Effects induced by 4.7 eV UV laser irradiation on pure silica core multimode optical fibers investigated by in situ optical absorption measurements

Full text

Page 1

Effects induced by 4.7 eV UV laser irradiation on pure silica core multimode optical
fibers investigated by in situ optical absorption measurements
Fabrizio Messina⁎, Francesco Comandè, Marco Cannas
Dipartimento di Scienze Fisiche ed Astronomiche, Università di Palermo, Via Archirafi 36, I-90123 Palermo, Italy
a b s t r a c ta r t i c l ei n f o
Article history:
Received 16 June 2010
Received in revised form 19 January 2011
Accepted 24 January 2011
Available online 9 March 2011
Keywords:
E′ centers;
Laser irradiation;
Hydrogen diffusion
We investigated by in situ optical absorption measurements the effects induced by 4.7 eV UV laser irradiation
on pure silica core optical fibers. Laser irradiation with 100 MW cm−2laser intensity generates in the fiber E′
centers which partially decay after irradiation due to their reaction with diffusing H2. An absorption band
peaked at 5.3 eV is observed to grow in the post-irradiation stage with a kinetics anti-correlated to the decay
of the 5.8 eV band of the E′ centers. The defect absorbing at 5.3 eV is proposed to be formed by trapping on
pre-existing precursors of hydrogen atoms made available by breaking of H2on E′. We also show by repeated
irradiation experiments that the 5.3 eV-absorbing center is photochemically destroyed by 4.7 eV laser light,
and we estimate the cross section of this process. Possible structural models for this defect are discussed.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
Although several works have addressed laser-induced generation
of defects in bulk SiO2[1–7], only a few of them have directly dealt
with optical fibers, one of the most important technological applica-
tions of SiO2. Indeed, it has generally been assumed that the
understanding obtained from the study of bulk SiO2 or of fiber
preforms could be applied as well to optical fibers. Thus, only a few
data concerning the effects induced by laser light on the optical
absorption (OA) properties of optical fibers have been reported in
literature so far [8–10], especially in the ultraviolet (UV) range where
the absorption bands of the main SiO2 point defects are located
[11–13]. This is an important issue especially for fibers specifically
designed to be used for the transmission of UV signals, that is multi-
mode fibers with large core diameters fabricated from OH-rich SiO2
[12,14]. One of the main deterioration mechanisms of amorphous SiO2
upon exposure to high intensity laser radiation is the generation of
optically absorbing point defects. The color center which usually
dominates the optical absorption (OA) spectrum of irradiated SiO2in
the UV range is the well known E′γcenter, often indicated simply as E′
absorbing at 5.8 eV [11–13]. The main structural feature of this defect
is a Si atom bonded to three oxygen atoms and hosting an unpaired
electron (≡Si•). Aside from the most common E′γ, other variants of the
E′ center (E′α,E′β, and E′δ) have been identified based on their different
electron spin resonance (ESR) signals [11,15–17]. Their absorption
properties are still a matter of debate.
Studies on γ and X-irradiation of optical fibers have shown that a
high concentration of OH groups generally brings about an increased
resistance to irradiation [18,19]. Studies on the response of high-OH
fibers to UV light have identified E′ centers as the main generated
defects, and have suggested them to arise from photochemical rupture
of Si\H precursors [9,10]. Also, the concentration of induced E′ centers
was found to be significantly reduced if the fiber is loaded with high
concentrations of H2[9]; this is likely due to passivation of induced E′
centers due to their reaction with diffusing hydrogen, as observed in
bulk SiO2[6,7,20–25]. Comparison of these studies with literature on
bulk SiO2actually suggests quite a similar phenomenology triggered by
laser irradiation in the two physical systems. Still, it cannot be taken for
granted that bulk SiO2is equivalent to fiber-SiO2as concerns resistance
to laser irradiation. In fact, it has been often suggested that the fiber
drawing process potentially creates in the fiber defective sites which
may be precursors for the generation of defects upon irradiation or
introduces unresolved stresses which make the material more prone to
be damaged by laser light [11,12]. A useful experimental approach to
investigate theseissues is themeasurementin situ of radiation-induced
OA, previously applied to investigate laser-induced kinetics of point
defect generation and annealing in bulk SiO2[6,7,25]. Kinetic data are
oftenabletoyieldinformationondefectgenerationthatareunavailable
bymeasuring radiation-induced effects in stationary conditions only. In
the present paper we apply this approach to study the generation of
point defects in UV-transmitting multimode optical fibers under high
intensity UV laser irradiation.
2. Materials and methods
We performed experiments on standard commercial pure SiO2
core/F-doped SiO2 cladding step-index multimode optical fibers
Journal of Non-Crystalline Solids 357 (2011) 1985–1988
⁎ Corresponding author. Tel.: +39 0916234218; fax: +39 0916162461.
E-mail address: fmessina@fisica.unipa.it (F. Messina).
0022-3093/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2011.01.040
Contents lists available at ScienceDirect
Journal of Non-Crystalline Solids
journal homepage: www.elsevier.com/ locate/ jnoncrysol

Page 2

provided by Avantes, with 200 μm core and 20 μm cladding
diameters. These are fibers specifically designed for UV transmission,
whose synthetic silica core material nominally features ∼1000 ppm
in weight of Si\OH impurities, metallic impurities below 10 ppb,
and an unstructured tail-shaped absorption spectrum in the UV
range with absorption coefficients of the order of ∼1 dB/m [26].
After removing the polyamide coating, fiber samples were irradiated
at room temperature by 4.7 eV photons emitted by a pulsed
(pulsewidth τ=5 ns) Q-switched Nd:YAG laser system. The laser
was weakly focused so as to irradiate only a 3 mm-long portion of
the fiber. Laser beam hit the sample perpendicularly to the fiber axis.
Data reported herewith were obtained by exposing the fiber to laser
intensities between 10 MW cm−2and 100 MW cm−2and by using a
pulse repetition rate of 5 Hz. Intensity values have been calculated
without taking into account refractive effects taking place at the
entrance of the beam within the fiber. During and after the end of
each irradiation session, we measured in situ the induced OA varia-
tions (in the 200–400 nm range) of the fibers by an AVANTES AVS-
S2000 spectrophotometer, equipped by a D2 lamp source and a
Charge Coupled Device (CCD) detector, and allowing a time resolu-
tion of 1 s. To this purpose, the fiber being irradiated was inserted in
the all-in-fiber transmission line used to bring light from the lamp to
the CCD, thus allowing to measure in real time the kinetics of in-
duced OA from the variations of the signal revealed by the detector.
This approach also guarantees that the measured signals are asso-
ciated only to features induced in the fiber core. It turns out that also
light from the D2lamp is able to induce OA variations in the fiber;
nevertheless, we verified that these effects are always several orders
of magnitude smaller than those induced by laser irradiation within
our experimental conditions.
3. Results
We report in Fig. 1 a typical kinetics of the absorption spectrum as
measuredinsituinanopticalfibersampleduringandaftertheendofan
irradiation session consisting in ∼104UV laser pulses of 100 MW cm−2
laser intensity. Absorption coefficients have been calculated based on
the 3 mm length of the irradiated portion of the fiber. The main
absorption signal generated by irradiation is the band peaked at 5.8 eV
characteristic of E′ centers [11–13], the peak absorption coefficient
α(5.8 eV) at the end of irradiation being (1.60±0.02)cm−1, while
minor contributions to the absorption spectrum for Eb5 eV cannot be
ruled out. The concentration of E′ centers, as deduced from α(5.8 eV), is
plotted against time in the inset of Fig. 1, revealing that laser-induced
growthof thedefects has saturated to a value of (2.6±0.1)×1016cm−3
at the end of irradiation. After the laser is switched off, the induced
5.8 eVbandundergoesspontaneousdecayonatimescaleofafew103s,
as evidenced by data acquired at several delays after the end of
irradiation(Fig.1-(b)).Ontheother hand,a progressiveincrease of αin
the post-irradiation stage is clearly observed in the spectral region
between 4.5 eV and 5.3 eV. Least-square fitting of a representative
differenceabsorptionspectrumbyalinearcombinationoftwoGaussian
bands(insetofFig.1-(b))revealsthattheobservedmodificationsofthe
absorption spectrum can be described as the growth of a band peaked
at (5.3±0.1)eV with (0.68±0.08)eV full width at half maximum
(FWHM)concurrenttothedecreaseofthemainpeakat(5.86±0.02)eV
with(0.71±0.02)eVFWHM.Satisfactoryfitswiththeseparameterscan
be obtained on the difference spectra calculated at all delay times. It is
worth noting that the apparent redshift in time of the peak of the main
band is actually an effect of the overlap with the growing contribution
at 5.30 eV.
In Fig. 2 the post-irradiation kinetics of α(5.8 eV)(t) detected on
the peak of the absorption band of the E′ centers is reported. We also
show the kinetics of α(4.9 eV)(t) as a measure of the amplitude of the
5.3 eV band growing in the post-irradiation stage evidenced by Fig. 1-
(b). The spectral position of 4.9 eV (as opposed to the peak position)
was chosen in order to avoid interference from the larger 5.8 eV band.
Comparison between the decay kinetics of α(5.8 eV)(t) and the
growth of α(4.9 eV)(t) clearly suggests an anticorrelation between
the behaviors of the two contributions. The correlation plot in the
(a)
(b)
Fig. 1. Absorption spectrum measured in pure silica core optical fibers after selected
numbers of pulses during (a) and at selected times after the end (b) of the Nd:YAG laser
irradiation session. Labels in panel (b) represent delays from the end of the irradiation
session, while the arrows highlight the modifications observed in the spectrum with
growing time. Inset of panel (a): Kinetics of the E′ center concentration as deduced from
the time dependence of the 5.8 eV absorption coefficient. Inset of panel (b): difference
between the spectrum measured at t=50 s after the end of irradiation and that
measured at t=0. The full red line is the result of least-square fitting of the difference
spectrum with a linear combination of two Gaussian bands (see text for further details).
Fig. 2. Absorption coefficientα(5.8 eV)(t) (fullsymbols) and α(4.9 eV)(t)(opensymbols)
as a function of time t. The origin of the time scale corresponds to the end of the irradia-
tion session. Inset: α(5.8 eV)(0)–α(5.8 eV)(t) plotted against α(4.9 eV)(t)–α(4.9 eV)(0).
The full line (r=0.97) was obtained by linear fitting data corresponding to tb2000 s.
1986
F. Messina et al. / Journal of Non-Crystalline Solids 357 (2011) 1985–1988

Page 3

inset confirms this result, while evidencing possible small deviations
at long times from anticorrelated time dependencies.
Finally, we performed a set of experiments aimed at investigating
theeffect of repeated irradiations: asamplewhich hadbeen irradiated
with 104pulses of 100 MW cm−2intensity was irradiated again ∼1 h
after the end of the first irradiation session by using a laser intensity
I2=75 MW cm−2. By performing in situ OA measurements on the
fiber being re-irradiated it turns out that the second irradiation
bleaches the absorption at 4.9 eV grown during the post-irradiation
stage. As a matter of fact, we observe a progressive reduction (Fig. 3)
of α(4.9 eV) during the second irradiation, the bleaching being almost
complete after a few 103laser pulses. This effect is also accompanied
by a regrowth of the 5.8 eV band which will not be dealt here. The
efficiency of this bleaching process can be estimated by the initial
annealing rate Γ=α−1(dα/dN)N=0
α(4.9 eV), as estimated by a linear fit on the first ∼10 points in Fig. 3,
normalized for the absorption coefficient at 4.9 eV at the beginning of
the re-irradiation experiment. By repeating the re-irradiation experi-
ment as a function of I2we found out that the dependence of Γ on laser
intensity is approximately linear (inset of Fig. 3). Preliminary data
show, instead, that the intensity dependence of the induced 5.8 eV
band is quadratic, in agreement with the previous finding on fused
SiO2samples,[27] where this result was interpreted as an evidence of
the E′ centers being generated by a two-step process.
E=4.9eV, i.e. the initial decrease slope of
4. Discussion
The growth of the 5.8 eV band due to laser-induced E′ centers is
commonly observed as the main outcome of UV laser irradiation of
bulk SiO2for intensities in the MW cm−2range, as demonstrated by
several previous works [1–7,13]. The efficiency of defect generation
observed in the fiber (Fig. 1), 2.5×1016cm−3defects after 104pulses,
turns out to be one or two orders of magnitude higher than previously
observed in bulk high-purity synthetic silica for comparable laser
intensity levels [2,5,7]. On the other side, it is close to that typical of
fused SiO2materials, where it was shown that UV-induced damage
processes are assisted by two-step processes via inter-band electronic
levels provided by impurities [27]. Still, the native absorption
coefficient of the fiber material, 2×10−3cm−1at 200 nm, is much
lower than that of fused SiO2, where impurity-related absorption is of
the order of 1 cm−1in the UV spectral range [27]. On one side, a
precise quantitative comparison is hindered by the fact that we
cannot exclude an enhancement of the defect generation efficiency in
this experiment due to partially focusing of the laser beam inside the
material due to the cylindrical profile of the fiber. On a qualitative
basis, however, such results suggest that SiO2in the form of an optical
fiber, although being very pure from the chemical point of view, is
intrinsically much more sensitive to laser-induced generation of point
defects than bulk SiO2due to stresses accumulated in the material
during the fiber drawing process leading to the formation of high
concentration of precursor sites for the formation of defects, e.g.
oxygen vacancies or strained Si\O\Si bonds [11–13].
Post-irradiation decay of E′ centers on a time scale of a few hours
has been previously interpreted as a consequence of its chemical
reaction with diffusing H2[6,7,20–25]:
≡Si•+ H2⇒≡Si−H + H
ð1Þ
where H2required for reaction (1) to take place may be already
present in the native fiber or be itself a consequence of irradiation via
breaking of Si\OH or Si\H bonds. An important difference between
present results and previous findings both on fibers and on bulk SiO2
is our evidence of a 5.3 eV band growing in the post-irradiation stage
concurrently to E′ centers decay. Data in Fig. 2 show that the growth
kinetics of the 5.3 eV band is quite closely correlated to that by which
E′ centers are annealed by diffusing H2. This leads to assumption that
the center responsible for the 5.3 eV band is formed by fast trapping
on some pre-existing precursor of very mobile diffusing hydrogen
atoms made available by the main reaction (1): in fact, within this
scheme the two kinetics would be strongly anticorrelated being both
governed by rupture of H2on E′ defects acting as dissociation centers.
The small deviations from a strict correlation observed at long times
are likely due to a minor additional annealing process of E′ that is
not hydrogen-related. As a consequence of this reasoning, the defect
absorbing at 5.3 eV is expected to contain a hydrogen atom in its
chemical structure. In the following, the defect absorbing at 5.3 eV is
referred to as 5.3 eV-center.
Data in Fig. 3 yield information on the response of the 5.3 eV-
center to UV laser light. Indeed, re-irradiation of a sample in which the
5.3 eV-centers have already formed in the post-irradiation stage leads
to the rapid annealing of the absorption band associated to these
defects. This demonstrates that 5.3 eV-centers can be destroyed by
laser light at 4.7 eV photon energy. Also, the linear dependence of the
annealing rate on laser intensity (inset of Fig. 3) shows that annealing
is triggered by a single photon process. Being the laser photon energy
4.7 eV within the band at 5.3 eV, the photo-induced annealing of the
defect can be interpreted as arising from the fact that absorption is
due to a transition from the ground electronic state to an unstable
excited state from which the defect has a significant dissociation
probability. The slope Λ obtained by linear fitting data in the inset of
Fig. 3 is proportional to the photo-dissociation cross section of 5.3 eV-
centers, σd=Λτ−1hν∼1×10−19cm2. The ability of 4.7 eV photons to
destroy 5.3 eV-centers also explains why they grow only after the end
of irradiation, as their formation during the irradiation session is
contrasted by their concurrent annealing photo-induced by each laser
pulse. Further studies are needed to get more detailed information on
the nature of this photochemical annealing process. It is worth noting,
however, that σdis comparable with typical photo-dissociation cross
sections measured for other defects in SiO2, e.g. the H(II) center which
is formed by H trapping on twofold coordinated Ge impurities [28,29].
While the identification of the microscopic structure of the 5.3 eV-
center cannot be conclusively addressed here, tentative models can be
put forward based on comparison of present data with literature. H(I)
center (_Si•\H), consisting in a Si atom bonded to two oxygen atoms
and one hydrogen and hosting an unpaired electron, was shown to be
destroyed by absorption of UV light [24], although no satisfactory
reconstruction of the shape of its absorption band in this spectral
region is available in the literature. Additionally, H(I) is supposedly
formed by trapping on twofold coordinated Si precursors (_Si••) of H
atoms made available by rupture of H2on paramagnetic centers such
Fig. 3. Variations of the absorption coefficient at 4.9 eV observed during irradiation with
75 MW cm−2laser intensity of a sample which had been previously irradiated with 104
pulses of 100 MW cm−2intensity. Inset: normalized annealing rate Γ of the absorption
coefficient at 4.9 eV as a function of laser intensity,least-square fitted by a linear function.
1987
F. Messina et al. / Journal of Non-Crystalline Solids 357 (2011) 1985–1988

Page 4

as E′ [24,30], with this being consistent in principle with the scheme
proposed here. However, ESR measurements (not reported) on the
irradiated fiber samples reveal E′-type signals but do not show the
characteristic 7.4 mT doublet associated to H(I) [30].
Another point defect in SiO2proposed to have similar character-
istics to those discussed here for the 5.3 eV-center is E′β, a variant of
the E′ center distinguishable from the more common E′γcenter by its
electron spin resonance (ESR) properties [11,15,31]. E′βwas demon-
strated to grow in X-irradiated samples as a consequence of the
reaction of H atoms with some unknown precursor [11,15,31]. The
most commonly accepted formation process of E′βis trapping of H
atoms on pre-existing oxygen vacancies (≡Si−Si≡) [11,31]:
h
The proposed structure of E′βcomprises a silicon dangling bond
and a nearby ≡Si\H group. The latter is assumed to be sufficiently far
from the dangling bond that no hyperfine interaction between the
proton and the unpaired electron can be revealed by ESR. This model
proposed for E′βis essentially identical with that developed for the E′2
center in crystalline α-SiO2[11]; as E′2was associated to an absorption
band at 5.4 eV [15], one could expect an absorption band with similar
characteristics to be measured also for the hypothetically isostructural
E′βembedded in amorphous SiO2. Hence, these evidence are con-
sistent with the tentative identification of the 5.3 eV-centers with E′β
forming on pre-existing oxygen vacancies in the fiber material. The
absence of absorption data for the fiber in the vacuum-UV range
hinders us from verifying the presence of the 7.6 eV absorption band
associated to oxygen vacancies [11], which should be expected in the
system for process (2) to be meaningful.
On the other side, this attribution cannot be considered a definitive
one at the moment, since neither the structural model of E′βnor the
reaction responsible for its formation has been conclusively clarified.
Indeed, other possibilities aside from Eq. (2) have been proposed.
According to one of them [32], E′βmay consist in an isolated silicon
dangling bond structure (i.e. ≡Si•) formed by reaction of H atoms with
Si\H impurities, that is by the inverse of reaction (1). This would
requirereaction(1)tobeendothermicanditsinversetobeexothermic.
This remarkably contrasts with several experimental evidence (includ-
ing those of Fig. 1 in the present paper) in favor of reaction (1) taking
place only from left to right at temperatures N200 K [6,7,20–25],
although requiring an activation energy of ∼0.4 eV which provides the
main bottleneck limiting the reaction rate [25].
≡Si−Si≡ + H⇒≡Si−H…≡Si•E′β
i
ð2Þ
5. Conclusions
We investigated by in situ optical absorption measurements the
modifications induced in a commercial multimode optical fiber by
exposure to pulsed UV laser light. The major absorption band induced
by laser irradiation is that at 5.8 eV associated to E′ centers, which are
generated by laser exposure of the fiber and partially destroyed in the
post-irradiation stage due to their chemical reaction with diffusing
hydrogen. The generation efficiency of E′ centers in the fiber is much
higher than that typically observed in high purity bulk SiO2, likely due
to the abundance of defect precursors formed during the fiber
drawing process. Post-irradiation decay of E′ centers is accompanied
by the concurrent formation of another center absorbing at 5.3 eV
supposedly driven by diffusion of H atoms produced as a side effect of
the reaction between E′ and H2. This defect is tentatively proposed to
be E′β, formed by trapping on oxygen vacancies of H atoms produced
as a side effect of the reaction between E′ and H2. Independent of its
microscopic structure, the 5.3 eV-absorbing defect turns out to be
photosensitive that it can be photochemically destroyed by absorp-
tion of 4.7 eV photons with a ∼10−19cm2cross section.
Acknowledgements
We are grateful to the members of the LAMP group (http://www.
fisica.unipa.it/amorphous) for support and enlightening discussions.
References
[1] T.E. Tsai, D.L. Griscom, Phys. Rev. Lett. 67 (1991) 2517.
[2] K. Arai, H. Imai, H. Hosono, Y. Abe, H. Imagawa, Appl. Phys. Lett. 53 (1988) 1891.
[3] H. Nishikawa, R. Nakamura, Y. Ohki, Y. Hama, Phys. Rev. B 48 (1993) 15584.
[4] N. Fukata, Y. Yamamoto, K. Murakami, M. Hase, M. Kitajima, Appl. Phys. Lett. 83
(2003) 3495.
[5] K. Kajihara, Y. Ikuta, M. Hirano, H. Hosono, J. Non-Cryst. Solids 322 (2003) 73.
[6] F. Messina, M. Cannas, J. Phys. Condens. Matter 17 (2005) 3837.
[7] M. Cannas, F. Messina, J. Non-Cryst. Solids 351 (2005) 1780.
[8] R.K. Brimacombe, R.S. Taylor, K.E. Leopold, J. Appl. Phys. 66 (1989) 4035.
[9] P. Karlitschek, G. Hillrichs, K.F. Klein, Opt. Commun. 155 (1998) 376.
[10] K. Saito, M. Ito, A.J. Ikushima, S. Funahashi, K. Imamura, J. Non-Cryst. Solids 347
(2004) 289.
[11] G. Pacchioni, L. Skuja, D.L. Griscom (Eds.), Defects in SiO2and Related Dielectrics:
Science and Technology, Kluwer Academic Publishers, USA, ISBN: 0-7923-6685-9,
2000.
[12] H.S. Nalwa (Ed.), Silicon-based Materials and Devices, Academic Press, USA, ISBN:
0-12-513909-8, 2001.
[13] L. Skuja, H. Hosono, M. Hirano, Proc. SPIE 4347 (2001) 155.
[14] G.P. Agrawal, Fiber-Optic Communication Systems, third editionJohn Wiley &
Sons INC9780471215714, 2002.
[15] D.L. Griscom, J. Ceram. Soc. Jpn. 99 (1991) 899.
[16] G. Buscarino, S. Agnello, F.M. Gelardi, Phys. Rev. Lett. 94 (2005) 125501.
[17] G. Buscarino, S. Agnello, F.M. Gelardi, Phys. Rev. Lett. 97 (2006) 135502.
[18] D.L. Griscom, J. Non-Cryst. Solids 349 (2004) 139.
[19] S. Girard, IEEE Trans Nucl Sci 54 (2007) 4.
[20] Z. Li, S.J. Fonash, E.H. Poindexter, M. Harmatz, F. Rong, W.R. Buchwald, J. Non-
Cryst. Solids 126 (1990) 173.
[21] A.A. Bobyshev, V.A. Radtsig, Kinet. Katal. 31 (1990) 931.
[22] H. Imai, K. Arai, H. Hosono, Y. Abe, T. Arai, H. Imagawa, Phys. Rev. B 44 (1991)
4812.
[23] N. Kuzuu, Y. Komatzu, M. Murahara, Phys. Rev. B 44 (1991) 9265.
[24] V.A. Radtsig, V.N. Bagratashvili, S.I. Tsypina, P.V. Chernov, O. Rybaltovskii, J. Phys.
Chem. 99 (1995) 6640.
[25] F. Messina, M. Cannas, J. Phys. Chem. C 111 (2007) 6663.
[26] Avantes Catalog, http://www.avantes.com/2006–2007.
[27] F. Messina, M. Cannas, J. Phys. Condens. Matter 20 (2008) 275210.
[28] F. Messina, M. Cannas, R. Boscaino, J. Non-Cryst. Solids 351 (2005) 1770.
[29] M. Cannas, F. Messina, J. Non-Cryst. Solids 345 (2004) 433.
[30] L. Skuja, J. Non-Cryst. Solids 239 (1998) 16.
[31] D.L. Griscom, J. Non-Cryst. Solids 68 (1984) 301.
[32] D.L. Griscom, J. Appl. Phys. 58 (1985) 2524.
1988
F. Messina et al. / Journal of Non-Crystalline Solids 357 (2011) 1985–1988