinates of
the ith electron of the center. The final term in ~1! is a re-
laxation operator with diagonal elements describing popula-
tion decay and off-diagonal elements describing coherence
relaxation ~dephasing!.
The applied field consists of a sequence of three pulses at
times t j with arbitrary wave vectors k j ~ j 51, 2, or 3!, of the
form

S D
3
1 k j •R
E~ R,t ! 5
2 ( ej E j
j51
t2t j 2
v
exp$ i @ v ~ t2t j ! 2k j •R

FIG. 1. Schematic energy-level diagram of a luminescent center 1 f j # % c.c.1, ~3!

with quasidegenerate levels a, a 8 , and a 9 in the ground state and b,
b 8 , and b 9 in the excited state. Population decay between sublevels where e j is the polarization vector of the jth field.
is denoted by L and between electronic levels by g. The dephasing In the basis of states shown in Fig. 1, equations of motion
rate at the optical resonant frequency v0 is G.
for individual density-matrix elements are

S(
pulses is tuned over a wide range, for a fixed interpulse spac- i
ing, evidence for energy splittings on a coarse scale can also ṙ b 8 a 8 5 ~ i v b 8 a 8 1G ! r b 8 a 8 2 V b8a9r a9a8
be obtained. In Sec. II B, the parametrization involved in \ a9

D
experimental determinations of dephasing mechanisms in
this work is examined. Detailed interactions of a center with
its surroundings may be deduced when the dephasing rate is
2 ( r b8b9V b9a8 , ~4!
b 9
measured as a function of temperature over a wide range. We
also present explicit expressions for the polarization depen-
i
dence of echo signals for a center of C 3 v symmetry in Sec.
II C. Polarization studies yield information on magnetic de-
ṙ a 8 a 9 5 ~ i v a 8 a 9 1L a 8 a 9 ! r a 8 a 9 2
\ ( ~ V a8b9r b9a9
b9
generacies.
2 r a8b9V b9a9 ! 1 d a8a9 ( g r b9b8 , ~5!
b9
A. Origin of quantum and polarization beats
Figure 1 depicts energy levels of a luminescent center 1
with ground- and excited-state manifolds a and b, respec- ṙ b 9 b 8 52 ~ i v b 9 b 8 1L b 9 b 8 ! r b 9 b 8 2
\ ( ~ V b9a9r a9b8
tively. The manifolds contain sublevels u a & , u a 8 & ,... and u b & , a 9
u b 8 & ,..., which are closely spaced in energy compared to the 2 r b9a9V a9b8 ! 2 d b9b8g r b9b8 , ~6!
frequency v0 of the optical transition and may be degenerate.
The optical frequency between a particular ground-state sub-
level u a & and an excited-state sublevel u b & is denoted by r a8b85 r *
b8a8 . ~7!
v ba 5 v b 2 v a . For simplicity, we assume that population
relaxation and dephasing on allowed transitions between ex- Here, g is the excited-state population decay rate and G is the
cited and ground-state manifolds does not depend on the dephasing rate.
magnetic substate quantum number. Additionally, intersub- To calculate the third-order response associated with
level dephasing and population decay rates are taken to be three-pulse photon echoes, a third-order perturbation expan-
constant within a given manifold ~L a 8 a 9 5L a for a 8 Þa 9 and sion for the density matrix corresponding to a power series in
L b 8 b 9 5L b for b 8 Þb 9 !. the perturbing field interaction lV is normally developed,
The equation of motion for the density matrix r describ-
ing the interaction of light with a luminescent center is given r 5 r ~ 0 ! 1l r ~ 1 ! 1l 2 r ~ 2 ! 1••• . ~8!
by17
The expansion parameter l is set equal to one at the end of

ṙ 52
i
\
@~ H 0 1V ! , r # 2S D
]r
]t relax
. ~1!
the calculation. By substituting Eq. ~8! into Eq. ~1! and
equating terms of the same order in l, a recursion relation
may be generated for r(n) .
53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 429

ṙ ~ 0 ! 5 @ H 0 , r ~ 0 ! # 1L relaxr ~ 0 ! ,

ṙ ~n!
5@H0 ,r ~n!
# 1V relaxr ~n!
1 @ V, r ~ n21 !
#.
~9a!

~9b!
ṙ ~b 8!a 8 52 ~ i v b 8 a 8 1G ! r ~b 8!a 8 1
3 3 i
\ S( b9
r ~b 8!b 9 V b 9 a 8
2

L relax describes linear decay according to L relaxr

52( d r / d t!relax and comprises relaxation, due to spontane-
ous emission and dephasing by various mechanisms. Assum-
2 (
a 9
~2!
V b8a9r a9a8 D . ~13!

ing that excited-state populations and system coherences are In first order, the only nonzero matrix elements correspond
to intermanifold coherences. In second order, intramanifold
initially zero, one obtains the following equations for system
coherence terms and populations appear. Finally, in third or-
evolution:
der, the off-diagonal term r~3! needed to calculate the optical
polarization P5Tr~mr! from second-order populations and
i Zeeman coherences appears.
ṙ ~b 8!a 8 52 ~ i v b 8 a 8 1G ! r ~b 8!a 8 1 r~ ! ,
1 1 0
V ~10!
\ b8a8 a8a8 The evolution of the optical polarization is determined by
decay constants of the system and relative timing and polar-
ization of the pulses, together with the pulse areas propor-
ṙ ~a 8!a 9 52 ~ i v a 8 a 9 1L a 8 a 9 ! r ~a 8!a 9 1 d a 8 a 9
2 2
( g r ~b29!b 8 tional to the magnitude of the complex field quantities,

i
b 9
Ẽ1 5e1 exp~ 2iki •R! E `
E i ~ t ! dt. ~14!
( ~1! ~1! 2`
2 ~ V a8b9r b9a92 r a8b9V b9a9 ! , ~11!
\ b 9 The index i ~51,2,3! in ~14! labels different input pulses.
When the bandwidth of the field greatly exceeds the dephas-
ing rate, the optical pulses can be approximated by d func-
ṙ ~b 9!b 8 52 ~ i v b 9 b 8 1L b 9 b 8 ! r ~b 9!b 8 2 d b 9 b 8 g r ~b 9!b 8
2 2 2
tions and the equations of motion may be integrated easily.
i Using the rotating wave approximation, the resulting third-
2
\ ( ~1! ~1!
~ V b9a9r a9b82 r b9a9V a9b8 ! , ~12! order coherence appropriate for homogeneously broadened
a 9 media is found to be

r ~b 8!a 8 ~ t ! 5
3
S D i
2\
3
e 2G ~ t2t 3 1t 2 2t 1 ! e 2i v b 8 a 8 ~ t2t 3 ! (
a9b9
$ r a 9 a 9 ~ 0 ! e 2L b ~ t 3 2t 2 ! e i v b 8 b 9 ~ t 3 2t 2 ! @ e 2i v b 9 a 9 ~ t 2 2t 1 ! ~ Ẽ*1 • m b 8 a 9 !

i v b 9 a 9 ~ t 2 2t 1 ! ~ 2L a 1i v a 8 a 9 !~ t 3 2t 2 !
3 ~ Ẽ2 • ma 9 b 9 !~ Ẽ*
3 • mb 9 a 8 ! 1e 2 • mb 8 a 9 !~ Ẽ1 • ma 9 b 9 !~ Ẽ*
~ Ẽ* 3 • mb 9 a 8 !# 1e

3@ r a 9 a 9 ~ 0 ! e i v b 9 a 9 ~ t 2 2t 1 ! ~ Ẽ*
3 • mb 8 a 9 !~ Ẽ1 • ma 9 b 9 !~ Ẽ*
2 • mb 9 a 8 ! 1 r a 8 a 8 ~ 0 ! e
2i v b 9 a 8 ~ t 2 2t 1 !
3 • mb 8 a 9 !~ Ẽ2 • ma 9 b 9 !
~ Ẽ*
2 g ~ t 3 2t 2 ! 2i v b 9 a 9 ~ t 2 2t 1 ! 2i v b 9 a 9 ~ t 2 2t 1 !
1 • mb 9 a 8 !# 2 r a 9 a 9 ~ 0 !~ 12e
3~ Ẽ* @e 3 • mb 8 a 8 !~ Ẽ2 • ma 9 b 9 !~ Ẽ*
~ Ẽ* 1 • mb 9 a 9 ! 1e

3 • mb 8 a 8 !~ Ẽ1 • ma 9 b 9 !~ Ẽ*
3~ Ẽ* 2 • mb 9 a 9 !# % . ~15!

m ab 5e^aurub& is the usual electric dipole transition moment F a 8 a 9 ;b 8 b 9 ~ T 8 !

between states a and b. For inhomogeneously broadened
media, only homogeneous groups of centers with positive
phase evolution factors exp[i v b 9 a 9 (t 2 2t 1 )] can be rephased 5e ~ 2L b 1i v b 8 b 9 ! T 8 ~ Ẽ*
2 • mb 8 a 9 !~ Ẽ1 • ma 9 b 9 !~ Ẽ*
3 • mb 9 a 8 !
by application of the third pulse. These groups generate a
macroscopic coherent polarization, which radiates at the 1e ~ 2L a 1i v a 8 a 9 ! T 8 ~ Ẽ3* • mb 8 a 9 !~ Ẽ1 • ma 9 b 9 !~ Ẽ2* • mb 9 a 8 !
echo time t echo5t 3 1T ~where T5t 2 2t 1 !. Only half the
terms in ~15! contribute at the time of the echo, yielding the 2 ~ 12e 2 g T 8 !~ Ẽ*
3 • mb 9 a 9 !~ Ẽ1 • ma 9 b 9 !~ Ẽ*
2 • mb 9 a 9 ! .
polarization
Unless intersublevel dephasing rates La or Lb are exception-
r ~b 8!a 8 ~ t5t echo!
3 ally high, coherence decay is normally determined by the
exponential prefactor exp~22GT! in ~16!, and is twice as
5
i
\ SD 3
e 22GT (
a9b9
r a 9 a 9~ 0 ! e i ~ v b 9a 92 v b 8a 8 !T
fast in inhomogeneous systems as it is in homogeneous sys-
tems.
To account for the effect of an inhomogeneous distribu-
3F a 8 a 9 ;b 8 b 9 ~ T 8 ! , ~16! tion g~D! of resonant frequencies, a spectral average is taken
with respect to the detuning D5vba 2v. Then, the echo po-
where T 8 5t 3 2t 2 , and larization is
13 430 A. LENEF et al. 53

The three terms in ~19! correspond, respectively, to ‘‘direct’’

one-phonon-assisted decay, indirect two-phonon-assisted Ra-
man decay, and two-phonon Orbach relaxation. The con-
stants a and b weight the different processes and \vs corre-
sponds to the spin state level splitting governing the direct
process. D gives the energy separation between nearly de-
generate spin states and a discrete electronic ~or vibrational!
state that governs the Orbach process.
From the functional form of ~19!, it can be recognized
that the direct process dominates at low temperatures. At
higher temperatures, other decay channels make sizeable
contributions. The theoretical exponent for the Raman term
is generally taken to be m55,18 but in this work it was left as
an additional fitting parameter to test this theoretical power
dependence. Hence, the comparison of ~19! with experimen-
tal dephasing rates measured versus temperature determines
FIG. 2. ~a! Diagram of the four-wave-mixing geometry, show- the relative importance of these potential contributions, as
ing the counter-propagating forward and backward pump beams k2 well as the three parameters vs , D, and m.
and k3 , respectively, and the relative timing between pulses.

SD 3
C. Polarization dependence and Zeeman coherence
i
^ r ~b 8!a 8 ~ t 8 !
3
& 52 p
\
e 22G ~ t 8 1T !
(
a9b9
r a 9 a 9 ~ 0 ! F a 8 a 9 ;b 8 b 9 ~ T 8 ! Transient four-wave-mixing relies on a tensorial interac-
tion between three incident optical waves and the third-order

3 E ei ~ v b 9 a 9 2D ! T e 2i ~ v b 8 a 8 2D ! t 8 g ~ D ! dD
dipolar response of the system. While most of our experi-
ments were performed with linearly polarized beams, addi-
tional information was obtained regarding magnetic proper-

52 p SD i
\
3
G ~ t 8 1 t ! e 22G ~ t 8 1T !
ties of the N-V center through observations with circularly
polarized beams. Predictions are developed below for Zee-
man coherence observable in the N-V center as the result of
its electronic-spin character. These calculations show that, in
3 (
a9b9
r a 9 a 9 ~ 0 ! F a 8 a 9 ;b 8 b 9 ~ T 8 ! e i v b 9 a 9 T e 2i v b 8 a 8 t 8 . addition to generating quantum beats, interference effects
among degenerate magnetic states introduce polarization de-
~17! pendence into photon echo signals.
Ratios of expected photon echo signal intensities for dif-
Here, we have introduced a convenient time variable ferent configurations of incident polarized fields can be cal-
t 8 5t2t 3 and G(t) denotes the inverse Fourier transform of culated from ~17! very simply ~in the short-time limit
g~D!. T,T 8 →0! by considering the expression for the third-order
It is evident from ~17! that two kinds of interference can polarization,
be observed through echo experiments on luminescent cen-
ters with discrete levels of the type considered here, depend-
ing on whether the delay time T or T 8 is varied ~Fig. 2!. If
the delay T between the first two pulses is varied, polariza-
P ~i 3 ! 5« 0 (
jkl
x ~i 3jkl! ~ Ẽ *
1 j E 2k Ẽ 3l 1Ẽ *
1 j Ẽ 3k Ẽ 2l ! , ~20!

tion beats occur at frequencies Dv5v b 9 a 9 2 v b 8 a 8 , which in-

clude combinations ~sums and differences! of the actual level and using known or assumed properties of the transition mo-
splittings. These beats arise from interference of polariza- ments comprising the susceptibility:
tions, which do not have a common level. On the other hand,
if the delay time T 8 between pulses two and three is varied, 1
quantum beats occur only at the splitting frequencies x ~i 3jkl! }
n (s a(,a s s s s
~ m a 8 b 9 ! k ~ m b 9 a 9 ! j ~ m a 9 b 9 ! l ~ m b 8 a 8 ! i . ~21!
8 9
Dv5vb 9 b 8 and Dv5va 9 a 8 themselves. In this case, the in- b 8 ,b 9
terfering polarizations have a common level.
Such a calculation is illustrated in the Appendix, where it is
B. Dephasing mechanisms specifically adapted to the N-V center. The notation ( m sab ) k
refers to the kth component of the dipole transition moment
Coherence relaxation is determined by radiative and non-
between states a and b. The index s ~s51...n! in ~21! ac-
radiative processes according to
counts for n possible different defect orientations. Construc-
G tot5 ~ g r 1 g nr! /2, ~18! tive and destructive interferences result from summing the
contributions to the third-order response in ~20!, as described
where the radiative contribution is gr 5g, and the non radia- in more detail in the Appendix. Among the possible se-
tive portion is given in the case of solids by15 quences of circularly polarized pulses, which satisfy momen-

S D
tum conservation, certain ones yield signals an order of mag-
\vs aD 3 nitude more intense than others. The main predictions are
g nr5a coth 1bT m 1 . ~19!
2k B T exp~ D/k B T ! 21 given in Table I.
53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 431

TABLE I. Comparison of observed and predicted echo properties for various polarization that conserve
angular momentum. Only modulation components with relative amplitudes of 15% or more are listed.

Polarization Theoretical Observed Frequencies Relative

sequence intensity intensity ~GHz! amplitude

s1s1s1;s1 1.0 1.0 0.6360.01 1.0

1.6360.01 0.77
2.2960.01 1.29
s1s1s2;s2 1.6 0.93 0.5660.01 1.0
1.6660.02 0.37
2.2860.01 0.81
2.8860.01 0.15
s1s2s1;s2 0.16 0.11 0.6260.03 1.0
1.6060.02 0.33
2.2760.02 0.49
s1s2s2;s1 1.0 0.93 0.4960.02 1.0
1.7060.01 0.26
2.2960.01 0.45
Linear 0.6360.02 1.0
1.6760.02 0.42
2.3060.01 0.95
2.8760.01 0.23

Signal intensities depend upon details of magnetic degen- faces were repolished using traditional methodology21 prior
eracies not immediately evident from the point-group sym- to subsequent optical measurements.
metry of the center. Suppression of experimental echoes by Next, the absolute concentration of N-V centers was de-
an order of magnitude for s1s2s1 polarization sequences termined in one sample and related to the magnitude of in-
compared to s1s1s1, for example, it is predicted on the duced IR absorption at 1332 cm21 ~prior to annealing!, as
basis of Zeeman coherence on A↔E optical transitions of well as the irradiation dosage. This was done by first deter-
the N-V center, whereas no such effect can occur on A↔A mining the relative density of N-V centers to singly substi-
transitions. Hence, observations of interference effects fur- tuted nitrogen from ratios of EPR signals shown in Fig. 3.
nish evidence for angular momentum degeneracy and The main lines at 3300 G correspond to paramagnetic ~sub-
complement more direct methods of assigning the spectro- stitutional! nitrogen and the two sets of four small lines on
scopic character of states involved in the optical transition. either side of the nitrogen spectrum correspond to distin-
guishable orientations of N-V centers. Taking identical oscil-
lator strengths for N and N-V spin transitions, the ratio of
III. EXPERIMENT integrated intensities for the four low-field N-V lines to the
intensity of the central N resonance was found to be 17.5%.
A. Sample preparation Thus, the concentration of centers was r~N-V!50.175r~N!
A total of 30 type-1b synthetic single-crystal diamonds of
approximate dimensions 1.733.533.5 mm3 were obtained
from the Sumitomo Electric Corporation for this study. Con-
centrations of N-V centers in the range 531016 –531018
cm23 were achieved in these samples using a simple proce-
dure. Infrared-absorption measurements were first used to
establish that all as-grown samples contained 200620 ppm
substitutional nitrogen atoms, based on the 1130-cm21
calibration.19 Next, vacancies were created by electron irra-
diation in a Van de Graaff accelerator. An energy of 1.7
MeV was used for these runs, somewhat above the 0.75-
MeV threshold for 1:1 defect creation.20 A current of 20 mA
was maintained in a beam of diameter 2 cm, delivering a
dose rate of 2.431015 e 2 cm22/min. Irradiation was per-
formed at the triple point of ice, and exposure times varied
from 20 min to 50 h. Following irradiation, the samples were FIG. 3. Electron-paramagnetic-resonance spectra in the region
placed in a quartz boat and inserted into a furnace filled with of the isolated nitrogen resonance. The four pairs of small features
argon at 820 °C for 4 h. Subsequent to this annealing step, it to the sides are due to four orientationally distinct groups of N-V
was noticed that, in addition to turning pink, sample surfaces center. A comparison of line strengths provides the relative concen-
dulled somewhat due to graphitization, so the large ~100! tration of N-V centers to paramagnetic nitrogen ~N! directly.
13 432 A. LENEF et al. 53

FIG. 5. Dependence of absorbance ~integrated between 1330–

1340 cm21! on electron dosage. The solid curve is a best fit to the
regression y5m 1 x1m 2 x 2 , with m 151.06310218, m2
58.68310 , after the subtraction of the background ~58.3 cm21!
238

in untreated samples.

B. Four-wave-mixing spectroscopy
Tunable subpicosecond pulses were generated using a
synchronously pumped DCM dye laser pumped by a mode-
locked 25-W Nd:YAG ~yttrium aluminum garnet! laser. An
external cavity dumper permitted variable pulse repetition
rates from 1 Hz–7.2 MHz. With a conventional output cou-
pler, the laser was capable of generating pulses with autocor-
relations yielding full widths at half maximum ~FWHM! of
800 fs at a repetition rate of 72 MHz with peak powers of
1–2 kW. Cavity-dumped pulses were in the 10–15-kW
range with FWHM autocorrelation widths of 1.02 ps
~FWHM Gaussian pulse widths of 700 fs!, as determined
from second-harmonic generation in a type-I BBO crystal
@Fig. 6~a!#. Measured emission spectra were nearly sym-
metrical Gaussians with a FWHM of 1.0 nm, yielding a
time-bandwidth product of 0.4, which is very close to the
transform limit @Fig. 6~b!#.
Degenerate four-wave mixing was performed in a stan-
dard counterpropagating pump geometry ~Fig. 7!. Most of
the measurements were performed on a sample with an
absorption-length product somewhat less than unity
FIG. 4. Infrared-absorption features in synthetic diamond after ~a0L,1!, to ensure linear probe response with maximum
irradiation. ~a! Growth of the 1332-cm21 peak versus 2-MeV elec- signal strength. Transmission was 50% at 637.3 nm in this
tron irradiation dosage. Bottom to top: 0, 1.031018, 2.531018, and sample at liquid-helium temperature. When the center wave-
531018 e/cm2. ~b! Effects of annealing: reduction of the 1332-cm21 length was tuned in the vicinity of 637 nm, a coherent signal
feature and growth of the 1450-cm21 peak. Bottom to top: as beam with an excitation spectrum matching the zero-phonon
grown, after irradiation ~531018 e/cm2! and after annealing. Sample absorption line of the N-V center was generated in a direc-
thickness was 1.67 mm throughout. tion opposite to that of the probe wave. Its intensity was
recorded versus temporal delay of the input beams at various
temperatures and for various polarizations.
for an electron dose of 1.031018 cm22, where r~N! is the To measure population decay ~T 1! in the E state, pulse
post-annealing nitrogen concentration. With r~N!5200 ppm/ durations were lengthened to tp 55 ps ~autocorrelation
~110.175!, the final concentration of centers was determined FWHM 8 ps! for maximum signal by turning off the satu-
to be r~N-V!530610 ppm for this sample. rable absorber jet in the dye laser. The delay between the first
In Fig. 4 spectral changes, which occur in the infrared as two pulses was set to zero and the third pulse delayed by as
the result of irradiation @Fig. 4~a!# and annealing @Fig. 4~b! much as 7 ns, using a double-pass optical delay line @Fig.
and 4~c!#, are shown. In Fig. 5, the relationship between 7~a!#. Precision alignment of the 60-cm-long, high-resolution
dosage and absorption by induced features at 1332 cm21 is mechanical stage used for this purpose was accomplished
presented for four samples, which had identical initial nitro- using a dual-axis silicon position sensor to ensure that the
gen concentrations ~221 ppm!, to within our ability to mea- delayed beam remained centered on the interaction region
sure. with measured deviations of less than a fiftieth of the focused
53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 433

FIG. 6. ~a! Second-harmonic autocorrelation of the experimental

optical pulses ~logarithmic scale! showing a FWHM of 1.02 ps. ~b!
The optical pulse spectrum corresponding to ~a!. A Gaussian fit
~solid curve! to the data points yields a FWHM of 1.0 nm.

beam diameter over the entire displacement range. The stage

was driven by a stepper motor, which typically advanced 3
mm per step. Cavity dumping rates for these measurements
were held below 10 kHz to avoid accumulating population in
the metastable excited state of the center.10 This avoided the
background signal caused by accumulated gratings, improv-
ing the signal-to-noise ratio for observation of weaker single-
pulse radiative decays.
The probe intensity was mechanically chopped and syn-
chronous detection applied together with spatial filtering to FIG. 7. Setup for ~a! T 1 and ~b! T 2 measurements.
discriminate against scattered light. Typically 50 scans were A study of echo intensity versus cavity-dumping rate was
averaged. Data were then fitted with single exponential de- performed to identify the maximum permissible repetition
cay functions superimposed on variable background levels. rate, which avoided population accumulation in the meta-
For T 2 ~dephasing! measurements, 1-ps pulses were used stable level of the N-V center.10 By varying cavity-dumping
and the interval between the second and third pulses was rate Q at a constant chopping frequency of 15 Hz, it was
fixed at 1.5 ns, while the delay between the first two pulses found that echo intensity was linear in Q below approxi-
was varied @Fig. 7~b!#. Complete decay curves were mea- mately 10 kHz at liquid-helium temperatures. Hence, all
sured in this way at sixteen temperatures between 5–75 K. population and dephasing decays were measured at rates
The signal magnitude and dephasing time were also mea- lower than this. However, for Q in the MHz range and av-
sured at fixed interpulse delays, while the excitation wave- erage input powers of 40–50 mW, the signal beam was vis-
ible to the eye. After verifying that quantum beat modulation
length was changed, looking for evidence of coarse excited-
frequencies did not depend on repetition rate, quantum beat
state structure. For investigation of polarization selection observations were conducted at more elevated rates for better
rules in the four-wave-mixing response, zero-order quarter- signal-to-noise ratio.
wave plates were inserted in the input beams to generate
circular polarization. Under these conditions, signals were IV. RESULTS AND DISCUSSION
verified to be circularly polarized and detected through an
A. Infrared spectroscopy
analyzer consisting of a quarter-wave Babinet-Soleil com-
pensator, linear polarizer, and spatial filter for improved sig- In this section, changes which occur in the infrared-
nal isolation. absorption spectrum of diamond during the formation of N-V
13 434 A. LENEF et al. 53

centers are described, inasmuch as they may relate to the

excited-state structure of these centers. A strong absorption
line appears at 1332 cm21 in proportion to electron irradia-
tion of diamond. Subsequent annealing @Figs. 4~a! and 4~b!#
causes single nitrogen impurities to associate with vacancies
to form N-V centers, and the strength of this line then dimin-
ishes. Simultaneously a sharp feature at 1450 cm21
emerges22 in proportion to dosage and hence ~directly or in-
directly! to N-V concentration ~Fig. 5!. This line has previ-
ously been attributed to vibrations of a charged nitrogen
interstitial,23,24 but in view of its appearance during the for-
mation of N-V centers, we briefly discuss the alternative pos-
sibility that it is an optical transition of the N-V center itself.
The center responsible for the 1450-cm21 line is known to
FIG. 8. Excited-state population ~T 1! decay of the N-V center
contain only a single nitrogen atom23 and exhibits no obtained with the picosecond resolution at l5637.2 nm and T577
uniaxial stress splitting along principal directions.25 Since the K by varying delay T 8. Low signal-to-noise ratio is the result of the
splitting of orientational degeneracies is too small to be ob- need for a low repetition rate to avoid population accumulation in
served by traditional methods, these results and the narrow- excited states.
ness of the transition could also be argued to be consistent
with absorption to a ~metastable! singlet state of an N-V
center. No EPR signal has been reported in association with estimate a pulse area of u5( m E/2\) t p 50.75 rad, which
the 1450-cm21 feature.26 This would seem odd if the feature confirms that the experiment was performed just within the
is due to an interstitial, since the impurity is paramagnetic ~perturbative! limit of applicability of theoretical results from
and a shifted resonance might be expected upon displace- Sec. III.
ment of the nitrogen from its substitutional position in the
lattice. However, this result has a natural explanation in the 2. Echo modulations
case of an N-V model, since these centers have a very weak At liquid-helium temperatures, echo dephasing behavior
ground-state EPR signal in the absence of light.10 exhibited modulations, the contrast of which depended mark-
The existence of an isotope shift of the 1450-cm21 edly on the pulse repetition rate. T 2 decay curves of unaccu-
absorption23 would seem to favor a vibrational origin of the mulated echo signals showed very weak modulation compo-
line. However, the ratio of absorption frequencies for differ- nents, whereas accumulated echo signals were nearly 100%
ent nitrogen isotopes does not agree very well with the in- modulated. Component frequencies, however, showed no de-
verse square root of their respective masses, as it should in a pendence on repetition rate, remaining constant within ex-
straightforward interstitial model. Consequently, the N-V perimental error over the full range. An example of an accu-
electronic singlet state, the decay signature of which was mulated photon echo coherence decay curve is shown in Fig.
previously reported by Redman,10 might profitably be sought 9~a!.
by using double-resonance techniques in the region of 180 At negative delays, a small nonzero signal, as well as
meV above the ground state. modulation, is evident in the figure. This signal resulted from
free-induction decay, and contributed a ~homogeneous! com-
B. Visible excited-state spectroscopy ponent symmetric about the time origin. It was detected to-
gether with the desired photon echo radiation in this experi-
1. Population decay
ment, because time-resolved detection was not used for most
Four-wave-mixing signals were measured versus delay measurements. However, it was verified that the true photon
between the second ~forward pump! and third ~backward echo signal accounted for 95% of the four-wave-mixing re-
pump! pulses at 77 K to obtain the population decay time of sponse at these temperatures in a separate experiment using
the zero-phonon transition. At high repetition rates a strong time-resolved, cross correlation.28
constant background and appreciable signal intensity were To optimize the Fourier analysis, in the face of rapid de-
observed at negative delay times. These effects were due to cay over a relatively small range, we made use of a Ham-
accumulated gratings, as outlined in the previous section. All ming window29 for numerical data processing. The power
decay data were, therefore, recorded under conditions such spectrum shown in Fig. 9~b! for the data of Fig. 9~a! was
that signal intensities were independent of repetition rate, obtained in this way. A least-squares fitting routine furnished
namely, below 10 kHz at 77 K and below 6 kHz at 6 K. spectral peak positions for the modulation components. At
The decay curve at 637.7 nm is shown in Fig. 8. The data 637.8 nm, strong components were identified at frequencies
yielded a characteristic time of t56.4860.07 ns at 77 K and of 0.6360.02 GHz, 1.60060.003 GHz, 2.29860.006 GHz,
a derived value of T 152t512.9660.14 ns. The only previ- and 2.8660.01 GHz. Farther to the red at 638.4 nm, a similar
ously published value of this decay time required instrumen- determination yielded frequencies of 0.5660.01 GHz, 1.72
tal deconvolution,22 but is nevertheless in good agreement 60.01 GHz, and 2.29060.002 GHz, showing less than a
with our measurement. 10% variation over a detuning range of 1.5 nm from line
From this radiative decay time, the dipole moment of the center.
A→E transition of the N-V center may be estimated,27 yield- These values are in excellent agreement with sideband
ing m53.7310227 cm C. In turn, this value permits us to frequencies reported earlier in persistent hole-burning
53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 435

FIG. 10. Echo intensity versus excitation wavelength at 6.1 K.

The two curves correspond to different fixed delays of T57 ps
~squares! and T50.5 ns ~circles!. The solid curve is the result of a
fit to the sum of two Gaussians with a resultant separation of 4662
cm21. Also shown are the dephasing times T 2 measured versus
wavelength for T57 ps ~diamonds!.

lished hole-burning data at lex5636.699 nm ~Fig. 6 of Ref.

9!. This indicates that a change in the electronic structure of
the excited state occurs at an energy corresponding roughly
to 637.3 nm, an unexpected result which is clarified in the
next section.
FIG. 9. ~a! Modulated echo decay versus delay T at a wave-
length of 637.8 nm and a temperature of 6.7 K. All beams were 3. Dephasing mechanisms
linearly polarized and delay time between pulses two and three was Dephasing measurements versus wavelength and tem-
fixed at T 851.5 ns. ~b! Power spectrum of accumulated echo data in perature revealed different aspects of the electronic structure
~a!. of the N-V center. The principal results are given in Figs. 10
and 11.
experiments.6,9 However, in the present experiments, the in-
In Fig. 10, the echo intensity recorded at constant delay
terferences responsible for echo modulation were established
across the zero-phonon transition clearly resolved itself into
within the duration of the optical pulses, showing in a direct
two components at different energies, centered around 637
manner that relaxed excited states are not involved in side-
nm. The higher energy component on the blue side of the
band generation. Consequently, the modulation frequencies
inhomogeneous distribution of transition wavelengths exhib-
can only be related to splittings in the ground or excited
ited very rapid dephasing. The red component dephased ten
states connected directly by the optical field. The presence of
times more slowly. A least-squares fit to the positions of two
a ground-state splitting is directly confirmed by the excellent
agreement between the highest-frequency component and the
ground-state resonant frequency measured in EPR.4,10 Be-
cause our measurements were performed by varying the de-
lay between first and second pulses, the observed frequencies
should include simple combinations of energy-level split-
tings, in addition to the pure splittings themselves ~see Sec.
III!. This is the basis for the analysis of the measured fre-
quencies in terms of two excited-state components and their
sum frequency in the next section. It provides independent
verification of the results and interpretation of hole-burning
experiments reported in Ref. 9.
Echo data to the blue side of 637.3 nm, the wavelength at
which hole-burning sidebands were reported to become
abruptly indistinct in Ref. 9, revealed dramatically increased
dephasing rates. This rapid decay unfortunately prevented FIG. 11. Temperature dependence of dephasing rate ~T 21 2 ! at
the use of the Hamming filter to analyze echo data in this wavelengths l5636.4 nm ~diamonds! and l5638.4 nm ~circles!.
spectral range. Consequently, a direct Fourier transform was The curves give the results of fitting Eq. ~18! to the data. For red-
performed in this region and only a weak component at 2.6 detuned data, elimination of Raman or direct dephasing terms re-
60.05 GHz was identified reliably at an incident wavelength sulted in distinctly poorer fits. All contributions were, therefore,
of 635.7 nm. However, this result is once again in accord included in the final analysis. For blue-detuned data, the Orbach
with persistent hole-burning results. A single homogeneously term significantly worsened convergence and did not improve the
broadened satellite with this detuning can be seen in pub- fit. In this case, it was excluded from the final fit ~dashed curve!.
13 436 A. LENEF et al. 53

TABLE II. Parameters yielding the observed temperature dependence of dephasing.

Wavelength a \vs /2 D b
21
~nm! ~MHz cm23! ~cm21! ~cm ! ~MHz cm23! m

638.4 0.1 60.04 106 4 3965 0.0760.01 3.260.2

636.4 0.0660.02 68616 0.0460.02 5.262.6

Gaussian lines yielded a relative splitting of 4662 cm21, tween J-T potential wells.15,31 This point is discussed further
directly confirming an earlier inference of such an electronic in the conclusions.
splitting within this line.30 Because the ground-state structure
does not admit such a splitting,10 the excited state itself nec- 4. Polarization dependence
essarily consists of two components u1& and u2&.
To obtain information regarding magnetic degeneracies
From the results of Fig. 11, it is clear that the mechanism
and selection rules for optical transitions of the N-V center,
of dephasing also differs on the two sides of the zero-phonon four-wave-mixing experiments were performed with circu-
transition. Dephasing rates versus temperature were found to larly polarized beams, in addition to the measurements with
be qualitatively and quantitatively different at representative linear polarization. The main results are given in Figs. 12
wavelengths on the red and blue sides of 637 nm, with a and 13, and Table I. Figure 12~a! presents a semilogarithmic
sharp transition between them. At 636.4 nm, the dephasing plot comparing echo intensity versus delay between the first
rate was constant below approximately 30 K and increased two pulses for two different polarization sequences, namely
rapidly above this temperature. At 638.4 nm no plateau was ~s1s1s1;s1! and ~s1s2s1;s2!. Measurements revealed
observed and long-wavelength data sustained moderate cur- that at 638 nm the overall signal intensity was ten times
vature over the whole temperature range. stronger for the ~s1s1s1;s1! sequence than for the
Least-squares fits of Eq. ~18! to this data yielded the re- ~s1s2s1;s2! sequence. The ratio of polarized signal inten-
sults given in Table II. Dephasing behavior on the blue side sities did not change much at wavelengths both longer and
of the line was dominated by a direct term with a character- shorter than 638 nm, and always exceeded one ~Fig. 13!.
istic energy interval of 68616 cm21 below 30 K and a Ra- Comparative best-fit dephasing times for s1 and s2 signals
man term with m55.262.6 above this temperature. The lat- were T 254tmeas52.7860.25 ns and 2.8260.09 ns, respec-
ter power dependence is characteristic of a non-Kramers
center ~even number of electrons!, in which there are two
~m57! or more ~m55! excited levels.18 The Orbach term
had to be omitted from the fit for stable convergence, an
indication that its degrees of freedom did not contribute to an
improved description of the data. Overall, this analysis was
consistent with a picture of one-phonon emission between
the two components u1& and u2& of the excited state ~sepa-
rated by 46 cm21! at low temperature and the appearance of
an m55 Raman dependence above 30 K.
On the red side of the line, a more complicated parametri-
zation of the temperature dependence emerged from similar
analysis. The radiative limit was approached only asymptoti-
cally at the lowest temperatures, and between 6–20 K an
Orbach term with an energy interval of 3965 cm21 made an
important contribution. But contributions from a T m power-
law dependence, with m53.260.2 and a direct term with a
large energy interval, were not negligible.
Our interpretation of these results is that up to tempera-
tures of 30 K the u A 1 & ↔u1& optical polarization dephases by
a direct u1&→u2& decay process, in which a high-frequency
phonon is emitted. Above this temperature, a T 5 Raman pro-
cess sets in. On the other hand, the nonradiative portion of
dephasing of the u A 1 & ↔u2& polarization occurs largely by a
two-phonon Orbach process between the u1& and u2& states,
at temperatures up to about 20 K. Some centers appear to
absorb single phonons to reach the upper excited state u1&,
whence they make direct radiative transitions to the ground FIG. 12. ~a! Logarithmic plots of the polarization dependence of
state. At higher temperatures, dephasing is determined by a accumulated echo decay curves as a function of T at a fixed value
T 3 process, which was recognized in previous work as in- of T 851.5 ns. l5637.8 nm and T55.5 K. Polarization sequences
dicative of dephasing in a Jahn-Teller (J-T) system with are s1s1s1; s1 ~solid! and s1s2s1; s2 ~dashed!. ~b! Power
strain energies that are large compared to the couplings be- spectra of ~a!.
53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 437

as from point to point within any given sample. For a hypo-

thetical sample with a permanent internal, uniaxial strain,
one would not expect internal strain to be universally ori-
ented along a unit-cell diagonal in synthetic and natural
samples, and it could not be parallel to the threefold axis of
more than one of the four distinguishable classes of N-V
center in diamond. Hence in the limit of large strain, we
point out that irregular, inhomogeneous distributions of N-V
center resonance frequencies would be expected in different
samples rather than a single well-defined splitting.
The T 3 dephasing behavior of u2&↔u A & dipole polariza-
tion at high temperature suggests a different origin for the
observed electronic structure, namely, a Jahn-Teller effect
FIG. 13. Echo intensity versus excitation wavelength at 6.1 K accompanied by strain. On the basis of temperature depen-
for different polarization sequences: s1s1s1; s1 ~circles! and dence of line positions and widths,32 as well as uniaxial
s1s2s1; s2 ~squares!. Interpulse delays were fixed at T50.2 ns strain measurements,5 it was concluded, in previous work,
and T 851.5 ns. that E vibrational modes couple only weakly to the N-V
center and do not produce a significant J-T distortion of the
tively, on the red side of the zero-phonon line. Modulation excited state. Mixing of u1& and u2& components of the
components were identical within experimental error, as in- strain-split line measured by polarized luminescence tech-
dicated in Fig. 12~b!, except for a slow background modula- niques in this earlier work agreed with proportions expected
tion at long times believed to be an artifact of acquisition at from random strain, with a magnitude corresponding to the
low signal levels. All results are summarized in Table I. inhomogeneous linewidth. The possibility of J-T coupling
was, therefore, dismissed. However, the experimental results
presented in the present paper are more consistent with the
V. CONCLUSIONS opposite conclusion. Moderate amounts of random strain
The variation of four-wave-mixing signal intensity with alone certainly do not explain our quantum beat results, since
wavelength clearly revealed the existence of two electronic a continuous strain distribution would merely broaden the
states u1& and u2& separated by 46 cm21 within the upper excited state. Well-defined quantum beats would never form.
state manifold of the 637-nm transition of the N-V center. We propose a model in which u1&→u2& coupling arises
Modulations at several GHz were detected in time-resolved from a strong Jahn-Teller interaction with a splitting between
signals and furnished ground and excited-state spin-spin vibronic levels ~46 cm21!, which slightly exceeds the inho-
splittings on a finer frequency scale. Measurements of coher- mogeneous linewidth ~FWHM 25 cm21!. This coupling
ence relaxation rates versus temperature were consistent with strength is large enough to quench spin-orbit coupling,33 so
u1&→u2& decay by one-phonon emission and dephasing in that a u1& and u2& manifold structure augmented by spin-spin
state u2& involved virtual two-phonon ~Orbach! excitation of splittings can be preserved overall, in apparent agreement
the type u2&→u1&→u2&. At relatively high temperatures, our with our experiment. This situation would explain why the
analysis identified a T 5 Raman decay process in state u1& and
a T 3 decay in u2& suggestive of a J-T effect accompanied by
strain.
Echo signal variations for different polarization sequences
agreed quantitatively with predictions on A 1 ,A 2 ↔E transi-
tions of a C 3 v center when Zeeman coherence in the spin
triplet ground state was taken into account. Suppression of
echoes was uniform over an excitation wavelength range that
probed different excited states. This indicated that this effect
depended on ground-state spin sublevel degeneracy. More-
over, only the first two pulses had a significant overlap
within the excited-state dephasing time T 2 . Interference ef-
fects with long coherence times could, therefore, only have
arisen from Zeeman coherence between 3A(m s 511) and
3
A(m s 521) ground states. This set of observations con-
firmed the residual m s 561 ground-state degeneracy already
known from EPR studies,18 as well as the E orbital angular
momentum symmetry of the excited state required to pro-
duce echo polarization dependence.
Our overall results are consistent with the excited-state
structure sketched in Fig. 14. We believe that neither the
coarse-grain excited-state splittings nor the finer GHz split- FIG. 14. Empirical energy-level structure deduced for the N-V
tings are determined by uniaxial strain. Internal strains are center from photon echo observations.
ordinarily randomly oriented from sample to sample, as well
13 438 A. LENEF et al. 53

Jahn-Teller interaction was not evident in earlier for each electron of the center in appropriate irreducible rep-
measurements.33 In our model, strain does not couple mag- resentations a and b of orbital angular momentum states uf&,
netic substrates together in first order, so that spin-spin inter- referenced to the coordinate system of the luminescent
actions are insensitive to both strong J-T coupling and center.14 r does not act on the spin of the center, so spin
strain. In the presence of quenching of the spin-orbit interac- matrix elements are not considered. Circularly polarized
tion, this model can explain why excited and ground-state fields can of course alter spin states, but for the purposes of
splittings generate quantum beat frequencies of similar mag- this discussion it is adequate to set all spin matrix elements
nitude. to be equal and nonzero. The components of r transform as
Detailed molecular-orbital calculations presenting the as- X, Y , and Z ~defect coordinates!, and, therefore, in C 3 v sym-
pects of this model for centers with different assumed num- metry can be expressed as A 1 % E operators according to
E E
bers of active electrons are presented in Ref. 14. A compari- r A 1 5Z, r X 1 5X and r Y 2 5Y . Similarly, irreducible forms of
son of these calculations with the present experimental the unit vectors in crystal coordinates are eA 1 5eZ , eE 5eX ,
results shows that unless a Jahn-Teller interaction is intro- and eE 5eY .
duced in the excited E state, the coarse splitting resolved in For Z oriented along a cube diagonal, transformations be-
this work and the existence in the excited state of the three tween defect coordinates (x,y,z) and irreducible crystal co-
fine rf splittings that we have observed are not satisfactorily ordinates are given by
explained. Hence, this research supports the conclusion that
the N-V center is a neutral, two-electron center governed r5r A 1 eA 1 1r EX eEX 1r EY eEY , ~A4!
primarily by a strong Jahn-Teller effect and weak spin-spin
interactions. where

ACKNOWLEDGMENTS 1
r A15 ~ x1y1z ! ,
This research was sponsored in part by the Air Force Of- )
fice of Scientific Research ~H. Schlossberg! and the National
Science Foundation Technology Center for Ultrafast Optical 1
Science ~STC PHY 8920108!. A. Lenef, D. Redman, and S. r EX 5 ~ 2z2x2y ! ,
A6
Brown gratefully acknowledge support from a DSO fellow-
ship from the Department of Education, an AT&T scholar- 1
ship, and a Rackham Dissertation Grant, respectively. The r EY 5 ~ x2y ! , ~A5!
authors are indebted to L. Haliburton, S. Yazu, S. Satoh, and &
K. Tsuji of Sumitomo Electric for electron irradiation of and
samples, and to Carol Stockton for assistance with infrared-
absorption spectroscopy. We also wish to thank R. Schwartz 1
and N. Manson for helpful discussions. eA 1 5 ~ ex 1ey 1ez ! ,
)
APPENDIX
1
To compare third-order responses to various input pulse eEX 5 ~ 2ez 2ex 2ey ! ,
polarization sequences for transitions between states of spe- )
cific orbital angular momentum symmetries, contributions to
the third-order polarization P~3! from all possible transitions 1
eEY 5 ~ ex 2ey ! . ~A6!
need to be considered. In these calculations, it is essential to )
account for the vector character of the optical field in the
scalar interaction of Eq. ~2!. We outline such calculations Representations of r for the other seven ~cube diagonal! ori-
below and show, from Eq. ~20!, for example, that a s1s1s1 entations in crystal coordinates can be found with transfor-
input sequence yields an x component of the four-wave- mations between octants of a sphere. Then, by applying the
mixing polarization given by Wigner-Eckart theorem, matrix elements between all combi-
nations of irreducible orbital angular momentum states can
P ~x3 ! 52« 0 E 0 u E 0 u 2 ~ x ~xxxx
3! 3!
1 x ~xxy ~3! ~3!
y 2 x xyxy 1 x xy yx ! , ~A1!
be found. These are listed below:
whereas a s1s2s1 input sequence yields ^ f A 1 u ru f A 1 & 5eA 1 ^ f A 1 i r A 1 i f A 1 & , ~A7!

P ~x3 ! 52« 0 E 0 u E 0 u 2 ~ x ~xxxx

3! 3!
2 x ~xxy ~3! ~3!
y 2 x xyxy 2 x xy yx ! , ~A2! ^ f A 2 u ru f A 1 & 50,
which is quite different from ~A1!. Such differences result in
striking changes of echo intensities for different polarization ^ f EX u ru f A 1 & 5eEX ^ f E i r E i f A 1 & ,
sequences.
For specified initial and final states, the calculation of ^ f EY u ru f A 1 & 5eEY ^ f E i r E i f E & ,
susceptibility proceeds by writing the dipole matrix ele-
ments, ^ f A 2 u ru f A 2 & 5eA 1 ^ f A 2 i r A 1 i f A 2 & ,

m ab 52e ^ f a u ru f b & , ~A3! ^ f EX u ru f A 2 & 52eEY ^ f E i r E i f A 2 & ,

53 ELECTRONIC STRUCTURE OF THE N-V CENTER IN . . . 13 439

3!
^ f EY u ru f A 1 & 5eEX ^ f E i r E i f A 2 & , x ~xxxy 5C $ ~ m A 1 ,E X ! k ~ m E X ,A 1 ! j ~ m A 1 ,E X ! l ~ m E X ,A 1 ! i

1 ~ m A 1 ,E X ! k ~ m E X ,A 1 ! j ~ m A 1 ,E Y ! l ~ m E Y ,A 1 ! i
1
^ f EX u ru f EX & 5eA 1 ^ f E i r A 1 i f E & 1 eEY ^ f E i r E i f E & ,
1 ~ m A 1 ,E Y ! k ~ m E Y ,A 1 ! j ~ m A 1 ,E X ! l ~ m E X ,A 1 ! i
&
1 ~ m A 1 ,E Y ! k ~ m E Y ,A 1 ! j ~ m A 1 ,E Y ! l ~ m E Y ,A 1 ! i % . ~A10!
1
^ f EY u ru f EY & 5 eEX ^ f E i r E i f A 2 & ,
& The constant C is given by C5(i/2\) 3 g(0) r A 1 ,A 1 (0) . Evalu-
ation of tensor components with an even number of repeated
1 indices then yields the expressions
^ f EY u ru f EY & 5eA 1 ^ f E i r A 1 i f E & 2 eEY ^ f E i r E i f E & .
&
3!
The remaining matrix elements may be obtained by taking x ~xxxx 5 x ~y3y!y y 5 x ~zzzz
3! 3!
5 x ~xxy ~3! ~3!
y 5 x y yzz 5 x zzxx
complex conjugates of the expressions above. The dipole
matrix element ^ f A 2 u r u f A 1 & is zero, because r transforms as 5 94 C u ^ f E i r E i f A 1 & u 4 , ~A11!
A 1 ^ E under C 3 v symmetry. The reduction of the product
representation for this matrix element does not contain the and
totally symmetric representation A 1 .
Application of these results to the echo susceptibility in
~21! determines third-order selection rules between any two 3!
x ~xyxy 3!
5 x ~yzyz 3!
5 x ~zxzx 5 91 C u ^ f E i r E i f A 1 & u 4 . ~A12!
states of specific symmetry. First of all, the susceptibility
tensor exhibits general symmetry relations, which are useful,
These values are, in fact, equal to the final susceptibility after
namely, x i jkl (3) 5 x jilk (3) and x i jkl (3) 5 x kli j (3) . Additionally,
averaging over all orientations of a C 3 v center with a transi-
other simplifications result from specific geometries or con-
tion moment along a cube diagonal, because transformations
figurations, like that of circularly polarized input fields cou-
between equivalent orientations do not change the signs of
pling nondegenerate states. For an A 1 →A 81 transition, with a
susceptibilities which have an even number of repeated indi-
transition moment lying along the @111# direction, one finds ces. Evaluation of the third-order susceptibility for an E↔E
transition yields results qualitatively similar to those for
A 1 ↔E, and will not be considered here.
3!
x ~xxxx ~@ 111# ! 5 u ^ f A 8 i r A 1 i f A & u 2 ~ eA 1 •ex !~ eA 1 •ex !~ eA 1 •ex ! In summary, the susceptibilities for different circularly
polarized pulse sequences on A 1 ↔A 1 or A 2 ↔A 2 transitions
3~ eA 1 •ex ! 5 91 u ^ f A 8 i r A 1 i f A & u 2 . ~A8! are the same. For instance,

In view of the totally symmetric form of the basis vector

eA 1 5~ex 1ey 1ez !/), all tensor components of x (3) xxxx ~@111#!
are equal on an A 1 →A 18 transition. Susceptibility contribu- P ~x3 ! ~ s 1 s 1 s 1 ; s 1 ! 5 P ~x3 ! ~ s 1 s 2 s 1 ; s 2 ! , ~A13!
tions from centers oriented along other cube diagonals can be
found by coordinate transformations equivalent to reflections where the first three s polarizations in parentheses indicate
in ~100!, ~010!, and ~001! planes. input fields and the last one indicates the output field. No
Also, because xi jkl transforms like the outer product dependence of the echo intensity is expected for transitions
r i r j r k r l , components with indices that are repeated an odd between nondegenerate states. On the other hand, if one state
number of times change sign when the orientation is re- ~or both states! of a transition is ~are! degenerate within the
flected in these planes. Consequently, when all eight possi- excitation bandwidth, a strong intensity dependence of echo
bilities for the orientation of N-V centers are taken into ac- response is predicted. For an E↔A 1 transition, we find
count, the total xxxy component of the susceptibility, for
example, is given by
P ~x3 ! ~ s 1 s 1 s 1 ; s 1 ! 5 109 « 0 C u ^ f E i r E i f A 1 & u 4 E 0 u E 0 u 2 ,
~A14!
3! 3! 3! 3!
x ~xxxy 5 81 @ x ~xxxy ~@ 111# ! 1 x ~xxxy ~@ 1̄11# ! 1 x ~xxxy ~@ 11̄1 # ! 1•••
3! whereas
1 x ~xxxy ~@ 1̄1̄1̄# !# 50. ~A9!

Consequently, only x components in which indices appear an

even number of times make nonzero contributions to the P ~x3 ! ~ s 1 s 2 s 1 ; a 2 ! 52 94 .« 0 C u ^ f E i r E i f A 1 & u 4 E 0 u E 0 u 2 .
total response. ~A15!
For degenerate states, the signal polarization depends
strongly on the polarization sequence of the incident pulses. Results for E↔A 2 transitions are similar, with A 1 replaced
For example, using ~21! to find the susceptibility for an by A 2 in the reduced matrix element. The intensities pre-
A 1 →E transition, one obtains dicted on the basis of squaring these signal polarizations P~3!
13 440 A. LENEF et al. 53

for the four possible sequences of s-polarized pulses, which defined total angular momenta. The method and results pre-
conserve angular momentum, are given in Table I. Polariza- sented here permit predictions, however, even when no in-
tion dependence of four-wave mixing has been described formation on quantum numbers is available, but wave-
previously34 for transitions between atomic states of well- function symmetries are known.

1
J. E. Shigley, E. Fritsch, C. M. Stockton, J. I. Koivula, C. W. 171, 213 ~1968!; J. P. Gordon, C. H. Wang, C. K. N. Patel, R. E.
Fryer, and R. E. Kane, Gems Gemol. 22, 192 ~1986!. Slusher, and W. J. Tomlinson, ibid 179, 294 ~1969!; D. Grisch-
2
A. T. Collins, J. Phys. C 16, 2177 ~1983!; S. C. Rand, in Tunable kowsky and S. R. Hartmann, Phys. Rev. B 2, 60 ~1970!; J.
Solid State Lasers II, edited by A. B. Budgor, L. Esterowitz, and Schmidt, P. R. Berman, and R. G. Brewer, Phys. Rev. Lett. 31,
L. G. DeShazer ~Springer-Verlag, New York, 1986!, pp. 276– 1103 ~1973!.
17
280. M. Sargent III, M. O. Scully, and W. E. Lamb, Laser Physics
3
L. duPreez, Ph.D. dissertation, University of Witwatersrand, ~Addison-Wesley, Reading, Massachusetts, 1974!.
1965. 18
M. B. Walker, Can. J. Phys. 46, 1347 ~1968!.
4
J. H. N. Loubser and J. A. Van Wyk, Diamond Research ~Indus- 19
R. M. Chrenko, H. M. Strong, and R. E. Tuft, Philos. Mag. 23,
trial Diamond Information Bureau, London, 1977!, Vol. 77, p. 313~1971!.
11. 20
J. A. Baldwin, Phys. Rev. Lett. 10, 220 ~1963!.
5
G. Davies and M. F. Hamer, Proc. R. Soc. London, Ser. A 348, 21
P. Grodzinsky, Diamond Technology ~N.A.G., London, 1953!.
285 ~1976!. 22
G. S. Woods and A. T. Collins, J. Phys. C 15, L949 ~1982!.
6
R. T. Harley, M. J. Henderson, and R. M. Macfarlane, J. Phys. C 23
A. T. Collins et al., J. Phys. D 20, 969 ~1987!.
17, L233 ~1984!. 24
A. Mainwood, J. Phys. C 12, 2543 ~1979!.
7 25
N. R. S. Reddy, N. B. Manson, and E. R. Krausz, J. Lumin 38, 46 W. A. Runciman and T. Carter, Solid State Commun. 9, 315
~1987!. ~1971!.
8
K. Holliday, N. B. Manson, M. Glasbeek, and E. Van Oort, J. 26
G. S. Wood, Philos. Mag. B 50, 673 ~1984!.
Phys. Condens. Matter 1, 7093 ~1989!. 27
A. T. Collins, M. F. Thomaz, and M. I. B. Jorge, J. Phys. C 16,
9
D. Redman, S. W. Brown, and S. C. Rand, J. Opt. Soc. Am. B 9, 2177 ~1983!.
768 ~1992!. 28
See, for example, J. Shah, in Spectroscopy of Semiconductor Mi-
10
D. A. Redman, S. W. Brown, R. H. Sands, and S. C. Rand, Phys. crostructures, edited by G. Fasol, A. Fasolino, and P. Lugli ~Ple-
Rev. Lett. 67, 3420 ~1991!. num, New York, 1989!, pp. 535. See, also, S. T. Cundiff and D.
11
A. T. Collins, J. Phys. C 13, 2641 ~1980!. G. Steel, IEEE J. Quantum Electron, QE-28, 2423 ~1992!.
12
W. B. Fowler, The Physics of Color Centers ~Academic, New 29
A. V. Oppenheim and R. W. Schafer, Digital Signal Processing
York, 1968!. ~Prentice-Hall, Englewood Cliffs, NJ, 1975!, pp. 242.
13 30
N. B. Manson, X.-F. He, and P. T. H. Fisk, Opt. Lett. 15, 1094 E. van Oort, B. van der Kamp, R. Sitters, and M. Glasbeek, J.
~1990!. Lumin. 48-49, 803 ~1991!.
14 31
A. Lenef and S. C. Rand, following paper, Phys. Rev. B 53, R. Pirc, B. Zeks, and P. Gosar, J. Phys. Chem. Solids 27, 1219
13 441 ~1996!. ~1966!.
15
A. Abragam and B. Bleaney, Paramagnetic Resonance of Tran- 32
G. Davies, J. Phys. C 7, 3797 ~1974!.
sition Metal Ions ~Clarendon, Oxford, 1970!. 33
F. S. Ham, Jahn-Teller Effects in Electron Paramagnetic Reso-
16
I. D. Abella, N. A. Kurnit, and S. R. Hartmann, Phys. Rev. 141, nance Spectra ~Plenum, New York, 1969!.
391 ~1966!; M. Scully, M. J. Stephen, and D. C. Burnham, ibid. 34
S. Saikan, J. Opt. Soc. Am. 72, 514 ~1982!.