PHYSICAL REVIEW B 95, 195204 (2017)

Electron-nuclear spin dynamics of Ga2+ paramagnetic centers probed

by spin-dependent recombination: A master equation approach
V. G. Ibarra-Sierra,1 J. C. Sandoval-Santana,1 S. Azaizia,2 H. Carrère,2 L. A. Bakaleinikov,3 V. K. Kalevich,3
E. L. Ivchenko,3 X. Marie,2 T. Amand,2 A. Balocchi,2 and A. Kunold4
1
Departamento de Física, Universidad Autónoma Metropolitana Iztapalapa, Av. San Rafael Atlixco 186,
Col. Vicentina, 09340 Cuidad de México, Mexico
2
Universitè de Toulouse, INSA-CNRS-UPS, LPCNO, 135 avenue de Rangueil, 31077 Toulouse, France
3
Ioffe Physical-Technical Institute, 194021 St. Petersburg, Russia
4
Área de Física Teórica y Materia Condensada, Universidad Autónoma Metropolitana Azcapotzalco,
Av. San Pablo 180, Col. Reynosa-Tamaulipas, 02200 Cuidad de México, Mexico
(Received 13 November 2016; revised manuscript received 3 April 2017; published 10 May 2017)

Similar to nitrogen-vacancy centers in diamond and impurity atoms in silicon, interstitial gallium deep
paramagnetic centers in GaAsN have been proven to have useful characteristics for the development of spintronic
devices. Among other interesting properties, under circularly polarized light, gallium centers act as spin filters that
dynamically polarize free and bound electrons reaching record spin polarizations (close to 100%). Furthermore,
the recent observation of the amplification of the spin filtering effect under a Faraday configuration magnetic
field has suggested that the hyperfine interaction that couples bound electrons and nuclei permits the optical
manipulation of the nuclear spin polarization. Even though the mechanisms behind the nuclear spin polarization
in gallium centers are fairly well understood, the origin of nuclear spin relaxation and the formation of an
Overhauser-like magnetic field remain elusive. In this work we develop a model based on the master equation
approach to describe the evolution of electronic and nuclear spin polarizations of gallium centers interacting with
free electrons and holes. Our results are in good agreement with existing experimental observations. In particular,
we are able to reproduce the amplification of the spin filtering effect under a circularly polarized excitation
in a Faraday configuration magnetic field. In regard to the nuclear spin relaxation, the roles of nuclear dipolar
and quadrupolar interactions are discussed. Our findings show that, besides the hyperfine interaction, the spin
relaxation mechanisms are key to understand the amplification of the spin filtering effect and the appearance
of the Overhauser-like magnetic field. To gain a deeper insight in the interplay of the hyperfine interaction and
the relaxation mechanisms, we have also performed calculations in the pulsed excitation regime. Our model’s
results allow us to propose an experimental protocol based on time-resolved spectroscopy. It consists of a
pump-probe photoluminescence scheme with the detection and the tracing of the electron-nucleus flip-flops
through photoluminescence measurements.

DOI: 10.1103/PhysRevB.95.195204

I. INTRODUCTION alloys, Ga2+i paramagnetic centers with only one bound

Negatively charged nitrogen-vacancy centers in diamond electron can selectively capture another conduction band (CB)
[1–5], phosphorous atom impurities in silicon [6–11], electron with the opposite spin orientation [26–30]. Due to
and other schemes based on point defects embedded in this mechanism, paramagnetic centers act as a spin filter that
semiconductors have been widely studied as alternatives to blocks the recombination of CB electrons with the same spin
develop quantum bits [12,13]. One of the necessary conditions and efficiently captures electrons whose spin is in the opposite
for quantum computing is long electron spin decoherence direction. In the centers, the bound and captured electrons form
times to ensure a minimum of fault tolerance [10,14]. In a singlet state that is destroyed as either one of the electrons
diamond’s nitrogen-vacancy centers [15], silicon vacancies in recombines to the valence band (VB). It is important to note
silicon carbide [16], silicon [17], and any III-V based quantum that while the lifetime of conduction electrons with the spin
dots [10,18] the fluctuating nuclear bath is the main source opposite to that of the bound electrons is a few picoseconds, the
limiting spin coherence time. The nuclear dipole-dipole lifetime of conduction electrons copolarized with the majority
interaction is believed to be the dominant mechanism of bound electrons may extend to nanoseconds. As a conse-
behind the diffusion-induced electron-spin decoherence quence the free photoelectron spin polarization can reach over
[19]. To protect the dynamics of the nuclear spins of point 80% under circularly polarized incident light. Additionally
defects from the decoherence induced by the environment, the photoluminescence (PL) intensity can be as high as 800%
semiconductors mainly composed of spin-zero isotopes such under circularly polarized optical excitation compared to a
as silicon and carbon are preferred over III-V semiconductors linearly polarized one [21,22]. The increase in CB electron
[5]. Even though the two stable isotopes of Ga, 69 Ga and population allows even for the detection of electron spin po-
71
Ga, have nuclear spin 3/2, in dilute nitride GaAsN point larization by electrical means due to a giant photoconductivity
interstitial defects give rise to paramagnetic centers that effect under circularly polarized light [24,31].
have very peculiar and useful properties. One of them is the Whereas in diamond and silicon the optical excitation acts
spin-dependent recombination (SDR) [20–27]. In Ga(In)NAs directly on the point defects, in GaAsN the bound electron is

2469-9950/2017/95(19)/195204(14) 195204-1 ©2017 American Physical Society

V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

dynamically spin polarized due to the recombination of spin- In this paper we examine the spin dynamics in GaAsN
polarized CB electrons on the paramagnetic centers. Although alloys. We propose a model based on the master equation
this mechanism is radically different, Ga2+i centers themselves for the density matrix that describes the main interactions
are very similar to nitrogen vacancies and phosphorous atoms. between CB electrons, VB holes, and paramagnetic traps. Our
Recent experiments on GaAsN subject to a weak magnetic approach overcomes most of the limitations of the previous
field in Faraday configuration [32–35] have shown consistently models. First, it considers Ga centers with 3/2 nuclear spin.
an enhancement in the spin-filtering mechanism in comparison Second, in contrast to older proposals, all of the off-diagonal
to the zero magnetic field. The general agreement among elements arising from the preexisting two-charge model
different models [28,35–37] is that the hyperfine interaction [20,21,27,29,39], are taken into account. This introduces the
between the bound electron and the nucleus in the centers is main mechanisms of SDR into the model. The major advantage
the key element behind this phenomenon. At low magnetic of this improvement is that it can be used to study experimental
fields, the spin-filtering effect is reduced due to spin-state conditions in which the magnetic field is tilted with respect to
mixing induced by the hyperfine interaction. For higher values the incident light. Finally, the spin relaxation mechanism was
of the magnetic field such that the Zeeman energy exceeds introduced through Wangsness, Bloch, and Redfield relaxation
the hyperfine interaction, the pure bound electron spin states theory [40–43]. Two interactions are explored as possible
and, consequently, the spin-filtering effect are recovered. Even candidates to produce the nuclear spin relaxation: dipolar
though the role of the hyperfine interaction is well established, interaction between neighboring nuclei [40] and quadrupolar
some aspects of these phenomena lack full understanding. [40,44] interaction with charge fluctuations in the environment.
Some observations point to the existence of an Overhauser-like This not only has allowed us to clearly identify the dipolar
effective magnetic field [32,33] whose origin is yet unclear. interaction as the dominant nuclear relaxation mechanism
It manifests as a shift in the band to band PL intensity or but has also shown that the Overhauser-like magnetic field
the degree of CB electron spin polarization as functions of originates from it.
the longitudinal magnetic field. These two curves shift to the To further explore the role of hyperfine interaction in
positive or negative regions of the magnetic field depending Ga centers, we have studied the time-resolved electronic
on the helicity of the circularly polarized light. Another aspect and nuclear spin polarizations under pulsed excitation. Using
that needs further study is the nuclear interaction between these results we outline a method based on a pump-probe
Ga2+i centers and adjacent Ga atoms that would lead to photoluminescence scheme to trace the coherent evolution of
nuclear spin relaxation. The Overhauser-like magnetic field coupled electron and nuclear spins as they flip-flop due to the
has been correctly reproduced in Ref. [36], however, nuclear hyperfine interaction.
spins are assumed to relax very rapidly and the relaxation This paper is organized as follows. In Sec. II we introduce
mechanism is purely phenomenological. Moreover, when a the master equation model that describes the key processes
trap captures an electron it is assumed to instantly lose all in the spin dynamics of electrons and nuclei in GaAsN.
its nuclear spin polarization. On the other hand, the model The role of the hyperfine interaction and the nuclear spin
presented in Ref. [37] considers two phenomenological and relaxation mechanism are discussed in this section. The
arbitrary nuclear spin relaxation times for traps with one and mathematical forms of the dipolar and quadrupolar dissipators
two bound electrons. Despite this improvement, the model are introduced in Sec. II D. In Sec. III A we establish the main
considers Ga centers with 1/2 nuclear spins instead of 3/2 in mechanism behind the nuclear spin relaxation in Ga2+ centers
order to simplify the kinetic equations. Even though the first by comparing the theoretical results issued by the model
version of this model can only address experimental conditions with existing experimental results. Simulations in the pulsed
in which the magnetic field is in Faraday configuration, it excitation regime are presented in Sec. III B. In Sec. IV we
was later generalized to consider tilted magnetic fields by summarize the main results.
introducing off-diagonal terms [38]. However, this model is
strictly valid for Ga paramagnetic traps with 1/2 nuclear spins.
II. MASTER EQUATION
Although the SDR mechanisms are partially integrated to the
rate equation approach [35], the off-diagonal elements of the The system mainly consists of four elements: VB holes,
master equation are neglected and therefore the hyperfine CB electrons, unpaired traps, and paired traps. Whereas
interaction is accounted for by phenomenological transition hole spin relaxes with a characteristic time below 1 ps [45],
rates. This analysis has two major limitations: first, it only CB electronic spin relaxes on a typical time in the range
allows for the study of situations in which the magnetic field 100–400 ps [25,46]. Therefore, for the sake of simplicity, we
is in Faraday configuration and, second, it does not take consider VB holes to be unpolarized. Gallium paramagnetic
into account the full complexity of the hyperfine interaction traps or unpaired traps can be understood as a 3/2 nuclear spin
off-diagonal terms. Even though this analysis gives the correct coupled to a 1/2 bound electron spin via hyperfine interaction.
behavior for the conduction band electron spin polarization On the other hand, when a CB electron is captured by an
degree as function of the longitudinal magnetic field, it fails unpaired trap, forming a paired trap, the bound and captured
to explain the origin of the Overhauser-like effective field. It electrons produce a singlet state that cannot interact with the
considers that the Overhauser-like field is proportional to the nuclear spin. Thus, the quantum state basis that describes this
nuclear degree of polarization regardless of the nuclear spin. system must have: (i) one state for holes in the VB; (ii) two
This contradicts the fact that for 1/2 nuclear spins the magnetic states for the spin-up and spin-down CB electrons; (iii) eight
field shift is strictly zero even for nuclear spin polarization states that account for the nucleus-bound electron system in
degrees close to 100% [37]. the unpaired traps; and (iv) four states for the nuclear spin in

195204-2
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

the paired traps. The complete quantum state basis is therefore G

given by
  
B = |1 = |h,|2 = | ↓,|3 = | ↑,|4 = − 32 , ↓ ,
     
|5 = − 12 , ↓ ,|6 =  12 , ↓ ,|7 =  32 , ↓ , (a) CB (b) CB
      Ds
|8 = − 32 , ↑ ,|9 = − 12 , ↑ ,|10 =  12 , ↑ , (1)
3   3 
|11 =  2 , ↑ ,|12 = − 2 , ↑↓ , SDR
     
|13 = − 12 , ↑↓ ,|14 =  12 , ↑↓ ,|15 =  32 , ↑↓ ,
D2
where |h is the VB hole state and | ↑ and | ↓ are the
spin-up and spin-down CB electron states, respectively. The UT PT
following eight states |− 32 , ↓, |− 12 , ↓, . . . , | 32 , ↑ are the
bound-electron and nuclear spin states projected along the z VB VB
axis corresponding to the unpaired trap. Finally, the paired
traps are described by the nuclear spin states |− 32 , ↑↓, . . . ,
| 32 , ↑↓.
The dynamics of the four parts of the system and their
interactions can be described through the master equation
dρ i
= [ρ,H ] + D(ρ), (2)
dt h̄ (d) CB (c) CB
where ρ is the density matrix, H is the Hamiltonian and
D(ρ) is the dissipator. The Hamiltonian contains the internal
interactions among the four components of the system. We
Dsc
are interested in the combined effect of an external magnetic HFI
field and the hyperfine interaction in Ga centers, therefore, the
Hamiltonian must contain Zeeman and hyperfine interaction D1
terms. The interactions among the different parts of the system UT UT
and the surroundings are accounted for by the dissipator. Rec. to VB
These are mostly interactions with the electromagnetic field, VB VB
occurring during recombination or excitation processes, or
interactions with the nuclear spin environment. The main
processes introduced in our model are schematized in Fig. 1.

A. Hamiltonian: Zeeman and hyperfine interactions FIG. 1. Schematic diagram of the processes involved in the
The Hamiltonian is given by nuclear spin polarization of Ga centers. The flow of angular
momentum is also shown. Following the selection rules for GaAs,
H = h̄ω · Ŝ + h̄ · Ŝc + A Î 1 · Ŝc . (3) three spin-up, one spin-down CB electrons, and four unpolarized
The first and second terms in the Hamiltonian correspond to VB holes are generated by four photons with 100% left circular
the Zeeman interaction for conduction and bound electrons, polarization (a). The angular momentum of photons is transferred to
respectively. In these terms ω = gμB B/h̄,  = gc μB B/h̄, the CB electrons. Traps whose bound electrons are spin polarized in
the opposite direction to the majority of the CB electrons, can capture
B is the external magnetic field, μB is the Bohr magneton
CB electrons with opposite spin orientation forming a spin singlet (b).
and, g and gc are the CB and bound electrons gyromagnetic
In a course of this process the angular momentum of the CB electrons
factors. The hyperfine interaction term, the third one on the
is transferred partly to the bound electrons in traps. Simultaneously,
right-hand side of Eq. (3), couples the bound electron and CB electrons’ spin and paired traps nuclear spin relax due to DS and
the nuclear spin in an unpaired trap. The hyperfine coupling D2 leading to loss of angular momentum by the system. When one
constant is given by A. Other possible interactions between of the trapped electrons recombines with the hole the spin singlet in
the electron spins and the many surrounding nuclear spins are the paired trap is dissociated and turns into an unpaired trap (c). At
neglected. The strong localization of the 4s electrons wave the same time, the bound electron and nuclear spins in the unpaired
function in Ga centers [47] in the vicinity of the nucleus trap relax due to the DSC and D1 dissipators (c). Again the system
|ψ4s |2 = 72.7×1024 cm−3 yields a large hyperfine parameter transfers to angular momentum to the environment. At this stage the
A ≈ 8.5 eV coupling bound electrons and the corresponding bound electron and the nucleus are able to interact via the hyperfine
nuclei. Conversely, CB electrons weakly interact with the interaction and exchange angular momentum through the series of
surrounding nuclear spins giving their small probability den- flip-flops (d). From (d) to (a) the center can capture a new electron
sity |ψCB |2 ≈ 1/V = 1×1013 cm−3 , where V is the sample and the cycle starts again.
volume. Therefore, the hyperfine interaction between CB elec-
trons and all the surrounding nuclei is neglected. Moreover, cause a noticeable shift of the optically detected electron spin
the nuclear spin polarization of the host atoms surrounding resonance lines, which was not observed in high-resolution
the defects should be ruled out, because otherwise it would experiments [35,47]. The dynamic polarization of the host

195204-3
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

nuclei as the source of an Overhauser field must also be acting on ρ. In this manner any operator can be expanded as a
excluded from the model. The remarkable signature of such a linear combination of the elements of (5) as
field is an additional maximum that appears in the Hanle curve
and shifts with respect to the zero magnetic field [48–50]. A 85
Tr[λ̂q Ô]
Ô =  λ̂q . (8)
very recent experiment performed in GaAsN subjected to tilted Tr λ̂2q
q=1
magnetic fields [38] clearly shows the absence of the additional
peak. A very convenient set of operators is the one formed by the
The spin operator for CB electrons that appears in the generators of the unitary groups U (1) (VB holes), U (2) (CB
first Zeeman term of the Hamiltonian is Ŝ = (Ŝx ,Ŝy ,Ŝz ). electrons), U (4) (paired traps), and U (8) (unpaired traps). The
In unpaired traps, the bound electron spin and nuclear operators forming this set are not only of physical significance
spin operators entering the hyperfine interaction term are but they are also linearly independent and orthogonal with
Ŝc = (Ŝcx ,Ŝcy ,Ŝcz ), and Î 1 = (Iˆ1x ,Iˆ1y ,Iˆ1z ), respectively. As respect to the trace. Explicitly, the set of operators in (5) is
the singlet state formed in the paired traps interacts nei- given by
ther with the external magnetic field nor with their nu-  = {p̂,Ŝi ,Ûk,j,i ,T̂j,i }, i,j,k = 0,1,2,3, (9)
clear spin Î 2 = (Iˆ2x ,Iˆ2y ,Iˆ2z ), it does not appear in the
Hamiltonian. where p̂, Ŝi , Ûk,j,i , and T̂j,i generate the VB, CB, unpaired
trap, and paired trap blocks. The VB hole population density
B. Density matrix operator space operator can be represented by the matrix
⎛ ⎞
The relaxation mechanisms of CB electrons and nuclei in 1
the Ga centers are described by the dissipator D(ρ). Also, ⎜ ⎟
⎜ 02×2 ⎟
the photoexcitation and recombination of electrons will be p̂ = ⎜ ⎟. (10)
⎝ 08×8 ⎠
accounted for by D.
Our approach to formulating a suitable dissipator consists 04×4
in expanding the relevant operators as linear combination of The operators that generate the CB block can be compiled in
the elements of an operator vector space. In principle this set the matrix
should be formed in the basis (1) by linearly independent ⎛ ⎞
15×15 Hermitian matrices. However, the density matrix 0
⎜ ⎟
structure is considerably simplified by assuming that the four ⎜ ŝi ⎟
Ŝi = ⎜ ⎟, i = 0,1,2,3, (11)
components of the system (CB, unpaired trap, paired trap, VB) ⎝ 08×8 ⎠
are interconnected only by the dissipator. As the four parts of 04×4
the system are exclusively coupled by the recombination or
excitation processes, this is a reasonable assumption. Thus, the where the electron population density in the CB is given by
density matrix operator can be presented in the block diagonal n̂ = Ŝ0 and their spin operators are Ŝi for i = 1,2,3. Here
form: ŝ0 = 1̂2×2 is the 2×2 identity matrix, and ŝi for i = 1,2,3
⎛ ⎞ are the standard Pauli spin matrices that fulfill the usual spin
ρVB
commutation relations
⎜ ρCB ⎟ 
ρ=⎝ ⎠, (4)
ρ1 [ŝi ,ŝj ] = i ij k ŝk . (12)
ρ2 k=1,2,3

where the four blocks ρVB (1×1), ρCB (2×2), ρ1 (8×8), and This definition allows us to write the matrices that generate
ρ2 (4×4) are the partial density matrices of VB holes, CB the unpaired trap and paired trap blocks in the compact forms
electrons, unpaired traps, and paired traps, respectively. Given ⎛ ⎞
0
that ρ takes the form of a block diagonal matrix since no ⎜ ⎟
coherences can arise between the four components, it suffices ⎜ 02×2 ⎟
Ûk,j,i = ⎜ ⎟, (13)
to consider the smaller vector space of 85 Hermitian matrices ⎝ ŝk ⊗ ŝj ⊗ ŝi ⎠
that generate the four blocks. 04×4
We start by finding an internal space of Hermitian matrices
and
 = {λ̂1 ,λ̂2 , . . . ,λ̂n }, (5) ⎛ ⎞
0
which spans the n = 85 relevant elements of the 15×15 matrix. ⎜ ⎟
⎜ 02×2 ⎟
The generators in this set can be chosen in such a way that they T̂j,i =⎜ ⎟. (14)
⎝ 08×8 ⎠
are Hermitian and orthogonal with respect to the scalar product
ŝj ⊗ ŝi
given by the trace
†  According to this scheme the population density of unpaired
λ̂i ,λ̂j  ≡ Tr[λ̂i λ̂j ] = δi,j Tr λ̂2i . (6) traps is N̂1 = Û0,0,0 , and the one for paired traps is N̂2 = T̂0,0 .
This choice conveniently links the inner product with the Similarly, the CB electrons’ spin operators are: Ŝx = Ŝ1 ,
expected value of a given operator Ô Ŝy = Ŝ2 , and Ŝz = Ŝ3 . We have the same case for the bound
electron spin operator components in unpaired traps where
O = Tr[Ôρ] = ρ,Ô, (7) Ŝcx = Û1,0,0 , Ŝcy = Û2,0,0 , and Ŝcz = Û3,0,0 . The operators of

195204-4
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

the nuclear spin of unpaired traps and paired traps can be In the above equations, the population densities of CB
expressed as linear combinations of the elements of  as electrons and VB holes are given by n and p respectively.
The density of unpaired traps is N1 , N2 is the density of
electron singlets hosted by the centers (paired traps), and
Î 1 = MÛ , (15)
consequently Nc = N1 + N2 is the total density of Ga centers.
The vectors S and Sc represent the average spin polarizations
Î 2 = MT̂ , (16)
of free and bound electrons. The spin-dependent capture of
where free electrons in the Ga centers is ensured by the recombination
rate terms cn (nN1 − 4S · Sc ) and cn (SN1 − Sc n) where cn is a

constant. Notice that these two terms vanish when the system
Û = (Û0,0,1 ,Û0,1,1 ,Û0,2,2 ,Û0,0,2 ,Û0,1,2 ,Û0,2,1 ,Û0,0,3 ,Û0,3,0 ), is fully polarized, i.e., Sz = n/2, Sx = Sy = 0, Scz = N1 /2,
(17) and Scx = Scy = 0. The recombination rate of one of the
electrons trapped in the Ga centers to the VB is given by the
terms cp N2 p where cp is a constant. We thus require that the
T̂ = (T̂0,1 ,T̂1,1 ,T̂2,2 ,T̂0,2 ,T̂1,2 ,T̂2,1 ,T̂0,3 ,T̂3,0 ), (18)
dissipator’s structure is such that the master equation reduces to
and Eqs. (22)–(27) when the hyperfine interaction is lifted (A = 0).
This may be achieved by identifying n, p, N1 , N2 , S, and
⎡√ ⎤
3 2 2 Sc with the quantum statistical average of the corresponding
√0 0 0 0 0
M=⎣ 0 0 0 3 −2 2 0 0⎦. (19) operators, namely n = Tr[ρ n̂], p = Tr[ρ p̂], N1 = Tr[ρ N̂1 ],
0 0 0 0 0 0 1 2 N2 = Tr[ρ N̂2 ], S = Tr[ρ Ŝ], and Sc = Tr[ρ Ŝc ]. As well, the
quantum statistical average of any generator of  is given
by λq = Tr[ρ λ̂q ] and therefore, the density matrix can be
C. Dissipator expanded as
The dissipator can be separated into six parts
85
Tr[ρ λ̂q ] 85
λq
ρ=  λ̂q =  λ̂q . (28)
D(ρ) = G + DSDR + DS + DSC + D1 + D2 . (20) q=1
Tr λ̂q2
q=1
Tr λ̂2q

Here G contains the terms, which correspond to VB hole and The SDR part of the dissipator DSDR can also be expanded in
CB electron photogeneration. The selective capture of CB terms of the elements of  as
electrons in unpaired traps according to their relative spin
85
C[λ̂q ] 85
Cq
orientation and the subsequent recombination to the VB are DSDR =  λ̂q =  λ̂q . (29)
described by the DSDR dissipator. CB and bound electron spin q=1
2
Tr λ̂q q=1
Tr λ̂2q
relaxations are accounted for by DS and DSC . Nuclear spin
relaxation of unpaired traps and paired traps are introduced To determine the coefficients Cq ≡ C[λ̂q ] = Tr[λ̂q DSDR ] we
through the dissipators D1 and D2 . insert (29) in the master equation (2) and multiply by n̂, p̂, N̂1 ,
The term G that models the generation of electrons and N̂2 , Ŝ, or Ŝc . By taking the trace of the resulting equation we
holes is given by readily find the coefficients
G = (G↑ + G↓ )(p̂ + n̂) + 2(G↑ − G↓ )ê · Ŝ, (21)
C[p̂] = −cp pN2 , (30)
where G↑ and G↓ are the spin-up and spin-down electron
generation rates. The unit vector e is collinear with the incident C[n̂] = −cn (nN1 − 4S · Sc ), (31)
light.
To build the DSDR part of the dissipator we resort to the C[N̂1 ] = −cn (nN1 − 4S · Sc ) + cp N2 p, (32)
two-charge model [20,21,27,29] given by the following kinetic
equations (h̄ = 1)
C[N̂2 ] = cn (nN1 − 4S · Sc ) − cp N2 p, (33)

ṅ = −cn (nN1 − 4S · Sc ) + G↑ + G↓ , (22) C[ Ŝ] = −cn (SN1 − Sc n), (34)

ṗ = −cp N2 p + G↑ + G↓ , (23)
C[ Ŝc ] = −cn (Sc n − SN1 ). (35)
Ṅ1 = −cn (nN1 − 4S · Sc ) + cp N2 p, (24) At this point we have considerable freedom since these
equations only define 10 of the 85 coefficients needed to
Ṅ2 = cn (nN1 − 4S · Sc ) − cp N2 p. (25) fully determine the DSDR dissipator. However, the choices
1 G↑ − G↓ get narrowed down by imposing the symmetry and invariance
Ṡ = −cn (SN1 − Sc n) − S+ω×S+ ê, (26) properties that the system is expected to satisfy. As the most
τs 2
basic requirement, the master equation must be invariant under
1
Ṡc = −cn (Sc n − SN1 ) − Sc +  × Sc . (27) any arbitrary rotation in accordance with the space’s isotropy.
τsc The tensors T̂j,i and Ûk,j,i must therefore transform by the

195204-5
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

corresponding laws. The complete set of coefficients is thus where s = −1/2,1/2 and m = −3/2, − 1/2,1/2,3/2 are the
given by bound electron spin and nuclear spin indices. This dissipator
is highly symmetrical.
C[p̂] = −cp pT0,0 , (36) Second, we consider the relaxation due to the dipolar
interactions between neighboring Ga nuclei. In this case, the
 

3 Hamiltonian (A1) only contains the irreducible spherical ten-
C[n̂] = −cn S0 U0,0,0 − 4 Sr Ur,0,0 , (37) sors of rank k = 1. The terms in this Hamiltonian correspond
r=1 to the angular momentum operators interacting with a random
local field. Substituting (A6) and (A7) in (A5) we obtain the
C[Ŝk ] = −cn (Sk U0,0,0 − S0 Uk,0,0 ), (38) dissipators for unpaired traps and paired traps in the form
 

3
1  ˆ ˆ
3
C[Û0,j,i ] = cp pTj,i − cn S0 U0,j,i − 4 Sr Ur,j,i , (39) D1 = − [I1i ,[I1i ,ρ]], (46)
r=1 3τn1 i=1
1  ˆ ˆ
3
C[Ûk,j,i ] = −cn (S0 Uk,j,i − Sk U0,j,i ), (40)
D2 = − [I2i ,[I2i ,ρ]], (47)
3τn2 i=1
 

3
C[T̂j,i ] = −cp pTj,i + cn S0 U0,j,i − 4 Sr Ur,j,i , (41) where Iˆ1i and Iˆ2i are the ith components of the nuclear spin
r=1 operators for unpaired traps and paired traps, respectively. The
where i,j = 0,1,2,3 and k = 1,2,3. The spin precession terms nuclear spin relaxation times for unpaired and paired traps are
ω×S and ×Sc , absent in Eqs. (36)–(41), are accounted for considered to be different in principle and therefore are set to
by the Zeeman terms in the Hamiltonian (3) since they concern τn1 and τn2 .
the coherent evolution of the system. Finally, we study the relaxation due to the quadrupole
The SDR part of the dissipator also excludes the spin interaction with random fluctuation of the local electric field
relaxation of CB and bound electrons. These terms enter the gradient. The Hamiltonian takes the form of (A1) where k = 2.
dissipator through DS and DSC as The substitution of the irreducible spherical tensors of the rank
k = 2 defined by Eqs. (A8)–(A10) in terms of the nuclear
angular momentum components (A5) leads to the following
1  λq 2 
5 3
DS = −  λ̂q = − Si Ŝi , (42) dissipators
τs q=3 Tr λ2q τs i=1
1 
5
1   Uk,j,i
3 3
D1 = − [Q̂1,i ,[Q̂1,i ,ρ]], (48)
DSC =−  2 Ûk,j,i . (43) 2τn1 i=1
τsc k=1 j,i=0 Tr Ûk,j,i
1 
5

The CB electron spin relaxation time due to the Dyakonov- D2 = − [Q̂2,i ,[Q̂2,i ,ρ]]. (49)
2τn2 i=1
Perel mechanism is given by τs , while τsc is the phenomenolog-
ical bound electron spin relaxation time in Ga centers [36,37]. Here, the operators Q̂1i and Q̂2i are related to the rank k = 2
The dissipators (42) and (43) yield the CB and bound electron irreducible spherical tensors and therefore can be expressed in
spin relaxation terms S/τs and Sc /τsc in the two-charge model. terms of the nuclear spin operators as
D. Nuclear spin relaxation
1  2 
To get an insight into the role of the possible mechanisms Q̂n,1 = √ Iˆnx − Iˆny
2
, (50)
2 3
involved in nuclear spin relaxation we consider three different
1
models: nonselective, dipolar, and quadrupolar. As a reference Q̂n,2 = √ (Iˆnx Iˆny + Iˆny Iˆnx ), (51)
we also study the effects in the absence of spin relaxation. 2 3
The dipolar and quadrupolar interactions are dealt through the 1
Q̂n,3 = √ (Iˆnx Iˆnz + Iˆnz Iˆnx ), (52)
Wangsness, Bloch, and Redfield relaxation theory [40–43] 2 3
summarized in Appendix A. 1
First, we study the nonselective [37] dissipators for un- Q̂n,4 = √ (Iˆny Iˆnz + Iˆnz Iˆny ), (53)
2 3
paired traps and paired traps given by 1 2 
⎛ ⎞ Q̂n,5 = 2Iˆnz − Iˆny
2
− Iˆnx2
. (54)
6
δm,m 
3/2
1 ⎝
(D1 )sm,s  m = − ρ1;s,m;s  ,m − ρ1;s,m ;s  ,m⎠, The explicit forms of the dissipators corresponding to the
τn1 4 m =−3/2 dipolar interaction (46)–(47) and quadrupole interaction
(48)–(49) are presented in Appendix B.
(44)
⎛ ⎞

3/2 III. RESULTS AND DISCUSSION
1 δm,m
(D2 )m,m = − ⎝ρ2;m,m − ρ2;m ,m⎠, (45) The model developed above is used in this section to
τn2 4 m =−3/2
examine the interplay of the hyperfine interaction and the

195204-6
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

nuclear spin relaxation mechanisms in the continuous wave

and pulsed excitation regimes. First, the theoretical results
are compared with previous experimental observations under
continuous wave excitation in order to identify the main
interaction behind nuclear spin relaxation and hyperfine
interaction in Ga centers. Then, we analyze spin dynamics
of the bound electrons and nuclei in the pulsed excitation
regime. We outline a method for detecting the bound electron
and nuclear spin coherent oscillations induced by hyperfine
interaction by means of a pump-probe PL scheme.
In order to extract information from the model, we start
by building the system of kinetic equations that follow from
the master equation (2). By multiplying both sides of Eq. (2)
by λ̂q , inserting the density matrix in the form (28) into the
resulting expression and taking the trace we obtain a set of
n = 85 differential equations of the form
i FIG. 2. PeCB calculated as a function of a magnetic field in the
λ̇q = Tr[[H,λ̂q ]ρ] + Tr[Dλ̂q ] Faraday configuration for different pump powers and nonselective
h̄
nuclear spin relaxation under σ + (solid lines) and σ − (dashed
= Fq (λ1 ,λ2 , . . . ,λn ,t), q = 1,2, . . . ,n. (55)
lines). The nuclear spin relaxation times are τn1 = 5405 ps and
Unlike the two-charge model that only considers the SDR τn2 = 446 ps. Dotted horizontal lines present the behavior of CB
mechanism, Zeeman interaction, and electron spin relaxation, electron polarization in the absence of nuclear spin relaxation.
these new kinetic equations also take into account the hyperfine
interaction and nuclear spin relaxation. As we stated above, the two key features behind the
We study the spin dynamics of electrons and nuclei hyperfine interaction in Ga centers are a growth in the PL
by numerically solving the system of ordinary differential intensity J and degree of circular polarization PeCB [32,34–37]
equations (55). The relevant parameters are then extracted from and an Overhauser-like magnetic field [32,33,35–37]. They are
the thus obtained λq functions, which in turn are the quantum observed under circularly polarized excitation and a Faraday
statistical averages. We assume that before the optical excita- configuration magnetic field Bz .
tion (t = 0) the unpaired traps are equally populated and that More specifically for the first feature, PeCB (Bz ) and J (Bz )
the electrons as well as the nuclei are completely unpolarized. exhibit a minimum shifted from Bz = 0. As |Bz | increases,
Therefore, initially N1 (0) = λ6 (0) = Nc and λq = 0 for q = 6. PeCB and J (Bz ) saturate at values above Bz ≈ 25 mT where
Notice that these initial conditions also imply that at this stage Zeeman energies are comparable to the hyperfine interaction.
there are no paired traps, namely N2 (0) = λ70 (0) = 0. In this region the hyperfine interaction has been completely
exceeded by the Zeeman interaction and therefore the bound
A. Nuclear spin relaxation: Continuous wave regime electrons and nuclei in the Ga centers are effectively decou-
pled. The degree of circular polarization of the CB to VB
Under continuous wave stimulation, the generation of spin
photoluminescence as a function of the Faraday configuration
up and spin down electrons is given by the smooth step function
magnetic field can be described by an inverted Lorentzian-
  
1 ± Pi t − t0 like curve. In Fig. 2 we have calculated the CB electron
G↑↓ = W G 1 + tanh , (56) spin polarization degree PeCB = 2Sz /n as a function of the
4 σ
magnetic field in Faraday configuration Bz under a circularly
where W is the excitation power, G= polarized excitation. In this case we have chosen the spin
2.3×1023 mW−1 s−1 cm−3 is the power to generated electron relaxation times τn1 = 5405 ps and τn2 = 446 ps that give
ratio, Pi = ±0.15 is the spin polarization degree of the good quantitative agreement with the experiment. The PeCB =
optically generated CB electrons, t0 = 100 ps is the onset 2Sz /n dependence for the nonselective mechanism is shown in
time of the excitation and σ = 10 ps is the duration of the Fig. 2 for different pump powers W . As a reference, Fig. 2 also
onset. The system is allowed to evolve for a sufficiently long presents the behavior observed in the absence of nuclear spin
time (200 ns) to reach steady-state conditions. relaxation (D1 = D2 = 0) as thick dotted lines. These plots
Some of other parameters as Nc = 3×1015 cm−3 , show that despite the hyperfine interaction, in the absence of
τs = 110 ps, τsc = 1700 ps, τ ∗ = 1/cn Nc = 4.4 ps, τh∗ = spin relaxation, PeCB (Bz ) does not display any sign of the spin
1/cp Nc = 12 ps, g = +1, and gc = +2 where estimated from filtering enhancement. These results are exactly the same as
previous experimental results [20,21,32,51]. For the nuclei those obtained with the two charge model that does not contain
at the Ga centers, the hyperfine parameter was estimated the effects of the hyperfine interaction. Therefore, in order for
to be A = 6.9×10−2 cm−1 = 8.5 μeV, the average hyperfine the effects of the hyperfine interaction to be visible, nuclear
parameter of the two stable isotopes of Ga atoms [22,32,36]. spin relaxation is essential.
The nuclear spin relaxation times τn1 and τn2 are determined The second experimental feature of this phenomenon is
below by comparing the theoretical calculations with the a shift of the minimum of the plots PeCB (Bz ) and J (Bz )
experimental results. with respect to Bz = 0 that points to the existence of an

195204-7
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

FIG. 3. Conduction band electron spin polarization PeCB as a function of the magnetic field in the Faraday configuration for pumping power
W = 25, 75, 150, and 250 mW. Theoretical results for the (a) dipolar and (b) quadrupolar models under σ + (solid lines) and σ − (dashed lines)
light excitation. (c) Experimental results under σ + (circles) and σ − (squares) light excitation [32,37], solid lines are the guides for the eye. The
insets in (a) and (b) show the detail of the displacement of the PeCB curves for σ + (solid lines) and σ − (dashed lines) that is induced by the
Overhauser-like magnetic field at W = 75 mW.

Overhauser-like magnetic field [33]. The PeCB (Bz ) and J (Bz ) yield nonvanishing Beff as it can be seen in Figs. 3(a)–3(b)
curves are shifted to the positive or negative magnetic field and 4(a)–4(b). In Figs. 4(a)–4(b) the PL intensity is calculated
depending on the helicity of the circularly polarized exciting as J (Bz ) = cr np where cr is the bimolecular recombination
light. Thereby, under σ − and σ + excitation the minimum is rate.
located at Bz = Beff < 0 and Bz = Beff > 0, respectively. The Even though the calculated shifts produced by both mech-
experimental results show that the Overhauser-like magnetic anisms qualitatively agree with the experimental observations
field Beff grows with the excitation power W until it apparently [see Figs. 3(c) and 4(c)], only the dipolar one is able to accu-
saturates at approximately 25 mT. The nonselective dissipator rately fit the experimental measurements as we discuss below.
yields vanishing Beff as no shift is observed for PeCB in Fig. 2. In addition to the Overhauser-like magnetic field, two features
This dissipator is too symmetric to be able to produce an that strongly depend on the nuclear spin relaxation mechanism
Overhauser-like magnetic field and hence must be ruled out as are the depths of the inverted Lorentzian-like PeCB (Bz ) and
the leading nuclear spin relaxation mechanism. J (Bz ) curves, which parametrized by ξ = PeCB (∞)/PeCB (0)
The dipolar and quadrupolar mechanisms are quite dif- and ζ = J (∞)/J (0), respectively. To discern which of the two
ferent from the nonselective one. Figures 3 and 4 show the mechanisms is the dominant one, we compare our theoretical
effect of the dipolar and quadrupolar nuclear spin relaxation calculations with the experimental observations of Beff , ξ ,
mechanisms on the CB electron spin polarization PeCB (Bz ) and ζ [32,33,37]. The power dependence of Beff , ξ , and
and the PL intensity J (Bz ) as functions of the longitudinal ζ is determined by finding the minima, PeCB (0) and J (0),
magnetic field at different excitation powers. Contrary to the and maxima, PeCB (∞) and J (∞), of PeCB (Bz ) and J (Bz ) for
nonselective mechanism, the dipolar and quadrupolar ones different excitation powers.

FIG. 4. PL intensity J as a function of the magnetic field in the Faraday configuration for pumping power W = 25, 75, 150, and 250 mW.
Theoretical results for the (a) dipolar and (b) quadrupolar models under σ + (solid lines) and σ − (dashed lines) light excitation. (c) Experimental
results under σ + (circles) and σ − (squares) light excitation [32,37], solid lines are the guides for the eye. The insets in (a) and (b) show the
detail of the displacement of the J curves for σ + (solid lines) and σ − (dashed lines) that is induced by the Overhauser-like magnetic field at
W = 75 mW.

195204-8
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

FIG. 5. Isolines of the Overhauser-like magnetic field Beff and

depths ξ = PeCB (∞)/PeCB (0) and ζ = J (∞)/J (0) at fixed power
W = 120 mW for the dipolar relaxation mechanism. The isolines for
Beff = 24 mT, ξ = 1.46 and ζ = 2.1 under an excitation power of FIG. 6. Overhauser-like magnetic field Beff , ξ , and ζ as functions
W = 120 mW are shown. These three lines cross at the point marked of the excitation power W for the dipolar nuclear spin relaxation. In
with the triangle in τn1 = 212.4 ps = 5405 ps and τn2 = 28.8 ps = (a), the experimental data (circles) present the shift of the free electron
446 ps. polarization dependence PeCB (Bz ) and the simulated curve (line). In
(b), shows the experimental (circles) and theoretical (line) results for
ξ . The experimental data (circles) in (c) present the shift drawn from
In Fig. 5 we plot the isolines for Beff = 24 mT, ξ = 1.46, the PL intensity dependence J (Bz ) and the corresponding theoretical
and ζ = 2.1 as functions of the nuclear spin relaxation times curve (line). In (d), presents the experimental (circles) [32,33,37] and
τn1 and τn2 . These three correspond to the experimental theoretical (line) results for ζ .
results observed for an excitation power of W = 120 mW
[32]. The three isolines intersect at τn1 = 5405 ps and τn2 =
446 ps. In accordance with these results, the Beff , ξ , and excitation powers reported experimentally and therefore we
ζ isolines at other excitation power coincide at similar τn1 must also rule out the quadrupolar mechanism.
and τn2 values. Collecting the intersecting points of all the We observe that the calculated widths (≈ 0.3 T) of the
experimental results we find that the nuclear spin relaxation inverted Lorentzian curves for either mechanism do not
times must fall within the ranges 4000 ps < τn1 < 9500 ps coincide with the experimental ones (≈ 0.2 T). The calculated
and 400 ps < τn2 < 700 ps. The best quantitative agreement
with the experimental data is accomplished by using the
nuclear spin relaxation times τn1 = 5405 ps and τn2 = 446 ps
consistent with the ranges above.
Plots of Beff , ξ , and ζ as functions of the excitation power
are shown in Figs. 6(a)–6(d). Whereas Fig. 6(a) shows the
magnetic field Beff extracted from the minima of the ξ (Bz )
curves, Fig. 6(c) plots the Beff extracted from the ζ (Bz )
minima. It is worth noting that both Beff obtained from the
ξ (Bz ) and ζ (Bz ) minima yield almost identical experimental
and theoretical results. The calculated ξ and ζ as functions
of the power present very good quantitative agreement with
the experimental data. Likewise, the theoretical values of
the Overhauser-like magnetic field Beff remarkably coincide
with the experimental results within the range of power
values shown in Figs. 6(a) and 6(c). Nevertheless, while
the experimental results suggest that Beff as a function of
the excitation power saturates at approximately 25 mT, the
computed Beff vanishes for powers above 250 mW [not shown FIG. 7. Isolines of the Overhauser-like magnetic field Beff and
in Figs. 6(a) or 6(c)]. depths ξ = PeCB (∞)/PeCB (0) and ζ = J (∞)/J (0) at fixed power for
Contrary to the dipolar mechanism, the quadrupolar one the quadrupolar relaxation. The isolines for Beff = 24 mT, ξ = 1.46,
yields systematically nonintersecting isolines regardless of the and ζ = 2.1 under an excitation power of W = 120 mW are shown.
excitation power value used to calculate them. In Fig. 7 we Even though these three lines do not cross we have marked with a
present the Beff = 24 mT, ξ = 1.46, and ζ = 2.1 isolines that triangle the point τn1 = 5405 ps and τn2 = 446 ps that yield the best
clearly do not intersect. This behavior is observed for all the fit.

195204-9
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

width remains constant for a wide range of parameters and thus

it is not related to the nuclear spin relaxation mechanisms. This
discrepancy will be discussed elsewhere.
It is important to verify that the nuclear bath is fast enough
for the the Redfield theory and the master equation approach
to be valid for the dipolar mechanism. The nonperturbed
Hamiltonian H0 contains mainly the hyperfine interaction
and the Zeeman energy and therefore |H0 | = h̄ω0 ≈ 10 μeV.
On the other hand, the perturbed Hamiltonian corresponds to
the dipolar interaction between neighbouring nuclei and thus
|H1 | ≈ h̄ω1 ≈ 0.4–0.6 peV. Equation (A12) then yields
τc  ω1−1 ≈ 1.14–1.84 ms. (57)
FIG. 8. Time dependence of the spin polarization of bound
The correlation time is obtained from Eq. (A13) electrons Scz (solid lines) and nuclei I1z (dashed lines) after being
1 excited by the pump pulse. The pump pulse is left circularly polarized
τc = ≈ 1 ps. (58) and therefore the majority of the electrons are spin polarized in the
τn ω02 +z direction. Two extreme situations are illustrated. In the probe
Consequently the time range of validity for the master equation pulse 1 bound electrons are spin polarized in the same direction
approach in this case is given by as CB electrons and in the probe pulse 2 bound electrons are spin
polarized in the opposite direction to CB electrons. In the first
1 ps  t  1 ms, (59) situation CB electrons with the opposite spin polarization to the
majority are rapidly recombined through the Ga centers enhancing the
in accordance with the results obtained here. spin filtering effect. In this case a large SDRr is expected. In contrast,
in the second situation, CB electrons whose spin polarization is that
B. Coherent oscillations of electronic and nuclear spins of the majority are efficiently recombined lowering the SDRr .
in Ga centers: Pulsed excitation regime
A good estimate of the electron and hole population in either
Having identified the nuclear spin relaxation mechanism situation is the time-resolved SDR ratio given by
and the corresponding relaxation times we are in a position
to consider time-resolved simulations. Our aim here is to J− nσ (t)pσ− (t)
SDRr (t) = = − , (61)
develop a method to observe the coherent oscillations of bound JX nπX (t)pπX (t)
electrons and nuclei in Ga centers. To do so we propose a
pump-probe scheme that we describe below. where the PL intensity under circularly polarized light J− ∝
In pulsed excitation regime, the generation terms are given nσ− (t)pσ− (t) is proportional to the CB and VB density
by populations nσ− (t) and pσ− (t). Similarly JX ∝ nπX (t)pπX (t)
where nπX (t) and pπX (t) are the density populations of CB
T W G 1 ± Pi  − t 22 (t−δt)2 electrons and holes under linearly polarized light. If CB
G↑↓ = √ e 2σ + ηe− 2σ 2 , (60)
σ 2π 2 electrons are captured spin dependently by the Ga centers then
nσ− (t) > nπX (t) and pσ− (t) > pπX (t) and therefore SDRr > 1.
where W is the pulse average power, σ = 1 ps is its width In accordance with the above considerations, SDRr (δt1 ) >
and T = 12 ns is the period between repeated pulses [31]. The SDRr (δt) > SDRr (δt2 ) where δt1 < δt < δt2 . Thus, it is
pump pulse originates at t = 0 and the probe is delayed δt. possible to trace the oscillations of bound electrons and
Figure 8 is an outline of the proposed method. The pump
pulse is left circularly polarized and therefore most of the
electrons are spin polarized in the +z direction. Likewise, the
probe pulses are left circularly polarized Fig. 8 also shows a
plot of the bound electron and nuclear spin polarizations as a
function of time after being excited by the pump pulse. The
pump and probe pulse’s widths are exaggerated to make them
visible in the given time scale. In this plot it is possible to
observe the electron-nucleus flip-flops as oscillations of I1z
(dashed lines) and Scz (thick lines) that are phase shifted by
π/2. As indicated in this diagram, the second pulse at time
delays t1 and t2 probes two extreme situations. In the first one
CB electrons and bound electrons are mostly spin polarized
in opposite directions. The center is therefore more likely to
capture a CB electron whose spin is oriented in the opposite FIG. 9. Trace of the coherent electron-nuclear spin oscillations
direction to the majority therefore rising the population of via the SDRr . The solid circles correspond to maxima of the time-
electrons in the CB. In contrast, in the second situation centers resolved SDRr as a function of the time delay δt between the pump
are more likely to capture electrons whose direction is parallel and the probe pulses. Below, Scz (solid line) and I1z (dashed line) are
to the majority diminishing the electron population in the CB. presented for reference.

195204-10
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

nuclei by successively measuring the time-resolved SDRr for This work was partially supported by Programme Investisse-
different probe pulse delays. ments d’Avenir under the program ANR-11-IDEX-0002-
By determining the maxima of the time-resolved SDRr 02, reference ANR-10-LABX-0037-NEXT. A.K. gratefully
for different probe pulse delays we have obtained the plot appreciates the financial support of “Departamento de Ciencias
displayed in Fig. 9 setting η = 1. Here the SDRr maxima Básicas UAM-A” Grants No. 2232214 and No. 2232215.
are plotted as a function of their corresponding time delays J.C.S.S. and V.G.I.S. would like to acknowledge the support
δt as closed circles. Similar results (not shown here) are received from the “Becas de Posgrado UAM” Scholarships
obtained by calculating the SDRr from the integrated PL. No. 2151800745 and No. 2112800069. We are indebted
Below, the spin polarization of bound electrons Scz and to Professor J. Grabinsky for the careful reading of the
nuclear spin polarization I1z are shown for reference. This plot manuscript.
demonstrates that it is possible to trace the coherent oscillations
of the spin polarization of bound electrons interacting with the APPENDIX A: REDFIELD RELAXATION THEORY
nuclei by means of the time-resolved SDR ratio.
According to the Wangsness, Bloch, and Redfield relax-
ation theory [40–43] the interaction of a nucleus with its
IV. SUMMARY surroundings can be accounted for by the Hamiltonian
We have analyzed the spin dynamics of electrons and  k
∗
nuclei in GaAsN by developing a model based on the H(t) = γ fk,r (t)T k,r , (A1)
master equation approach. The main mechanisms behind r=−k
the spin-dependent recombination are considered as well as where γ is a constant, T k,r is a rth component of the rank k
the hyperfine interaction in Ga paramagnetic traps. We have irreducible spherical tensor and fk,r (t) is a random function
demonstrated that the nuclear spin relaxation in centers plays that describes the interaction with the surroundings.
an essential role in reproducing the two most significant To second order, the average fluctuations of the surround-
signatures of the hyperfine interaction in Ga centers. First, ings with the nucleus are given by the following dissipator
the amplification of the spin filtering effect under a Faraday
configuration magnetic field is visible only if some nuclear  t
1
spin relaxation mechanism is present. Second, the features of D=− 2 dt  [H(t),[H(t  ),ρ̄]], (A2)
h̄ −∞
the Overhauser-like magnetic field not only depend on the
hyperfine interaction but also strongly rely on the nature of the where ρ̄ is the average density matrix. By substituting the
∗
nuclear spin relaxation mechanism. We have tested the dipolar Hamiltonian (A1) in (A2) and using the fact that fk,r (t) =
†
interaction between neighboring Ga atoms and the quadrupolar (−1) fk,−r (t) and T k,r = (−1) T k,−r we get the general form
r r

interaction of Ga centers with random charge distribution of the dissipator

background. We have proven that the dipolar is the only
1 
k
†
mechanism compatible with the experimental observations. D=− 2 [T k,s ,[T k,r ,ρ̄]]
Indeed, a scenario where large charge distribution variations h̄ s,r=−k
are present in the vicinity of the Ga nuclei is difficult  t 
to imagine. Although most of the experimental results are × dt  γ 2 fk,s (t)fk,r
∗
(t  ) . (A3)
correctly reproduced by our model some aspects remain −∞
elusive. Two of these are the discrepancies between the The above dissipator describes the interaction of a nucleus
saturation values of the Overhauser-like magnetic field Beff with the fluctuations of a random electromagnetic field. The
in the high-power regime and the widths of the PeCB (Bz ) and ∗
functions fk,s (t) and fk,r (t  ) satisfy the relation
J (Bz ) curves. 

To further explore the effects of the hyperfine interaction

∗
γ 2 fk,s (t)fk,r (t  ) = δs,r ξ e−|t−t |/τc , (A4)
and the nuclear spin relaxation we have tested the model in 
where ξ e−|t−t |/τc is a correlation function, τc is the correlation
the pulsed excitation regime. In particular, we have proposed time of the fluctuating field and ξ is the correlation amplitude
a pump-probe scheme that allows us to trace the coherent when t = t  . With (A4), the dissipator D (A3) is simplified to
oscillations of the bound electron spin interacting with its
1  †
k
nucleus through the hyperfine interaction.
Even though in principle this model is conceived for Ga DNSR = − [T ,[T k,r ,ρ̄]], (A5)
2τn r=−k k,r
centers, it can be easily adapted for other type of centers where
dipolar or quadrupolar interactions are the leading mechanisms where τn = h̄2 /2ξ τc is the nuclear spin relaxation time.
of nuclear spin relaxation. In the case of magnetic dipole interactions, only the
irreducible spherical tensors of rank k = 1 participate in the
ACKNOWLEDGMENTS Hamiltonian (A1). They can be expressed in terms of the
nuclear spin components as
French and Russian authors acknowledge funding from LIA
CNRS-Ioffe RAS ILNACS. L.A.B. E.L.I. thanks the Russian T 1,0 = Iˆz , (A6)
Foundation for Basic Research (Grants No. 17-02-00383 and
No. 17-52-16020). V.K.K. acknowledges the financial support 1
T 1,±1 = ∓ √ (Iˆx ± i Iˆy ). (A7)
of the Government of Russia (Project No. 14.Z50.31.0021). 2

195204-11
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

In the case of electric quadrupole interaction k is equal to 2 1  

and the components of irreducible spherical tensors of rank 2 (D2 ) 21 , 32 = − 3ρ2; 12 , 32 − ρ2;− 32 ,− 12 , (B8)
2τn2
are 1
(D2 ) 21 ,− 12 = − ρ2; 12 ,− 12 , (B9)
 2 τn2
T 2,0 = 1
2Iˆz − Iˆy2 − Iˆx2 , (A8)
6 1
(D2 )− 1 , 1 = − ρ2;− 1 , 1 , (B10)
1 2 2 τn2 2 2
T 2,±1 = ∓ √ [Iˆx Iˆz + Iˆz Iˆx ± i(Iˆy Iˆz + Iˆz Iˆy )], (A9)
2 6 1  
(D2 ) 23 ,− 12 = − 3ρ2; 32 ,− 12 + ρ2; 12 ,− 32 , (B11)
1   2τn2
T 2,±2 = √ Iˆx2 − Iˆy2 ± i(Iˆx Iˆy + Iˆy Iˆx ) . (A10)
2 6 1  
(D2 ) 21 ,− 32 = − 3ρ2; 12 ,− 32 + ρ2; 32 ,− 12 , (B12)
In order for the Redfield theory and the master equation 2τn2
approach to be applicable the following conditions must be 1  
fulfilled: (i) the bath’s correlation time τc should be sufficiently (D2 )− 32 , 12 = − 3ρ2;− 32 , 12 + ρ2;− 12 , 32 , (B13)
2τn2
short to satisfy the inequality
1  
(D2 )− 1 , 3 = − 3ρ2;− 1 , 3 + ρ2;− 3 , 1 , (B14)
τc ω1  1, (A11) 2 2 2τn2 2 2 2 2

and the density matrix should be valid in the time scale t given 1
(D2 ) 23 ,− 32 = − ρ 3 3, (B15)
by τn2 2; 2 ,− 2
1
τc  t  ω1−1 , (A12) (D2 )− 32 , 32 = − ρ 3 3. (B16)
τn2 2;− 2 , 2
where ω1 is a characteristic frequency of the perturbation In a short form the above equations can be presented as
Hamiltonian |H1 | [43,52,53]; (ii) furthermore, the relaxation
times should be longer than the correlation time 1  (EQ)
(D2 )m,m = − Qm,m ;m1 ,m ρ2;m1 ,m1 .
τc  τn . (A13) τn2 
1
m1 ,m1

The correlation time in the previous expression may be Then the quadrupolar dissipator for unpaired trap is given by
estimated from [54]
1  (EQ)
1 (D1 )s,m;s  ,m = − Qm,m ;m1 ,m ρ1;s,m1 ;s  ,m1 .
τc ≈ , (A14) τn1 
m1 ,m1
1

τn ω02
where h̄ω0 = |H0 | is the characteristic energy of the unper- The dipolar dissipator for paired traps is given explicitly by
turbed Hamiltonian.
1  
(D2 ) 32 , 32 = − ρ2; 32 , 32 − ρ2; 12 , 12 , (B17)
τn2
APPENDIX B: EXPLICIT FORM OF THE DIPOLAR 1  
AND QUADRUPOLAR INTERACTIONS (D2 )− 3 ,− 3 = − ρ2;− 3 ,− 3 − ρ2;− 1 ,− 1 , (B18)
2 2 τn2 2 2 2 2

The quadrupolar dissipator for paired trap is given explicitly 1  

(D2 ) 12 , 12 = − 7ρ2; 12 , 12 − 3ρ2; 32 , 32 − 4ρ2;− 12 ,− 12 , (B19)
by 3τn2
 1  
1 ρ2; 12 , 12 + ρ2;− 12 ,− 12  (D2 )− 12 ,− 12 = − 7ρ2;− 12 ,− 12 − 3ρ2;− 32 ,− 32 − 4ρ2; 12 , 12 ,
(D2 ) 32 , 32 = − ρ 3 3− , (B1) 3τn2
τn2 2; 2 , 2 2
 (B20)
1 ρ2; 12 , 12 + ρ2;− 12 ,− 12 
(D2 )− 32 ,− 32 = − ρ2;− 32 ,− 32 − , (B2) 2  √ 
τn2 2 (D2 ) 32 , 12 = − 3ρ2; 32 , 12 − 3ρ2; 12 ,− 12 , (B21)
 3τn2
1 ρ2; 32 , 32 + ρ2;− 32 ,− 32  2  √  
(D2 ) 1 , 1 = − ρ2; 1 , 1 − , (B3) (D2 ) 21 ,− 12 = − 4ρ2; 12 ,− 12 − 3 ρ2; 32 , 12 + ρ2;− 12 ,− 32 ,
2 2 τn2 2 2 2 3τn2
 ρ2; 3 , 3 + ρ2;− 3 ,− 3 
1 (B22)
(D2 )− 12 ,− 12 = − ρ 1 1− 2 2 2 2
, (B4)
τn2 2;− 2 ,− 2 2 √
2  
1   (D2 )− 12 ,− 32 = − 3ρ2;− 12 ,− 32 − 3ρ2; 12 ,− 12 , (B23)
(D2 ) 3 , 1 = − 3ρ2; 3 , 1 − ρ2;− 1 ,− 3 , (B5) 3τn2
2 2 2τn2 2 2 2 2
2  √ 
(D2 ) 12 , 32 = − 3ρ2; 12 , 32 − 3ρ2;− 12 , 12 , (B24)
1   3τn2
(D2 )− 32 ,− 12 = − 3ρ2;− 32 ,− 12 − ρ2; 12 , 32 , (B6) √ 
2τn2 2  
(D2 )− 12 , 12 = − 4ρ2;− 12 , 12 − 3 ρ2; 12 , 32 + ρ2;− 32 ,− 12 ,
1   3τn2
(D2 )− 12 ,− 32 = − 3ρ2;− 12 ,− 32 − ρ2; 32 , 12 , (B7) (B25)
2τn2

195204-12
ELECTRON-NUCLEAR SPIN DYNAMICS OF Ga2+ . . . PHYSICAL REVIEW B 95, 195204 (2017)

2  √  4
(D2 )− 3 ,− 1 = − 3ρ2;− 3 ,− 1 − 3ρ2;− 1 , 1 , (B26) (D2 ) 3 ,− 3 = − ρ 3 3, (B31)
2 2 3τn2 2 2 2 2 2 2 τn2 2; 2 ,− 2
1   4
(D2 ) 3 ,− 1 =− 3ρ2; 3 ,− 1 − ρ2; 1 ,− 3 , (B27) (D2 )− 32 , 32 = − ρ2;− 32 , 32 . (B32)
2 2 τn2 2 2 2 2 τn2

1   If the above equations are presented as

(D2 ) 21 ,− 32 = − 3ρ2; 12 ,− 32 − ρ2; 32 ,− 12 , (B28) 1  (MD)
τn2 (D2 )m,m = − Qm,m ;m1 ,m ρ2;m1 ,m1 ,
τn2 1
1   
m1 ,m1
(D2 )− 32 , 12 = − 3ρ2;− 32 , 12 − ρ2;− 12 , 32 , (B29)
τn2 then the dipolar dissipator for unpaired trap is given by
1  (MD)
1   (D1 )s,m;s  ,m = − Qm,m ;m1 ,m ρ1;s,m1 ;s  ,m1 .
(D2 )− 12 , 32 = − 3ρ2;− 12 , 32 − ρ2;− 32 , 12 , (B30) τn1 
1
τn2 m1 ,m1

[1] A. Auer and G. Burkard, Phys. Rev. B 93, 035402 (2016). [18] P.-F. Braun, X. Marie, L. Lombez, B. Urbaszek, T. Amand, P.
[2] T. van der Sar, Z. H. Wang, M. S. Blok, H. Bernien, T. H. Renucci, V. K. Kalevich, K. V. Kavokin, O. Krebs, P. Voisin,
Taminiau, D. M. Toyli, D. A. Lidar, D. D. Awschalom, R. and Y. Masumoto, Phys. Rev. Lett. 94, 116601 (2005).
Hanson, and V. V. Dobrovitski, Nature (London) 484, 82 [19] E. Abe, A. M. Tyryshkin, S. Tojo, J. J. L. Morton, W. M. Witzel,
(2012). A. Fujimoto, J. W. Ager, E. E. Haller, J. Isoya, S. A. Lyon,
[3] N. Mizuochi, P. Neumann, F. Rempp, J. Beck, V. Jacques, M. L. W. Thewalt, and K. M. Itoh, Phys. Rev. B 82, 121201
P. Siyushev, K. Nakamura, D. J. Twitchen, H. Watanabe, S. (2010).
Yamasaki, F. Jelezko, and J. Wrachtrup, Phys. Rev. B 80, 041201 [20] V. K. Kalevich, E. L. Ivchenko, M. M. Afanasiev, A. Y.
(2009). Shiryaev, A. Y. Egorov, V. M. Ustinov, B. Pal, and Y. Masumoto,
[4] J. R. Maze, J. M. Taylor, and M. D. Lukin, Phys. Rev. B 78, JETP Lett. 82, 455 (2005).
094303 (2008). [21] V. K. Kalevich, A. Y. Shiryaev, E. L. Ivchenko, A. Y. Egorov, L.
[5] J. J. Pla, F. A. Mohiyaddin, K. Y. Tan, J. P. Dehollain, R. Rahman, Lombez, D. Lagarde, X. Marie, and T. Amand, JETP Lett. 85,
G. Klimeck, D. N. Jamieson, A. S. Dzurak, and A. Morello, 174 (2007).
Phys. Rev. Lett. 113, 246801 (2014). [22] X. J. Wang, Y. Puttisong, C. W. Tu, A. J. Ptak, V. K. Kalevich,
[6] B. E. Kane, Nature (London) 393, 133 (1998). A. Y. Egorov, L. Geelhaar, H. Riechert, W. M. Chen, and I. A.
[7] J. J. Pla, K. Y. Tan, J. P. Dehollain, W. H. Lim, J. J. L. Morton, Buyanova, Appl. Phys. Lett. 95, 241904 (2009).
D. N. Jamieson, A. S. Dzurak, and A. Morello, Nature (London) [23] H. M. Zhao, L. Lombez, B. L. Liu, B. Q. Sun, Q. K. Xue, D. M.
489, 541 (2012). Chen, and X. Marie, Appl. Phys. Lett. 95, 041911 (2009).
[8] T. D. Ladd, D. Maryenko, Y. Yamamoto, E. Abe, and K. M. Itoh, [24] F. Zhao, A. Balocchi, A. Kunold, J. Carrey, H. Carrre, T. Amand,
Phys. Rev. B 71, 014401 (2005). N. Ben Abdallah, J. C. Harmand, and X. Marie, Appl. Phys. Lett.
[9] A. Laucht, R. Kalra, J. T. Muhonen, J. P. Dehollain, F. A. 95, 241104 (2009).
Mohiyaddin, F. Hudson, J. C. McCallum, D. N. Jamieson, [25] F. Zhao, A. Balocchi, G. Truong, T. Amand, X. Marie, X. J.
A. S. Dzurak, and A. Morello, Appl. Phys. Lett. 104, 092115 Wang, I. A. Buyanova, W. M. Chen, and J. C. Harmand,
(2014). J. Phys.: Condens. Matter 21, 174211 (2009).
[10] W. Yao, R.-B. Liu, and L. J. Sham, Phys. Rev. B 74, 195301 [26] Y. Puttisong, X. J. Wang, I. A. Buyanova, H. Carrre, F. Zhao,
(2006). A. Balocchi, X. Marie, C. W. Tu, and W. M. Chen, Appl. Phys.
[11] V. Ivády, K. Szász, A. L. Falk, P. V. Klimov, D. J. Christle, Lett. 96, 052104 (2010).
E. Janzén, I. A. Abrikosov, D. D. Awschalom, and A. Gali, [27] E. L. Ivchenko, V. K. Kalevich, A. Y. Shiryaev, M. M. Afanasiev,
Phys. Rev. B 92, 115206 (2015). and Y. Masumoto, J. Phys.: Condens. Matter 22, 465804 (2010).
[12] T. D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. Monroe, [28] D. Paget, Phys. Rev. B 30, 931 (1984).
and J. L. O’Brien, Nature (London) 464, 45 (2010). [29] C. Weisbuch and G. Lampel, Solid State Commun. 14, 141
[13] J. Wrachtrup, S. Yang, and D. D. B. Rao, SPIE Newsroom (1974).
(2016). [30] C. T. Nguyen, A. Balocchi, D. Lagarde, T. T. Zhang, H. Carrre,
[14] W. M. Witzel, M. S. Carroll, A. Morello, L. Cywiński, and S. S. Mazzucato, P. Barate, E. Galopin, J. Gierak, E. Bourhis,
Das Sarma, Phys. Rev. Lett. 105, 187602 (2010). J. C. Harmand, T. Amand, and X. Marie, Appl. Phys. Lett.
[15] R. Hanson, V. V. Dobrovitski, A. E. Feiguin, O. Gywat, and 103, 052403 (2013).
D. D. Awschalom, Science 320, 352 (2008). [31] A. Kunold, A. Balocchi, F. Zhao, T. Amand, N. B. Abdallah,
[16] D. Simin, V. A. Soltamov, A. V. Poshakinskiy, A. N. Anisimov, J. C. Harmand, and X. Marie, Phys. Rev. B 83, 165202 (2011).
R. A. Babunts, D. O. Tolmachev, E. N. Mokhov, M. Trupke, [32] V. K. Kalevich, M. M. Afanasiev, A. Y. Shiryaev, and A. Y.
S. A. Tarasenko, A. Sperlich, P. G. Baranov, V. Dyakonov, and Egorov, Phys. Rev. B 85, 035205 (2012).
G. V. Astakhov, Phys. Rev. X 6, 031014 (2016). [33] V. K. Kalevich, M. M. Afanasiev, A. Y. Shiryaev, and A. Y.
[17] J. J. Pla, K. Y. Tan, J. P. Dehollain, W. H. Lim, J. J. L. Morton, Egorov, JETP Lett. 96, 567 (2012).
F. A. Zwanenburg, D. N. Jamieson, A. S. Dzurak, and A. [34] Y. Puttisong, X. J. Wang, I. A. Buyanova, and W. M. Chen,
Morello, Nature (London) 496, 334 (2013). Phys. Rev. B 87, 125202 (2013).

195204-13
V. G. IBARRA-SIERRA et al. PHYSICAL REVIEW B 95, 195204 (2017)

[35] Y. Puttisong, X. J. Wang, I. A. Buyanova, L. Geelhaar, H. [45] D. J. Hilton and C. L. Tang, Phys. Rev. Lett. 89, 146601 (2002).
Riechert, A. J. Ptak, C. W. Tu, and W. M. Chen, Nat. Commun. [46] V. Kalevich, A. Y. Shiryaev, E. Ivchenko, M. Afanasiev, A. Y.
4, 1751 (2013). Egorov, V. Ustinov, and Y. Masumoto, Physica B 404, 4929
[36] C. Sandoval-Santana, A. Balocchi, T. Amand, J. C. Harmand, (2009).
A. Kunold, and X. Marie, Phys. Rev. B 90, 115205 (2014). [47] X. Wang, I. A. Buyanova, F. Zhao, D. Lagarde, A. Balocchi,
[37] E. L. Ivchenko, L. A. Bakaleinikov, and V. K. Kalevich, X. Marie, C. Tu, J. Harmand, and W. Chen, Nat. Mater. 8, 198
Phys. Rev. B 91, 205202 (2015). (2009).
[38] E. L. Ivchenko, L. A. Bakaleinikov, M. M. Afanasiev, and V. K. [48] M. D’yakonov, V. Perel, V. Berkovits, and V. Safarov, Sov. J.
Kalevich, Phys. Solid State 58, 1539 (2016). Exp. Theor. Phys. 40, 950 (1975).
[39] M. Dyakonov and V. Perel, Sov. J. Exp. Theor. Phys. 36, 995 [49] B. Zakharchenya, V. Kalevich, V. Kul’kov, and V. Fleisher,
(1973). Sov. Phys. Solid State 23, 810 (1981).
[40] R. K. Wangsness and F. Bloch, Phys. Rev. 89, 728 (1953). [50] F. Meier and B. P. Zakharchenya, Optical Orientation (Elsevier,
[41] A. Redfield, in Advances in Magnetic Resonance, edited by Amsterdam, 2012), Vol. 8, Chap. 5.
J. S. Waugh (Academic Press, New York, 1965), Vol. 1, [51] D. Lagarde, L. Lombez, X. Marie, A. Balocchi, T. Amand, V. K.
pp. 1–32. Kalevich, A. Shiryaev, E. Ivchenko, and A. Egorov, Phys. Status
[42] G. W. Leppelmeier and E. L. Hahn, Phys. Rev. 142, 179 Solidi A 204, 208 (2007).
(1966). [52] A. G. Redfield, IBM J. Res. Dev. 1, 19 (1957).
[43] J. Kowalewski and L. Maler, Nuclear Spin Relaxation in [53] H.-P. Breuer and F. Petruccione, The Theory of Open Quantum
Liquids: Theory, Experiments, and Applications, Series in Systems (Oxford University Press on Demand, Oxford, 2002),
Chemical Physics (CRC Press, Boca Raton, 2006), Chap. 4. Chap. 3.
[44] N. A. Sinitsyn, Y. Li, S. A. Crooker, A. Saxena, and D. L. Smith, [54] M. Dyakonov, in Spin Physics in Semiconductors (Springer,
Phys. Rev. Lett. 109, 166605 (2012). Berlin, 2008), pp. 1–28.

195204-14