Radiative decay of vacuum-ultraviolet excitation of silica synthesized by molecular precursors of Si-Si sites: An indicator of intracenter relaxation of neutral oxygen vacancies

Article abstract of DOI:10.1103/PhysRevB.71.075101

A well defined type of oxygen deficiency has been induced in the silica network by introducing Si-Si groups through a molecular doping based on the sol-gel synthesis. In this material, we have investigated the visible and UV photoluminescence (PL), up to 7 eV, arising from localized states excited in the energy range 3-12 eV by synchrotron radiation. The PL excitation spectrum is dominated by one main sub-band-gap excitation band, peaked at 7.1 with a full width of half-maximum of 0.8 eV. No PL arises exciting at 5 eV, specifically no PL band at 4.4 and 2.7 eV are observed, ruling out the formation of twofold coordinated -Si-sites. These data confirm all previous theoretical and experimental assignments of transitions at about 7-8 eV to localized excitations of neutral-oxygen-vacancy (NOV) sites. Exciting in the vacuum-UV (VUV) above 7 eV, the investigated material does not show the typical 4.4 and 2.7 eV PL observed in fused silica. The peculiar NOV configuration appears to inhibit the photoconversion process responsible in fused silica for the VUV excitation of the twofold coordinated silicon emission. The main emission is peaked at about 3.7 eV, with a PL lifetime of about 1 μs. A minor component with a much slower lifetime (800 μs) has been detected at 2.9 eV, with a further minor excitation channel at about 6 eV. All emissions excited at 7.1 eV show an intensity decrease with the temperature, but a negligible thermal change of the lifetime. The results give an indication of the possible energy level structure of NOV and evidence an efficient nonradiative decay mechanism of the excited state, caused by a strong electron-phonon interaction during the VUV excitation of the defect. The analysis of the data suggests a large Si-Si bond relaxation of about 0.1 nm, giving definite experimental confirmation of previous theoretical calculations. 2005 The American Physical Society.

Full text of DOI:10.1103/PhysRevB.71.075101

PHYSICAL REVIEW B 71, 075101 ͑2005͒
Radiative decay of vacuum-ultraviolet excitation of silica synthesized by molecular precursors
of Si-Si sites: An indicator of intracenter relaxation of neutral oxygen vacancies
A. Paleari,* N. Chiodini, D. Di Martino, and F. Meinardi
Istituto Nazionale Fisica della Materia, Dipartimento di Scienza dei Materiali, Università di Milano-Bicocca, via Cozzi 53,
I-20125 Milano, Italy
͑
Received 21 May 2004; revised manuscript received 27 September 2004; published 1 February 2005͒
A well defined type of oxygen deficiency has been induced in the silica network by introducing Si-Si groups
through a molecular doping based on the sol-gel synthesis. In this material, we have investigated the visible
and UV photoluminescence ͑PL͒, up to 7 eV, arising from localized states excited in the energy range
3
–12 eV by synchrotron radiation. The PL excitation spectrum is dominated by one main sub-band-gap
excitation band, peaked at 7.1 with a full width of half-maximum of 0.8 eV. No PL arises exciting at 5 eV,
specifically no PL band at 4.4 and 2.7 eV are observed, ruling out the formation of twofold coordinated -Si-
sites. These data confirm all previous theoretical and experimental assignments of transitions at about 7–8 eV
to localized excitations of neutral-oxygen-vacancy ͑NOV͒ sites. Exciting in the vacuum-UV ͑VUV͒ above
7
eV, the investigated material does not show the typical 4.4 and 2.7 eV PL observed in fused silica. The
peculiar NOV configuration appears to inhibit the photoconversion process responsible in fused silica for the
VUV excitation of the twofold coordinated silicon emission. The main emission is peaked at about 3.7 eV,
with a PL lifetime of about 1 ␮s. A minor component with a much slower lifetime ͑800 ␮s͒ has been detected
at 2.9 eV, with a further minor excitation channel at about 6 eV. All emissions excited at 7.1 eV show an
intensity decrease with the temperature, but a negligible thermal change of the lifetime. The results give an
indication of the possible energy level structure of NOV and evidence an efficient nonradiative decay mecha-
nism of the excited state, caused by a strong electron-phonon interaction during the VUV excitation of the
defect. The analysis of the data suggests a large Si-Si bond relaxation of about 0.1 nm, giving definite
experimental confirmation of previous theoretical calculations.
DOI: 10.1103/PhysRevB.71.075101
PACS number͑s͒: 78.55.Qr, 81.20.Fw
I. INTRODUCTION
bands at 5 and 6.9 eV PL, respectively, while PL at 4.4 and
2
.7 eV are assigned to S →S and T →S transitions, re-
1
0
1
0
Silicon-silicon covalent bonds in the 4:2 coordination
structure of silicon dioxide are among the key defect con-
figurations responsible for defect-related processes in SiO2,
many of them with a demonstrated technological relevance.
Point defects play an important role in several optical appli-
cations of silica, such as in photorefractive writing of optical
waveguides and in vacuum-ultraviolet ͑VUV͒ photo-
spectively. In this scheme, the 7.6 eV band was initially in-
terpreted as caused by transitions to a further excited state of
8
the same defect, with a wSi-Siw configuration. However,
successive studies, with detailed consideration of the kinetics
of the PL decay processes excited at 5, 6.9, and 7.6 eV,
resulted in a new scenario with two kinds of oxygen-
deficient centers ͑ODCs͒, named ODC͑I͒ and ODC͑II͒, ab-
1
–3
9
lithography.
Nevertheless, the attribution of the optical
sorbing at 7.6 and 5 eV, respectively. A few studies inves-
4
,5
features arising from point defects are still debated. Quite
surprisingly, a major uncertainty concerns the identification
of transitions pertaining to the neutral oxygen vacancy
tigated the possible attribution of ODC͑I͒ and ODC͑II͒ to
relaxed wSi-Siw and unrelaxed wSi¯Siw sites,
1
0
11–14
respectively, also by means of theoretical calculations.
1
5
͑
NOV͒ in the form of wSi-Siw sites, which is expected to
More recently, during the last decade, Skuja proposed a
completely different ODC͑II͒ model, now widely accepted,
where the phenomenological scheme of the 5 and 6.9 eV
excited PL, with all implications on lifetimes, temperature
dependence of relative intensities, and cation-substitution
be the basic and more abundant intrinsic defect in the silica
network. Optical absorption at 7.6 eV in oxygen deficient or
5
neutron irradiated silica is almost consensually attributed to
excitation of this site from its ground-state S singlet to an
0
excited S state. This attribution is based on experimental
effects on transition rates, is well accounted for by a
1
4
4,16
and theoretical evidence. Alternative assignments were pro-
twofold-coordinated -Si- configuration,
whose C2v local
5
6,7
posed earlier and more recently. Further complications
come from controversial assignments of the emission bands
observed by excitation at 7.6 eV: a fast 4.4 eV ͑few ns decay
time͒ and a slow ͑about 10 ms of decay time͒ 2.7 eV photo-
luminescence ͑PL͒ bands. These emissions are excited within
PL excitation ͑PLE͒ bands at 5 and 6.9 eV, and also in agree-
ment with a four-level intracenter scheme comprising three
singlet states S , S , and S , and a triplet T level. The S
symmetry is also consistent with polarization-sensitive PL
1
7
measurements. Ab initio calculations have recently con-
firmed this proposal.1
8,19
However, the upgraded interpreta-
tion of ODC͑II͒ further complicated the assignment of the
ODC͑I͒ optical activity. In fact the problem arose of explain-
ing the coincident PL features excited at 7.6 and 5 eV. A
model of defect conversion upon 7.6 eV excitation from
wSi-Siw sites to -Si- configuration, with a possible nonra-
0
1
2
1
0
3
→
S and S →S transitions are responsible for excitation
diative decay towards two EЈ centers ͑i.e., unpaired sp elec-
1
0
2
1098-0121/2005/71͑7͒/075101͑7͒/$23.00
075101-1
©2005 The American Physical Society

PALEARI et al.
PHYSICAL REVIEW B 71, 075101 ͑2005͒
trons in threefold coordinated wSi· sites͒ was proposed by
dynamic ͑MD͒ simulations. From an experimental point of
view, detailed measurements21 have recently confirmed that
an intersystem process is involved in the 7.6 eV excited
ment, especially within a highly hydrated low-density ma-
trix. Specifically, care should be taken to account of the pos-
sible occurrence of reactions, in principle chemically
feasible, such as
1
5
Skuja and then theoretically demonstrated by molecular
2
0
¨
4.4 eV PL. On the other hand, recent numerical calculations
wSi - Si w ꢀ v Si + v Si v ,
͑1͒
by embedded density-functional theory of configuration en-
ergy suggest possible intracenter PL from T and S states of
wSi - Si w + w Si - OH ꢀ w Si - H
+ w Si - O - Si w ,
1
1
wSi-Siw sites in the same energy range of emissions from
1
2–14
͑2͒
͑3͒
-
Si- sites, although with a large energy uncertainty.
In
2
0
addition, the previously quoted MD study gives a non-null
activation energy for the defect conversion from wSi-Siw
to -Si- sites. Even if its value ͑0.72 eV͒ is probably overes-
timated, this suggests a possible intracenter decay, and it is
not clear whether it is superimposed on the -Si- emissions or
strongly inhibited by the occurrence of a largely more fa-
vored photoconversion of the wSi-Siw site. In fact, no in-
tracenter radiative decay from ODC͑I͒ has yet been identified
as free from interference with optical activity ascribable to
stable or transient ODC͑II͒. As a consequence, the energy
structure and the occurrence itself of NOV centers have re-
mained elusive.
In this work, we have found a way to force wSi-Siw
sites into a configuration that prevents the occurrence of
ODC͑II͒, either as native or transient defects from NOV pho-
toconversion. This result gives evidence for the sensitivity of
the photoconversion process to the defect environment, and
gives also an insight into the possible NOV transitions and
energy structure of this basic defect in silica.
wSi - Si w + H O ꢀ w Si - H + w Si - OH.
2
Reaction ͑1͒ expresses the possibility of the thermally ac-
tivated reaction of defect conversion between ODC͑I͒ and
ODC͑II͒, analogous to the VUV excited defect conversion of
ODC͑I͒ decaying in ODC͑II͒ proposed to explain the excita-
15,20
tion of ODC͑II͒ PL within the ODC͑I͒ absorption band.
Occurrence of this reaction would compromise the selectiv-
ity of the synthesis method in doping with ODC͑I͒ only.
Checking the absence of PL from ODC͑II͒, we verified that
the equilibrium of reaction ͑1͒ is largely unbalanced towards
ODC͑I͒ throughout the synthesis route, confirming that
wSi-Siw sites from EMODS are stable as regards reaction
͑
1͒. Reactions ͑2͒ and ͑3͒ may instead be really efficient
during the thermal treatment in removing a major fraction of
the wSi-Siw sites introduced by the synthesis, because of
the great amount of OH groups and adsorbed water in the
first steps of the sol-gel preparation method. In fact, hydroxyl
groups and water molecules are highly reactive in silicon
dioxide, as experimentally and theoretically proved.2 . Re-
action ͑3͒ has been recently studied by first-principles
calculations.25 To verify that the equilibrium of reactions ͑2͒
and ͑3͒ is not completely unbalanced towards the formation
of wSi-H sites, a parallel synthesis of samples from TEOS
with addition of triethoxysilane ͑TREOS͒ Si͑OCH CH ͒ H,
3–25
II. EXPERIMENT
A. Material synthesis and treatments
The choice of the kind of material synthesis was crucial in
this work to obtain a system enabling the investigation of a
particular kind of oxygen deficiency. To ensure the introduc-
tion of the specific NOV configuration, synthesis of silica by
the sol-gel method was used. This method starts from room
temperature hydrolysis of molecular precursors of the silica
2
3 3
containing wSi-H groups, was carried out. In fact, if a frac-
tion of wSi-Siw sites is stable as regards reaction ͑2͒
throughout the synthesis route, the material resulting from
addition of TREOS should contain wSi-Siw sites as well.
Anyway, amounts of EMODS or TREOS of few mol % were
used to assure a less than complete reaction with hydroxyl
groups in the final high-temperature steps of the treatment,
when OH content drops at smaller values. Therefore, oxygen
deficient silica samples were prepared by co-gelling TEOS
and 1 or 5 mol % of EMODS or TREOS in ethanol, as a
solvent, by adding H O ͑TEOS/H O 1:8 molar ratio, TEOS/
ethanol 1:3 volume ratio͒. Gelation occurred in a few days at
0 °C. Xerogel was then obtained by slowly evaporating the
solvent. The final transparent glass has been produced by
heating in inert atmosphere N or Ar ͑4 °C/h͒ up to com-
plete densification at 1050 °C. Reference samples from
TEOS were also synthesized for comparison.
SiO tetrahedral units, specifically Si͑OCH CH ͒ ͑TEOS͒,
4
2
3 4
forming a low-density skeleton that is successively purified
from solvents and organic radicals, as well as dried and den-
sified, through a slow thermal treatment in controlled atmo-
sphere ͑usually in a flux of oxygen͒ up to about 1000 °C.22
In this way, ethyl and methyl groups are removed ͑even be-
low 500 °C͒, as well as possible carbonyl residuals. Pure
silica with the typical density of vitreous silica is finally
obtained, with a content of OH groups of the order of
2
2
4
0.1 at. %.
In the first step of the synthesis, other molecular reagents
2
may be introduced together with TEOS to achieve complex
compositions or doping of the silica matrix. The hexam-
ethoxydisilane ͑EMODS͒ Si ͑OCH ͒ molecule, containing
2
3 6
the wSi-Siw group, can be used to introduce NOV sites in
the silica network, provided that a suitable inert atmosphere
prevents complete oxidation of the material, comprising all
Si-Si bonds, during the thermal treatment. However, even in
an inert atmosphere, wSi-Siw sites may react and convert
into other coordination structures during the thermal treat-
B. Optical measurements
PL measurements were carried out using synchrotron ra-
diation at the SUPERLUMI experimental station of HASY-
LAB at DESY ͑Hamburg, Germany͒, in the temperature
range between 9 and 300 K. The excitation spectral band-
075101-2

RADIATIVE DECAY OF VACUUM-ULTRAVIOLET…
PHYSICAL REVIEW B 71, 075101 ͑2005͒
FIG. 2. 10 K PL spectrum excited at 7.1 eV in NOV-doped
silica produced by the sol-gel method, and PL excitation spectrum
of PL at 3.7 eV.
FIG. 1. ͑a͒ Raman spectra at 300 K of silica produced by the
sol-gel method with addition of EMODS ͑curve 1͒ and commercial
synthetic pure silica ͑Tetrasil B, Haereus͒ ͑curve 2͒. Spectra are
shifted for clarity. ͑b͒ Difference spectrum ͑curve 2−curve 1͒. In-
set: infrared spectrum of silica produced by the sol-gel method with
addition of EMODS.
residuals of the solgel synthesis in form of ethylic or meth-
ylic groups, whose vibrational modes should give narrow
peaks in the range 2800–3000 cm, but only a relevant con-
centration of Si-OH sites. Visible and near-UV absorption
spectra from 800 to 200 nm and EPR measurements of
samples of the final material did not evidence any detectable
optical or paramagnetic activity from metallic or other impu-
rities or intrinsic color centers, showing completely flat spec-
tra.
width was 0.3 nm with photon energy ranging between 5 and
12 eV. PL signals were detected by a photomultiplier
coupled with monochromators with 3 nm of emission band-
width up to 5 eV, and 0.5 nm up to 7.5 eV in a vacuum
chamber. Data were corrected for the spectral dependence of
the excitation intensity. Time resolved measurements in the
nanosecond domain were carried out with a time window of
200 ns and 0.8 ns overall time response. Slow PL decays
B. Photoluminescence
were measured by chopping the excitation with a time win-
dow of 20 or 500 ␮s and 1 or 25 ␮s of pulse rise time,
respectively. Lifetime values were calculated by fitting decay
curves as a convolution of contributions from exciting pulse
and emission decay.
Analysis of PL activity was carried out collecting PL
spectra at several excitation energy values in the range
5
–8 eV to identify the defect-related sub-band-gap excited
emissions of the silica glass investigated. Excitation at en-
ergy higher than the gap of silica was also carried out at
1
1 eV. This analysis, throughout a set of different samples
III. RESULTS
synthesized from TEOS, TEOS and EMODS, and TEOS and
TREOS, evidenced one kind of PL activity, efficiently ex-
cited at around 7 eV. This activity is completely absent in
undoped silica samples produced via the same route from
TEOS only. No sample exhibits PL excited at around 5 eV,
ruling out the formation of ODC͑II͒.
The observed emission is peaked at 3.7 eV, with a shoul-
der at about 2.9 eV, and an excitation band peaked at
7.12 eV with a bandwidth of about 0.8 eV ͓full width at
half-maximum ͑FWHM͔͒ ͑Fig. 2͒. Possible PL components
in the VUV up to 7 eV were also looked for, using a different
detection line, but no other emission was found.
Significant intensity changes were observed changing the
temperature, as shown in Fig. 3͑a͒. Specifically, the PL in-
tensity, irrespective of the emission energy, grows following
the temperature dependence shown in Fig. 3͑b͒. By contrast,
distinct PL lifetime ␶ values pertain to the PL components. A
longer PL decay ͑␶Ϸ800 ␮s͒ is observed for the minor com-
ponent at 2.9 eV, whereas a lifetime of about 1 ␮s has been
estimated for the emission at 3.7 eV. No relevant change of
lifetime was observed increasing the temperature.
The bandwidth of the excitation spectrum ͑0.78 eV at
8 K͒ shows a small thermal broadening of 0.07 eV when
heating the sample to 300 K ͓Fig. 4͑a͔͒. Such a large
temperature-independent bandwidth indicates either a strong
electron-phonon interaction or a large inhomogeneous con-
A. Material characterization
Analysis of the main physical and structural features of
the final material was carried out through measurements of
refractive index ͑using the prism-coupler technique͒, poros-
ity ͑with the Bnanauer-Emmett-Taylor method͒, Raman
analysis, visible and IR absorption, and electron paramag-
netic resonance ͑EPR͒ measurements. We found a refractive
−
5
index value 1.45705±8ϫ10 at 633 nm, to be compared
−
5
26
with 1.45702±3ϫ10 of commercial pure silica. This
value is consistent with the transformation of the xerogels
into glass after the thermal treatment at 1050 °C, as expected
from this kind of route.22 Measurements of surface area after
the thermal treatment did not give values above the sensitiv-
2
ity limit ͑1 m /g͒. The Raman spectrum in the phonon en-
−
1
ergy range 200–1300 cm ͓curve 1 in Fig. 1͑a͔͒, with broad
−
1
structures centered at 440, 800, and 1050–1200 cm ͑␻1,
␻3, and ␻ modes, respectively͒ and peaks at 495 and
4
1
6
06 cm− ͑D and D peaks, respectively͒, reproduces the
1 2
2
7
typical features of pure silica. These data reflect the forma-
tion of a glassy network without crystallization or spurious
2
8,29
phases.
The comparison with the spectrum of a commer-
cial pure silica ͓curve 2 in Fig. 1͑a͔͒ does not show signifi-
cant deviations ͓difference spectrum in Fig. 1͑b͔͒. The IR
spectrum ͓inset of Fig. 1͑a͔͒ rules out the presence of organic
075101-3

PALEARI et al.
PHYSICAL REVIEW B 71, 075101 ͑2005͒
produced by analogous sol-gel preparation method, with
comparable probability of contamination from impurities due
to either the reagents or the process. Therefore, the 7.1 eV
PLE cannot be ascribed to impurities or extrinsic residuals of
the synthesis process. The analysis of the physical and struc-
tural properties ͑Sec. III A͒, in agreement with other studies
on sol-gel silica,22 confirms that the sol-gel route, after a
thermal treatment at temperature higher than 1000 °C, gives
rise to the formation of a fully inorganic matrix of amor-
phous SiO with a purity level comparable with commercial
2
synthetic pure silica. Raman data, together with refractive
index and surface area measurements, show typical bulk fea-
tures without detectable evidence of a porous matrix.2
Therefore, surface defects are unlikely to be responsible for
8,29
FIG. 3. ͑a͒ PL spectra excited at 7.1 eV in NOV-doped silica
produced by the sol-gel method and densified up to 1050 °C at
different temperatures from 8 K to 300 K ͑with intensity decreas-
ing by increasing the temperature͒ in steps of about 50 K. ͑b͒ Tem-
perature dependence of the PL intensity. The curve is the best fit
from Eq. ͑7͒.
the observed PL activity. As a matter of fact, PL emissions of
porous silica or silica xerogels3
0–33
show PL bands in the
spectral region 2–4 eV, but with small Stokes shift
0.5–2 eV͒ from the excitation and fast PL decay time in the
͑
nanosecond domain, in contrast with the excitation and life-
time features of our data. Anyway, the optical activity of
porous silica-based materials typically vanishes after thermal
treatment at about 1000 °C.33
tribution. To investigate this aspect, we have collected exci-
tation spectra at different emission energies ͓Fig. 4͑b͔͒,
showing no significant modification of the spectral shape as
instead expected from an inhomogeneously broadened band.
Figure 4͑b͒ also shows a much less efficient excitation chan-
nel at about 6 eV, with a relatively larger yield for the 2.9 eV
PL component. No relevant change of intensity appears by
changing the temperature at this excitation energy.
In summary, the strategy of synthesis, the material char-
acterization, the measurements of reference samples, and the
consideration of previous works on sol-gel silica, point to the
assignment of the 7.1 eV optical transition to the excitation
of intrinsic defects, specifically to wSi-Siw sites introduced
by molecular doping in an otherwise pure silica network.
This result is in agreement with all previous theoretical
and experimental attributions of VUV excitation transitions
above 7 eV to NOV’s. Nevertheless, the approach we have
followed to induce NOV sites gives rise to a PL phenomenol-
ogy with three main peculiarities: ͑i͒ no optical activity is
observed from ODC͑II͒ decay, neither native nor generated
from photoactivated conversion of VUV excited ODC͑I͒; ͑ii͒
VUV excitation gives rise to previously unobserved radiative
decay channels; and ͑iii͒ the VUV PLE band related to NOV
is slightly different in energy with respect to the band usually
attributed to ODC͑I͒ and found in the range 7.4–7.6 eV.
The observed peculiarities are evidence of the possibility
of obtaining a NOV environment significantly different from
that arising in silica produced at high temperature. As re-
gards the lowering of the excitation energy, the shift is com-
parable with that encountered in other silica defects occur-
ring in different variants as a result of environment
modifications, such as bulk and surface EЈ centers ͑from
IV. DISCUSSION
A. Origin of the PL activity
Our data show that a VUV PLE activity at 7.1 eV is in-
duced in silica as a result of the molecular doping with pre-
cursors of wSi-Siw groups. The careful consideration of
the possible reactions involving wSi-Siw precursors during
the networking and densification of the silica matrix ͑Sec.
II A͒ suggests to ascribe the PL activity to wSi-Siw sites in
form of intrinsic defects. This is supported by the compari-
son between doped and undoped samples, and by the physi-
cal and structural analysis of the final material. In fact, the
7.1 eV PLE is absent in undoped ͑from TEOS only͒ silica
5
.8 to 5.4 eV͒ and native/extrinsic and radiation-induced
4
,34
ODC͑II͒ ͑from 5.2 to 4.9 eV͒.
Indeed, theoretical cluster
calculations35 of the NOV energy structure showed that a
slight modification of configuration constraints in the shell of
the six coordinating oxygen atoms, with a resulting change
−
3
of the SiuSi distance of only 5ϫ10 nm, is sufficient to
decrease the S →S transition energy of 0.4 eV. Therefore,
0
1
particular structural constraints, deriving from the molecular
introduction of Si-Si sites via the sol-gel method, can really
justify a significant shift of the NOV excitation.
FIG. 4. ͑a͒ Excitation spectra of the 3.7 eV emission in NOV-
The lowering of transition energy from 7.6 down to
7.1 eV is consistent with an excited NOV level deeper in the
gap and a more localized charge density of the excited con-
figuration. Therefore, the strong Rydberg feature of the VUV
doped silica at 8 and 300 K. Solid curves are Gaussian fit of the
data giving the indicated FWHM values. ͑b͒ 8 K excitation spectra
of PL at 2.7 ͑circles͒ and 3.7 eV ͑triangles͒.
075101-4

RADIATIVE DECAY OF VACUUM-ULTRAVIOLET…
PHYSICAL REVIEW B 71, 075101 ͑2005͒
respect to the ground level of the site. Indeed, previous the-
oretical studies on the NOV configuration in silica suggested
a triplet state of the site with excitation energy of about
4
2
–6 eV and relaxation energy of the excited state of about
eV, resulting in a triplet-to-singlet transition energy rang-
ing from 1 to 4 eV. This was found by MD, Hartree-Fock, or
configuration interaction calculations.1
1,12,14,20
Our data in-
deed confirm the possibility of intracenter radiative decays of
NOV sites and give an experimental basis for a quantitative
refinement of further calculations.
FIG. 5. Schematic of transitions induced by VUV excitation
involving singlet ͑S͒ and triplet ͑T͒ states of NOV sites. Radiative
͑
nonradiative͒ transitions are indicated by solid ͑dashed͒ arrows
C. Effects of the temperature
with the respective rate constants ͓Eqs. ͑4͒–͑6͔͒.
As discussed in the previous section, the thermal behavior
of IPL reflects the temperature dependence of the competitive
S →S transition, evidenced by different theoretical
0
1
nonradiative decay of S to the ground state, which in turn is
1
11,12,20
studies,
should be reduced in molecularly induced
related to the vibrational properties of the excited state. To
investigate this aspect, the thermal behavior of the PL inten-
sity ͑inset of Fig. 3͒ may be analyzed starting from the ki-
netics of excitation and decay of the NOV site:
NOV. A significantly higher potential barrier is thus expected
for configuration changes, in agreement with the lack of
NOV-to-ODC͑II͒ photoconversion in our samples. Prelimi-
nary MD calculations36 confirm the possibility of NOV con-
figurations with a barrier higher than 2 eV.
d
n = k n − ͑k + kISC͒ns,
͑4͒
͑5͒
s
0
0
s
dt
B. Scheme of radiative transitions
d
The results in Figs. 2 and 3, together with lifetime ␶ val-
ues, give an indication of the decay paths of this NOV vari-
ety. The lifetime values indicate a small transition probability
nt = kISCn − k n ,
s
t t
dt
where n , n , and n are the populations of NOV in the S , S ,
0
s
t
0
1
from the emitting levels ͑T states͒ towards the ground state,
1
and T states, respectively, while k , k , and k are the total
1
s
t
ISC
consistent with partially forbidden triplet-to-singlet T →S
1
0
rate constants ͑accounting for both radiative and nonradiative
transitions. In contrast, no fast S →S radiative decay, ex-
1
0
transitions͒ of S and T decays towards the ground state and
1
1
pected in the nanosecond domain, is observed, indicating
that strongly competitive decay channels are active from the
of the IST transition from S to T levels, respectively, as
1
1
indicated in Fig. 5. In this scheme, in the stationary state, the
intensity of PL emission is proportional to the population nt
of the radiative T levels, and dependence of n upon the rate
7
.1 eV excited S level. One of these is directed towards the
1
T levels responsible for the 2.9, 3.7 eV PL emissions ͑Fig.
1
1
t
5͒. However, another competitive process is active, consist-
constants can be obtained:
ing of an efficient nonradiative decay to the ground state,
favored by the temperature, and competitive with the inter-
system crossing ͑ISC͒ process S →T as well. In fact, the
kISC k0
nt ϰ
n .
0
͑6͒
1
1
ks + kISC kt
PL intensity from T →S turns out to be strongly dependent
1
0
As discussed in the previous section, the lifetime mea-
surements and the decrease of PL at high temperature sug-
gest that the thermal dependence of k and k is negligible.
on the temperature without a comparable effect on the PL
lifetime. The PL intensity decreases by more than a factor of
10 heating from 10 to 300 K, but no detectable change of ␶
t
ISC
Therefore, in Eq. ͑6͒, the temperature dependence of n , de-
is observed, ruling out that nonradiative decays of T states
t
1
termining the thermal behavior of PL intensity, arises from
could have a relevant role in the change of emission inten-
ks. The explicit dependence on the temperature can be evi-
denced through the expression of the temperature depen-
sity. In addition, the ISC S →T process does not appear
1
1
significantly phonon assisted, unlike the ODC͑II͒ defect con-
figuration, since no evidence is observed for a positive con-
tribution to the PL intensity by increasing the temperature.
dence of the rate k dominated by the nonradiative decay of
s
the S level to the ground state. The temperature dependence
1
of the rate constant k may be approximately described by an
Emission from T levels by exciting at 6 eV might reflect the
s
1
3
7
expression derived by Mott’s activation-energy law
ks
radiative decay of the forbidden S →T excitation. The very
0
1
ϰexp͑−E /kT͒ with a temperature dependent activation en-
low intensity did not allow us a detailed spectral analysis and
no reliable value of PL lifetime could be estimated by excit-
ing at 6 eV. In any case the attribution to PL excited by the
S →T transition is supported by the thermal behavior of the
a
ergy E =B−͗n͘ where B is the potential barrier for the non-
a
T
radiative process from the minimum energy of the excited
state and ͗n͘ is Planck’s thermal average phonon number
T
0
1
−
1
PL intensity, which does not show relevant changes in the
investigated temperature range, as expected on the basis of
the lack of temperature dependence of ␶ by exciting at
͓exp͑h␯/kT͒−1͔ to take into account the thermal depen-
dence of the phonon contribution to the excited state energy
͑with effective phonon energy h␯͒ in a configurational-
coordinate approach.38 As a consequence, the temperature
dependence of PL intensity I , proportional to n given by
7
.1 eV.
In this scheme, the PL emissions around 2.5–4 eV give
an indication of the energy position of the triplet state with
PL
t
Eq. ͑6͒, takes the form
75101-5
0

PALEARI et al.
PHYSICAL REVIEW B 71, 075101 ͑2005͒
1
sites, specifically the Franck-Condon offset ⌬d between the
configurational coordinates of the two states in their relaxed
configurations. In fact a=2␭ ͑with a value of about 16͒ is the
dimensionless measure of the offset between the minima of
S and S in units of zero-point-vibration amplitude x . Since
the relevant phonon energy resulting from our analysis
matches the energy of the stretching mode of the Si-Si bond,
we may suppose that the relevant configurational-coordinate
is approximately related to the stretching of the Si-Si bond,
IPL
ϰ
.
͑7͒
1
+ A exp͓− ͑B − ͗n͘ ͒/kT͔
T
The fit of data in Fig. 3͑b͒ using Eq. ͑7͒ gives an energy
barrier B for the nonradiative process of 0.04 eV and an
effective energy h␯ of the phonon modes of about 410 cm .
1
0
v
−
1
−
1
This value matches well the energy of about 430 cm of the
stretching mode of the Si-Si bond.39
D. Electron-phonon interactions
while the vibrational amplitude xv depends on the reduced
2
mass ␮ of the Si-Si group through the expression x
v
The values of phonon energy and activation barrier ob-
2
=
ប /h␯␮. From x and a, the excitation-related relaxation of
v
tained from the kinetic model of I ͑T͒ can furnish an ap-
PL
NOV site results in an offset ⌬d
of about 0.12 nm,
SiuSi
proximate evaluation of the electron-phonon interactions re-
sponsible for the peculiar nonradiative processes at NOV
defects. The lack of data on the energy of the radiative decay
prevents a quantitative analysis in terms of a detailed
configurational-coordinate model. In any case, the B value of
the barrier for nonradiative decays may be viewed as the
energy from the minimum of the excited level to the cross-
which is a large distortion compared with the equilibrium
Si-Si bond length of 0.24–0.26 nm.39 Our approximate
evaluation of ⌬d
calculations: MD calculations
is indeed consistent with theoretical
SiuSi
2
0
found a relaxation of
0
.07 nm consequent to excitation to the triplet NOV state,
suggesting even larger relaxation of the excited S state. In
1
fact, cluster calculations11 of the excited S state found that
3
8
1
over of the potential curves of the two states. Taking the
Si atoms move 0.36 nm apart after NOV excitation, resulting
−
1
phonon energy h␯ in the range 410–430 cm and the exci-
tation energy Eexc at 7.1 eV, the energy B for the crossover
between the levels may be related to an effective Huang-
in ⌬d
Ϸ0.11 nm, in good agreement with our analysis.
SiuSi
2
V. CONCLUSIONS
Rhys factor ␭ that gives a measure of the relaxation energy
of the excited site in phonon energy units h␯:
The controversial question of the energy structure and in-
tracenter transition scheme of NOV in silica, complicated
until now by the proven occurrence of concomitant photoac-
tivated intercenter processes, has been experimentally disen-
tangled by forcing molecular Si–Si precursors in a network,
preventing the ODC͑I͒→ODC͑II͒ photoconversion. Evi-
dence for the intracenter energy structure of the NOV con-
figuration has been obtained, giving an experimental basis
for all previous calculations of the S →S and T →S NOV
2
2
2
1/2 2
Eexc − ␭ h␯ + B = ␭ h␯͓1 + ͑B/␭ h␯͒ ͔ ,
͑8͒
where the left side describes the excited state potential at the
crossover, while the right side is the crossover energy along
the ground-state potential curve. Equation ͑8͒ gives ␭ Ϸ60,
2
reflecting a large electron-phonon interaction. In this case,
broad bands are expected from a large site relaxation upon
excitation. The bandwidth ͑FWHM͒ is given by ⌬E
0
1
1
0
=
2.35h␯␭͑1+␤͒/͑1−␤͒, where ␤=exp͑−h␯/kT͒ is the Bolt-
transitions, as well as of the environment sensitivity of the
2
zmann factor. From the estimated values ␭ and h␯, we ob-
tain ⌬EϷ0.9 eV in the low-temperature limit ͑␤=0͒, with an
expected thermal broadening of about 0.1 eV at 300 K. This
is close to the experimental bandwidth of 0.8 eV and the
small thermal broadening of 0.07 eV barely observed just
above the experimental uncertainty ͑Fig. 4͒. The agreement,
even though within the crude approximation of the model,
suggests a relevant relaxation of the excited configuration. If
the nonradiative decay were not so efficient as the small B
ODC͑I͒-to-ODC͑II͒ interconversion mechanism. A promi-
nent role of the site relaxation upon excitation to the S state
1
is strongly supported by the analysis of our data. The huge
excitation-induced relaxation represents the main peculiarity
of NOV sites, probably the crucial feature for several aspects
involving defect photoactivated processes in silicon dioxide.
ACKNOWLEDGMENTS
value shows, a PL emission with a relevant Stokes shift
The authors thank Simone Agnello for fruitful comments
and discussion. They also thank Marco Bernasconi and
Marco Cannas for making available theoretical and experi-
mental unpublished results, respectively. This work is part of
a National Project partially supported by the Italian Govern-
ment. Experiment at HASYLAB was supported by the IHP
Contract No. HPRI-CT-1999-00040/2001-00140 of the Euro-
pean Commission.
2
2
␭ h␯Ϸ6 eV would be expected. Experiments to detect pos-
sible infrared emissions at lower temperature by exciting at
.1 eV might clarify this aspect, but the competitive decay of
the S state through T levels is expected to prevent obser-
7
1
1
vation of the S →S radiative decay. In any case, another
1
0
important parameter can be obtained from the semiquantita-
tive evaluation of the electron-phonon interaction of NOV
075101-6

RADIATIVE DECAY OF VACUUM-ULTRAVIOLET…
Electronic address: alberto.paleari@mater.unimib.it
PHYSICAL REVIEW B 71, 075101 ͑2005͒
*
͑1995͒.
1
20
21
J. Albert, MRS Bull. 23, 36 ͑1998͒.
M. J. Dejneka, MRS Bull. 23, 57 ͑1998͒.
W. Margulis, F. C. Garcia, E. N. Hering, L. C. Guedes Valente, B.
D. Donadio, M. Bernasconi, and M. Boero, Phys. Rev. Lett. 87,
2
195504 ͑2001͒.
3
S. Agnello, R. Boscaino, M. Cannas, F. M. Gelardi, M. Leone,
and B. Boizot, Phys. Rev. B 67, 033202 ͑2003͒.
C. J. Brinker and G. W. Scherer, Sol-Gel Science ͑Academic,
New York, 1989͒.
Lesche, F. Laurell, and I. C. S. Carvalho, MRS Bull. 23, 31
2
2
2
3
͑
1998͒.
4
L. Skuja, in Defects in SiO and Related Dielectrics: Science and
2
Technology, edited by G. Pacchioni, L. Skuja, and D. L.
Griscom ͑Kluwer Academic, Dordrecht, 2000͒, p. 73.
D. L. Griscom, J. Ceram. Soc. Jpn. 99, 899 ͑1991͒.
K. M. Davis and M. Tomozawa, J. Non-Cryst. Solids 201, 177
͑1996͒.
5
6
24
25
J. C. Mikkelsen, Appl. Phys. Lett. 39, 903 ͑1981͒.
T. Bakos, S. N. Rashkeev, and S. T. Pantelides, Phys. Rev. B 69,
195206 ͑2004͒.
G. Ghosh, Handbook of Thermo-Optic Coefficient of Optical Ma-
terials with Applications ͑Academic, New York, 1998͒.
A. E. Geissberger and F. L. Galeener, Phys. Rev. B 28, 3266
͑1983͒.
N. Chiodini, F. Meinardi, F. Morazzoni, A. Paleari, R. Scotti, and
G. Spinolo, Solid State Commun. 109, 145 ͑1999͒.
V. Gottardi, M. Guglielmi, A. Bertoluzza, C. Fagnano, and M. A.
Morelli, J. Non-Cryst. Solids 63, 71 ͑1984͒.
A. N. Trukhin, in Defects in SiO and Related Dielectrics: Sci-
2
ence and Technology, edited by G. Pacchioni, L. Skuja, and D.
L. Griscom ͑Kluwer Academic, Dordrecht, 2000͒, p. 235.
H. J. Fitting, T. Barfels, A. N. Trukhin, and B. Schmidt, J. Non-
Cryst. Solids 279, 51 ͑2000͒.
R. Tohmon, H. Mizuno, Y. Ohki, K. Sasagane, K. Nagasawa, and
Y. Hama, Phys. Rev. B 39, 1337 ͑1989͒.
2
2
2
2
3
6
7
8
9
0
7
8
9
0
H. Nishikawa, E. Watanabe, D. Ito, and Y. Ohki, Phys. Rev. Lett.
72, 2101 ͑1994͒.
1
H. Imai, K. Arai, H. Imagawa, H. Hosono, and Y. Abe, Phys. Rev.
B 38, 12 772 ͑1988͒.
M. A. Garcia, S. E. Paje, M. A. Villegas, and J. Llopis, Mater.
Lett. 43, 23 ͑2000͒.
11
G. Pacchioni and G. Ieranó, Phys. Rev. Lett. 79, 753 ͑1997͒.
V. Sulimov, S. Casassa, C. Pisani, J. Garapon, and B. Poumellec,
Modell. Simul. Mater. Sci. Eng. 8, 763 ͑2000͒.
M. Busso, S. Casassa, C. Pisani, and V. Sulimov, Modell. Simul.
Mater. Sci. Eng. 10, 21 ͑2002͒.
V. B. Sulimov, P. V. Sushko, A. H. Edwards, A. L. Shluger, and
A. M. Stoneham, Phys. Rev. B 66, 024108 ͑2002͒.
L. Skuja, J. Non-Cryst. Solids 239, 16 ͑1998͒.
L. Skuja, J. Non-Cryst. Solids 149, 77 ͑1992͒.
L. Skuja, A. N. Streletsky, and A. B. Pakovich, Solid State
1
1
1
2
3
4
31
J. Lin and K. Baerner, Mater. Lett. 46, 86 ͑2000͒.
A. Anedda, C. M. Carbonaro, F. Clemente, R. Corpino, F. Raga,
and A. Serpi, J. Non-Cryst. Solids 322, 95 ͑2003͒.
N. Chiodini, F. Meinardi, F. Morazzoni, A. Paleari, R. Scotti, and
D. Di Martino, Appl. Phys. Lett. 76, 3209 ͑2000͒.
F. Meinardi and A. Paleari, Phys. Rev. B 58, 3511 ͑1998͒.
G. Pacchioni and G. Ieranó, J. Non-Cryst. Solids 216, 1 ͑1997͒.
D. Donadio, Master’s thesis, University of Milano-Bicocca, 2000.
N. F. Mott, Proc. R. Soc. London, Ser. A 167, 384 ͑1938͒.
C. W. Struck and W. H. Fonger, J. Lumin. 10, 1 ͑1975͒.
O. A. Zhikol, A. F. Oshkalo, O. V. Shishkin, and O. V. Prezhdo,
Chem. Phys. 288, 159 ͑2003͒.
3
3
3
2
3
4
1
1
1
5
6
7
35
36
37
3
8
Commun. 50, 1069 ͑1984͒.
B. L. Zhang and K. Raghavachari, Phys. Rev. B 51, 7946 ͑1995͒.
V. B. Sulimov and V. O. Sokolov, J. Non-Cryst. Solids 191, 260
1
1
8
9
39
075101-7