Defect-related photoluminescence of hexagonal boron nitride
Laboratoire de Physique des Lasers-LPL, CNRS UMR 7538, Institut Galilée, Université Paris 13, 93430 Villetaneuse, France
Institute of Physics, University of Tartu, 142 Riia Street, 51014 Tartu, Estonia
Laboratoire d’Ingénierie des Matériaux et des Hautes Pressions-LIMHP, CNRS, Institut Galilée, Université Paris 13,
93430 Villetaneuse, France
?Received 19 May 2008; revised manuscript received 2 July 2008; published 16 October 2008?
Photoluminescence of polycrystalline hexagonal boron nitride ?hBN? was measured by means of time- and
energy-resolved spectroscopy methods. The observed bands are related to donor-acceptor pair transitions,
impurities, and structural defects. The excitation of samples by high-energy photons above 5.4 eV enables a
phenomenon of photostimulated luminescence ?PSL?, which is due to distantly trapped conduction band
electrons and valence band holes. These trapped charges are metastable and their re-excitation with low-energy
photons results in anti-Stokes photoluminescence. The comparison of photoluminescence excitation spectra
and PSL excitation spectra allows band analysis that supports the hypothesis of Frenkel-type exciton in hBN
with a large binding energy.
DOI: 10.1103/PhysRevB.78.155204 PACS number?s?: 78.55.Cr, 71.35.?y, 71.55.Eq
Optical and fluorescence properties of hexagonal boron
nitride ?hBN? deserved particular interest during the last de-
cade since the first observation of an intense far-UV exciton
emission,1,2making this material a candidate for use in new
light-emitting devices. A successful synthesis of high-purity
temperature1,2and later at atmospheric conditions2enabled
this achievement. Accordingly, understanding of the elec-
tronic band and exciton structure of hBN becomes even more
an important issue as it serves a basic system for
single-wall3,4and multiwall boron nitride nanotubes.5–9
Despite of many efforts in the past, the electronic proper-
ties of hBN remain largely unknown. They have been earlier
studied by luminescence,1,10–17
elastic x-ray scattering,26–28x-ray absorption,25,29,30and elec-
tron energy loss31–33spectroscopy. After all, a large spread of
band-gap energies reported in literature, ranging from 3.1 to
7.1 eV,14is currently explained by sample quality and related
to experimental methods used. Recently, arguing on gener-
ally high luminosity of the free exciton luminescence, Wa-
tanabe et al.1assumed that hBN is a direct band-gap mate-
rial. They have measured the band-gap energy of 5.971 eV
and inferred an exciton binding energy of 0.149 eV that cor-
responds to the Wannier exciton model. However this result
is in a large disagreement with the most recent theoretical
approximation.34,35They predict hBN to be an indirect band-
gap material with a band-gap energy of 5.95 eV and a lowest
direct interband transitions at 6.47 eV.34Moreover, Arnaud et
al.34have deduced a huge exciton binding energy of 0.72 eV
and predicted that the low-lying exciton in hBN belongs to
the Frenkel type. The intense free exciton luminescence ob-
served in single-crystal hBN is explained by a large oscilla-
at highpressure and
optical reflectance and
tor strength of merged excitonic transitions.34
In view of many disagreements further confrontation be-
tween experiment and theory will continue. In these condi-
tions, more experimental data concerning electronic and re-
lated optical properties of hBN are highly required. In
particular, energy- and time-resolved photoluminescence
methods provide valuable information about excitonic and
interband transitions. Moreover, in polycrystalline samples
energy transfer to impurities and defects may inhibit these
intrinsic transitions and strongly affect the fluorescence spec-
tra. However, the energy transfer is specific to excitation en-
ergy and the relevant excited states can be experimentally
In the present paper we report on detailed analysis of the
defect-related intraband luminescence of hBN. Important in-
formation about intrinsic properties of hBN is obtained by
combining two experimental approaches: time- and energy-
resolved photoluminescence and photostimulated lumines-
cence. The discussion is organized as follow. The experiment
is described in Sec. II and experimental results are presented
on the Sec. III. In Sec. IV A we discuss the nature of the
observed luminescence bands. Finally, in Sec. IV B, careful
comparison between photoluminescence excitation ?PLE?
and photostimulated luminescence ?PSL? excitation spectra
brings more precision to exciton and band-gap transitions of
The samples were prepared from commercial hexagonal
BN powders ?Alfa 99.5%? compacted in pellets of size 8
?8?1 mm3at a hydrostatic pressure of 0.6 GPa. The grit
size of the hBN powder has been estimated by means of
granulometry and transmission electron microscopy ?JEM
100C JEOL?. It ranged from 0.3 to 10 ?m with an average
PHYSICAL REVIEW B 78, 155204 ?2008?
©2008 The American Physical Society155204-1
particle size of 3.1 ?m corresponding to the maximum in
the mass distribution curve. The samples were then heated at
800 K under vacuum for 12 h to avoid organic impurities and
traces of water.
The luminescence properties of the samples were studied
using vacuum ultraviolet ?VUV? synchrotron-radiation ?SR?
source of DORIS storage ring at HASYLAB ?DESY, Ham-
burg?. The facility of the SUPERLUMI station used in the
experiments is described in details elsewhere.36Briefly,
samples were cooled down to 8 K and irradiated by
monochromatized SR ???=3.3 Å? under high vacuum
??10−9mbar?. The measurements of luminescence spectra
were carried out using a visible 0.275-m triple-grating ARC
monochromator equipped with a CCD detector or a photo-
multiplier operating in the photon-counting mode. The pulse
structure of SR ?130 ps and 5 MHz repetition rate? enables
time-resolved luminescence analysis at time scale of 200 ns
with subnanosecond temporal resolution. Spectra were re-
corded within a time gate ?? delayed after the SR excitation
pulse. Typically two time gates have been used simulta-
neously: a fast one with ??1=1–4 ns and a slow one with
curves were measured at fixed excitation and luminescence
photon energies. The recorded spectra were corrected for the
primary monochromator reflectivity and SR current.
The PSL excitation spectra of pd-hBN samples were mea-
sured at the BL 52 beamline of MAX-laboratory synchrotron
?Lund, Sweden?. The experiment built up by the Tartu group
is described in Refs. 37 and 38. At each excitation energy
Eexcthe sample was irradiated by a fixed number of UV or
VUV photons ?2.5?106counts? in the energy region of
Eexc=5–15 eV. After completing the dose, the irradiation
was stopped and the phosphorescence decay was measured.
When the phosphorescence intensity drops until almost zero
?photomultiplier ?PM? noise level? that typically happens af-
ter a few minutes, the sample was activated by a bright spec-
troscopic source at h?=2.0?0.5 eV through a double-prism
monochromator DMR-4 and the time dependence of PSL
intensity was recorded. The integral PSL intensity was taken
as a measure of the number of recombined electron-hole
Under photoexcitation above 6 eV the luminescence spec-
tra of hBN are composed of several luminescence bands
?Fig. 1?. Basically, one broadband and one structured emis-
sions between 3 and 4 eV and two relatively narrow near
band-gap emissions at 5.3 and 5.5 eV were distinguished.
Two high-energy emissions were discussed in details in our
recent publication.17They were assigned, respectively, to
quasi-donor-acceptor pairs ?DAPs? ?q-DAPs? ?5.3 eV band?
and bound excitons ?5.5 eV band?. Below we complete the
assignment of low-energy luminescence.
A. Luminescence around 4 eV in hBN powder samples
The hBN powder sample is a strongly luminescent mate-
rial, whose luminescence spectra depend on excitation en-
ergy. The low-temperature luminescence spectra of hBN
samples recorded with different excitation energies varying
from 4 to 6.5 eV are presented in Fig. 1. Under excitation
below 5.0 eV a strong structured UV emission is observed.
Four peaks, labeled ???, ???, ???, and ???, are clearly visible.
Amultiple Gaussian fit procedure results in the peak energies
of 4.099 ???, 3.912 ???, 3.731 ???, and 3.539 eV ???. These
fourpeaks are equallyspaced
=186?1.4 meV. At room temperature these peaks are
broadened but no appreciable shift was detected. When the
excitation energy exceeds 5.0 eV, a very broad band ??E
?1 eV? centered at 3.9 eV appears and superimposes with
the structured emission ?Figs. 1?c? and 1?d??. However, the
peaks of the structured emission can always be observed on
the top of the broad band.
At excitation energy of 4.27 eV only the structured UV
emission can be observed. The luminescence decay curves of
its four peaks ???–??? are shown in Fig. 2. Within the ex-
perimental error bars, all the decay curves are single expo-
nential with short characteristic decay time of ?=1.1 ns. The
typical decay curve of the broad emission measured at Elum
=3.91 eV and Eexc=5.96 eV is plotted in Fig. 2 by curve
???. At this excitation energy the broad band dominates the
luminescence spectra. Although the probed energy Elum
=3.91 eV is in coincidence with the peak ??? of the struc-
tured emission, the decay curve appears to be multiexponen-
FIG. 1. ?Color online? Luminescence spectra of polycrystalline
hBN. The excitation energy is indicated on each spectrum. The
phonons replicas are labeled in ?a?. The star in ?c? and ?d? indicates
the second order of the 5.5 eV band.
MUSEUR, FELDBACH, AND KANAEVPHYSICAL REVIEW B 78, 155204 ?2008?
tial and of longer lifetime than that of the peak ???.
The time-resolved photoluminescence method allows
separating several contributions of superposed luminescence
bands if their respective lifetimes are noticeably different.
This is the case of the structured and continuous broadband
emissions around 4 eV. Figure 3?a? shows the total PLE
spectrum measured at Elum=3.73 eV. The last corresponds
to the peak ??? of the structured emission and is close to the
maximum of the broadband emission. Two principal features
can be distinguished in this spectrum: ?i? a sharp onset at
4.09 eV followed by a decrease in the luminescence intensity
until nearly zero at Eexc=4.9 eV and ?ii? a strong peak at
5.95 eV preceded by a weak structured footer that begins at
5.0 eV. The PLE spectra obtained with short- and long-time
windows are displayed in Figs. 3?b? and 3?c?. The long-time
window of ??2=22–200 ns selects long-lived excited states,
which are apparently responsible for the broad emission
band ?Fig. 3?c??. The peak maximum at 5.95 eV strongly
contributes to excitation of this band. This peak is blue-
shifted with respect to that of the exciton absorption at 5.82
eV in hBN single crystal.1Nevertheless, a weak lumines-
cence at 5.9 eV is also reported in hBN single crystal at low
temperatures and assigned to another exciton series. The
Stokes shift with respect to the excitation peak at 5.95 eV
could result from inhomogeneities of the local field or dislo-
cations such as stacking faults.1Therefore we tentatively as-
sign the peak at 5.95 eV to the excitation of higher exciton
series. This excitonic peak is preceded by a weak and un-
structured footer ranging from 5 to 5.5 eV and is followed by
a broad continuum until 7.6 eV. Conversely, the PLE spectra
obtained with the short-time window ???1=1–4 ns? show
contributions from both short- and long-decay luminescence
components ?Fig. 3?b??. Nevertheless, a comparison with the
PLE spectra of the long component ?Fig. 3?c?? allows firmly
assigning the structured emission with the onset at 4.09 eV to
a short-lived excited state. Moreover, we note a second weak
excitation onset of the structured emission at Eexc=5.2 eV.
Similar luminescence spectra to those depicted in Fig. 1
have been recently reported on photoluminescence and
powder.10,39–48However, in these publications, no distinc-
tions were made between the broad band and structured
emissions. We now assign the structured UV emission to
impurities ?probably C? and the broadband emission to deep
DAP of strongly localized center. These assignments will be
discussed below in Sec. IV A.
experiments of commercial hBN
B. Photostimulated luminescence in hBN powder samples
The photostimulated luminescence arises from the trap-
ping of free charge carriers in distant lattice sites with sub-
sequent recombination of carriers released from these traps
by a visible light.38The charge separation and trapping can
result from different excitation processes listed below and
depicted in Fig. 4:
A−+ D++ h?exc→ A−+ D++ e−+ h+→ Ao+ Do,
A−+ D++ h?exc→ Ao+ D++ e−→ Ao+ Do,
A−+ D++ h?exc→ A−+ Do+ h+→ Ao+ Do,
A−+ D++ h?exc→ Ao+ Do.
The band-to-band transitions ?Eq. ?1?? or impurity-band ion-
ization ?Eqs. ?2? and ?3?? are the most likely contributions to
PSL. In these cases at least one of photoexcited charge car-
FIG. 2. ?Color online? Luminescence decay curves at different
photon energies. Curves ?, ?, ?, and ? correspond to the peaks of
the structured emission. Curve ? corresponds to the broad emission
FIG. 3. ?Color online? ?a? PLE spectra ?Elum=3.72 eV? of poly-
crystalline hBN and its ?b? short- and ?c? long-lived components.
DEFECT-RELATED PHOTOLUMINESCENCE OF HEXAGONAL…
PHYSICAL REVIEW B 78, 155204 ?2008?
riers ?e−or h+? becomes free and can migrate away from the
point of excitation before being trapped by the acceptor or
donor defects. Spatially closely trapped charges then annihi-
late giving rise to phosphorescence. In contrast, at rather
large distance their lifetime becomes infinite and their re-
combination can only be possible after reactivation by light
or thermally. Figure 5 shows the phosphorescence decay
curves ?Elum=3.91 eV? for different photon excitation ener-
gies ranging from 5.4 to 5.9 eV. At Eexc?5.50 eV the phos-
phorescence decay curves are characterized by an almost
monoexponential decay with characteristic time of ?=5.8 s
?the dotted line in Fig. 5 shows the PM dark noise?. How-
ever, at Eexc?5.60 eV another extremely long-lived compo-
nent of low intensity appears, which looks like a plateau in
Fig. 5. We ascribe this long component to the recombination
of charges trapped at large distance tending to infinity.
The observed phosphorescence might also be related to
the “dark” ?dipole forbidden? exciton states theoretically pre-
dicted in hBN by Wirtz et al.35The authors suggested a
coupling between the “white” ?dipole allowed? and dark ex-
citon states, which makes them appear in the absorption
spectra of hBN monocrystals reported by Watanabe et al.1
However, we disregard their contribution in PSL experiments
since the phosphorescence is already observed at Eexc
=5.1 eV ?Fig. 5?, which is well below the exciton energy in
hBN. Moreover, PSL is related to the states with infinite
lifetime ?excited above 5.5 eV? whose depopulation is only
triggered by light. In contrast, the hypothetic dark states be-
come accessible when the crystalline symmetry is broken
through defects or limited sample quality. This is our case
since no free exciton emission has been observed. Accord-
ingly, the singlet and triplet excitons are expected to be
strongly coupled in our polycrystalline hBN samples and
their lifetime is short ?subnanosecond?. In this case no long-
lived exciton states are expected.
The PSL excitation spectrum is represented by open
circles in Fig. 6. The full measured spectrum is shown in the
insert of the figure for indicative purposes. In the following,
we will discuss the PSL phenomenon specific to excitation
energies ranging from 5 to 7 eV. The signal shows up appre-
ciably at Eexc?5.50 eV in correlation with the appearance
of a long component of the phosphorescence decay ?Fig. 5?.
The PSL grows almost exponentially with the excitation en-
ergy until Eexc?Ei1=5.7 eV where it locally attains a maxi-
mum. At this energy the slope suddenly changes and the
growth of the signal becomes much slower. Another local
maximum of the PSL spectrum appears at Eexc?Ei2
=6.10 eV. Finally, for excitation energies above 6.2 eV the
PSL signal increases progressively until 7 eV where a pla-
teau is reached.
For the sake of the discussion we have superimposed in
Fig. 6 the PLE spectrum of the near band-edge luminescence
at Elum=5.3 eV. This emission was discussed in details in
our previous publication17and has been assigned to the ra-
diative recombination of the so-called q-DAP.
A. Luminescence of polycrystalline hBN samples
The structured UV luminescence ?Figs. 1?a? and 1?b?? has
been reported in the past by numerous groups either in
FIG. 4. Schematic representation of electronic transitions rel-
evant to PSL.
FIG. 5. ?Color online? Phosphorescence decay curves for differ-
ent photon excitation energies.
FIG. 6. ?Color online? PSL excitation spectrum ?open circles?
and PLE spectrum of q-DAP luminescence ?5.3 eV? of polycrystal-
line hBN. The total measured PSL excitation spectrum is shown in
MUSEUR, FELDBACH, AND KANAEVPHYSICAL REVIEW B 78, 155204 ?2008?
experiments.10,14,39–42,44,45,47,48Some authors have assigned it
to intrinsic luminescence of hBN molecular layers47or to
phonon-assisted band-edge luminescence.42However, recent
experiments1and theoretical calculations34indicate the hBN
band-gap energy in the range from 6 to 6.5 eV that disables
these interpretations. Others authors45,49,50claimed that the
structured luminescence is due to transition between the con-
duction band ?CB? and acceptor carbon atom at substitu-
tional N site. Recently, experimental evidence of the correla-
tion between carbon and oxygen contents in hBN samples
and intensity of the structured luminescence has been
In a recent publication39the structured luminescence of
hBN has been attributed to defects or impurities. The peaks
???, ???, and ??? in Fig. 1?a? were ascribed to phonon repli-
cas of the zero-phonon line ??? involving TO phonons
?wTO=169 meV?. Due to the low sample temperature in
present experiments, the luminescence peaks become well
resolved, allowing more precise determination of their spec-
tral positions than in room-temperature experiments.39,42The
obtained phonon energy of wg=186 meV, which falls be-
tween the known energy of LO ?199 meV? and degenerated
TO and LO ?169 meV? phonons at the ? point,18,19,26,52cor-
responds more probably to a local phonon mode around the
impurity involved in the luminescence process. Moreover,
the pronounced red shoulder observed on each phonon rep-
lica peak shows that multiphonon processes play a major role
in the energy relaxation of the impurity. The coupling be-
tween the defect and the lattice is weak as it is shown by the
Huang-Rhys factor S=1.3 obtained from the normalized
peak intensities. This assignment is consistent with fast lu-
minescence decay of the peaks ???, ???, ???, and ??? ?Fig. 2?.
The PLE spectrum shown in Fig. 3?b? supports our inter-
pretation of the defect-related luminescence. Indeed, it re-
flects characteristic features of such transitions. The lumines-
cence intensity steeply increases and then decreases
progressively with the excitation energy. The sharp onset in
the PLE spectrum at 4.09 eV corresponds to the minimum
energy required to excite the impurity. As in the case of weak
electron-phonon coupling, this absorption onset coincides
with the zero-phonon luminescence peak ???. Interestingly,
two phonon replicas separated by we=170 meV that is close
to the energy of the TO- and LO-degenerated phonons26can
be observed in the PLE spectra. They are indicated by arrows
in Fig. 3?a?. We see that the energies of phonons involved in
the luminescence ?wg? and excitation ?we? processes are dif-
ferent. This results from the electronic state of the defect,
which is not the same in both cases. Consequently, lattice
deformations around the defect are different, which affects
the local phonon frequencies. The fact that ?g??esignifies
a stronger matrix interaction with the impurity in the excited
We do not evidence the impurity involved in the lumines-
cence process in current experiments. However, we can
guess about the influence of carbon with more or less confi-
dence. Recent experiments reveal that the structured lumi-
nescence strongly shows up in hBN samples contaminated
by carbon and oxygen.51This result is consistent with our
previous results39and is supported by our complementary
photoluminescence experiments carried out with a pyrolytic
BN ?pBN? sample. These experiments will be described in
details in a forthcoming article.53pBN is a high-purity ma-
terial, free of carbon compound, obtained by gas-phase reac-
tions between BCl3and NH3at 2300 K and deposition on a
Si substrate ?chemical vapor deposition ?CVD? method?. Ac-
tually, pBN sample exhibits no structured emission around 4
eV whatever the excitation energy is. Its excitation by energy
photon above 4.5 eV uniquely results in a continuous emis-
sion band similar to that observed in hBN.53Moreover, the
PLE spectrum of hBN ?Fig. 3? is rather complicated with a
first edge at 4.09 eV and a second one at 5.2 eV. This may
indicate that the defect involved in the radiative process has
at least two excited levels separated by 1.1 eV positioned
within the band gap. Since the substitutional carbon impurity
at nitrogen site ?CN? is supposed to introduce two energy
levels split by ?0.8 eV in the energy range of 3.2–4.9 eV
below the conduction band of hBN,50its participation in the
luminescence process is most likely.
We discuss now the nature of the broad luminescence
band observed in hBN powder sample under photoexcitation
above 5 eV. Due to its long and multiexponential decay
?curve ??? in Fig. 2?, we assign this luminescence to radiative
recombination of deep DAPs:
Ao+ Do→ A−+ D++ h?DAP.
We remark that this deep DAP is different from that related
to the 5.3-eV emission ?q-DAP?.17As Fig. 3?c? shows, the
correspondent PLE spectrum of the long-lived luminescence
component is dominated by the excitonic peak at 5.95 eV.
This fact indicates that the energy transfer to the DAP recom-
bination channel via free excitons is efficient,
hBN+ h?exc→ exciton,
exciton+ A−+ D+→ A0+ D0.
This excitonic peak is preceded by a weak shoulder for ex-
citation energies ranging from 5 to 5.5 eV. We have assigned
this shoulder to a direct excitation of the deep DAP ?Eq. ?4??.
The PSL spectrum displayed in Fig. 6 shows that no notice-
able photostimulated luminescence can be observed follow-
ing hBN irradiation between 5 and 5.5 eV that indicates no
efficient population of distant traps. In contrast to the direct
ionization of the donor or acceptor ?Eqs. ?2? and ?3??, mecha-
nism ?4? does not lead to the charge injection into CB ?e−? or
valence band ?h+? and does not contribute significantly to
A very large bandwidth of the deep DAP band ??1 eV?
and its nearly symmetrical shape suggest that at least one of
defects involved in the emission is strongly coupled to the
lattice.54This assumption is supported by a comparison of
the luminescence spectra obtained at room and low ?9 K?
temperatures. When the temperature is increased from 9 to
300 K ?Fig. 1?c?? thermally activated quenching is intensi-
fied. The intensity of the broad DAP band then drops and the
band shifts to the blue. Blueshifts induced by the rises of the
temperature have been reported for deep, the so-called “self-
activated,” DAP luminescent bands of several semiconduc-
tors: ZnS,55,56GaAs,57and GaN.58The blueshift is generally
DEFECT-RELATED PHOTOLUMINESCENCE OF HEXAGONAL…
PHYSICAL REVIEW B 78, 155204 ?2008?
observed when localized complex centers with strong
electron-phonon coupling are involved in the luminescence
process. The configurational coordinate ?CC? model, which
takes into account the interaction of such localized center
with matrix, predicts a linear shift of the band position with
temperature.56The relevant blueshift that results from ther-
mal occupation of vibrational levels associated with the ex-
cited state and thermal quenching is due to radiationless re-
combination of e−-h+pairs.
The acceptor complex center involved in such emissions
is usually formed by an acceptorlike vacancy adjacent to a
donor impurity atom,55–58and it is often called “A center.”
The emission results from the electron transition from a rela-
tively distantdonor tothe
complex.59According to theoretical calculations most stables
defects in hBN are supposed to be boron vacancy VBfor
n-type conditions and nitrogen vacancy VNfor p-type condi-
tions. To the best of our knowledge there are no calculations
concerning vacancy-impurity complexes in hBN. Neverthe-
less, results obtained in cubic BN ?cBN? show that VB-CB
and VB-ONcomplexes form deep acceptors. Such center in
association with the shallow CBor ONdonor may be respon-
sible for DAP transitions at 3.9 eV observed in hBN poly-
B. Photostimulated luminescence and band-gap transitions of
As we have already mentioned, the PSL excitation onset
cannot be a measure of band-gap energy. However, in com-
bination with PLE spectra, PSL excitation spectra can pro-
vide valuable information concerning exciton and band-edge
energy positions. Below we will discuss the PSL excitation
in the framework of relevant processes depicted in Fig. 4.
Several interesting details can be remarked from the com-
parison of PLE and PSL excitation spectra displayed in Fig.
Several free exciton transitions in hBN merge in the
dominant peak of the PLE spectrum ?Fig. 3?c?? in the range
of photon energies between 5.8 and 6.0 eV.34One remarkable
feature in Fig. 6 is the noncontribution of these excitons to
the PSL excitation. In fact, dissociation of excitons is re-
quired for a storage of distant charges, which results in PSL.
This dissociation is only possible if the gain in energy due to
the charge localization is greater than the exciton binding
energy. Using this reasoning, we have previously set up the
lower limit to the exciton binding energy De?0.4 eV.17Two
factors additionally contribute to the inhibition of exciton
dissociation: its short lifetime and limited spatial extent. We
have not observed free exciton luminescence in our hBN
polycrystalline sample. The free excitons are rapidly bound
to defects where a part of them recombine radiatively. The
lifetime of free excitons in hBN is then small and that re-
duces their dissociation probability. In a similar way, small
exciton radius decreases the dissociation probability. The
noncontribution of the free exciton of hBN to the PSL pro-
cesses may be an indication of its small radius and tight-
The PSL excitation spectrum in Fig. 6 follows the expo-
nential growth of the PLE spectrum of q-DAP luminescence
at 5.3 eV below 5.7 eV.17The q-DAP states apparently con-
tribute to the PSL signal that complements the general
scheme in Fig. 4 by the type IV transition corresponding to
the direct q-DAP excitation ?Eq. ?4??. This fact seems sur-
prising since no free charge carriers are created in such pro-
cess required for a distant charge trapping. Indeed it can be
understood if we consider the q-DAP transition energy as a
function of the pair distance. Closer charge pairs possess a
higher transition energy and distant pairs lower. The charge
diffusion within the manifold of q-DAP levels below ioniza-
tion can therefore take place toward more distant and hence
long-lived states, which contribute to PSL. The existence of
the exchange between q-DAP traps excited according to Eq.
?4? was suggested in our previous paper.17
In the same publication we have reported a change in the
q-DAP population mechanism at 5.7 eV. At this energy the
q-DAP ionization takes place following Eqs. ?2? and ?3? and
results in the regime crossover from Raman to photolumines-
cence. This can be an ionization of either acceptor or donor
states, whichever is higher lying. Higher-energy photons be-
tween 5.7 and 6.0 eV efficiently produce hBN excitons; how-
ever, they do not appreciably contribute to PSL. This is seen
in Fig. 6 in both PLE ?Elum=5.3 eV? and PSL spectra. How-
ever, another sharp PSL excitation maximum is observed in
the high-energy wing of the exciton absorption band at 6.1
eV. We can relate a high efficiency of the distant charge traps
population at this energy to the ionization of acceptor or
donor states, whichever is lower lying. Resuming, we at-
tribute two spectral features observed at Ei1=5.7 eV and
Ei2=6.1 eV in the PSL excitation spectrum to the direct ion-
ization of donor and acceptor levels ?or vice versa? involved
in the q-DAP luminescence ?processes II and III of Fig. 4
and Eqs. ?2? and ?3??:
EA− = Eg− Ei1 or i2,
ED+ = Eg− Ei2 or i1.
Keeping in mind this assignment, we can estimate the band-
gap energy of hBN. Indeed, the energy of the q-DAP lumi-
nescence Eq-DAP=5.3 eV is given by the relation
Eq-DAP= Eg− EA− − ED+.
Combining Eqs. ?7? and ?8? we then obtain
Eg= Ei2+ Ei1− Eq-DAP? 6.5 eV.
Alternatively, we can assign the spectral feature observed
at Ei2=6.1 eV to the dissociation of excitons on the high-
lying acceptor or donor state of energy Ei1=5.7 eV.17In the
framework of this hypothesis the dissociation of lower-
energy excitons is energetically forbidden. Indeed, the en-
ergy gain from the charge localization on donor or acceptor
is Eex−Ei1, where Eex?Ei2is the exciton energy. For the
efficient charges separation, this gain in energy has to be
higher than the exciton binding energy Eg−Eex. This allows
the energy balance equation relative to the process threshold,
Eg− Eex= Eex− Ei1,
MUSEUR, FELDBACH, AND KANAEVPHYSICAL REVIEW B 78, 155204 ?2008?
Eg= 2Ei2− Ei1= 6.5 eV.
Interestingly, both alternatives in identification of the spectral
feature at Ei2=6.1 eV result in the same band-gap energy of
Our estimation of the hBN band-gap energy is in good
agreement with that obtained from all-electron GW calcula-
tions recently performed by Arnaud et al.34Moreover, taking
the exciton transitions at 5.8 eV, we obtain the exciton bind-
ing energy of 0.7 eV. This disagrees with the Wannier-type
exciton suggested by Watanabe et al.1and supports Arnaud’s
prediction of the Frenkel-type exciton with strong binding
The shape analysis of PLE and PSL excitation spectra
supports the obtained value of hBN band-gap energy. In Fig.
7 we have plotted the PLE spectra of the bound excitons
luminescence ?Elum=5.5 eV? together with the PLE spectra
of broad DAP luminescence ?Elum=3.73 eV?. The PLE
?Elum=5.5 eV? peak at 5.8 eV fits well the position of the
strongest excitonic transitions predicted theoretically,34,60,61
as represented by vertical bar in Fig. 7.As well the plateau of
the PLE spectra in the energy range between 6.1 and 6.5 eV
is related to low-intensity excitonic transitions converging to
the dissociation limit. Interestingly, both PLE spectra coin-
cide in the high-energy spectra range above 6.1 eV and fol-
low the PSL spectrum until 6.5 eV. Above 6.5 eV the PSL
grows faster while the intensity of the PLE spectra decreases.
The PSL is expected to increase at h??Egwhere, according
to Fig. 4, process I efficiently contributes to the distant traps
population. This difference in shapes between the PLE and
PSL spectra is therefore explained by interband transitions
and supports well our band-gap energy estimation Eg
Our final remark concerns the shape of PLE spectrum.
From a general point of view, as long as the hBN can be
considered as optically thin material, the shape of the PLE
spectrum is expected to follow the intrinsic absorption spec-
trum. At higher absorbance, a saturation or even a decrease
in PLE intensity appears as an intensification of nonradiative
recombination processes. Interestingly, we observe the indi-
cation of saturation effect around 6.5 eV in good agreement
with our band-gap estimation.
Photoluminescence of hexagonal boron nitride has been
studied by means of the time- and energy-resolved photolu-
minescence spectroscopy methods. Depending on the excita-
tion energy several luminescence bands have been observed.
?i? A strongly structured band in the energy range between
4.1 and 3.3 eV is assigned to phonon replicas of an impurity
luminescence. ?ii? A very broad band ??E?1 eV? centered
at 3.9 eV is assigned to DAP transitions involving a strongly
localized acceptor complex center. Moreover, the intensity
ratio between these two emissions strongly depends on exci-
tation photon energy. ?iii? q-DAPs are responsible for the
5.3-eV emission and ?iv? emission of excitons bound to de-
fects is observed at 5.5 eV.
The excitation of samples by high-energy photons above
5.4 eV enables another phenomenon called photostimulated
luminescence, which is due to distantly trapped photoin-
duced charges. PSL is observed in hBN following interband
and acceptor- and donor-band transitions. Moreover, we
show that in contrast to DAP, PSL can also result from a
direct q-DAP excitation. The comparison of photolumines-
cence excitation and PSL spectra allows band-gap energy
estimation of 6.5 eV and supports the hypothesis of Frenkel-
type exciton in hBN with large binding energy of 0.7 eV.
most recent theoretical
This work was supported by the IHP Contract No. HPRI-
CT-1999-00040 of the European Commission. The authors
are grateful to G. Stryganyuk for assistance in conducting
experiments at HASYLAB ?synchrotron DESY? and to P.
Jaffrennou and F. Ducastelle for helpful discussions.
1K. Watanabe, T. Taniguchi, and H. Kanda, Nat. Mater. 3, 404
2Y. Kubota, K. Watanabe, O. Tsuda, and T. Taniguchi, Science
317, 932 ?2007?.
3R. S. Lee, J. Gavillet, M. Lamy de la Chapelle, A. Loiseau, J.-L.
Cochon, D. Pigache, J. Thibault, and F. Willaime, Phys. Rev. B
64, 121405?R? ?2001?.
4R. Arenal, O. Stephan, J. L. Cochon, and A. Loiseau, J. Am.
FIG. 7. ?Color online? PSL excitation spectrum ?open circles?
and PLE spectra of bound exciton luminescence ?5.5 eV, dashed
line? and deep DAP luminescence ?3.73 eV, solid line?. Theoretical
oscillator strengths ?Refs. 34, 60, and 61? of free exciton transitions
in hBN are shown by straight vertical bars.
DEFECT-RELATED PHOTOLUMINESCENCE OF HEXAGONAL…
PHYSICAL REVIEW B 78, 155204 ?2008?
Chem. Soc. 129, 16183 ?2007?. Download full-text
5A. Loiseau, F. Willaime, N. Demoncy, G. Hug, and H. Pascard,
Phys. Rev. Lett. 76, 4737 ?1996?.
6N. G. Chopra, R. J. Luyken, K. Cherrey, V. H. Crespi, M. L.
Cohen, S. G. Louie, and A. Zettl, Science 269, 966 ?1995?.
7P. Jaffrennou, F. Donatini, J. Barjon, J. S. Lauret, A. Maguer, B.
Attal-Tretout, F. Ducastelle, and A. Loiseau, Chem. Phys. Lett.
442, 372 ?2007?.
8P. Jaffrennou, J. Barjon, J. S. Lauret, A. Maguer, D. Golberg, B.
Attal-Trétout, F. Ducastelle, and A. Loiseau, Phys. Status Solidi
B 244, 4147 ?2007?.
9P. Jaffrennou, J. Barjon, T. Schmid, L. Museur, A. Kanaev, J.-S.
Lauret, C. Y. Zhi, C. Tang, Y. Bando, D. Golberg, B. Attal-
Tretout, F. Ducastelle, and A. Loiseau, Phys. Rev. B 77, 235422
10M. G. Silly, P. Jaffrennou, J. Barjon, J.-S. Lauret, F. Ducastelle,
A. Loiseau, E. Obraztsova, B. Attal-Tretout, and E. Rosencher,
Phys. Rev. B 75, 085205 ?2007?.
11K. Watanabe, T. Taniguchi, T. Kuroda, and H. Kanda, Appl.
Phys. Lett. 89, 141902 ?2006?.
12K. Watanabe, T. Taniguchi, and H. Kanda, Phys. Status Solidi A
201, 2561 ?2004?.
13C. A. Taylor II, S. W. Brown, V. Subramaniam, S. Kidner, S. C.
Rand, and R. Clarke, Appl. Phys. Lett. 65, 1251 ?1994?.
14V. L. Solozhenko, A. G. Lazarenko, J. P. Petitet, and A. V. Ka-
naev, J. Phys. Chem. Solids 62, 1331 ?2001?.
15D. A. Evans, A. G. McGlynn, B. M. Towlson, M. Gunn, D.
Jones, T. E. Jenkins, R. Winter, and N. R. J. Poolton, J. Phys.:
Condens. Matter 20, 075233 ?2008?.
16P. Jaffrennou, J. Barjon, J.-S. Lauret, B. Attal-Trétout, F. Ducas-
telle, and A. Loiseau, J. Appl. Phys. 102, 116102 ?2007?.
17L. Museur and A. V. Kanaev, J. Appl. Phys. 103, 103520 ?2008?.
18D. M. Hoffman, G. L. Doll, and P. C. Eklund, Phys. Rev. B 30,
19R. Geick, C. H. Perry, and G. Rupprecht, Phys. Rev. 146, 543
20S. N. Grinyaev, F. V. Konusov, V. V. Lopatin, and L. N. Shiyan,
Phys. Solid State 46, 435 ?2004?.
21G. Chen, X. Zhang, B. Wang, X. Song, B. Cui, and H. Yan,
Appl. Phys. Lett. 75, 10 ?1999?.
22Z. Remes, M. Nesladek, K. Haenen, K. Watanabe, and T. Tan-
iguchi, Phys. Status Solidi A 202, 2229 ?2005?.
23A. Mansour and S. E. Schnatterly, Phys. Rev. B 36, 9234
24R. D. Carson and S. E. Schnatterly, Phys. Rev. Lett. 59, 319
25J. B. MacNaughton, A. Moewes, R. G. Wilks, X. T. Zhou, T. K.
Sham, T. Taniguchi, K. Watanabe, C. Y. Chan, W. J. Zhang, I.
Bello, S. T. Lee, and H. Hofsäss, Phys. Rev. B 72, 195113
26J. Serrano, A. Bosak, R. Arenal, M. Krisch, K. Watanabe, T.
Taniguchi, H. Kanda, A. Rubio, and L. Wirtz, Phys. Rev. Lett.
98, 095503 ?2007?.
27J. A. Carlisle, E. L. Shirley, L. J. Terminello, J. J. Jia, T. A.
Callcott, D. L. Ederer, R. C. C. Perera, and F. J. Himpsel, Phys.
Rev. B 59, 7433 ?1999?.
28J. J. Jia, T. A. Callcott, E. L. Shirley, J. A. Carlisle, L. J. Ter-
minello, A. Asfaw, D. L. Ederer, F. J. Himpsel, and R. C. C.
Perera, Phys. Rev. Lett. 76, 4054 ?1996?.
29B. M. Davies, F. Bassani, F. C. Brown, and C. G. Olson, Phys.
Rev. B 24, 3537 ?1981?.
30F. C. Brown, R. Z. Bachrach, and M. Skibowski, Phys. Rev. B
13, 2633 ?1976?.
31C. Tarrio and S. E. Schnatterly, Phys. Rev. B 40, 7852 ?1989?.
32R. D. Leapman, P. L. Fejes, and J. Silcox, Phys. Rev. B 28, 2361
33R. D. Leapman and J. Silcox, Phys. Rev. Lett. 42, 1361 ?1979?.
34B. Arnaud, S. Lebègue, P. Rabiller, and M. Alouani, Phys. Rev.
Lett. 96, 026402 ?2006?.
35L. Wirtz, A. Marini, M. Gruning, and A. Rubio, arXiv:cond-mat/
36G. Zimmerer, Nucl. Instrum. Methods Phys. Res. A 308, 178
37V. Nagirnyi, M. Kirm, A. Kotlov, A. Lushchik, and L. Jönsson,
J. Lumin. 102-103, 597 ?2003?.
38V. Nagirnyi, P. Dorenbos, E. Feldbach, L. Jönsson, M. Kerikmäe,
M. Kirm, E. v. d. Kolk, A. Kotlov, H. Kraus, A. Lushchik, V.
Mikhailik, R. Sarakvasha, and A. Watterich, in Proceedings of
the Eighth International Conference on Inorganic Scintillators
and their use in Scientific and Industrial Applications SCINT 05
?Alushta, Crimea, Ukraine, 2005?.
39L. Museur, D. Anglos, J. P. Petitet, J. P. Michel, and A. V. Ka-
naev, J. Lumin. 127, 595 ?2007?.
40B. Berzina, L. Trinckler, R. Krutohvostov, R. T. Williams, D. L.
Caroll, R. Czerw, and E. Shishnonok, Phys. Status Solidi C 2,
41A. V. Kanaev, J. P. Petitet, L. Museur, V. Marine, V. L. Solo-
zhenko, and V. Zafiropulos, J. Appl. Phys. 96, 4483 ?2004?.
42J. Wu, W.-Q. Han, W. Walukiewicz, J. W. Ager III, W. Shan, E.
E. Haller, and A. Zettl, Nano Lett. 4, 647 ?2004?.
43A. I. Lukomskii, V. B. Shipilo, and L. M. Gameza, J. Appl.
Spectrosc. 57, 607 ?1992?.
44M. Kawaguchi, Y. Kita, and M. Doi, J. Mater. Sci. 26, 3926
45K. Era, F. Minami, and T. Kuzuba, J. Lumin. 24-25, 71 ?1981?.
46S. Larach and R. E. Shrader, Phys. Rev. 102, 582 ?1956?.
47S. Larach and R. E. Shrader, Phys. Rev. 104, 68 ?1956?.
48L. Museur, J. P. Petitet, J. P. Michel, W. Marine, D. Anglos, C.
Fotakis, and A. Kanaev, J. Appl. Phys. ?to be published?.
49A. Katzir, J. T. Suss, A. Zunger, and A. Halperin, Phys. Rev. B
11, 2370 ?1975?.
50A. Zunger and A. Katzir, Phys. Rev. B 11, 2378 ?1975?.
51T. Taniguchi and K. Watanabe, J. Cryst. Growth 303, 525
52G. Kern, G. Kresse, and J. Hafner, Phys. Rev. B 59, 8551
53L. Museur and A. Kanaev ?unpublished?.
54L. Pavesi and M. Guzzi, J. Appl. Phys. 75, 4779 ?1994?.
55H. Samelson and A. Lempicki, Phys. Rev. 125, 901 ?1962?.
56S. Shionoya, T. Koda, K. Era, and H. Fujiwara, J. Phys. Soc. Jpn.
19, 1157 ?1964?.
57E. W. Williams, Phys. Rev. 168, 922 ?1968?.
58M. A. Reshchikov and H. Morkoc, J. Appl. Phys. 97, 061301
59J. W. Allen, Semicond. Sci. Technol. 10, 1049 ?1995?.
60L. Wirtz, A. Marini, M. Grüning, C. Attaccalite, G. Kresse, and
A. Rubio, Phys. Rev. Lett. 100, 189701 ?2008?.
61B. Arnaud, S. Lebègue, P. Rabiller, and M. Alouani, Phys. Rev.
Lett. 100, 189702 ?2008?.
MUSEUR, FELDBACH, AND KANAEVPHYSICAL REVIEW B 78, 155204 ?2008?