AG-BASED THICK-FILM FRONT METALLIZATION OF SILICON SOLAR CELLS

The evolution of microstructure and electrical properties of silver-based thick-film metallizations of silicon solar cells prepared by infrared firing processes has been investigated. The performance of the cells are shown to be dependent on several dynamical and diffusive phenomena. In particular, the sintering of silver grains, silver diffusion into the glass, and the flow of glass at the metal/silicon interface strongly affect important characteristics of the cells such as sheet and contact resistivities and the adhesion of fingers and bus bars. The existance of an optimum value of the peak firing temperature is observed and explained in terms of competitive phenomena occurring at the metal/silicon interface. Moreover it is shown that IR firing treatments require a careful consideration of the sequence of printing and firing steps. The features of heat treatments performed in conveyor belt furnances using Joule and infrared sources are compared.


INTRODUCTION
The advantages of thick film technology for solar-cell manufacture are well known1-3. Virtually all the processes required to prepare a silicon solar cell (dopant diffusion to create a junction, metal deposition for front and back contacts, AR coating, etc.) can be performed using this technology although contact-deposition is the most commonly adopted1. Main advantages of TFT for this purpose are low cost of materials, automation of processes, and stability and reliability of the contacts3. Nevertheless, thick-film metallized cells sometime show a lower efficiency than those prepared by thin film processes. On the other hand, the knowledge of basic phenomena involved in thick-film metallization is still unsatisfactory, and con- A thick-film-metallization ink contains the main metal and other constituents such as a glass frit and sometime other metal oxides. Silver is the most commonly used main constituent of inks for front contacts while Ag/A1 based inks are usually employed for the back surface. These choices are due to compromises among basic requirements of good conductivity, low diffusivity into silicon, and good solderability.
Other constituents of the paste (glass frit and metal oxides) play a relevant role in the contact performance, particularly in achieving low series, high parallel resistance, and good adhesion with the silicon substrate.
The choice of the firing cycle for a given ink is usually based on instructions suggested by the ink manufacturer. Nevertheless, the final choice is determined by the particular facilities available and cell design by a trial and error approach. As a consequence information in the literature on the evolution of microstructure and electrical properties of thick film metal contacts on Si is fragmentary and par-37 tially contradictory-. This paper present the results of a study of front contacts prepared from Ag-based inks. The results illustrate the role played by different constituents of the ink and by the preparation processes in the evolution of microstructure and electrical properties of the cells.

SAMPLE PREPARATION
Solar cells were prepared from (100) single crystal wafers of silicon p-type (boron-doped) having a resistivity into the range 0.5-5 ohm.cm. Although results of this study equally apply to polysilfcon wafers, single-crystal cells were more suitable for analytical investigation. The original wafers, 4" in diameter and 300 tm thick were POC13 diffused at 850C, 25' in order to obtain a n+/p junction at about 3000 below the surface.
Thick-film metallizations were prepared using a silver/aluminumbased ink (4902 DuPont) on the back and a Ag-based ink (TFS 3347) on the front of the cells. The first step was back-metallization printing. After drying at about 200C, the front metallization was deposited and the wafers were fired in a conveyor infrared furnace (Radiant Technology Corporation, C-500 A). Figure 1 shows a typical firing profile. Peak temperature was changed as a parameter in the range 450C to 750C, by steps of 50C.
Samples were also prepaked without back metallization.

EXPERIMENTAL METHODS
A scanning electron microscope (Philips PSEM-500 PX) equipped with an X-ray energy dispersive spectrometer (EDAX-9100) was used for metallurgical and elemental investigation. The  Some samples were cleaved after a quick extraction from a liquid nitrogen bath in order to obtain very sharp cross sections enabling us to investigate metal-Si interactions and depth morphology.
Metallization-thickness measurements were performed using a Taylor-Hobson Talysurf 10. Standard photovoltaic response of cells under AM1 illumination was measured. Electrical properties of metal layer,s (sheet resistance) and devices (diffusion layer sheet resistivity, contact resistance) were studied using the test pattern illustrated in Figure 28 Finally, contact adhesion was evaluated in terms of the force required to peel off from the silicon surface small pads (5 x 5 mm2) of contacts on which a tinned copper strip was soldered.

Layers
Thermal treatments induce changes in the microstructure and morphology of printed layers due to the following processes: shrinking of the layers, evident from a decrease in thickness with an increase of the peak temperature Tf ( Figure 4); sintering of the Ag particles, as observed by SEM ( Figure 5); glass flow down to the silicon surface ( Figure 6). Particles of glass frits are visible on the printed layers up to Tf 450C, but can not identified on samples fired at 550C. These results suggest that the transition temperature of the glass lays in the range 5 between 500 to. 50 C. Sintering of Ag grains and flakes, starts in this range of temperatures, in standard processes with back deposition present. most of metal grains keep their individuality and the mean sizes are unchanged. Also observed are glassy flakes on metal particles that are roundish in appearance. Heat treatment at 600C allows sheet formation: necks between particles increase and grains lose their individuality to form a continuous layer.
A positive correlation exists between the behavior of thickness change and the sintering of grains as observed by SEM analysis.

DISSOLUTION OF SILVER IN GLASS BINDER
In a parallel investigation 9 we studied diffusion and solubility of silver in seven lead glasses of different compositions. We found that dissolution of Ag occurs via thermally activated diffusion processes characterized by relatively high diffusion coefficients D, in the range of 10 -12 to 10 -11 m 2 sec -1 at 850C and activation energies in the range from 1 to 1.5eV (according to the composition of the glass). Higher boron content and/or lower silicon content were associated with higher diffusivities of Ag.
Moreover We did not try to get quantitative data on solubilities in samples prepared at lower temperatures because this analysis would require analytical methods of higher sensitivity and better spatial resolution than that available. However, the extrapolation of the data collected at the temperatures of interest for the present case (600C to 700C), also shows that in these conditions, a firing process lasting for a few minutes is able to saturate a layer less than a micrometer thick with several hundreds ppm of silver atoms. As a consequence the glassy layer interfacing the silver metallization with the silicon substrate has to be considered a "heavily doped" glass.

SUBSTRATE/LAYER INTERFACE
The buried layer/silicon interface is exposed to analysis by means of the above mentioned chemical etch, which dissolves away Ag-metal but does not affect the glass binder. A typical micrograph of the wafer surface under a finger prepared in this way is shown in Figure 7. The wavelike reliefs pertain to a glassy film covering all the Si wafer surface, since elemental analysis detects Pb, the main constituent of the glass binder. The thickness of this film, for Tf 600C in standard conditions, is about 1 1m on the highest "hillocks" that cover a large fraction of the underlaying silicon surface.
Both the area covered by "thick glassy film" and the heights of the "hillocks" increase by increasing the firing temperature. A qualitative evaluation of these two quantities is illustrates in Figure 8. In this interfacial glassy film, EDS analysis shows the presence of dissolved Ag, in agreement with the results presented in the previous section.
A correlation between the formation of the glassy interfacial film and the adhesion of the contact is found. Figure 9 shows that the b) ELECTRICAL PROPERTIES Figure 10 shows the dependence of contact resistance Rc and metal sheet resistance R on the peak firing temperature. RI monotonically decreases in the low temperature range and becomes nearly constant at higher temperatures. This result can be related to the behavior of sintering and shrinking of the metal layer in a comparison of Figures 9, 4 and 5. On the other hand, the contact resistance as a function of the peak firing temperature exhibits a minimum close to 650C. The I-V characteristics of the cells under AM1 exposure are shown in Figure 11. The high series resistance of the cells prepared at low

DISCUSSION AND CONCLUSIONS
We have analyzed the evolution of the microstructure of Ag-based thick-film front contacts of Si solar cells as well as the related evolution of electrical and adhesion properties. Evidence has been found for: the formation of a glassy interracial layer due to flow of the glass binder; dissolution of Ag-metal into this glass; a minimum of the contact resistance; and a considerable effect of the processes steps on the formation of the front contacts.
These findings enable us to describe and predict, in a general way, how preparation processes and material properties affect the cell performance.
In thermal treatments in conveyor belt furnances operating by the Joule effect, the heat exchanges are mainly through thermal convection and conduction, while radiation plays a negligible role. In fact, radiation is involved only during the short-time interval which lasts between the entrance of the material into the hot zone of the muffle and the time when it has been heated up at a temperature nearly equal to that of the muffle walls. If Tc (e.g. 300C) represents the temperature of the material preheated in the drying zone and Tp (e.g.. 650C) the peak temperature of hot (firing) zone, the temperature gradient (Yp Tc) between the source (muffle) and the body is relatively low, and so is the heat Q exchanged by radiation. Moreover, as long as the body heats up by means of thermal convection (the furnace atmosphere) and conduction (being in contact with the heated belt), Q decreases to a null value.
The process is quite different in IR furnaces. Here the temperature Ts of the source is considerably higher (e.g. 1300C) than both Tc and Tp 1. Hence, the radiation-transferred energy Q is much higher, and the exchange occurs during the whole path of the sample into the furnace. These high Q values enable fast firing cycles, which in turn make the heat exchange by conduction and convention negligible. On the other hand, the absorption of radiant heat is dependent on the "optical" properties of the body, which should be a "black body" in order to have maximum absorption. A Si wafer, 300-400 ktm thick, largely differs from a black body in the range of radiation wavelengths generated by a IR source. In fact, typical IR furnaces emit a spectrum of radiation between .5 to 4.5/m, with a maximum power near 1.8 m 11. In these spectral regions, the absorption coefficient of Si is very low12, so that a large part of the incident radiative energy is lost (Figure 12a). However, when the silicon wafer is covered by the back contact, the contact traps the radiative heat on the cell, since it behaves like a "mirror" for IR radiation ( Figure  12b). In our experimental conditions, we found that both Ag and AI (the main constitutents of the back contact) exhibit a high reflectivity at the radiation wavelengths of interest. In this way, the radiation heat is dispersed, and a large fraction of thermal energy contributes : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : For the front metal formation, we have to take into account the following cooperative phenomena, generally occurring in the indicated temperature ranges: a) evaporation and pyrolysis of the organic vehicle (T < 400C) b) sintering of Ag particles (500-650C) c) glass percolation through the porous metal layer and coverage of the silicon surface with a thin glass film (T 500-650C) d) Ag dissolution in the glass binder The phenomenon at point b) is mainly responsible for the sheet resistance of fingers and bus bars (R in Figure 10). The adsorbed heat promotes the sintering up to a maximum stage, where the minimum value of R1 is reached. The phenomena at point c) are mainly responsible for the contact adhesion (Figures 6 and 9) but greatly affect also the contact resistance ( Figure 10). We have found that for adequate adhesion, heat treatments at relatively low temperatures are sufficient, and the pull strength does not increase by increasing Tf over 550C, or in other words, by increasing thickness and area of the interfactial glassy layer on the Si wafer. Conversely the contact resistance Re is greatly affected by the interracial glassy film properties AG-BASED THICK-FILM FRONT METALLIZATION 149 which are related to the phenomena outlined in c) and d). The experimental evidence of a minimum value of Rc is consistent with the evolution of the microstructure of the front contact if we assume that electron tunneling is responsible for the electrical transport from Ag metal particles to Si through the interfacial glassy layer. At low temperatures, the interfacial layer is thinner on average but less transparent to electrons because of a small content of dissolved Ag. At intermediate temperature the best compromise between thickness and dissolution of Ag occours. At higher firing temperatures the increased "doping" of the glass is not enought to compensate for the higher thickness of the glass film. This analysis explains why the glass binder is so critical for the cell performance. The glass composition affects the transition temperature and the viscosity vs temperature dependence, as well as the diffusivity and solubility of Ag into the glass. Moreover, the amount of glass binder affects the Ag wetting and sintering as well as the thickness of the glassy film at the Si/layer interface.
In conclusion, our study has pointed out the main phenomena responsible for the formation of Ag-based thick-film metallizations and the interplay between material properties, preparation processes, and cell performance. Our findings could form the basis for further improvements of inks and processes for high-efficiency, reliable, low-cost solar cells.