Glass Frit Dissolution Influenced by Material Composition and the Water Content in Iodide/Triiodide Electrolyte of Dye-Sensitized Solar Cells

Katrine Flarup Jensen, Md. Mahbubur Rahman, Welmoed Veurman, Henning Brandt, Chan Im, Jürgen Wilde, Andreas Hinsch, and Jae-Joon Lee 1 Fraunhofer Institute for Solar Energy Systems (ISE), HeidenhofstraBe 2, 79110 Freiburg, Germany 2 KonkukUniversity-Fraunhofer Next Generation Solar Cell Research Center (KFnSC), 120Neungdong-ro, Gwangjin-gu, Seoul 143-701, Republic of Korea 3 Department of Advanced Technology Fusion, Konkuk University, 120 Neungdong-ro, Gwangjin-gu, Seoul 143-701, Republic of Korea 4Department of Microsystems Engineering-IMTEK, University of Freiburg, Georges-Köhler-Allee 103, 79110 Freiburg, Germany 5 Nanotechnology Research Center and Department of Applied Chemistry, Konkuk University, 268 Chungwondaero, Chungju 380-701, Republic of Korea


Introduction
The dye-sensitized solar cell (DSC) is a photoelectrochemical cell which imitates nature's photosynthesis [1].In its classical configuration, see Figure 1, the cell consists of 2 glass substrates coated with transparent conductive oxide (TCO) (usually SnO 2 : F on soda-lime glass) on which the electrodes are deposited.The photoelectrode consists of a mesoporous layer of semiconductive TiO 2 with a monolayer of chemically adsorbed dye.The transparent counter electrode is coated with a very thin catalytic layer of platinum.A liquid redox electrolyte containing the redox couple iodide/triiodide (I − /I 3 − ) is encapsulated between the two electrodes by a sealant agent, resulting in a plate distance of 20-50 m.When light strikes the solar cell, the photon is absorbed by the dye.The excited dye injects an electron into the TiO 2 conduction band which diffuses through the TiO 2 network.The electron enters the external circuit through the TCO and is transferred to the counter electrode.The photo-oxidized dye injects an electronic hole into the electrolyte which, in the form of I 3 − diffuses to the counter electrode.At the counter electrode, I 3 − is reduced back to I − .
In order to make the DSC attractive to the building industry, long-term stable, large modules on single substrates have to be realized.A thermally and chemically stable sealant is inevitably required to avoid degradation of the DSC.The degradation of the DSC can be caused either by intrinsic instability of the cell components or mechanically by electrolyte leakage or intrusion of H 2 O or O 2 from the outside environment [2,3].
To achieve long-term stable DSCs, H 2 O in the electrolyte should be prevented.Although H 2 O in electrolyte can create an initial increase in photovoltage [4], faster device degradation has been reported [4].This is assumed to be due to loss of iodine when H 2 O is present in the electrolyte [4] as well as dye detachment from the TiO 2 [5].
Many materials have been explored as sealant material such as thermoplastic hotmelt foils like Surlyn 1702 (DuPont) and Bynel 4702 (Dupont), UV curable glues, and glass frit (GF) [6,7].GF is used as sealing material at Fraunhofer ISE.GF is believed to be a strong candidate as sealant material in glass-based DSC modules as it possesses the same properties as the substrates [8][9][10], is nonpermeable, and is UV-and temperature stable in the temperature range which can occur in hot climates.But it is not trivial to find a suited Pb-free [11] GF with low melting temperature that is chemically stable against the oxidative redox electrolyte.It has previously been reported that GF was seen to leach elements into electrolyte [12,13], which caused depletion of I
The GFs were used as received from the manufacturers, and the main compositional elements of the GFs are given in Table 1.The melting temperature,   , was experimentally determined by heating the GF paste to various temperatures Table 1: Main composition of glass frits for sealing.
: temperature where glass powder initiates to melt.
and by investigating the particles by SEM.The onset of melting was determined as   .

The Test Cell.
In the working DSC, the sealing material will only be exposed to the electrolyte at the typically 20-50 m high cell edges.But with the aim to perform accelerated testing, we created a system with a larger (approx.30 times) openly exposed GF area.The total cell area was 5 cm × 0.6 cm.An area of 2 cm × 0.4 cm GF was doctor bladed on one TCO glass (Pilkington TEC-8) and dried in a convection furnace at 150 ∘ C for about 10 min.The samples were sintered with the manufacturer's recommended sintering profile.An intermediate step in the heating curve assured that the organic binder was burnt out of the paste, before melting of the glass particles [2,9].The peak sintering temperature  max was approximately 50 ∘ C higher than   (given in Table 1).
The GF layer had a height of 12 m after sintering.This TCO glass was sealed with another TCO glass by a Surlyn gasket (45 m) as described in [7], creating encapsulated cells with a plate distance of 20 m; see Figure 2. The widely used Surlyn is chemically stable, but it has a softening point of 65 ∘ C, so it is not suitable for tests exceeding 60 ∘ C [7].Surlyn was used as encapsulation to isolate the effect of GF on the confined area inside the cell and was suitable as the storage and exposure temperatures of the cells were mild (room temperature).The cells were filled with electrolyte, and the filling holes were sealed with Surlyn and a thin slide glass.The reference cells did not contain a GF layer.

SEM and EDX Sample Preparation.
After 100 days, the electrolyte samples were extracted from the cells and placed on a TCO (SnO 2 : F) glass, which is classified as chemically inert and thermally stable up to 600 ∘ C [14].Solvents were evaporated, and the organic iodide salt was burnt out at 350 ∘ C, so precipitated residues from the electrolyte could be examined with SEM and EDX.The samples were sputtered with Au to enhance the conductivity.

Instrumentation.
UV-visible measurements were carried out with a spectrophotometer (Scinco UV-Vis spectrophotometer, Neosys 2000) with a detection limitation for transmittance lower than 10 −2 .Absorbance of the electrolyte inside the cell was measured from 350 nm to 500 nm.The baseline was taken with a test cell containing H 2 O.A scanning electron microscope, SEM, and energy dispersive X-ray analysis, EDX, (FESEM JSM 6700F) were used to investigate the morphology and analyze the leached GF elements.
The GF layer thickness was determined with a Dektak profilometer.

Absorbance Spectrum of 𝐼 3
− .The absorbance spectrum of I 3 − in acetonitrile has specific absorbance peaks at around 290 nm and 360 nm [15].I 3 − absorbs until approximately 500 nm.Calibration measurements with known I 3 − concentration in redox electrolyte showed a slight red-shift of the absorbance peak to 366 nm, Figure 3(a), due to the difference in polarity of the ionic liquid PMIM-I compared with acetonitrile [16].A good linear dependence between absorbance and I 3 − concentration according to Beer-Lambert's law [17] was observed at 366 nm as well as 430 nm, Figure 3(b).The linearity at 430 nm will be used to correlate the glass frit interaction with I 3 − loss since the test cells showed absorbance saturation at 366 nm.

Detection of 𝐼 3
− Change in Electrolyte by UV-Vis Spectrophotometry.Absorbance measurements on the test cells were performed during 100 days.The absorbance spectra of the cells with redox electrolyte were saturated for most samples.This was due to the high triiodide concentration in the electrolyte.Therefore, the absorbance value at 366 nm could not be determined.
The absorbance spectra for the reference cell (no GF) and the test cells with GF1-GF4 containing redox electrolyte with 5 and 10 wt% H 2 O are shown in Figures 4(a)-4(e).Only selected representative data are shown.The largest change in absorbance over time was seen for the cells containing GF2 and GF3 when H 2 O was added to electrolyte.
Figure 4(a) displays the initial absorbance spectra of the reference cell (no GF) with water-containing electrolyte initially and after 99 days.No changes in the spectra are seen.This was also the case for the test cells with GF1, Figure 4(b), and GF4, Figure 4(e), which showed constant absorbance spectra over time.Therefore, a long-term reaction between H 2 O and I 3 − under the mild storage conditions (dark, room temperature) can be excluded.Figures 4(c) and 4(d) clearly show that the I 3 − concentration decreases over time in the test cells with GF2 and GF3.The reason for I 3 − depletion can thus be assigned to the GF material and the addition of H 2 O in the system.
The absorbance spectra for the cell with GF2 changed over time and displayed a new absorbance peak at 475 nm, Figure 4(c).This will be further discussed in Section 4.1.
For the cells with GF3 and H 2 O-containing electrolyte, a continuous decrease of the absorbance was seen.After 71 days, the characteristic absorbance peak of I 3 − at 366 nm − concentration with absorbance at, respectively, 366 nm and 430 nm.The linearity at 430 nm is important since the absorbance spectra for the test cells are saturated at 366 nm.Therefore, the decrease in absorbance at 430 nm for the test cells is correlated with the loss of I 3 − .became visible due to the significant decrease in I 3 − concentration, Figure 4(d).
The linearity between I 3 − concentration and absorbance at 430 nm was shown in Figure 3(b).Due to measurement saturation at 366 nm, the absorbance at 430 nm as a function of time for the cells with GF2 and GF3 is shown in Figure 5.
As it can be seen from Figure 5(a), a decrease in I 3 − absorbance for the GF2-containing cell is already visible when 1 wt% H 2 O is added to the redox electrolyte.The absorbance decrease, when respectively 5 wt% and 10 wt% H 2 O are added to electrolyte, follows a similar decay trend and becomes linear after approximately 70 days.
In Figure 5(b), the results for the cell with GF3 are shown.A decrease in absorbance is seen when respectively 5 and 10 wt% H 2 O are added to the electrolyte, with a less strong decrease for the cell with 5 wt% H 2 O in the electrolyte.The absorbance decrease for the cell with 10 wt% H 2 O becomes linear after 70 days, as it was also observed for the cells with GF2.

Diffusion Limitation of 𝐼 3
− in the Cell Design.The decrease in I 3 − content of the cell with GF3 and aqueous electrolyte was clearly visible, as the electrolyte bleached rapidly in the area of the GF layer.The diffusion of I 3 − from the uncoated area to the area with GF, therefore, seems to be of importance for the depletion time of I 3 − .The largest I 3 − concentration gradient will be from the end of the GF area to the end of the noncoated area (∼ 3 cm).The relation between I − ] = 1.5 ⋅ 10 −6 cm 2 /s and  = 3 cm.This is in good correspondence with the initiation of the observed linear decay in absorbance as seen in Figure 5.

EDX Investigation of Leached GF Elements in Redox
Electrolyte with 10 wt% H 2 O.After 100 days, the electrolyte was extracted from the cells, placed on a TCO glass, and heat treated.The leached GF elements could be determined by EDX.As the TCO consists of SnO 2 : F, Sn and O elements were detected in all EDX spectra.
The EDX spectrum of the precipitate from the reference cell with nonaqueous redox electrolyte is displayed in Figure 6(a).The EDX analysis was performed on the area shown in the inserted SEM image.The detected Sn, O, and F originated from the TCO substrate, C and Na from trace amounts in the electrolyte [18], and Au from the Au sputtering for better conductivity.It should be noted that no iodine was detected, so the heat treatment had successfully burnt out the iodine from the redox electrolyte.
The EDX spectra from the remains of the aqueous electrolyte (10 wt% H 2 O) in contact with GF2 and GF3 are shown in the Figures 6(b) and 6(c).The insets show the SEM images of the precipitate.The spectra were obtained by performing spot analysis on the precipitates.
The precipitate from the redox electrolyte with 10 wt% H 2 O extracted from the test cell with GF2 contained bismuth and iodide.For the test cell with GF3, zinc and iodide were detected as the precipitate.In order to ensure that zinc was detected, the acceleration voltage was increased to 20 keV for the sample with GF3.This caused the electron beam to go through the precipitate and the TCO and reach the soda-lime substrate.Hence, silicon was further detected.
The detected metals originated from the GF, and the iodine was from the electrolyte.Since iodine was burnt out during the heating of electrolyte, the detected bismuth and iodine for the GF2 sample and the detected zinc and iodine for the GF3 sample were bound in a thermally stable form.

Discussion
For DSC cells and modules, the I 3 − concentration is one of the possible current-limiting parameters.Therefore, significant I 3 − depletion will lead to a decrease in electrical performance [13].

Test Cell with GF2.
The absorbance measurements revealed a significant decrease in I 3 − absorbance over time for the samples in contact with GF2 and aqueous redox electrolyte.
The EDX analysis of leached elements from GF2 revealed a precipitate of bismuth and iodine.Even though the absorbance measurements showed decaying absorbance at 430 nm over time (Figure 5(a)), the cells did not visually bleach, which can be explained by the occurrence of a new absorbance peak at 475 nm as seen in Figure 4(c).This peak could be assigned to the BiI 4 − formation in aqueous redox electrolyte, as the absorbance peak of BiI 4 − is reported at 460 nm in acetonitrile [19].EDX revealed that zinc was leached from the GF in aqueous redox electrolyte and formed a thermally stable compound with iodine.At present, the formed zinc-iodine complex is not known, but could be present in the form of Zn(I 3 ) 2 (aq), [ZnI 3 ] − (aq), or [ZnI 3 (H 2 O)] − (aq) [20].The formed complex did not change the absorbance spectrum in the range of 350-500 nm over time.Despite the missing determination of the formed compound, the experimental results (Section 3.2) clearly showed a continuous decrease in I 3 − absorbance over time.This will be detrimental for the DSC operation.

Influence of GF Composition on Stability.
Glass dissolves by two mechanisms: selective leaching of glass modifier elements or complete glass dissolution [21].It is known that soda-lime glass dissolves in H 2 O with a dissolution rate in the order of 10 −6 cm 2 /s [21].The process is controlled by the exchange of Na + ions with H + , which breaks up the glass matrix [22].
GF2 and GF3 became unstable when H 2 O was added to the redox electrolyte.GF2 exhibited a low melting temperature compared with GF1, which consists of same major elements.Evidently the solubility of GF2 is higher than GF1.This can be the result of a low bond strength so it is easier to break up the glass structure.The effect of H 2 O in electrolyte led to leaching of bismuth ions that reacted with iodide and formed the BiI 4 − complex, as detected by spectrophotometry.
The composition of GF3 has a high content of ZnO compared with SiO 2 and Al 2 O 3 .As SiO 2 and Al 2 O 3 have a strong tendency to form a tetrahedral structure which forms the glass matrix, ZnO will act as a glass modifier by obtaining interstitial positions in the glass structure.Zinc ions can then readily be exchanged with H + if present in the electrolyte, which leads to leaching of zinc ions [23].
GF3 contained the same major oxides as GF4, but GF4 had a higher melting temperature, which explains the higher chemical stability towards H 2 O.

Accelerated Tests.
In order to test different kinds of GFs, an accelerated test would be preferred.Based on the experiences with the experiment carried out, reflections are made towards applicability and design of accelerated testing.Bi and I are detected due to GF2 and electrolyte interaction.(c) Spot analysis of precipitate from cell with 10 wt% H 2 O-containing electrolyte in contact with GF3.Zn and I are detected due to GF3 and electrolyte interaction.Si is detected due to increased acceleration voltage, so the soda-lime glass was reached by the electron beam.
The ageing conditions of the samples were mild as the samples were stored at room temperature and in the dark.Interaction between GF and redox electrolyte with H 2 O could be accelerated by using a cell design where diffusionlimitation of I 3 − is negligible.Furthermore, the storage temperature could be increased, but this would require a different hole-and cell sealing material than the present Surlyn.
In a GF-sealed DSC, the GF exposed to electrolyte is only present at the sealing edges compared with the investigated cells with an openly exposed GF layer.The GF area in these cells is approximately 30 times larger than in the case of a cell with a plate distance of 25 m sealed with GF, so the ratio between GF area and electrolyte volume is significantly larger, accelerating a possible reaction between the cell components.

Conclusions
In this work, interaction of glass frit (GF) with I − /I 3 − electrolyte and H 2 O present in the electrolyte was investigated for four different GFs.Thin cells with a confined layer of GF were fabricated to enable in situ monitoring of the I 3 − absorbance change of the redox electrolyte.The change in I 3 − absorbance was assigned to a reaction between the GF and the electrolyte and was used to evaluate the chemical stability of the different GFs.
All GFs were stable in nonaqueous redox electrolyte.When H 2 O was added to the electrolyte, two of the investigated GFs became unstable.GF2 based on SiO 2 -B 2 O 3 -Bi 2 O 3 leached bismuth, and a bismuth iodide complex was formed.This was confirmed by the new absorbance peak assigned to BiI 4 − and the detection of bismuth in the electrolyte by EDX.
International Journal of Photoenergy GF3, based on ZnO-SiO 2 -Al 2 O 3 , became unstable when 5 and 10 wt% H 2 O were added to the redox electrolyte.The instability was due to the high ZnO content, where the addition of H 2 O in electrolyte caused zinc ion leaching.The leached zinc ions reacted with I 3 − and caused gradual I 3 − depletion.EDX analysis on the extracted electrolyte confirmed the formation of a compound containing zinc and iodine.
To maintain long-term electrical performance and stability of the dye-sensitized solar cell, a nonreactive and watertight sealing is needed.The methodology of the study can be used to test various GF candidates in a system comparable with the working DSC.
3 − and electrochemical degradation of the DSC.The present work investigates the chemical stability of four low-temperature melting GFs in I − /I 3 − redox electrolyte.The stability of the GFs were investigated individually in nonaqueous I − /I 3 − redox electrolyte and in I − /I 3 − redox electrolyte containing 1, 5, and 10 wt% H 2 O.The design of experiments will be helpful to evaluate a GF for its compatibility and stability in the DSC environment.

3 −Figure 1 : 3 −
Figure 1: Working principle and device structure of a dye-sensitized solar cell (DSC).An organic electrolyte with the redox couple I − /I 3 − transports the positive charge between the mesoscopic photoelectrode and the counter electrode.

Figure 2 :
Figure 2: Cell design for investigating the accelerated influence of GF and H 2 O in the DSC electrolyte.Restricted area with a thin layer of GF and an area for absorbance measurements through the cell.The photo displays the test cells with 10 wt% H 2 O in the redox electrolyte after 100 days.R indicates the reference cell (no GF).

Figure 3 :
Figure 3: (a) Absorbance spectra of I 3 − in PMIM-I at various concentrations.45 m light path length, H 2 O as baseline.The absorbance peak of I 3 − is seen at 366 nm.(b) Linear dependence of I 3− concentration with absorbance at, respectively, 366 nm and 430 nm.The linearity at 430 nm is important since the absorbance spectra for the test cells are saturated at 366 nm.Therefore, the decrease in absorbance at 430 nm for the test cells is correlated with the loss of I 3 − .

Figure 4 :
Figure 4: Development of absorbance spectra during ageing of the reference cell (no GF) and the test cells with 5 and 10 wt% H 2 O added to the redox electrolyte.(a) Reference cell with no GF; no change in absorbance over time, (b) test cell with GF1 showing stable absorbance spectra over time, (c) test cell with GF2; notice the appearance of a new peak at 475 nm and the decrease in absorbance at lower wavelengths, (d) test cell with GF3 showing continuous decrease in absorbance over time, and (e) test cell with GF4 which is relatively stable over time.

Figure 5 :
Figure 5: Absorbance at 430 nm as a function of time for the test cells filled with solvent, nonaqueous redox electrolyte, and redox electrolyte with 1, 5, and 10 wt% H 2 O. (a) Test cell with GF2.Absorbance decreases over time for cells with 1, 5, and 10 wt% H 2 O in electrolyte.(b) Test cell with GF3.Absorbance decreases over time for cells with 5 and 10 wt% H 2 O in electrolyte.

4. 2 .
Test Cell with GF3.The GF3 test cells with 5 and 10 wt% H 2 O-containing redox electrolyte showed significant decrease in I 3 − absorbance over time as seen in Figure 5(b).The electrolyte bleached due to I 3 − depletion.

Figure 6 :
Figure6: EDX spectra of precipitates on TCO after evaporation of electrolyte.Samples were sputtered with Au to increase conductivity.(a) Area analysis as indicated from the SEM inset of nonaqueous electrolyte precipitate.Sn, O, and F originate from the TCO substrate.C and Na are trace amounts from the electrolyte.(b) Spot analysis of precipitate from cell with 10 wt% H 2 O-containing electrolyte in contact with GF2.Bi and I are detected due to GF2 and electrolyte interaction.(c) Spot analysis of precipitate from cell with 10 wt% H 2 O-containing electrolyte in contact with GF3.Zn and I are detected due to GF3 and electrolyte interaction.Si is detected due to increased acceleration voltage, so the soda-lime glass was reached by the electron beam.