Effect of glass frit composition on reliability of silver paste metallization in crystalline silicon solar cells

Glass frit used in conductive silver (Ag) pastes has a significant impact not only on the electrical performance but also on the long-term reliability of metallized electrodes in crystalline silicon (c-Si) solar cells. Here, we investigated the role of compositional changes on the metallization process of silver pastes by adjusting the SiO2, ZnO, Li2O, or Bi2O3 in lead borate glass melts, and performed damp heat (DH) tests in an acidic damp heat environment. It was found that the addition of Bi2O3 resulted in a decrease in conversion efficiency (Eta) of only 6.44% after the cell was treated with dilute acetic acid. Under scanning electron microscopy (SEM), it was observed that the cell with this glass frit had minimal changes in the microstructure of its silver-silicon contacts and silver electrodes. This finding helps to improve the performance and stability of solar cells in harsh environments.


Introduction
The development of photovoltaic (PV) industry has been constantly pursuing more efficient, cheaper, and more reliable solar cells [1][2][3].With the development and application of advanced manufacturing technology, more complex and efficient crystalline silicon (c-Si) cell, such as tunnel oxide passivated contact (TOPCon) cells and back contact (BC) cells [4] are being pushed to the market and become mainstream products [5][6][7].Through measures such as large-scale manufacturing, improving production efficiency, and using cheaper raw materials, the cost of PV power generation has approached or been lower than that of traditional energy generation [8].However, all these innovations pose significant challenges to the reliability of the cells.
In order to ensure a service life of 20 ∼ 30 years, the PV modules must pass the damp heat (DH) test before being installed in the power stations, that is, they must reach the set efficiency indicators after being placed in an environment of 85 °C and 85% relative humidity for several hours [9][10][11].In addition to heat and moisture, acidic substances (acetic acid) decomposed from organic packaging materials (ethylene-vinyl acetate) may attack the cells [12,13].Therefore, the cell manufacturers need to randomly select samples for DH testing by exposing them to acetic acid.This test is essential especially when new Ag pastes are used [14][15][16] as a significant proportion of failures in PV module reliability failure cases have been found to be related to Ag paste metallization [17,18].
Previous studies were almost unanimous in concluding that the mechanism responsible for the decrease in cell efficiency in the DH test was the corrosion of the metal electrode contacts by acetic acid, as the remarkable phenomenon observed was the dimming of the electroluminescent image of the cells and a substantial increase in the resistance of the metallization contacts [19].Subsequent in-depth studies have shown that the inorganic glass phase that binds the silver electrode to the Si emitter is highly susceptible to attack by acid vapor, which disrupts the metallization contact structure and impedes current transfer between Si and Ag [20,21].
In previous p-type cells, such as aluminum back-surface-field (Al-BSF) cells and passivated emitter rear contact (PERC) cells [22,23] lead silicate, lead tellurite, and lead-free tellurite systematic glasses have received focused attention on their performance and characterization in DH testing [24,25].The reaction of PbO in the glass phase with acetic acid is considered to be one of the key mechanisms.It was observed that the migration of lead from the glass phase accelerated the dissolution of the glass, leading to the formation of gaps between the Ag-Si contacts.Another study reported the formation of alternate lead-containing layers at the Ag-Si interface, which inhibited carrier transport [26,27].In n-type TOPCon cells, the front-side Ag paste employs lead borate glass.Its resistance to acetic acid corrosion has yet to be greatly improved.How to inhibit or isolate the reaction between PbO and acetic acid is a subject of great interest.However, this subject has not been clearly recognized and effectively addressed.
Herein, this paper reports the results of a study on the correlation between acid vapor erosion and the composition of lead borate glass frit for metallized Ag paste electrodes, and discusses the effect of SiO 2 , ZnO, Li 2 O and Bi 2 O 3 components in the glasses.

Experiment
Analytical grade PbO, Bi 2 O 3 , SiO 2 , B 2 O 3 , ZnO and Li 2 O were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd and not further purified.Firstly, the chemicals were weighted according to the molar ratios listed in table 1 and mixed thoroughly.Then, they were put into an alumina crucible and heated in an electric furnace at 1000 °C for 30 min.The glass melt was quickly poured into deionized water at room temperature.The solidified glass bulk was ground into small particles with D50 size of about 2 μm using a ball mill.
The Ag paste was composed of 88 wt% Ag powder (DOWA Electronics Materials, Japan), 3 wt% glass frit, 1 wt% Al powder (Toyo Aluminum K.K., Japan) and 8 wt% home-made organic carrier.The glass frit was mixed firstly with organic carrier by hand, then Ag and Al powder were added to them.Subsequently, a three-roll mill was used for mechanical mixing to prepare Ag paste.
The as-prepared paste was screen printed on n-type monocrystalline Si cells (figure 1).The Si wafer was sized 18.2 cm × 18.2 cm and had a resistivity of 1-3 Ω•cm.After screen printing the as-prepared Ag paste and commercial rear-side Ag paste, the cells were baked at 300 °C and sintered in an industrial tunnel furnace with a peak temperature of 750 °C.
The damp heat (DH) test was conducted in a sealed box with a fan mounted on the side wall.First, a saturated aqueous solution of potassium chloride containing 3% acetic acid was prepared by dissolving 115 g of potassium chloride and 6 ml of acetic acid into 194 ml of deionized water.The solution was poured into the box.The test cells were suspended above the solution.The box was then sealed and placed in an oven at 85 °C and 85% relative humidity for 6 h.
The Ag-Si contact resistance and electrode resistance were measured using the transfer length method (TLM), while the thin-layer resistance was measured using the four-probe method.The electrical properties of  the cells were obtained by I-V testing (light source: tungsten iodine lamp; power density of about 1 kW m −2 , sample capacity of 20).The morphology of the Ag-Si interface was observed using field emission scanning electron microscopy (FE-SEM) and energy dispersive x-ray spectroscopy (EDS).The glass transition temperature (Tg) was measured using a differential scanning calorimeter (DSC; Q20, TA Corp.) at a heating rate of 10 °C /min.The infrared absorption spectra of the glass frit samples were measured in the range of 400-2000 cm −1 using a Fourier transform infrared spectrometer (FT-IR; Thermo Scientific Nicolet iS20).A pH meter (PHS-3E, Lei Chi, Shanghai, China) was used to monitor the pH change of the acetic acid solution containing the glass frit.The acid solution impregnated with the glass frit was centrifuged well and the supernatant was taken for ICP-OES testing on an Agilent 5110 (USA).A broadband dielectric impedance spectrometer (Concept 40) was used to test the dielectric constant of the glass at 10-10 7 Hz.

Results and discussions
Table 2 lists the electrical performance of the cells when the glasses in table 1 were used in the front-side Ag pastes.Among them, the cell with #3 glass frit had the highest conversion efficiency (Eta) of 24.91%.However, the maximum open-circuit voltage (V oc ) of 721.2 mV was obtained with #5 glass frit and the highest filling efficiency (FF) of 83.73% was obtained with #4 glass frit.The short-circuit currents (I sc ) of the cells were not significantly different, indicating that the characteristic line sizes of the printed Ag pastes were basically the same when different glasses were used.V oc and FF are the most critical factors in determining the efficiency of the cells, which are closely related to the characteristics of the glasses [28].
The microstructure of metallized Ag paste electrodes on c-Si cells was observed using a field emission scanning electron micrography (FE-SEM).As an example, figure 2(a) presents a typical cross-sectional SEM photograph of Ag electrodes, for which #1 glass frit was used in the Ag paste.It was clearly observed that Ag and Si was bonded together by a layer of glass phase medium from the images of FE-SEM and EDS (figures 2(b), (d)-(f).This indicates that the formation of Ag paste metallization contacts is closely related to the behavior of glasses.During the sintering process of the printed Ag paste, the glass melt flows onto the Si surface and erodes the passivation layer (Si 3 N x /Al 2 O x ), bringing Ag and Si into close contact.If the glass is highly erosive and at the same time has good wettability and flowability, then Ag and Si can make full contact and obtain a low contact  resistance.This is critical to improving FF of the cell.However, if the glass is so aggressive that it damages the p + -Si emitter or even the p-n junction, then the V oc value of the cell drops dramatically.In addition, figure 2(c) shows that the Ag electrode body consists of sintered grains.There is presence of small amounts of glass in the Ag grain boundaries, which has an effect on the sintering of Ag grains [29].
Based on the electrical parameters of the cells listed in table 2, it can be deduced that the #4 glass frit had the strongest corrosion capability, resulting in the lowest V oc value of 717.8 mV.The #5 glass frit had the weakest corrosion capability, and therefore the highest V oc value was obtained when it was used.However, due to its inability to satisfy good Ag-Si contact, the cell's FF was quite low.The #3 glass frit had a relatively moderate corrosion and balanced V oc and FF well, resulting in the highest Eta of the cell.In contrast, although the #1 and #2 glass frits also had a corrosion level that was between that of the #4 and #5 glass frits, the cell's performance was slightly poorer when they were used.This is because the Ag paste metallization contact is related not only to the corrosion ability of the glass, but also to the rheological and wettability properties of the glass [30].
The relationship between the thermal properties and composition of glass frit has been investigated.The #1 glass frit, only consisting of PbO and B 2 O 3 , was used as a reference.By adding SiO 2 , ZnO, Li 2 O or Bi 2 O 3 , the thermal properties of the glass frit were changed, as shown in figure 3. The DSC results (figure 3(a)) shows that compared to the #1 glass frit, the other four glass frits had their transition temperatures (T g ) increased to varying degrees.There was a crystallization peak at 440 °C in the #1 glass frit.When SiO 2 (#2 glass frit) or ZnO (#3 glass frit) was added, the crystallization peak disappeared, indicating an increase in the viscosity of the glass melt.The #4 glass frit had a crystallization peak at around 400 °C.This is due to the strong dissociation effect of Li 2 O on the glass network [31] The infrared spectra in figure 3 The rheological behavior of glass frits during the heating process is also closely related to their composition.As shown in figure 3(c), the initial shrinkage temperature varied from one glass frit to another.When the glass frit reached its maximum shrinkage, the liquid phase was generated.At the hemispherical temperature, the glass frit began to soften and exhibited fluidity.Unlike the change in transition temperature, the #4 glass frit was the last to undergo liquefaction, and due to its strong crystallization behavior, it had the highest softening and melting temperatures.Although the transition temperature and softening temperature of the #3 glass frit were all higher than those of the #1, #2, and #5 glass frits, its melting temperature was the lowest.This was attributed to the melting aid effect of Zn ions at high temperatures.After melting, the wetting properties of the #4 glass frit was improved compared to the #1 glass frit, forming a smaller contact angle of 26.2°on Si surface (figure 3(d)).The contact angle of the #1 glass melt was 35.9°.In contrast, the contact angles of the #2, #3 and #5 glass melts were larger than that of the #1 glass melt.This implies that Li 2 O can enhance the affinity between glass melt and silicon, while SiO 2 , ZnO and Bi 2 O 3 can increase the viscosity and surface tension of the glass melt [32].
In the Ag paste metallization, the glass frit composition had a significant impact on the electrical performance of the cell.Usually, in PbO-B 2 O 3 glass system, SiO 2 and ZnO (e.g.#2 and #3 glass frits) can increase the transition and softening temperatures of the glass frit and attenuate the corrosion on Si emitter, thus improving V oc of the cell.In addition, at high temperature, ZnO may also reduce the viscosity and improve the wettability of the glass melt, beneficial for obtaining higher FF.Li 2 O caused a significant increase in softening and melting temperatures of the glass frit (figure 3), which was due to the precipitation of crystals during the heating process.This indicates that the amount of the #4 glass melt flowing towards the Ag-Si interface was very small.However, the corrosiveness of the #4 glass frit increased instead, because Li 2 O split the glass network and weakened the bonding with Pb ions.It was the redox reaction between Pb ions and Si 3 N x passivation layer that determined the formation of Ag-Si metallization contact.Therefore, when the #4 glass frit was used for Ag paste, the electrical resistance of the metallized Ag-Si contact was low.Bi 2 O 3 can not only reduce the viscosity of the glass melt and improve its fluidity, but also inhibited the corrosive property of Pb ions.As a result, both the Ag-Si contact resistance and V oc of the cell were high when the #5 glass frit was used.
The cells were further evaluated by DH testing.Table 3 lists the electrical performance of the cells after the test.It is clear that the electrical parameters have deteriorated significantly.The bar graph in figure 4(a) shows the percentage change in the various electrical parameters.In comparison, when using the #5 glass frit, the cell maintained a relatively high Eta of 23.26%, which only decreased by 6.44%.In contrast, the cell using the #4 glass frit was the most severely damaged, thus its Eta was only 17.18%, having a decay of 30.98%.When other three glass frits #1, #2, and #3 were used, Eta of the cells decreased by 17.47%, 13.12%, and 17.78%, respectively.The results indicate that the cell reliability was closely related to the composition of the glass frit.In the PbO-B 2 O 3 glass system, SiO 2 (#2 glass frit) and Bi 2 O 3 (#5 glass frit) can enhance the acid resistance of the cell, while ZnO (#3 glass frit) and Li 2 O (#4 glass frit) would weaken the acid resistance.By comparison, Bi 2 O 3 showed a more enhancing effect than SiO 2 , and Li 2 O showed a more weakening effect than ZnO.
Among the electrical parameters of the cells, the most variable is the series resistance (R s ), as shown in figure 4(a).Even with the #5 glass frit, Rs increased the least, but by 226%.When the #4 glass frit was used, Rs surprisingly increased by 3311%.It was found that the larger R s was, the smaller V oc , I sc and FF were.In contrast,   The chemical stability of the glass frit is considered to be a decisive factor in the reliability of Ag paste metallization.It was found that in weakly acidic aqueous solutions, dispersion of the glass frit resulted in an increase in pH, which reflected the amount of alkali released.As shown in figure 6(a), after dispersion of the glass frit, the pH of the solution rises rapidly within the first 30 min due to the Lewis base effect on the glass surface.The #5 glass frit showed good stability and the pH of the solution stayed below 7.However, during the dispersion of #4 glass, the pH of the solution continued to rise, implying a continuous release of lithium ions.The main glass components, such as PbO and B 2 O 3 , were also dissolved in the aqueous solution, leading to the decomposition of the glass network structure.ICP measurements were made on the solutions after the glass frits had been dispersed for three hours.The results (figure 6(b)) showed that #1 and #4 glasses released more than 0.5% boron, which indicated that their networks were easily broken down.At the same time, a certain amount of elemental lead was dissolved in the solution, which caused a significant change in the electrical properties of the glasses.For example, as shown in figure 6(c), the dielectric constant of the glasses decreased after dispersing in the solutions.
It can be reasonably inferred that the dissociation of the glass network structure and the leaching of metal ions from the glass frits will lead to the destruction of the Ag/Si contacts and grain boundaries inside the electrode body.To further reveal the destruction mechanism, the microstructure of glass layer at Ag-Si contact interface was investigated using FE-SEM.For this observation, the Ag electrode was removed from the cell to expose the glass layer on the Si wafer.Prior to acid vapor treatment, it can be seen in figure 7(a) that there was a usual glass layer, in which a large number of nanoscale Ag colloidal particles were dispersed (figure 7(b)).These Ag colloidal particles are believed to facilitate the transmission of electric current.After acid vapor treatment, it was observed that not only the glass layer was eroded to form pits, but also many rod-shaped crystals were grown (figure 7(c)).The EDS analysis showed that these rod-shaped crystals were mainly composed of Pb and Cl elements (figures 7(d)-(f)).Therefore, it can be speculated that in an acidic vapor environment, as the glass network dissociated, Pb ions migrated out of the glass network and combined with the Cl ions in the acidic vapor to grow crystals.This eventually led to the disconnection of the Ag electrode from the Si surface, resulting in a dramatic increase in the Ag-Si contact resistance.In addition, the Ag grain boundaries in the electrode body were disrupted, resulting in an increase in resistivity.

Conclusions
Lead borate glass composition affects the metallization contact and acid corrosion of Ag paste electrodes on c-Si cells, as well as, the cell efficiency and electrical parameters, because of the changes of the network structure, thermal properties, fluidity and chemical stability by the addition of SiO 2 , ZnO, Li 2 O or Bi 2 O 3 components.These changes also made the DH test performance of the cells in an acidic vapor environment highly dependent on the glass frit composition.
The addition of SiO 2 and Bi 2 O 3 components to lead borate glass can reduce the excessive corrosion of Si emitter, which was beneficial to the cell to obtain a higher V oc value.With the addition of ZnO and Li 2 O, the glass can completely erode the passivation layer, leading to the full contact between the Ag and Si emitters, as a result the FF value of the cell increases.In order to achieve higher cell efficiency, it is necessary to optimize the glass composition to better balance the V oc and FF.It was also found that the SiO 2 and Bi 2 O 3 components can improve the chemical stability of the glass, while the ZnO and Li 2 O components have the opposite effect.In addition to glass decomposition, the reaction of migrating lead with acids can also seriously degrade the  performance of Ag electrodes.This study found that the Bi 2 O 3 component has a greater potential to inhibit glass network decomposition and lead migration.

Figure 2 .
Figure 2. SEM photographs of (a) cross-section of a metallized Ag electrode, (b) a layer glass at Ag-Si contact interface, (c) grain boundaries inside the Ag electrode, and (d)-(f) EDS of Ag-Si contact layer Si, Ag, Pb.
(b) show the vibrational characteristics of the B-O bonds from [BO 3 ] (∼1260 cm −1 and 710 cm −1 ) and [BO 4 ] (∼910 cm −1 ) groups.It is clear that Li 2 O in the #4 glass network can significantly increase the number of [BO 4 ] groups.Also, ZnO in the #3 glass frit may slightly increase the number of [BO 4 ] functional groups.Comparatively, SiO 2 or Bi 2 O 3 showed a relatively slightly impact on the structure of the glass network.

Figure 3 .
Figure 3. Characterizations of glass frit samples.(a) DSC curves, (b) Infrared spectra, (c) feature temperatures and (d) contact angle of glass melts on Si surface.

Figure 4 .
Figure 4. Chart diagrams showing the changes in the various electrical parameters of the cells and Ag electrodes after DH test.(a) Change ratios of electrical parameters of the cells; (b) contact resistivities (ρc) and (c) conductivities of the Ag electrodes.
the sheet resistance (R sh ) of the cells increased slightly.This implies that the Ag-Si contacts of the cells was damaged by the acid vapor.As shown in figure 4(b), the initial contact resistivity (ρ c ) was ranged from 2.74 to 3.98 mΩ•cm 2 , but after the DH test, it changed to 7.29-32.39mΩ•cm 2 .Consistent with the change in the series resistance, the #4 glass frit caused a maximum increment in ρ c of 1080%, while the #5 glass frit had a minimum increment of 83%.In addition, the Ag electrode conductivity also decreased significantly.As shown in figure4(c), the conductivity (σ) of the Ag electrode decreased from around 4.7 ∼ 4.8 × 10 7 S m −1 before DH test to 3.2 ∼ 4.2 × 10 7 S m −1 .The results showed that the cell with the more increase in Rs also had the more decrease in σ.This indicates that acid vapor also damaged the silver electrodes.After DH test, the microstructural changes at the Ag-Si contact interface (figures 5(a)-(e) and inside the Ag electrode body (figures 5(f)-(j) were investigated by SEM.In the cell using #4 glass frit, the contact interface was damaged severely (figure 5(d)), resulting in large areas of Ag electrodes detaching from the Si wafer.It can be found that the connection of Ag electrodes with Si wafer were damaged to a lesser extent with the use of #1 (figure 5(a)) and #3 (figure 5(c)) glass frits, with some voids appearing at the Ag-Si contact interface.In comparison, the Ag-Si contact formed with #2 (figure 5(b)) and #5 (figure 5(e)) glass frits suffered the least damage, where fewer voids were observed.Additionally, the Ag electrode body was also corroded by acid vapor, as shown in figures 5(f)-(j).From the cross-sectional SEM photographs of Ag electrodes, a large number of voids were observed inside the electrode body, forming a clear contrast with the dense sintered structure before DH test (figure 2).The internal structure of the Ag electrodes containing #5 glass frit were minimally damaged in comparison to the other glass frits.As a result, silver electrodes prepared from #5 glass frit (Bi 2 O 3 ) have superior resistance to acetic acid, which was in accordance with the change in conductivity.

Figure 7 .
Figure 7. SEM observation and characterization: (a)-(b) contact surface with removal of the Ag electrode before DH test; (c) the contact surface after DH test; (d) a rod-like product of lead and acid vapor; (e)-(f) EDS mapping images of the rod-like product.

Figure 6 .
Figure 6.(a) pH values of the solutions in which various glass frits were dispersed; (b) leaching proportion of lead and boron from glass frits in the solutions; (c) dielectric constant of the glass frits before and after DH test.

Table 2 .
Electrical performance of the cells.

Table 3 .
Electrical performance of the cells after DH test.