PHYSICAL REVIEW B 98, 195202 (2018)
Excitation properties of the divacancy in 4H-SiC
Björn Magnusson,
1,2
Nguyen Tien Son,
1
András Csóré,
3
Andreas Gällström,
4
Takeshi Ohshima,
5
Adam Gali,
3,6
and Ivan G. Ivanov
1
1
Linköping University, Department of Physics, Chemistry and Biology, S-581 83 Linköping, Sweden
2
Norstel AB, Ramshällsvägen 15, SE-602 38 Norrköping, Sweden
3
Department of Atomic Physics, Budapest University of Technology and Economics, Budafoki út. 8, H-1111, Hungary
4
Saab Dynamics AB, SE-581-88 Linköping, Sweden
5
National Institutes for Quantum and Radiological Science and Technology, 1233 Watanuki, Takasaki, Gunma 370-1292, Japan
6
Wigner Research Center for Physics, Hungarian Academy of Sciences, PO. Box 49, H-1525, Hungary
(Received 23 May 2018; revised manuscript received 18 September 2018; published 5 November 2018)
We investigate the quenching of the photoluminescence (PL) from the divacancy defect in 4H-SiC consisting
of a nearest-neighbor silicon and carbon vacancies. The quenching occurs only when the PL is excited below
certain photon energies (thresholds), which differ for the four different inequivalent divacancy configurations in
4H-SiC. An accurate theoretical ab initio calculation for the charge-transfer levels of the divacancy shows very
good agreement between the position of the (0/−) level with respect to the conduction band for each divacancy
configuration and the corresponding experimentally observed threshold, allowing us to associate the PL decay
with conversion of the divacancy from neutral to negative charge state due to capture of electrons photoionized
from other defects (traps) by the excitation. Electron paramagnetic resonance measurements are conducted in
the dark and under excitation similar to that used in the PL experiments and shed light on the possible origin of
traps in the different samples. A simple model built on this concept agrees well with the experimentally observed
decay curves.
DOI: 10.1103/PhysRevB.98.195202
I. INTRODUCTION
Defects in silicon carbide (SiC) have attracted attention
during the past few years from the point of view of possi-
ble applications in quantum technologies. The latter include
single-photon emitters applicable in quantum information
processing [1–5], magnetic sensors based on the Si vacancy
in 4H-SiC [6–8], and quantum bits (qubits) [9]. Owing to its
maturity, the 4H-polytype of silicon carbide (4H-SiC) is the
most studied polytype so far. The divacancy in 4H-SiC, which
is the main subject of this work, consists of carbon (V
C
) and
silicon (V
Si
) vacancies positioned on neighboring lattice sites
(denoted hereafter as VV for brevity). In 4H-SiC there exist
four inequivalent configurations of V
Si
-V
C
due to the presence
of two inequivalent lattice sites for each vacancy in the
unit cell [10]. The two inequivalent sites are usually termed
hexagonal (denoted h) and cubic (k) depending on whether
the arrangement of the next-nearest neighbors mimics that in
the hexagonal würtzite or in the cubic zinc blende structure.
Using the order V
Si
V
C
for the site positions of the vacancies,
the four inequivalent configurations are denoted further as hh,
kk, hk, and kh. When different charge states of any defect
must be specified, we will use superscripts explicitly denoting
the charge state, e.g., VV
0
denotes the neutral charge state of
the divacancy, while V
Si
2−
, for instance, denotes the double-
negatively charged state of V
Si
,etc.
The attractive properties of the divacancy in the 4H-SiC
for a solid-state quantum bit (qubit) have been demonstrated
recently [9]. In this work, the authors associate the four lines
observed in photoluminescence (PL) previously known as
PL1–PL4 with divacancies in each of the four inequivalent
configurations: hh (PL1), kk (PL2), hk (PL3), and kh (PL4).
Ab initio calculations [11] and electron paramagnetic res-
onance (EPR) measurements [10,12] show that the neutral
divacancy has an electron spin S = 1 and the ground and
excited states are triplets with A
2
symmetry in C
3v
(
3
A
2
).
The defect can be spin polarized using optical pumping [11]
with millisecond spin coherence times [4] and a high-fidelity
infrared spin-photon interface has been demonstrated [13].
In the present work we address the quenching of the PL
from the divacancies when certain photon energies are used
for excitation. The quenching phenomenon is not unique for
the divacancy and has been observed for several other defects
at resonant as well as nonresonant excitation. Resonant ex-
citation in this context means that the PL is excited at the
zero-phonon transition energy and registered by observing the
phonon sideband accompanying the zero-phonon line at lower
photon energies. The quenching of the divacancy has been
mentioned, e.g., in Ref. [14] (see the Supplementary Informa-
tion in this reference). It has been shown that application of
a second laser (repump laser) with higher photon energy and
much lower power completely recovers the PL intensity of
the divacancy (and other defects exhibiting quenching) [14].
The quenching effect is also the most probable reason for the
observation of the so-called blinking behavior exhibited by
single photon sources [1]. Although systematic investigation
of the quenching effect has been missing until recently, re-
searchers who have observed quenching tend to associate it
with change of the charge state of the investigated defect [14].
2469-9950/2018/98(19)/195202(15) 195202-1 ©2018 American Physical Society
BJÖRN MAGNUSSON et al. PHYSICAL REVIEW B 98, 195202 (2018)
At the time of writing of this paper we have found two other
papers [15,16] treating the quenching of the divacancy in 4H-
SiC. The main difference between these works and ours from
an experimental point of view is the use of a microscope-based
setupinRefs.[15,16], while in our work we use macroscopic
lenses for focusing the lasers and collecting the signal. This
leads to observation of much slower quenching dynamics
in our case than when a microscope objective is used. In
our work, we combine photoluminescence time-decay mea-
surements with photoluminescence excitation spectroscopy
(PLE) and EPR measurements conducted at similar illumina-
tion conditions as the PL. This approach makes it possible
to observe signatures in the EPR spectrum which change
upon the infrared (IR) or repumping excitation on a similar
timescale as observed in the PL. Using PLE we are also able
to distinguish the threshold laser energies for “switching” on
and off the quenching effect, which are found to differ for
the divacancies with different inequivalent configurations. The
experimentally obtained thresholds are considered in the light
of a new accurate theoretical calculation from first principles
for the energy positions of the charge-transfer states of the
divacancy. The improved accuracy of this calculation allows
direct comparison of the results for the individual VV con-
figurations with the experimental data. The good agreement
between theory and experiment suggests that the observed
quenching of the VV PL is due to conversion of the neutral
divacancies to their negative charge state. Finally, we propose
a dynamical model of the quenching which is simpler than
the one reported in Ref. [15] and shows good agreement with
experiment. Whenever appropriate, we compare our results
and their interpretation with the results and discussion in the
above-mentioned Refs. [15,16].
In Sec. II we describe the samples and the experimental
details. For the rest of the scope we choose a style of presenta-
tion in which the experimental data is presented first (Sec. III)
in order to set up foundation for the following discussion.
In Sec. IV we present the new results of first-principles
calculations on the charge transition levels of the divacancy,
which are discussed further in Sec. V in conjunction with the
experimental data in order to elucidate the underlying physics
of the quenching effect. Finally, in Sec. VI we describe a
dynamical model built on the ground of the notions developed
in Sec. V, and test its viability in describing the quenching
phenomenon. Section VII summarizes the conclusions.
All measurements in this work are done on ensemble of
divacancies.
II. SAMPLES AND EXPERIMENTAL DETAILS
In this study we present results from three 4H-SiC samples,
all exhibiting strong luminescence from the VV center (the
PL1–PL4 lines). However, the quenching dynamics of the di-
vacancy photoluminescence (VV PL) in these samples is quite
different; in particular, one of the samples does not exhibit
quenching at all when excited with the same photon energies
for which the other two samples do exhibit quenching. These
differences in the quenching behavior will be explained within
the physical model treated later in Secs. V and VI.
One of the samples has two counterparts cut from the
same high-purity semi-insulating (HPSI) 4H-SiC wafer, but
irradiated to different doses, 10
17
and 10
18
cm
−2
, with elec-
trons of energy 2 MeV. The samples are subsequently an-
nealed at a temperature of 800 °C for half an hour in order to
create the divacancies. Most results presented here are from
the piece irradiated to a dose of 10
17
cm
−2
, referred to further
as SI1, although the other piece referred to as SI1 (10
18
cm
−2
)
has also been measured and shows similar results. The EPR
results are obtained from this latter sample.
Another sample is n-type (N-doped) bulk 4H-SiC, re-
ferred to further as n-SiC. The N-doping level is in the low
10
17
cm
−3
range. The specimen presented in this work is
irradiated to 2 × 10
18
cm
−2
at 800 °C. Both SI1 and n-SiC
samples exhibit strong quenching of the VV spectrum upon
excitation with appropriate photon energy (e.g., 1.2 eV), but
at quite different rates.
The third sample is an as-grown HPSI 4H-SiC substrate
which, however, exhibits strong divacancy spectrum without
any irradiation/annealing. This particular sample does not
show any quenching in the PL lines from the divacancy at the
excitation photon energies at which the VV-PL in the rest of
the samples quenches. We refer to this sample further as SI2.
Most PL measurements are performed using a double
monochromator (SPEX 1404) on the detection side. The
monochromator is equipped with 600 grooves/mm gratings
blazed at 1000 nm and an InGaAsP photomultiplier, which
ensures optimum sensitivity in the emission region of the
divacancy (wavelengths ∼1078–1130 nm). The samples are
mounted in a variable-temperature cryostat operated with liq-
uid helium. We use a tunable Ti-sapphire laser as an excitation
source (referred to further as IR laser, or IR excitation),
which can be combined with a coincident green laser (532
or 514.5 nm). The repump green laser (∼10 mW power)
is not focused on the sample and we estimate its power
density to ∼0.1W/cm
2
. Even at several orders of magnitude
lower power than the IR excitation, the green laser has strong
effect on the intensity of the divacancy emission and the
quenching properties, although the green excitation alone is
very inefficient in exciting the divacancy spectrum.
Using an IR-laser power of ∼30 mW and moderate focus-
ing to ∼1mm
2
spot on the sample, one estimates power den-
sity about 3 W/cm
2
(photon flux density ∼1 × 10
19
cm
−2
s
−1
at photon energy hν = 1.2 eV). This is the main difference be-
tween our experimental conditions and these in Refs. [15,16],
where the exciting laser has been focused by a microscope
objective to a spot of the order of 1 μm
2
. Thus, even if the
laser power used with an objective is less than a milliwatt,
the power density at the sample can be at least four orders of
magnitude higher than ours due to the smaller spot. We will
see later that this high-power excitation influences strongly
the quenching dynamics by making it much faster than that
observed in our work.
III. EXPERIMENTAL RESULTS
In this section we present our experimental results, which
will be discussed later in the light of the theoretical calcula-
tions presented in the Sec. IV. The following discussion in
Sec. V describes the concepts used for interpretation of the
experimental data and uses the theoretical results of Sec. IV
to develop physical understanding of the quenching effect,
195202-2
EXCITATION PROPERTIES OF THE DIVACANCY IN … PHYSICAL REVIEW B 98, 195202 (2018)
which is used in Sec. VI to build a model for the PL decay.
It will be shown that the experimental data together with the
theoretical calculation strongly favor the model identifying
the negatively charged state of the divacancy as the “dark”
state into which the divacancy is converted during quenching.
This result agrees with [15], but is in contrast with the model
presented in [16] suggesting the positively charged state as the
dark state instead.
A. Optical properties
The general appearance of the PL spectra of the samples
excited with photon energies 1.528 eV (811 nm) and 1.333 eV
(930 nm) is shown in Fig. 1. Both these excitations are above
the energy threshold for quenching for all divacancy configu-
rations, but the former is chosen also above the threshold for
excitation of the negatively charged Si vacancy V
Si
−
(the zero
phonon lines are V1 at 861.6 nm and V2 at 916.5 nm) [17]in
order to examine its appearance in the samples.
All samples exhibit strong divacancy emission (the PL1–
PL4 lines), however, the appearance of the rest of the lines
differs substantially between the spectra. Thus, the V
Si
−
spec-
trum (V1, V2, and associated phonon sidebands) appears only
in the samples exhibiting quenching, SI1, and n-SiC, but is
missing in the SI2 sample. The latter sample (SI2) shows
weak contribution of a doublet around 1280 nm denoted as
VinFig.1(b). This doublet is identified as the PL signature of
vanadium in 4H-SiC [18], and its presence may have impact
on the lack of quenching of the VV PL in this sample, as
discussed later in Sec. V. We notice also the appearance of the
PL5 and PL6 lines observed previously in work investigating
the divacancy properties [9]. These lines are well visible in the
SI1 and SI2 samples, but not present at all in the n-SiC sample.
Two more lines denoted here as PL5´ and PL6´ with intensity
similar to that of PL5 and PL6 appear in all the three samples,
but at different excitation conditions, as seen in Fig. 1.The
origin of PL5, PL6, PL5´, and PL6´ is not known and these
lines will not be discussed further.
Finally, we make a detour to comment on the four lines
in the range ∼1175–1245 nm denoted N-V in Fig. 1, which
are observed only in the n-SiC sample. In an early work
[21], these lines were tentatively associated with the carbon
vacancy–carbon antisite pair (CAV), in analogy with a work
on 6H-SiC [22]. This latter work assigns six so-called P6/P7
lines observed in 6H-SiC in the range ∼0.99–1.07 eV to
the CAV defect in its six crystallographic-inequivalent con-
figurations. Four inequivalent configurations are expected in
4H-SiC, hence the association made in [21] for the four
lines appearing in 4H-SiC in the same spectral region as in
6H-SiC. Recent work [23], however, suggests that the lines
are instead due to the N-V pair, a defect expected to appear
in SiC and to have properties similar to those of the well-
studied N-V pair defect in diamond [24]. Our spectrum is
identical to that of Ref. [21], and since these lines are observed
only in the highly N-doped sample, we tend to confirm the
association with the N-V center made in Ref. [23]. However,
the interpretation of the spectrum observed in [23] does not
seem to be entirely correct, and the following corrections are
needed, in our opinion. First of all, the mutual disposition of
the lines observed in [23] and ours is the same, apart from a
Wavelength (nm)
Photon Energy (meV)
PL Intensity (arb. units)
PL2
(kk)
PL3
(kh)
PL4
(hk)
PL6
PL5
PL1
(hh)
PL5'
PL6'
exc
= 811 nm
T = 3.5 K
900 1000 1100 1200
100012001400
V2
V1
V1
V2
10
UD3
Raman
PL5'
PL6'
(a)
1050 1100 1150 1200 1250 1300
95010001050110011501200
16
16
16
6
Photon Energy (meV)
PL Intensity (arb. units)
n-SiC
SI2
SI1
PL6
PL5
PL5'
PL6'
{
V
(b)
PL5'
PL6'
PL6
PL5
Wavelength (nm)
n-SiC
SI2
SI1
NV4
NV3
NV2
NV1
N-V
NV4
NV3
NV2
NV1
N-V
exc
= 930 nm
T = 3.5 K
PL2
(kk)
PL3
(kh)
PL4
(hk)
PL1
(hh)
FIG. 1. PL spectra of the three investigated samples recorded
with two different laser excitations, (a) 811 nm (1.528 eV) and
(b) 930 nm (1.333 eV). Both excitations do not cause quenching of
the VV-related lines (PL1–PL4), but excitation with 811 nm allows
also observation of the silicon-vacancy (V
Si
−
) spectrum (the V1 and
V2 lines and associated phonon sidebands), whenever present. “Ra-
man” denotes Raman lines. UD3 denotes an unidentified defect [19]
which according to our recent unpublished data is most likely related
to tantalum [20]. The SI2 sample exhibits weak contribution from
vanadium (V), whereas the n-SiC sample shows strong contribution
from a defect tentatively associated in recent work [23] with the
nitrogen-vacancy pair in 4H-SiC.
common 0.6 meV shift for all lines, which can be attributed to
slight miscalibration of their measurement and/or ours. Thus,
there is no doubt that the same spectrum denoted N-V in
Fig. 1 is observed also in Ref. [23]. However, the low-energy
peak denoted PLX1/PLX2 in Ref. [23] is claimed to comprise
two lines, which are not resolved in their spectrum due to
low spectral resolution (estimated to ∼2 meV from their
figures). In our measurement the low energy peak NV1 is a
195202-3
BJÖRN MAGNUSSON et al. PHYSICAL REVIEW B 98, 195202 (2018)
TABLE I. List of the lines observed in PL in the three samples, in
order of ascending wavelengths (descending energies). The associa-
tion of UD3 with tantalum is tentative [20]. In the case of vanadium,
only the stronger low-temperature components α
1
and α
3
are given.
We follow the notations used for the components of the vanadium
related α line in 6H-SiC [27]. The α
2
and α
4
components are also
weakly visible at 3.5 K (blueshifted by 0.8 meV from α
1
and α
3
,
respectively), but are omitted in the table.
Line Peak position in nm (meV) Sample
V1 (V
Si
)
a
861.6 (1438.6) SI1, n-SiC
UD3 (probably Ta)
b
914.5 (1355.4) SI2
V2 (V
Si
)
a
916.5 (1352.4) SI1, n-SiC
PL6
c
1037.7 (1194.5) SI1, SI2
PL5
c
1041.9 (1189.6) SI1, SI2
PL6´ 1042.6 (1188.9) SI1, n-SiC
PL5´ 1047.3 (1183.5) SI1, n-SiC
PL4 (VV
hk
)
c
1078.5 (1149.3) All samples
PL3 (VV
kh
)
c
1107.6 (1119.1) All samples
PL2 (VV
kk
)
c
1130.5 (1096.4) All samples
PL1 (VV
hh
)
c
1132.0 (1095.0) All samples
NV4
d
1176.4 (1053.6) n-SiC
NV3
d
1180.0 (1050.4) n-SiC
NV2
d
1223.2 (1013.3) n-SiC
NV1
d
1242.8 (997.3) n-SiC
Vanadium (α
3
line)
e
1278.6 (969.4) SI2
Vanadium (α
1
line)
e
1281.5 (967.2) SI2
a
From [17].
b
See [20].
c
From [9].
d
New assignment based on our results and [23].
e
Labeled in analogy with the α line in 6H-SiC [27].
single symmetric line with apparent linewidth of 0.5 meV (our
resolution is ∼0.3 meV). The origin of the structure visible
in the PLX1/PLX2 peak in Ref. [23] is not known, but such
structure may arise, for instance, as a consequence of deterio-
rated focusing on different parts of the array detector used in
this work. Furthermore, the high-energy peak denoted NV4 in
Fig. 1 is observed also in Ref. [23], but incorrectly attributed
to tungsten (W). In fact, their spectrum does display tungsten
contribution, but the two W peaks separated by 1 meV accord-
ing to the original work [25] are not resolved and constitute
their highest-energy peak at ∼1059 meV, whereas the line
at ∼1054 meV seen in their spectra actually belongs to the
N-V spectrum, not to W. With these corrections in mind,
our N-V spectrum is identical to that of [23]. We notice that
with the above corrections the accuracy of agreement between
the N-V lines and the theoretical results presented in [23]
remains very good, within 50 meV for all the lines. A more
recent calculation [26] provides even better agreement with
our assignment for the NV1–NV4 lines. Since the N-doping
level in our sample (low 10
17
cm
−3
) is similar to that of the
sample used in Ref. [23](2× 10
17
cm
−3
), and since these
lines are not observed at all in the two HPSI samples, our
result can be seen as a confirmation of the association of the
considered lines with the N-V center in 4H-SiC.
The PL peaks observed in the three samples are summa-
rized in Table I.
Time (s)
5000 1000 1500 2000
Normalized PL
Intensity
1080110011201140116011801200
Wavelength (nm)
Photon Energy (meV)
PL Intensity (arb. units)
1020 1040 1060 1080 1100 1120 1140
20
PL1
PL2
PL3
PL4
PL5´
PL6´
PL5
PL6
(a) SI1
T= 3.5 K
=1000 nm
exc
P30 mW
1
P145 mW
2
T = 3.5 K
=1000 nm
exc
after ~3 h quenching
1000 200 300 400
Time (s)
Normalized PL Intensity
8
1080110011201140116011801200
Wavelength (nm)
Photon Energy (meV)
PL Intensity (arb. units)
1020 1040 1060 1080 1100 1120 1140
PL3
PL4
PL1
PL2
PL5´
PL6´
T = 3.5 K
=1000 nm
exc
after ~ 1/2 h quenching
PL5´
PL6´
P
2
P
1
experiment
fit
rescaled
experiment
experiment
rescaled
experiment
(b) n-SiC
T= 3.5 K
=1000 nm
exc
P30 mW
1
P145 mW
2
P
2
P
1
repumping 514 nm ON
repumping 514 nm ON
FIG. 2. Normalized decay curves illustrating the quenching of
the PL4 line upon 1000 nm excitation at two different laser powers
in (a) the SI1 and (b) the n-SiC sample. The insets show the
corresponding spectra with repump laser at 514.5 nm (top curves)
and after quenching with 1000 nm excitation for the indicated time.
The intensities of the rest of the lines present in the spectra are
not affected by the repump laser. Notice the different timescales
and the different scaling factors for the bottom curves in the insets.
The thick gray curves underlying the fast decay at short elapsed
times in both panels (rescaled experiment) are obtained by scaling
uniformly along the time axis the corresponding slow-decay curves
corresponding to lower-power excitation, thus illustrating the scaling
property discussed in text. The bold curves (magenta and blue in the
online version) in (a) are fits obtained from the model presented in
Sec. VI.
B. Quenching and recovery of the PL1–PL4 lines
Figure 2 displays a summary of the quenching behavior of
the divacancy lines observed in the SI1 and n-SiC samples
with laser excitation at 1000 nm (1.24 eV). The SI2 sample
does not show any quenching with this excitation. The PL
spectra of SI1 and n-SiC displayed in the insets of Figs. 2(a)
and 2(b) are obtained with and without repump laser at 514 nm
(2.41 eV). When the repump laser is present no quenching is
195202-4
EXCITATION PROPERTIES OF THE DIVACANCY IN … PHYSICAL REVIEW B 98, 195202 (2018)
observed in any of the PL1–PL4 lines. However, if the repump
laser is switched off, the quenching starts immediately, as
illustrated by the decay curves in Fig. 2 (zero time corresponds
to the switch-off moment). The figure illustrates also that the
quenching rates are very different for the two samples SI1 and
n-SiC (notice also the different timescales in the two panels).
Thus, the spectrum for the SI1 sample in Fig. 2(b) is taken
after ∼3 h of quenching, but it still shows contribution from
the divacancy spectrum. On the other hand, all divacancy-
related emission of the n-SiC sample quenches completely
within about 10 min of irradiation with the 1.24 eV laser,
as seen from the spectrum and shown for the PL4 line in
Fig. 2(b).
In comparing our results with those presented in
Refs. [15,16] we notice that the timescale for quenching of
the VV PL is very different from ours. In both references the
quenching requires subsecond times (milliseconds to tens of
microseconds) for reaching a steady state (usually, close to
zero PL intensity), whereas typical quenching times under the
excitation conditions used in our work are tens of seconds and
up to hours, depending on the sample. As already mentioned
in Sec. II, the main difference between our measurements
and these made in Refs. [15,16] is the laser power density
on the sample. Therefore, we have investigated the power
dependence of the quenching decay. The two decay curves
in each panel of Figs. 2(a) and 2(b) represent the normalized
decay of the PL4 line with time obtained at two different
exciting-laser power levels, P
1
≈ 30 and P
2
≈ 145 mW, thus
P
2
/P
1
≈ 4.8. Apparently, higher excitation power speeds up
the quenching. However, we observe an interesting property
concerning the power dependence of the decay. Namely, if
the curve obtained at lower power is rescaled along the time
axis by a certain factor, it coincides almost exactly with the
curve obtained at higher power. The thick gray curves in
Figs. 2(a) and 2(b) marked “rescaled experiment” are the
slow-decay curves rescaled by a factor ∼1/4.55 [for Fig. 2(a)]
and ∼1/3.85 [for Fig. 2(b)]; the rescaled curves match the
corresponding fast-decay curves almost exactly. Since both
rescaling factors are close to 1/4.8, which is the ratio between
the low and high powers used, this experiment suggests that
the quenching rate is roughly proportional to the exciting
power. We notice that our simple model of the quenching
dynamics discussed in more details later in Sec. VI reproduces
quite accurately the just-described “scaling” property, as well
as the shape of the decay for the SI1 sample. This is illustrated
in Fig. 2(a) where the bold lines overlapping the decay curves
in the range 0–1050 s represent the fits obtained from the
model. We notice, however, that a good fit for the n-SiC
sample [Fig. 2(b)] could not be obtained with the simple
model considering only one type of traps, as discussed later in
Sec. VI [see also the Supplemental Material (SM file)] [28].
Further properties of the quenching effect showing that
in darkness the quenching state (partially or fully quenched
PL level), as well as the recovering action of the repump
laser are preserved (“remembered”) for a long time at low
temperatures, are described in more detail in the SM file
(memory effects I and II) [28].
From the general behavior of the VV PL in the three
investigated samples it can be concluded that the observed
quenching (or its lack) is associated with interaction with
other defects, which are referred to as “traps” in the de-
scription of the dynamical model of the quenching presented
later on. During this interaction effective under excitation the
divacancy accumulates in a different charge state (positive
or negative) and the luminescence from the neutral charge
state (PL1–PL4 lines) decays. Furthermore, we will provide
arguments that actually the negative charge state is the one that
accumulates during quenching, which is in agreement with
[15] and disagrees with [16].
C. Temperature dependence
The quenching-recovery properties of the VV PL in the
samples exhibiting quenching are preserved also at higher
temperatures. Let us consider first the temperature depen-
dence of the PL spectra displayed in Fig. 3. Apparently,
the PL3 line dominates the spectrum at temperatures above
∼60 K and above approximately 150 K all contribution from
the sharp PL1–PL4 lines becomes indistinguishable. It is also
interesting to compare the higher-temperature spectra of the
three samples. The spectra of the three samples excited with
930 nm at 200 K are compared in the upper inset of Fig. 3.
It is quite obvious that the only sample the high-temperature
emission of which stems from the divacancy is the SI2 sample,
which does not exhibit any quenching. In the samples exhibit-
ing quenching, on the other hand, we see that the spectra at
200 K are dominated either by the silicon vacancy V
Si
−
(SI1),
or by the N-V-associated band at lower energies (n-SiC),
whereas the VV PL contribution is negligible. The lower inset
in Fig. 3 illustrates that the VV emission in the SI1 sample
can become dominant at 200 K when lower-energy excitation
is used (1000 nm), but this contribution still vanishes at room
temperature (293 K).
We notice also the obvious up-conversion expressed in the
fact that the V
Si
−
spectrum dominates the emission of the SI1
sample despite the fact that the IR excitation energy in both
insets of Fig. 3 is below the energy of both zero phonon lines
of V
Si
−
, V1 and V2. The up-conversion becomes apparent at
temperatures above ∼80 K. The up-conversion mechanism is
discussed later in Sec. V, where we show that the observed
up-conversion also supports the identification of the negative
charge state of the divacancy as the dark state. However, it is
quite clear that in both samples exhibiting quenching the VV
spectrum nearly vanishes above ∼150 K and the remaining
dominant emission is due to other defects (V
Si
−
in sample
SI1, or the N-V pair in sample n-SiC). This observation may
explain why the effect of repumping diminishes and vanishes
at higher temperatures above ∼150 K with the excitation of
976 nm (1.27 eV) reported in [15]. This effect may simply be
due to vanishing contribution of VV in the spectrum, but no
spectral data is given for elevated temperatures in Ref. [15].
We will discuss more on this subject in Sec. V.
Further details on the quenching-recovery behavior of the
VV PL at different temperatures are given in the SM file [28].
D. EPR results
We turn now to summarizing the EPR results. In order to
reproduce closely the conditions of the PL measurements we
have recorded EPR spectra in darkness, or under illumination
195202-5
![FIG. 4. EPR spectra of the samples in darkness and after illumination with 1030-nm laser for various durations, as indicated for each curve. The spectra in (a) are from the SI2 sample, in (b) from the n-SiC sample, and in (c) from the electron-irradiated HPSI 4H-SiC (same substrate as SI1, but irradiation dose 1 × 1018 cm−2). The middle lines for the spectra in darkness in (a) and (c) are from the positively charged C vacancy (VC+) [29,30], and in (b) from the negatively charged one (VC−) [31–33]. The inset in (b) illustrates the appearance of the N shallow donor spectrum in the beginning of the IR excitation; this spectrum vanishes in synchrony with the neutral divacancy quenching. The signals of axial (kk and hh) and basal (kh, hk) configurations of the neutral divacancy (VV0), as well as EI-4 [33] and another unidentified defect with spin one are indicated.](/figures/fig-4-epr-spectra-of-the-samples-in-darkness-and-after-132z1g04.webp)
![FIG. 1. PL spectra of the three investigated samples recorded with two different laser excitations, (a) 811 nm (1.528 eV) and (b) 930 nm (1.333 eV). Both excitations do not cause quenching of the VV-related lines (PL1–PL4), but excitation with 811 nm allows also observation of the silicon-vacancy (VSi−) spectrum (the V1 and V2 lines and associated phonon sidebands), whenever present. “Raman” denotes Raman lines. UD3 denotes an unidentified defect [19] which according to our recent unpublished data is most likely related to tantalum [20]. The SI2 sample exhibits weak contribution from vanadium (V), whereas the n-SiC sample shows strong contribution from a defect tentatively associated in recent work [23] with the nitrogen-vacancy pair in 4H-SiC.](/figures/fig-1-pl-spectra-of-the-three-investigated-samples-recorded-183tf8pd.webp)
![TABLE I. List of the lines observed in PL in the three samples, in order of ascending wavelengths (descending energies). The association of UD3 with tantalum is tentative [20]. In the case of vanadium, only the stronger low-temperature components α1 and α3 are given. We follow the notations used for the components of the vanadium related α line in 6H-SiC [27]. The α2 and α4 components are also weakly visible at 3.5 K (blueshifted by 0.8 meV from α1 and α3, respectively), but are omitted in the table.](/figures/table-i-list-of-the-lines-observed-in-pl-in-the-three-1h1eht90.webp)
