Ion exchange doping of solar cell coverglass for sunlight down-shifting

E. Cattaruzza a,n, M. Mardegan a, T. Pregnolato a, G. Ungaretti a, G. Aquilanti b, A. Quaranta c,G. Battaglin a, E. Trave a

a Molecular Sciences and Nanosystems Department, Università Ca’ Foscari Venezia, via Torino 155/b, 30172 Venezia-Mestre, Italyb Elettra-Sincrotrone Trieste S.C.p.A., s.s. 14 km 163.5, 34149 Basovizza, Trieste, Italyc Department of Industrial Engineering, University of Trento, via Mesiano 77, 38123 Trento, Italy

a r t i c l e i n f o

Article history:Received 11 February 2014Received in revised form14 July 2014Accepted 15 July 2014

Keywords:Ion exchangeGlass cover solar cellDown-shifting properties

a b s t r a c t

In this work, silicate glass slides were doped with silver or copper by ion exchange in molten salts inorder to obtain luminescent sheets suitable as down-shifters to be used for covering and protecting solarcells. The samples were treated at different temperatures, in nitrate, sulfate and chloride salts, fordifferent times. Post-exchange thermal treatments were performed on some of them in order to tailorthe ion aggregation states and their luminescence features. Optical absorption and near ultraviolet–visible luminescence analyses were performed in order to test the suitability of the samples as down-shifters and, moreover, X-ray absorption near edge structure (XANES) analyses were performed oncopper doped samples in order to better identify the oxidation states of the incoming ion as a function ofthe synthesis parameters. The performance of the best samples was tested by measuring the outputpower of a GaAs solar cell covered with the treated glass slides and illuminated with a solarsimulator lamp.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

Around the world, an increasing interdisciplinary effort isgrown to lower the material costs of solar cells, to increase theirenergy conversion efficiency, and to create innovative and efficientnew products for application based on photovoltaic (PV) technol-ogy. While the classical efficiency limit of a solar cell is currentlyestimated to be around 29%, detailed balance calculations showthat this could improved to approximately 37% for a silicon solarcell by using a reliable spectrum modification of sunlight radiationby luminescence down-shifting (LDS), up-conversion (UC), down-conversion (DC) and/or their combination [1]. In order to exploit asmuch as possible the solar spectrum with standard or highefficiency solar cells, a lot of work has been done on materialsfor the conversion of part of the solar spectrum into a rangesuitable for the solar cells [2–4]. The spectral conversion usuallyoccurs by means of luminescent layers deposited above (forDC and LDS) or below (for UC) existing solar cells made with wellknown high-efficiency manufacturing approaches. With the exist-ing techniques, neither modification of the active core nor buildingof complicated structure are needed, thus no considerable manu-facturing costs are added.

Down-shifting systems can be also obtained by doping thecover glass of the solar cells with suitable luminescent transitionmetal ions [5,6]. This approach is still quite unexplored. In the pastCe3þ ions were dispersed in cover glasses of GaAs solar devices forspatial applications, but for the protection of both the glass andthe encapsulation material from the UV radiation [4]. Recently,LSD glass plates have been obtained by dispersing Cuþ ions orluminescent Ag aggregates in the glass during the melting fabrica-tion process.

In this work we investigate for the first time the capability ofthe ion exchange process for the production of luminescent coverglasses with down-shifting properties suitable for solar cells. Theion exchange permits to modify the glass surface with a concen-tration of doping ions higher than the solubility limits of thefabrication processes from melts. Moreover, ion exchange can beused as a treatment procedure for commercial cover plates,avoiding the fabrication of special glasses. Several papers weredevoted to the investigation of this technique, since the firsttreatment for the surface toughening of silicate glasses [7] and,subsequently, in the application field of the planar optical wave-guides [8,9 and refs. therein]. Briefly, the process is realized byimmersing the glass substrate in a molten salt bath containing thedesired metallic ions: as a consequence of the chemical potentialgradient occurring between the bath and the glass surface, themetallic ions in the bath exchange with the alkali ion species ofthe glass. This process gives rise to a diffusion profile of the new

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/solmat

Solar Energy Materials & Solar Cells

http://dx.doi.org/10.1016/j.solmat.2014.07.0280927-0248/& 2014 Elsevier B.V. All rights reserved.

n Corresponding author. Tel.: þ39 041 2346716; fax: þ39 041 2346713.E-mail address: cattaruz@unive.it (E. Cattaruzza).

Solar Energy Materials & Solar Cells 130 (2014) 272–280

ion species, whose shape and depth depend on the diffusioncoefficients of the two exchanging ions, on the bath temperatureand on the salt composition. For these reasons, ion exchange givesa good control of the surface properties of the glass, allowing tooptimize the procedure for the production of good LSD systems.

In this work, we studied the feasibility of this approach byusing molten salt baths containing two species usually employedfor the production of optical waveguides in silicate glasses, e.g.Agþ and Cuþ . These ions are characterized by broad luminescencebands in the UV–blue–green region of the spectrum, whoseintensity and position strongly depends on the ion concentrationand glass composition [10–12]. Since the excitation spectra ofthese features are located in the UV part of the spectrum, it can beargued that these doping ions can act as luminescent down-converters for solar cells whose responsivity curves are peakedin the visible range.

In this paper we present the first results on the down-converting capabilities of float glass slides doped with Agþ andCuþ by ion exchange at different temperatures in different bathsand, moreover, post-exchange thermal treatments were carriedout on the samples. The optical absorption and luminescenceproperties were extensively studied and the first tests on PV cellsare presented.

2. Materials and methods

Ion exchange was performed on samples cut from a commer-cial low-iron extra-clear float soda-lime silicate glass (PilkingtonOptiwhite, 2.85 mm thick), commonly used as photovoltaic (PV)panel cover glass. The weight % composition declared by themanufacturer is 72 SiO2, 13.4 Na2O, 8.9 CaO, 4.2 MgO, 0.6 Al2O3,0.4 K2O, 0.2 ZrO2, 0.02 TiO2, 0.013 Fe2O3 and other elements intraces (calculated atomic % composition: 60.2 O, 24.9 Si, 9.0 Na,3.3 Ca, 2.2 Mg, 0.2 Al, 0.2 K, 0.03 Zr, 0.005 Ti, 0.003 Feþtraces). Thenominal composition is compatible with that obtained by Ruther-ford backscattering spectrometry (RBS) analysis (61 O, 25 Si, 9 Na,3 Ca, 2 Mg; other elements in traces).

Before the ion exchange, the glass slides were prepared by afour-step cleaning process in ultrasonic baths (deionized H2O;trichloroethylene; acetone; isopropyl alcohol). Agþ–Naþ ionexchange was realized by immersing the glass in a molten saltbath of AgNO3:NaNO3, with 1% molar concentration of silvernitrate. The treatment was performed at 320 1C, for 60 minfollowing two different procedures: some samples were comple-tely immersed in the bath (“dipped” samples) while some otherswere suspended floating on the bath surface (“floating” samples),with the tin-free side in contact with the salt. After the ionexchange, a heat treatment in air was performed on some Ag-doped samples at 350 1C or 400 1C for 1 h or 16 h.

The ion exchange procedure for the “floating” samples waschosen due to the characteristic surface contamination of com-mercial float glasses with tin. In fact, glass slides are manufacturedby floating molten glass on a bed of molten tin. This method givesvery flat surfaces as well as a uniform thickness [13]. However, onthe glass side put into contact with the molten tin, a detectableamount of this metal is observed usually for a thickness of somemicrons below the surface and up to few at% of concentration,depending on the glass composition and on the heating para-meters [14–17]. About the valence, in the first few microns mainlySn2þ is present; at larger depths, a large fraction of Sn4þ can befound [14–17]. This side effect is negligible for most applications,but it is quite important in the case of the glass doping with metalions, for the possibility to influence the oxidation state of theincoming ion during the ion exchange process. In fact, Sn2þ can bea reducing agent for exchanged metal ions, like Agþ and Cuþ [18],

giving rise to the unwanted precipitation of metal particles whichlower the transparency of the glass slide. Actually, for all theinvestigated glasses, RBS measurements showed that one of thetwo sample side is always characterized by the presence of few at%of tin in the whole probed surface layer (namely, up to 2 μm) witha profile as a function of the depth indicating that the metal canreach several microns of depth below the glass surface. In theother glass side only a very small signal was detected, localized onthe surface, due to the surface contamination induced by theprocess on float glass [14].

Copper ion exchange was done by using two different saltbaths. The first was a CuSO4:NaSO4 eutectic with a CuSO4 molarconcentration of 46%, where the samples were completelyimmersed for 20 min at the temperature of 550 1C. The secondbath was CuCl:ZnCl2, with a CuCl molar concentration of 11%. Inthis bath the ion exchange was performed at 350 or 400 1C, fordifferent times, in the range between 20 min and 16 h, by fullimmersion of the samples. No systematic heat treatments wereperformed on copper doped samples.

About the flatness of the glass before and after the processing,we measured with a profilometer a vertical deviations around71 nm in the pure glass, around 74 nm in all the ion-exchangedsamples, around 710 nm in all the ion-exchanged samples sub-jected to the thermal annealing.

As reference glass, we used a pure glass slide heated in air for1 h at 350 1C, which is a temperature similar to that of the ionexchange process we used to dope the glass. This reference sampleexhibits the same optical properties of the untreated glass and ofundoped glass slides treated at different temperatures and forlonger times.

UV–vis optical absorption (OA) spectra were acquired with aJASCO UV–VIS dual beam spectrophotometer, in the 250–800 nmregion with spectral resolution of 2 nm.

Photoluminescence (PL) spectra were recorded with a Fluorolog-3(Horiba-Jobin Yvon) modular system. The excitation of thesamples was obtained using a 450W Xe lamp coupled to a doublemonochromator for wavelength selection (λexc�1¼260 nm, λexc�2¼340 nm). The samples were placed with the tin-free side facing thelamp. Photoluminescence excitation (PLE) spectra were alsorecorded: the emission wavelengths were chosen according to theposition of emission bands collected in PL spectra. Photolumines-cence quantum yield (PLQY) was measured for two representativesamples by using an integrating sphere coupled to Horiba ScientificFluoroLog spectrofluorimeter.

RBS measurements were performed at INFN National Labora-tories of Legnaro, Italy (LNL) with a 2.0 MeV Heþ beam at ascattering angle of 1601. The diffused atoms concentration wascalculated by simulating the RBS spectra with the RUMP code [19].

X-ray absorption spectroscopy (XAS) measurements were per-formed at Elettra Synchrotron, in Basovizza, Trieste (Italy), at theXAFS beamline [20]. The storage ring operated at 2.0 GeV in top-up mode with a current of 300 mA. The Cu K-edge spectra wererecorded in fluorescence mode at RT, using a large area Si driftdiode detector. Samples of metallic copper, Cu, Cu2O and CuO weremeasured in transmission mode. The energies were defined byassigning the first inflection point of the spectra of the metalliccopper to 8979.0 eV. The monochromator was equipped with acouple of Si(1 1 1) crystals, and the harmonic rejection wasachieved by detuning the second crystal of the monochromatorby 30% of the maximum. A metallic copper reference sample wasused for energy calibration in each scan.

The capabilities of the treated glass slides as PV cover glasseswere measured by illuminating the samples with a 300 W Xenonsolar simulator equipped with filters to approximate the 1.5 airmass solar spectrum (Lot-Oriel). The samples were put withoutmatching fluids on a GaAs (CESI, active area 7�5 mm2) and the

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280 273

power-against-voltage P–V curves of the cell were obtained bymeasuring the flowing current I and the cell voltage V (P¼V� I)with a multimeter and by selecting different shunt resistanceswith a rheostat. For each sample, the curve was measured severaltimes after removing and putting again the glass slide on the cell:this procedure was repeated a number of times during thedifferent measurement runs, to take into account also the effectof the possible fluctuation of the solar simulation power. The finaluncertainty was found to be around 0.5% of the measured powervalues.

3. Results and discussion

3.1. Ag-doped samples

3.1.1. Optical absorptionSilver containing glasses can show different kind of optical

absorption features related to the oxidation state of silver [21–24]:the band of Agþ ions is located at around 300 nm, while around350 nm is peaked the band due to silver dimers (Ag2)þ or similaraggregates. A completely different feature, called surface plasmonresonance (SPR) band, is a well-known characteristic that ascer-tain the presence inside a glass of Ag metallic clusters in the nmsize range and it is due to resonant collective oscillations of freeelectrons in metal nanoparticles. This band has been extensivelystudied in many papers [see for instance 25,26] and for silicateglasses it is located around 400–420 nm.

The surface silver concentration in the tin-free side wasachieved by RBS analysis (spectra not reported). The ion exchangein a bath with 1 mol% of AgNO3 for 1 h allowed to reach a silverdoping up to 50% of the surface sodium concentration with apenetration depth of around 10 μm. Actually, in these samples thesilver concentration is enough high to guarantee the presence ofAgþ�Agþ pairs and to allow a controlled aggregation processwith mild thermal treatments [27].

Fig. 1 shows the UV–vis optical absorption spectra for some ofthe as-exchanged and annealed samples. Considering the as-exchanged samples (dipped and floating), they show a red-shiftof the UV absorption edge, if compared to the reference glass,together with a slower decrease of absorbance in the near UVregion (350–380 nm). For floating samples, at wavelengths higherthan 400 nm, the transmittance is very similar to the standardglass slide. It is well known that in Agþ ion exchanged silicateglasses a red shift of the UV edge occurs owing to the opticaltransitions of the 4d10-4d95s related levels, while dimers orsimple aggregates give rise to optical transitions in the near UVrange [22,23,28]. The edge shift is higher in dipped samples withrespect to floating ones, due both to the contribution of twosurfaces to the total absorbance and to the presence of smallaggregates, like (Ag3)2þ , grown after red-ox reactions betweenSn2þ and Agþ ions and characterized by absorption features in thenear UV region. Moreover, in dipped samples a higher absorbancearound 400–500 nm can be observed, related to the formation ofsub-nanometer metal aggregates which can be detrimental bothto the transmission of the visible sun light and to the lumines-cence suitable for the down conversion.

A more remarkable difference between dipped and floatedsamples is pointed out after the thermal annealing treatments.The spectra of the floating samples annealed after the doping(Fig. 1a) are characterized by the absence of the SPR band, asexpected for temperatures lower than 500 1C [25,28]. Only thesample annealed at the highest temperature (400 1C) shows therising of a broad faint feature located around 400–450 nm whichcan be attributed to the SPR of very small aggregates. On the otherhand, in the case of the dipped samples a well-defined SPR band at

430 nm is present even after annealing at 350 1C for 1 h (Fig. 1b)owing to the reduction of ionic silver by Sn2þ .

A simulation of the optical absorption spectrum by means ofthe Mie model, applied to a composite glass system made by silvermetal nanospheres embedded in soda-lime glass, gets informationabout the mean size of the silver nanoparticles [25,26]. The Mieapproach considers for metal particles the free electron Drude-likedielectric function that is related to the particle dimension throughthe dependence of the electron relaxation time on the particleradius. For the bulk silver dielectric function we used the expres-sion proposed by Quinten [29]. From the fitting of the experi-mental SPR bands shown in Fig. 1b we obtained a mean particlediameter of 1.6 nm, 1.8 nm and 2.0 nm for the samples 350 1C–1 h,350 1C–16 h, 400 1C–1 h, respectively. The uncertainty on thediameters was evaluated 0.1 nm from the fitting accuracy. Insummary, for the use as down-shifting systems we must take intoaccount that a glass transparency lowering usually occurs as aconsequence of the silver doping, mainly due to the formation ofmetallic nanoaggregates.

Taking into account of the optical absorption of the dippedsamples, in both as-exchanged and thermally treated samples, thefollowing analyses and tests were performed only on floatingsamples.

3.1.2. PhotoluminescenceSilver in glass matrices is usually characterized by three different

luminescence bands. The first is located at about 300–350 nm and itis attributed to the electronic transition 4d10-4d95s1 of isolatedAgþ ions in the glass network [10,11]. The second characteristicemission band, usually detected for high Agþ concentrations, iscentered between 400 and 500 nm and it is due to Agþ–Agþ pairs[10,11,25,28]. The third emission band is detected around 600 nmor more and it has been ascribed to small silver aggregates in theglass matrix. The structure of these aggregates is still not well

0.05

0.1

0.15

0.2

0.25

0.3

Ref. glassF (as-exc)F (350°C-1h)F (350°C-16h)F (400°C-1h)

Ag-doped glass

0

0.5

1

1.5

2

2.5

3

3.5

300 400 500 600 700 800

Ref. glassD (as-exc)D (350°C-1h)D (350°C-16h)D (400°C-1h)

Wavelength (nm)

Opt

ical

den

sity

Fig. 1. Optical absorption spectra of Ag-doped samples (1% AgNO3 bath;Ti.e.¼320 1C; ti.e.¼1 h). The annealing parameters are also reported. Different scalesare used for floating (a) and dipped (b) samples to better evidence some spectrafeatures.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280274

understood: they could be silver trimers, like (Ag3)2þ [25,28] and/ormixed species of Agþ and Ag0 such as (Ag2)þ , (Ag3)þ , (Ag3)2þ , andso on [30,31]. At present few data are reported in literatureconcerning the photoluminescence quantum yield of silver dopedglasses. Kuznetsov and co-workers [6] claim a yield of 20% at theexcitation of 366 nm and even higher at lower wavelengths.However, the same authors found a PLQY as low as 2% at 340 nm[32]. Of course, the result strongly depend on the network andaggregation states of silver; moreover, in ion exchanged glasses ithas to be taken into account that only few microns are involved inthe doping, so the absolute luminescence intensity is very low,giving rise to high relative errors.

The emission bands at the higher wavelengths are excited inthe UV–blue range and they can be exploited for increasing theyield of a silicon solar cell to the solar spectrum. Starting from theAg-doped glass, in which mainly Agþ ions are present, the basicidea is to tune the silver aggregation in order to maximize theemission in the 400–600 nm wavelength range. It is well-knownthat the aggregation degree of silver in doped glasses can becontrolled by thermal annealing [25 and Refs. therein]. Therefore,the treatment conditions have to be carefully chosen in order bothto allow the formation of dimers or trimers emitting at highwavelengths [6] and to avoid the formation of metal nanoparticlescharacterized by absorption bands which are detrimental to theglass transparency and which re-absorb the luminescencefeatures.

Photoluminescence spectra in Fig. 2a, recorded by usingλexc¼260 nm, show the typical emission band located around430 nm due to Agþ–Agþ pairs [27]. The band, observed also inas-exchanged samples, increases its intensity after thermal treat-ments. The same analyses, performed on the tin-rich side of thedipped samples, did not show this feature, since the reduction ofAgþ and the aggregation of metal particles leave a very lowamount of silver pairs in the glass network.

For the annealing at 350 1C, a treatment for longer times doesnot induce a higher intensity of the “silver pairs” band: this effectis not clear for the moment, but it can be related to the counter-acting effect of the silver diffusion at higher depths during thetreatment, living a lower silver concentration at the surface as thetime increases. At this regards, recently Simo et al. [33] proposed amechanism of formation of Ag small aggregates (below 1 nm ofsize, mainly Ag dimers) or Ag nanoparticles that depends on theannealing temperature, for which below a threshold temperature(around 400 1C) only the formation of Ag small aggregates takesplace, while the formation of metal nanoparticles is hindered.In Fig. 2b PLE spectra are reported, obtained by recording theluminescence intensity at the PL band maximum. It can beobserved that the emission around 400–500 nm is excited mainlyin the UV region, with a broad feature in the 300–425 nm rangewhich shifts at higher wavelengths after thermal annealing.

With λexc¼340 nm, annealed samples exhibit the PL bandaround 600 nm due to small silver charged aggregates, whoseintensity increases with the annealing temperature (Fig. 3a). Inparticular, at 350 1C the intensity do not depend on the annealingtime, as previously observed for the pair band, while it is higherfor the sample treated at 400 1C for 1 h. The PLE related spectra arereported in Fig. 3b, where a broad and intense feature locatedaround 350 nm can be observed. The reported PLE spectra outlinehow both the pair and the aggregate signals can convert thewavelength of the light in the 300–400 nm range which is quiteinteresting for the modification of part of the solar spectrum forcommercial photovoltaic panels.

It is well-known that absolute PLQY measurements are usuallyvery complicated and awkward, with several critical steps (see forinstance [34]): so, variations in the estimated values of 1 order ofmagnitude are often obtained simply by changing experimentalset-ups and protocols. Being aware of that, we estimated thephotoluminescence quantum yield of sample F (the as-exchangedfloating one): taking into account the very low absorbance of thissample as a consequence of the small amount of doping silver ion

0

1 107

2 107

3 107

4 107

5 107

6 107 Ref. glassF (as-exc)F (350°C-1h)F (350°C-16h)F (400°C-1h)

300 350 400 450 500 550 600

PL in

tens

ity (a

rb. u

nits

)

Ag-doped glass

(λexc=260 nm)

Wavelength (nm)

0

1 107

2 107

3 107

260 280 300 320 340 360 380 400 420

Ref. glass @ 520nmF (as-exc) @ 465nmF (350°C-1h) @ 450nmF (350°C-16h) @ 420nmF (400°C-1h) @ 420nm

Wavelength (nm)

PLE

inte

nsity

(arb

. uni

ts)

0

1 10

2 10

3 10

300 320 340 360 380 400Wavelength (nm)

PLE

inte

nsity

(a

rb. u

nits

)

Fig. 2. (a) PL spectra of Ag-doped “floating” samples recorded with λexc¼260 nm.The strong peak centered at λ¼2λexc is an instrumental artefact. (b) PLE spectra ofthe same samples. The monitored emission wavelength is reported in legend andthe inset better shows the 300–400 nm region. The annealing parameters areindicated in the labels.

0

2 106

4 106

6 106

8 106

1 107Ref. glassF (as-exc)F (350°C-1h)F (350°C-16h)F (400°C-1h)

400 450 500 550 600 650 700

PL in

tens

ity (a

rb. u

nits

)

Ag-doped glass

(λexc=340 nm)

Wavelength (nm)

0

5 106

1 107

1.5 107

2 107

300 350 400 450

Ref. glass @ 700nmF (as-exc) @ 600nmF (350°C-1h) @ 580nmF (350°C-16h) @ 600nmF (400°C-1h) @ 560nm

Wavelength (nm)

PLE

inte

nsity

(arb

. uni

ts)

Fig. 3. (a) PL spectra recorded with λexc¼340 nm of Ag-doped “floating” samples.(b) PLE spectra of the same samples. The monitored wavelength is reported inlegend. The annealing parameters are indicated in the labels.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280 275

diffused inside (see Fig. 1a), we chose 280 nm as the excitationwavelength and we selected the fluorescence emission regionbetween 370 and 540 nm. The small signals involved allowed todetermine only a realistic upper limit for the quantum yield: forthe sample F we obtained PLQY r0.5%.

3.2. Cu-doped samples

3.2.1. Optical absorptionCopper in doped glasses can be present in three different oxidation

states, namely Cuþ , Cu2þ and metallic nanoparticles. Cuþ ions arecharacterized by luminescent bands in the blue–green range whoseexcitation maximum is located around 280 nm, leaving glassestransparent in the visible range [12,35–37], whereas both Cu2þ ionsand metallic Cu nanoclusters dye the glass with broad absorptionbands in the visible range, around 800 nm and 560 nm, respectively[38,39].

From the point-of-view of photovoltaic applications, the suit-ability of Cu-doping of glass comes mainly from Cuþ ions, whichconvert the 250–320 nm range radiation into a visible fluorescencein the 400–650 nm range [5]. However, the presence of Cu2þ givesrise to a strong and broad absorption centered around 800 nm thatworsen the both the overall PL efficiency of Cu-doped glass andthe optical transparency of the glass in the near infrared region. Asa consequence, the treatment conditions have to be properlychosen in order both to increase the Cuþ/Cu2þ ratio inside thedoped glass and to avoid the precipitation of Cu metallic (nano)particles originating the SPR absorption band around 550 nm [26].

Samples dipped in CuSO4:NaSO4 salt bath were transparent butshowing a red coloration, due to the SPR band of copper metalnanoparticles which lowers the glass transparency in the visibleregion with a detrimental effect for solar applications. On theother hand, samples exchanged in the “floating” way to avoid theSn influence on the red-ox equilibrium are characterized by ayellow–orange coloration. In this case the absence of the reducingagent avoids the precipitation of copper; however, the formationof Cu2þ ions colors the glass preventing its use as LSD. As reportedin literature, in Cu–Na ion exchanged glasses the control of thecopper oxidation state inside the matrix is a difficult task, as wellas the understanding of the copper diffusion process [40]. Never-theless, lots of experimental data are available on the treatmentconditions and salt baths. For instance, as proposed by the Praguegroup [41,42], ion exchange in CuCl:ZnCl2 salt bath should favorthe synthesis of a doped glass containing mainly monovalentcopper ions. For this reason we performed ion exchange treat-ments on several samples in the chloride mixture.

After immersion in CuCl:ZnCl2 salt bath the Cu atomic fractionat the sample surface, measured by RBS, is around 1% and the Cu-doped region at the exchange temperature of 350 1C extends atdepths higher than 2 μm (the typical sampling depth of the RBS)just after 20 min of ion exchange. In order to increase the totalamount of Cu we increased the process temperature: actually, inthe sample exchanged for 4 h at 400 1C the Cu depth profileobtained by simulation of the RBS spectrum is quite flat within theprobed region, suggesting that the whole exchanged layer shouldbe of some tens of microns.

The RBS analysis, performed on different points of the surface,evidenced for some samples a surface accumulation of either Cu orZn. Scanning electron microscopy analysis clarified this aspect,showing the formation of large islands of Zn (most probably zincchloride, as suggested by the energy-dispersion spectrometryanalysis detecting Cl) on the glass surface, some hundreds of μmwide and more and more present by increasing the exchange timeor temperature. These islands could neither be removed by rinsingthe samples with distilled water nor with a friction with opticpaper. This phenomenon occurs by means of a reaction between

the salt and the glass surface and it lowers the surface opticalquality of long-time exchanged glasses, which at the naked-eyeexhibit opaque shadows on the surface. Moreover, the formationof salt islands lowers the contact surface between glass and salt,thus partially hindering the penetration of Cu ions into the glass.

Fig. 4 shows the UV–vis optical absorption spectra for somerepresentative Cu-doped samples obtained by ion exchange inCuCl:ZnCl2: as can be observed, neither SPR band (λE560 nm) norCu2þ band (λE800 nm) are present and the glass absorption edgehides the typical monovalent ion band (λE280 nm). In spite of thepresence of tin in our samples, the copper diffused inside the glassseems to be mostly present in þ1 oxidation state. The higherabsorbance of samples treated for the longest times (8 and 16 h)and at the highest temperature (400 1C) is due to the lightscattering on zinc compound islands grown on the sample surface.

3.2.2. PhotoluminescenceIn literature several examples of luminescent Cu-doped glasses

are described. Studies by Debnath et al. [35,43] report that a nearpure porous silica glass doped with monovalent copper ions byimpregnation shows two types of emitting centers, characterizedby PL bands at 490 nm and 430 nm, related to Cuþ embedded intoa distorted octahedral site and a cubic site, respectively. Borsellaet al. [12] report a green emission band in ion exchanged soda limeand borosilicate glasses, attributed to the presence of Cuþ ionslocated in distorted octahedral sites. It is also reported that PLintensity decreases with the Cuþ/Cu2þ ratio. This effect can beattributed either to an electronic excitation energy transfer fromCuþ to Cu2þ ions, or to an UV absorption due to a charge transferfrom Cu2þ to O2– ions that interferes with Cuþ excitation. Inphosphate glasses, Boutinaud et al. [44] report the occurrence ofblue and red fluorescence Cuþ features of increasing intensity asthe copper concentration increases, while Cu-doped BK7 glassshows a blue–green emission band regardless of the preparationparameters [45]. Ion exchange performed in a special Na2O-richsilicate glass by CuCl-containing salt baths resulted in a strongblue–green Cuþ luminescence, while the presence of Cu2þ ionswas inferred by a lowering of the luminescence yield [41]. Thequantum yield of Cuþ ions is generally higher than Agþ , andFujimoto and Nakatsuka [46 and Refs. therein] obtained a value ofquantum yield around 80% at 270 nm of excitation in Cuþ dopedsol–gel glasses. In those samples, the maximum of PL band wasaround 550 nm and a strong absorption was observed between200–400 nm in all the samples, attributed to monovalent copperion, together with the absorption band of divalent copper ion at800–900 nm.

In PL spectra reported in Fig. 5a, recorded with λexc¼260 nm,an emission band peaked around 500 nm can be observed, out-lining the down-conversion capability of Cuþ ions. In PLE spectra

0.04

0.08

0.12

0.16

0.2

Ref. glass350°C-20m350°C-3h350°C-4h350°C-8h350°C-16h400°C-4h

300 400 500 600 700 800

Opt

ical

den

sity

Cu-doped glass

Wavelength (nm)

Fig. 4. Optical absorption spectra of Cu-doped samples (CuCl:ZnCl2 salt bath). Theexchange parameters are reported in the label.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280276

(Fig. 5b), the typical Cuþ excitation band around 250–310 nm[12,45] is detected. For exchange times higher than few hours, thePL intensity decreases probably due to the formation of bivalentCu ions quenching the Cuþ luminescence [12]. By lowering theexcitation photon energy (λexc¼340 nm) we obtained the PLspectra shown in Fig. 6. Here we observe an interesting featureof the sample exchanged at 400 1C: an emission band centeredaround 600 nm. This band can be related to the presence of Cuþ

dimers. Copper tends to form dimers (Cuþ)2 in several soliddielectric hosts: their presence has been reported both fromtheoretical calculation and from measurements of Cu–Cu distancein different networks [47,48]. These dimers are characterized by abroad emission band centered around 600 nm, whatever the glasshost [45,49]. So, it can be concluded that in ion exchanged glasseshigh process temperatures promote the formation of copper

dimers. The same broad emission band was observed for a sampleexchanged at 350 1C and subjected to a post-exchange thermalannealing at 500 1C for 1 h, while no fluorescence features at600 nm were observed in samples heated at temperatures lowerthan 500 1C.

We estimated the photoluminescence quantum yield of thesample 350 1C–3 h: we chose an excitation wavelength of 280 nmfor a selected fluorescence emission region between 390 and540 nm. As for the Ag-doped sample F, the small signals involvedallowed to determine only an upper limit for the quantum yieldestimation: for the sample 350 1C–3 h we obtained PLQY r0.4%.The results obtained for λexc¼300 nm and λexc¼340 nm were inagreement with this estimation.

3.2.3. XAS and XANES analysesIn order to shed light on the copper oxidation states, we

performed XAS analyses on some Cu-doped glasses exchangedfor long times. Previous XAS studies of soda-lime glasses dopedwith a CuI:CuCl salt bath and for both process and annealing timesquite long (few days) [50] showed that Cu ions in the glassnetwork are mostly in the monovalent form but with a variablecoordination number, in same cases similar to Cu2þ . XAS analysesperformed with a grazing incidence approach showed that insoda-lime glasses doped with a CuSO4:NaSO4 salt bath (for processtimes of few minutes), the Cuþ/Cu2þ ratio is not constant in theexchanged region, changing with the depth from 30/70 at thesurface to 85/15 in the second half of the Cu-doped region [40,51].In Fig. 7 we plot the X-ray absorption near-edge structure (XANES)spectra of four samples (Ti.e.¼350 1C for 4, 8, and 16 h durationtime; Ti.e.¼400 1C for 4 h duration time), recorded at the CuK-edge, together with the spectra of Cu, Cu2O, and CuO standardsamples.

In samples exchanged at 350 1C, a clear modification of the Cu K-edge is detected by increasing the exchange time: the Cuþ char-acteristic strong pre-peak, centered around 8982 eV for the Cu2Ostandard and clearly visible in the doped sample after 4 h of ionexchange, decreases its intensity for 8 h of process duration and isalmost absent in the 16 h sample. The onset of the absorption for the4 h and 8 h samples is more similar to Cu2O rather than to CuOstandard, but the opposite stands for the 16 h sample. Actually, bytheoretical calculations of the Cu K-edge XANES spectra for Cuþ insilicate glasses and assuming that copper is coordinated with two

0

5 106

1 107

1.5 107

2 107

350°C-20m350°C-3h350°C-4h350°C-8h350°C-16h400°C-4h

360 400 440 480 520 560 600

PL in

tens

ity (a

rb. u

nits

)

Wavelength (nm)

Cu-doped glass(λexc=260 nm)

0

2 106

4 106

6 106

8 106

1 107350°C-20m @ 500nm350°C-3h @ 520nm350°C-4h @ 510nm350°C-8h @ 510nm350°C-16h @ 514nm400°C-4h @ 520nm400°C-4h @ 600nm

260 280 300 320 340

PLE

inte

nsity

(arb

. uni

ts)

Wavelength (nm)

Fig. 5. Luminescence spectra of samples doped by ion exchange in the CuCl:ZnCl2bath. The exchange parameters are reported in the labels. (a) PL spectra recordedwith λexc¼260 nm. The peak centered at λ¼2λexc is an instrumental artefact.(b) PLE spectra of the same samples. The emission wavelength is reported inthe label.

(λexc=340 nm)Cu-doped glass

2 105

4 105

6 105

350°C-20m350°C-3h350°C-4h350°C-8h350°C-16h400°C-4h350°C-20m*

400 450 500 550 600 650 700

PL in

tens

ity (a

rb. u

nits

)

Wavelength (nm)

Fig. 6. Photoluminescence spectra of Cu-doped samples (CuCl:ZnCl2 bath)recorded with λexc¼340 nm. The exchange parameters are reported. For compar-ison, the spectrum of a sample subjected to a post-exchange annealing in air atT¼500 1C for 1 h is also shown (indicated by *).

8970 8980 8990 9000 9010 90200

1

2

3

4

400°C-4h

Cu

350°C-8h

350°C-4h

Cu2O

350°C-16h

Abs

orpt

ion

coef

ficie

nt (a

rb. u

nits

)

Energy (eV)

CuO

Fig. 7. Cu K-edge normalized XANES spectra. Continuous lines: Cu-doped glasssamples obtained by CuCl:ZnCl2 salt bath (the process conditions are reported).Dashed lines: standard spectra recorded from a metallic Cu plate, a Cu2O powder, aCuO powder.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280 277

oxygen atoms in a linear configuration, Maurizio et al. [52] showedthat the intensity of the pre-peak can be influenced by the O�Cu�Oangle. The value is maximum when the angle is 1801 and decreasesgoing to 1201. In addition, at 1201 the white line exhibits a markedshift towards lower energies. Taking into account all these remarks,the significant change of the Cu K-edge observed by increasing thedoping time for Ti.e.¼350 1C is an evidence of the evolution, drivenby the process duration, of either the copper chemical state or thegeometric site structure around Cuþ . We are inclined towards achange of oxidation state of part of Cuþ ions. In fact, since thepresence of Cu2þ quenches the Cuþ PL intensity at 500 nm [12 andRefs therein], the lowering of the luminescence spectra shown inFig. 5a can be ascribed to the presence of the former species. Theabsence of the typical absorption band of Cu2þ ions around 800 nm(Fig. 4) could be explained by considering that the total amount ofcopper inside our 350 1C samples is lower than 1�1017 atoms/cm2,as obtained by RBS data. On the other hand, the sample exchanged at400 1C for 4 h exhibits a spectrum very similar to that of the 350 1C–4 h sample, indicating a preponderance of Cuþ .

Following the approach described by Gonella et al. [40], wesimulated the XANES spectra of the Cu-doped samples as a linearcombination of the standard spectra for Cu0, Cuþ and Cu2þ , inorder to obtain an estimation of the fraction of the different Cuoxidation state. In Table 1 we report both the different oxidationstate fractions evaluated from the simulation and the relativeamount of copper of the different samples evaluated from the CuK-edge position and amplitude. We observe that the higher theprocess time and temperature, the higher the copper doping level.Moreover, by increasing the process time the Cu2þ amountincreases at the expense of the Cuþ , while an opposite trendoccurs by increasing only the exchange temperature. Actually,during ion exchange the host glass structure does not accommo-date Cu ions as simple glass modifiers (as in a glass melt): thestructural and chemical rearrangements in the Cu-doped glassstructure can be strongly influenced by the process time andtemperature as a consequence of the formation of defects [53]acting as oxidizing agents by trapping electrons and changing thecopper oxidation state from Cuþ to Cu2þ .

3.3. PV cell tests

The performance test of the doped silicate glasses as PV cellcover was done by placing the slides on a GaAs small solar cell andby measuring the output power under the illumination withthe solar simulator. Fig. 8 shows the GaAs cell output power inthe maximum region, plotted as a function of the voltage, with theuntreated glass, a silver containing slide and two copper dopedsamples treated at 350 1C for different times (20 min and 3 h). Forthe cell without cover we measured Vopen circuit¼0.91 V, Ishortcircuit¼9.83 mA.

The Ag-doped sample giving the best results was the floatingas-exchanged, while all the dipped samples (because of the tinoxo-reduction action) and the annealed ones (because of theinduced silver precipitation) gave quite poor results with respect

to the untreated glass owing to a too marked optical absorption inthe visible region. Nonetheless, the floated slide gives a measuredpower slightly lower with respect to the untreated glass. As amatter of fact, the floating as-exchanged sample is characterizedby the lower luminescence yield in the visible region underexcitation at 340 nm and by the enhanced reflectance due to therefractive index increase at the glass surface. In fact, it is wellknown that in Agþ ion exchanged glass slides the refractive indexcan increase of Δn¼0.01–0.1, depending on the treatment condi-tions [8].

As far as Cu-doped samples are concerned, the two best-performing ones were the 350 1C–3 h and 350 1C–20 min (CuCl:ZnCl2 salt bath), as shown in Fig. 8. In particular with the sampletreated for 3 h we observed an increase of the maximum outputpower of 2% with respect to the reference glass. This result is quitesatisfying if we take into account that even in copper ion-exchanged samples a reflection enhancement can occur owing toa refractive index increase of around 0.006 [54]. Cu-doped glassesexchanged for times longer than 3 h give rise to curves slightlybetter than the reference one but quite similar within the experi-mental errors, as it happens for the 350 1C–20 min sample. Thisshould be due to the detrimental effect of the presence of Cu2þ

inside the exchanged region. In fact, in spite of the increasing totalamount of Cuþ (see Table 1), the contemporary increase of theCu2þ/Cuþ fraction is observed. The Cu containing glass treated at400 1C for 4 h gave a power curve similar to the 350 1C–20 minsample, even if it is characterized by a large amount of Cuþ , a lowCu2þ/Cuþ fraction, and an intense PL signal around 500 nm. Thelow performance can be due to poor optical quality of the slidesurface, due to the presence of salt islands, evidenced by thehigher absorbance shown in Fig. 4.

In the light of the experimental findings, for the Agþ dopingthe next step should be an investigation addressed to the max-imization of the whole PL emission in the visible range, taking careto avoid the precipitation of silver into nanoparticles (the mainresponsible of the optical absorption increase). Starting from theAg cluster formation model depicted in [33], the first investigationshould be to determine the exact threshold temperature TS for theformation of metal particles in the Ag-doped cover glass, since justbelow TS it should be possible – by means of suitable thermaltreatments – to maximize the PL light yield in the visible rangewithout lowering the glass optical transmittance.

As far as the copper doping is concerned, the nodal point seemsto be the control of the Cu2þ content in the exchanged glass: to dothat, it is necessary both to choose suitable molten salt baths andto avoid long exchange times. The use of a CuCl:ZnCl2 bath is

Table 1Cu oxidation state fractions evaluated from the K-edge XANES simulation results. Inthe last column the amount of copper normalized to the first sample is reported, asevaluated by the experimental Cu K-edge feature.

Cu0 Cuþ Cu2þ Cu total amount(relative)

350 1C–4 h o0.2 0.6�0.8 0.1�0.3 1350 1C–8 h o0.2 0.5�0.7 0.2�0.4 7350 1C–16 h o0.1 0.2�0.4 0.6�0.8 12400 1C–4 h o0.2 40.8 o0.1 30

GaAs celloutput power

6.2

6.3

6.4

6.5

6.6

6.7

660 680 700 720 740 760

Cu-doped glass(350°C-3h)

Cu-doped glass(350°C-20min)

Ag-doped glass(floating)

Referencepure glass

Pow

er (m

W)

Voltage (mV)

Fig. 8. Zoom on the maximum GaAs cell output power region of the power-voltagecurve. The samples used as cell cover glass are reported. The Cu-doped sampleswere prepared with CuCl:ZnCl2 salt bath.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280278

actually promising: even if in some cases we observed a detri-mental effect on the glass surface quality, work is in progress tocheck at the same time both temperature and duration of theprocess, to find the right parameters aiming to improve the finalcell efficiency. The use of CuI-based molten salt baths can be alsoinvestigated.

As a final remark, it is well-known that the use of ion exchangeto dope the cell cover soda-lime silicate glass increases the glassstrength by putting the outer surfaces into compression [55]: thiscould allow to replace the otherwise needed thermal temperingprocess by the intrinsic chemical toughening.

4. Conclusions

In this work, we studied how glass slides used for covering andprotecting solar cells can acquire down-shifting properties bymeans of ion exchange doping with luminescent metal ions. Inparticular, we inserted Ag and Cu ions into commercial silicateglass slides. Some silver doped samples were also thermallyannealed in order to control the aggregation state of Agþ andthe luminescence features of the samples. The down-shiftingsuitability of the prepared samples was tested on a GaAs PV cellilluminated with a solar simulator.

Due to the presence of tin on one sample side, Agþ ionexchange was performed only on the surface free from tin bymeans of the floating method, since in dipped samples tin gaverise to the formation of silver in metal nanoparticles with a stronglowering of the surface transparency. Silver doped samples exhib-ited broad luminescence bands peaked at 430 nm, ascribed toAgþ�Agþ pairs, whose intensity increase with the annealingtemperature. Moreover, a broad band peaked around 600 nmappears after post-exchange thermal treatments owing to theformation of (Ag3)2þ multimers. All these bands are excited inthe UV and near-UV region. Nonetheless, in spite of their evidentdown-shifting potential, in as-exchanged glass slides the surfacereflectance counteracts the luminescence yield, lowering the out-put power of the solar cell. Samples which underwent thermaltreatments after the ion exchange procedure gave the poorerresults, even if they were characterized by broad luminescencebands in the visible range, owing to the absorption effects of silvermultimers and aggregates. Further improvements can beattempted by lowering the AgNO3 molar ratio in the salt bath inorder to attain a lower concentration at the glass surface. By thisway, after thermal treatments below the critical temperature forthe formation of metal clusters, a better control of the glassluminescence can be achieved avoiding high concentrations ofmultimers, which are detrimental to the glass transparency.

Samples doped with Cuþ ions gave results better than Agþ ,demonstrating to be suitable for improving the efficiency of a solarcell, with an increasing of the 2% of the maximum output powerwith respect to the PV cell covered by the untreated glass, even ifthe evaluated PLQY of the tested samples was quite low (r 0.4%).In fact, copper ion exchanged samples are characterized by intenseand broad luminescence bands, centered at 500 nm, due to Cuþ

ions, excited by UV light and suitable for down-shifting applica-tions. The best samples were obtained by means of a CuCl:ZnCl2bath mixture, giving rise to samples more transparent with respectto slides treated in the CuSO4:NaSO4 eutectic, even if in some caseswe observed the formation of salt aggregates on the surface whichare detrimental to the glass transparency. The intensity of the Cuþ

band increases with the treatment time in samples exchanged at350 1C up to 3 h and then decreases due to the formation of Cu2þ

which gives rise to unwanted quenching effects. The formation ofCu2þ , even if at very low concentrations, is confirmed by XANESanalyses.

As a result, the main key role seems to be played by keeping theCu2þ concentration as low as possible in copper exchanged glass.In order to achieve this result, it is necessary to choose suitablemolten salt baths and to avoid long exchange times. Work is inprogress to check both temperature and duration of the process, tofind the right parameters aiming to improve the final cell effi-ciency. Another possibility could be to explore CuI-based moltensalt baths.

In the future, further tests will be done on different solar cellstoo, as well as by more realistic outdoor measurements. Actually,in the visible region the GaAs cell quantum efficiency is similar tothat of the crystalline-Si cell but different from that of theamorphous-Si cell: this last shows a decrease of the efficiencyfor wavelengths larger than 600 nm, in any case beyond the valuesof the photoluminescence bands evidenced in our samples.

Acknowledgements

The authors kindly thank Dr. N. Daldosso and Dr. V. Paterlini ofthe University of Verona (Italy) for the photoluminescence quan-tum yield measurements.

References

[1] C. Strümpel, M. McCann, G. Beaucarne, V. Arkhipov, A. Slaoui, V. Švrček, C. delCañizo, I. Tobias, Modifying the solar spectrum to enhance silicon solar cellefficiency – An overview of available materials, Sol. Energy Mater. Sol. Cells 91(2007) 238–249.

[2] B.S. Richards, Enhancing the performance of silicon solar cells via theapplication of passive luminescence conversion layers, Sol. Energy Mater.Sol. Cells 90 (2006) 2329–2337.

[3] X. Huang, S. Han, W. Huang, X. Liu, Enhancing solar cell efficiency: the searchfor luminescent materials as spectral converters, Chem. Soc. Rev. 42 (2013)173–201.

[4] E. Klampaftis, D. Ross, K.R. McIntosh, B.S. Richards, Enhancing the performanceof solar cells via luminescent down-shifting of the incident spectrum:a review, Sol. Energy Mater. Sol. Cells 93 (2009) 1182–1194.

[5] S. Gómez, I. Urra, R. Valiente, F. Rodríguez, Spectroscopic study of Cu2þ/Cuþ

doubly doped and highly transmitting glasses for solar spectral transforma-tion, Sol. Energy Mater. Sol. Cells 95 (2011) 2018–2022.

[6] A.S. Kuznetsov, V.K. Tikhomirov, V.V. Moshchalkov, UV-driven efficient whitelight generation by Ag nanoclusters dispersed in glass host, Mater. Lett. 92(2013) 4–6.

[7] D.K. Hale, Strengthening of Silicate Glasses by Ion Exchange, Nature 217 (1968)1115–1118.

[8] A. Tervonen, B.R. West, S. Honkanen, Ion-exchanged glass waveguide technol-ogy: a review, Opt. Eng. 50 (2011) 071107.

[9] A. Quaranta, E. Cattaruzza, F. Gonella, Modelling the ion exchange process inglass: phenomenological approaches and perspectives, Mater. Sci. Eng. B 149(2008) 133–139.

[10] E. Borsella, G. Battaglin, M.A. Garcìa, F. Gonella, P. Mazzoldi, R. Polloni,A. Quaranta, Structural incorporation of silver in soda-lime glass by the ion-exchange process: a photoluminescence spectroscopy study, Appl. Phys. A71 (2000) 125–132.

[11] E. Borsella, F. Gonella, P. Mazzoldi, A. Quaranta, G. Battaglin, R. Polloni,Spectroscopic investigation of silver in soda–lime glass, Chem. Phys. Lett.284 (1998) 429–434.

[12] E. Borsella, A. Dal Vecchio, M.A. Garcıa, C. Sada, F. Gonella, R. Polloni,A. Quaranta, L.J.G.W. van Wilderen, Copper doping of silicate glasses by theion-exchange technique: a photoluminescence spectroscopy study, J. Appl.Phys. 91 (2002) 90–98.

[13] L.A.B. Pilkington, Review lecture. The float glass process, Proc. R. Soc. Lond.A 314 (1969) 1–25.

[14] Y. Hayashi, K. Matsumoto, M. Kudo, The diffusion mechanism of tin into glassgoverned by redox reactions during the float process, J. Non-Cryst. Solids 282(2001) 188–196.

[15] G.H. Frischat, C. Müller-Fildebrandt, D. Moseler, G. Heide, On the origin of thetin hump in several float glasses, J. Non-Cryst. Solids 283 (2001) 246–249.

[16] G.H. Frischat, Tin ions in float glass cause anomalies, Comptes Rendus Chimie5 (2002) 759–763.

[17] Q. Zhang, Z.J. Chen, Z.X. Li, Simulation of tin penetration process in the surfacelayer of soda-lime-silica float glass, Sci. Chin. Technol. Sci. 54 (2011) 691–697.

[18] A. Quaranta, R. Ceccato, C. Menato, L. Pederiva, N. Capra, R. Dal Maschio,Formation of copper nanocrystals in alkali-lime silica glass by means ofdifferent reducing agents, J. Non Cryst. Solids 345–346 (2004) 671–675.

[19] M. Thompson, RUMP: Rutherford backscattering spectroscopy analysis pack-age, ⟨http://www.genplot.com⟩.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280 279

[20] A. Di Cicco, G. Aquilanti, M. Minicucci, E. Principi, N. Novello, A. Cognigni,L. Olivi, Novel XAFS capabilities at ELETTRA synchrotron light source, J. Phys.:Conf. Ser. 190 (2009) 012043.

[21] J. Qiu, M. Shirai, T. Nakaya, J. Si, X. Jiang, C. Zhu, K. Hirao, Space-selectiveprecipitation of metal nanoparticles inside glasses, Appl. Phys. Lett. 81 (2002)3040–3042.

[22] M.A. Villegas, J.M. Fernández Navarro, S.E. Paje, J. Llopis, Optical spectroscopyof a soda lime glass exchanged with silver, Phys. Chem. Glasses 37 (1996)248–253.

[23] S.E. Paje, M.A. García, J. Llopis, M.A. Villegas, Optical spectroscopy of silver ion-exchanged As-doped glass, J. Non-Cryst. Solids 318 (2003) 239–247.

[24] A. Nahal, K. Shapoori, Linear dichroism, produced by thermo-electric align-ment of silver nanoparticles on the surface of ion-exchanged glass, Appl. Surf.Sci. 255 (2009) 7946–7950.

[25] E. Cattaruzza, M. Mardegan, E. Trave, G. Battaglin, P. Calvelli, F. Enrichi,F. Gonella, Modifications in silver-doped silicate glasses induced by ns laserbeams, Appl. Surf. Sci. 257 (2011) 5434–5438.

[26] U. Kreibig, M. Vollmer, Optical Properties of Metal Clusters, Springer Series inMaterials Science, 25, Springer-Verlag, 1995.

[27] A. Quaranta, A. Rahman, G. Mariotto, C. Maurizio, E. Trave, F. Gonella,E. Cattaruzza, E. Ghibaudo, J.E. Broquin, Spectroscopic investigation of struc-tural rearrangements in silver ion-exchanged silicate glasses, J. Phys. Chem. C116 (2012) 3757–3764.

[28] E. Borsella, E. Cattaruzza, G. De Marchi, F. Gonella, G. Mattei, P. Mazzoldi,A. Quaranta, G. Battaglin, R. Polloni, Synthesis of silver clusters in silica-basedglasses for optoelectronics applications, J. Non-Cryst. Solids 245 (1999)122–128.

[29] M. Quinten, Optical constants of gold and silver clusters in the spectral rangebetween 1.5 eV and 4.5 eV, Z. Phys. B 101 (1996) 211–217.

[30] R.S. Varma, D.C. Kothari, R. Tewari, Nano-composite soda lime silicate glassprepared using silver ion exchange, J. Non-Cryst. Solids 355 (2009) 1246–1251.

[31] A.V. Podlipensky, V. Grebenev, G. Seifert, H. Graener, Ionization and photo-modification of Ag nanoparticles in soda-lime glass by 150 fs laser irradiation:a luminescence study, J. Lumin. 109 (2004) 135–142.

[32] A.S. Kuznetsov, J.J. Vélazquez, V.K. Tikhomirov, J. Mendez-Ramos,V.V. Moshchalkov, Quantum yield of luminescence of Ag nanoclusters dis-persed within transparent bulk glass vs. glass composition and temperature,Appl. Phys. Lett. 101 (2012) 251106.

[33] A. Simo, J. Polte, N. Pfander, U. Vainio, F. Emmerling, K. Rademann, Formationmechanism of silver nanoparticles stabilized in glassy matrices, J. Am. Chem.Soc. 134 (2012) 18824–18833.

[34] C. Würth, M. Grabolle, J. Pauli, M. Spieles, U. Resch-Genger, Relative andabsolute determination of fluorescence quantum yields of transparent sam-ples, Nat. Protoc. 8 (2013) 1535–1550.

[35] R. Debnath, S.K. Das, Site-dependent, luminescence of Cuþ ions in silica glass,Chem. Phys. Lett. 155 (1989) 52–58.

[36] D. Salazara, A. Villalobos, H. Porte, L.J. Villegas, H. Márquez, Propagation lossesof copper-ion-exchanged channel optical waveguides, J. Mod. Opt. 46 (1999)657–665.

[37] H. Márquez, J.M. Rincon, L.E. Celaya, Experimental study of CdCl2:CuClphotochromic coatings, Appl. Opt. 29 (1990) 3699–3703.

[38] B. Macalik, L. Krajczyk, J. Okal, T. Morawska-Kowal, K.D. Nierzewski,M. Suszyńska, Preparation and optical properties of soda-lime silicate glassespartially substituted by copper, Radiat. Eff. Def. Solids 157 (2002) 887–893.

[39] M. Cable, Z.D. Xiang, The optical-spectra of copper ions in alkali lime silicaglasses, Phys. Chem. Glasses 33 (1992) 154–160.

[40] F. Gonella, A. Quaranta, S. Padovani, C. Sada, F. D'Acapito, C. Maurizio,G. Battaglin, E. Cattaruzza, Copper diffusion in ion-exchanged soda–lime glass,Appl. Phys. A-Mater. 81 (2005) 1065–1071.

[41] P. Nebolova, J. Spirkova, V. Perina, I. Jirka, K. Mach, G. Kuncova, A study of thepreparation and properties of copper-containing optical planar glass wave-guides, Solid State Ion. 141–142 (2001) 609–615.

[42] J. Spirkova, P. Tresnakova, H. Malichova, M. Mika, Cuþ containing waveguidinglayers in novel silicate glass substrates, J. Mater. Sci.: Mater. Electron. 18 (2007)S375–S378.

[43] R. Debnath, S. Kumar, Dual luminescence of Cuþ in glass, J. Non-Cryst. Solids123 (1990) 271–274.

[44] P. Boutinaud, E. Duloisy, C. Pedrini, B. Moine, C. Parent, G. Le Flem, Fluores-cence properties of Cuþ ion in phosphate glasses of the BaLiPO4�P2O5

system, J. Solid State Chem. 94 (1991) 236–243.[45] Y. Ti, F. Qiu, Y. Cao, L. Jia, W. Qin, J. Zheng, G. Farrell, Photoluminescence of

copper ion exchange BK7 glass planar waveguides, J. Mater. Sci. 43 (2008)7073–7078.

[46] Y. Fujimoto, M. Nakatsuka, Spectroscopic properties and quantum yield ofCu-doped SiO2 glass, J. Lumin. 75 (1997) 213–219.

[47] P. Boutinaud, D. Garcia, C. Parent, M. Faucher, G. Le Flem, Energy levels of Cuþ

in the oxide insulators CuLaO2 and CuZr2(PO4)3, J. Phys. Chem. Solids 56 (1995)1147–1154.

[48] E. Fargin, I. Bussereau, G. Le Flem, R. Olazcuagal, C. Cartier, H. Dexpert, Firstevidence of CuI�CuI pairs in the Nasicon-type phosphate CuIZr2(PO4)3, Eur.J. Sol. State Inorg. 29 (1992) 975–980.

[49] T. Lutz, C. Estournès, J.C. Merle, J.L. Guille, Optical properties of copper-dopedsilica gels, J. Alloys Compd. 262 (1997) 438–442.

[50] J. Lee, T. Yano, S. Shibata, A. Nuhui, M. Yamane, EXAFS study on the localenvironment of Cuþ ions in glasses of the Cu2O–Na2O–Al2O3–SiO2 systemprepared by Cuþ/Naþ ion exchange, J. Non-Cryst. Solids 277 (2000) 155–161.

[51] F. Gonella, A. Quaranta, E. Cattaruzza, S. Padovani, C. Sada, F. D'Acapito,C. Maurizio, Cu-alkali ion exchange in glass: a model for the copper diffusionbased on XAFS experiments, Comp. Mater. Sci. 33 (2005) 31–36.

[52] C. Maurizio, F. d'Acapito, M. Benfatto, S. Mobilio, E. Cattaruzza, F. Gonella, Localcoordination geometry around Cuþ and Cu2þ ions in silicate glasses: an X-rayabsorption near edge structure investigation, Eur. Phys. J. B 14 (2000) 211–216.

[53] L. Skuja, Optically active oxygen-deficiency-related centers in amorphoussilicon dioxide, J. Non-Cryst. Solids 239 (1998) 16–48.

[54] F. Gonella, F. Caccavale, L.D. Bogomolova, F. D'Acapito, A. Quaranta, Experi-mental study of copper-alkali ion exchange in glass, J. Appl. Phys. 83 (1998)1200–1206.

[55] René Gy, Ion exchange for glass strengthening, Mater. Sci. Eng. B 149 (2008)159–165.

E. Cattaruzza et al. / Solar Energy Materials & Solar Cells 130 (2014) 272–280280