A screen printed silver/metallic glass (MG) paste formulated with Ag acetate resulted in a specific contact resistance in the range of 0.6-0.7 mΩ·cm2 on both the n- and p-type Si emitters of interdigitated back-contact solar cells. Silver nanocrystallites resulting from thermally decomposed Ag acetate prevented the Al MG frits from directly interacting with the Si emitter, thus reducing the amount of Al diffused into the Si emitters, and subsequently, the contact resistance. A photovoltaic conversion efficiency of 20.3% was achieved using this technique.
Bibliographical noteFunding Information:
Jun Kim Suk 1 a) Yun Kim Se 1 a) Man Park Jin 1 Hwan Park Keum 1 Ho Lee Jun 1 Mock Lee Sang 1 Taek Han In 1 Hyang Kim Do 2 Ram Lim Ka 2 b) Tae Kim Won 3 Cheol Park Ju 4 Soo Jee Sang 1 c) Lee Eun-Sung 1 c) 1 Electronic Materials Laboratory, Samsung Advanced Institute of Technology (SAIT) , San #14-1, Nongseo-dong, Giheung-gu, Yongin-si, Gyeonggi-do 446-712, South Korea 2 Center for Noncrystalline Materials, Department of Materials Science and Engineering, Yonsei University , 134 Shinchon-dong, Seodaemun-ku, Seoul 120-749, South Korea 3 Department of Optical Engineering, Cheongju University , 36 Naedock-dong, Cheongju 360-764, South Korea 4 Research Institute for Advanced Materials, Seoul National University , 599 Gwanak-ro, Gwanak-gu, Seoul 151–742, South Korea a) S. J. Kim and S. Y. Kim contributed equally to this work. b) Present address: Light Metal Division, Korea Institute of Materials Science, Changwon, Gyeongnam 642-831, South Korea c) Authors to whom correspondence should be addressed. Electronic addresses: email@example.com and firstname.lastname@example.org . Tel.: 82-31-280-9397. Fax: 82-31-280-9207. 05 08 2013 103 6 063903 05 06 2013 22 07 2013 07 08 2013 2013 AIP Publishing LLC 0003-6951/2013/103(6)/063903/4/ $30.00 A screen printed silver/metallic glass (MG) paste formulated with Ag acetate resulted in a specific contact resistance in the range of 0.6–0.7 mΩ·cm 2 on both the n- and p-type Si emitters of interdigitated back-contact solar cells. Silver nanocrystallites resulting from thermally decomposed Ag acetate prevented the Al MG frits from directly interacting with the Si emitter, thus reducing the amount of Al diffused into the Si emitters, and subsequently, the contact resistance. A photovoltaic conversion efficiency of 20.3% was achieved using this technique. crossmark Researchers and manufacturers of photovoltaic materials aspire to decrease the “dollar per watt” of solar cells. One of the best routes to achieve this is to fabricate high-efficiency cells such as interdigitated back contact (IBC), emitter-wrap-through, and metal-wrap-through solar cells using a low-cost printing process. 1 In a previous report, we demonstrated the feasibility of fabricating IBC solar cells by screen printing (SP) with a Ag paste formulated with Al metallic glass (MG) frit instead of an oxide glass frit. 2,3 Using the Ag/Al MG frit paste, the SP technique opens the way to replace the conventional electroplating (EP) method. The main drawback of the EP process is a multistep metallization process. Generally, the EP method consists of at least five steps: (i) preparing the seed layer, (ii) forming-gas annealing, (iii) applying the plating resist pattern, (iv) electroplating, and (v) etching, 4 whereas SP is a one-step process. Having fewer processing steps is definitely beneficial for reducing the fabrication cost and time. 5 Thus, the SP technique is more promising than the EP process for electrode formation for solar cells in a cost-effective way. However, the SP technique with the Ag/Al MG frit paste has one drawback: Al MG frit is a source of p-type dopants, 6 causing significantly higher contact resistance on n-type electrodes than on p-type electrodes. This problem could be solved by SP twice using two different Ag pastes, one in which the Al MG frit is present on the p-type emitter and the other in which the Al MG frit is absent on n-type emitters. Also, preparing an additional thin film to act as a diffusion barrier on the Si emitter can also be considered a solution to this problem. 7 However, increasing the number of processing steps undermines the benefits of the SP technique. Here, we report that we achieved a specific contact resistance ( ρ c ) in the range of 0.6–0.7 mΩ·cm 2 for both n- and p-type electrodes by printing a single paste just by adding Ag acetate into the Ag paste, without using additional processing steps. The thermally decomposed Ag acetate transforms into Ag nanocrystallites on the surface of the Al MG frit, forming an Al–Ag solid solution on the surface during firing. Thus, Al–Si interdiffusion could be reduced by preventing the Al MG frit from directly contacting the Si emitter, thus reducing both the interlayer thickness between the Ag electrode and Si emitter ( L ) and the degradation of n-type Si doping concentration ( N s ). By controlling the interdiffusion, thus optimizing L and N s , we were able to fabricate electrodes with a ρ c in the range of 0.6–0.7 mΩ·cm 2 , values that are comparable to those of electroplated electrodes (0.1–0.3 mΩ·cm 2 ). The gas atomization method was utilized for the fabrication of the Al 85 Y 8 Ni 5 Co 2 MG powders (PSI Ltd, U.K.). The composition of the atomized powders was carefully examined by inductively coupled plasma (ICP, Shimadzu, ICPS-8100) analysis because their composition can shift due to evaporation during induction melting followed by the atomizing process. The atomized powders were sieved using an air classifier (Donaselect 150, Aishin, Japan) to collect fine MG powders whose particle diameters D 90 was less than 3 μ m, where D 90 is the equivalent mass diameter at 90% cumulative mass. To investigate the effect of silver acetate on the Ag/MG paste, the paste was prepared by 3-roll milling the mixture of Ag powders (Changsung Corp., Korea), MG powders (D 90 < 3 μ m), silver acetate (Sigma-Aldrich), and organic components including butyl carbitol and ethyl cellulose. The optimum weight ratio of Ag acetate to the total solids in the paste was 0.04. The ratio was determined by measuring the ρ c of the electrodes, which were fabricated by printing a number of Ag/MG pastes formulated with various amounts of Ag acetate. To measure the ρ c using the transfer length method (TLM), electrode patterns were prepared by screen printing the Ag/MG frit paste on single crystalline Si wafers with a sheet resistance of 70 Ω/sq for boron-doped emitters (p-type) and 20 Ω/sq for phosphorous-doped emitters (n-type). The Si wafers with printed TLM electrode patterns and screen-printed IBC cells were dried at 200 °C to evaporate the organics in the paste and then fired in an infrared (IR) belt furnace (CDF-7214, Despatch). Total firing time did not exceed 60 s and the peak temperature of the process was limited to less than 610 °C. Five IBC cells were randomly selected among the 25 cells fabricated using the Ag/MG paste, and their photovoltaic efficiencies were measured once a week for six months to check their reliability. No degradation was observed. The conversion efficiency of the photovoltaic cell was measured using a solar simulator (WXS-155S-10, Wacom). We attempted to achieve a ρ c of less than l mΩ·cm 2 on both the n- and p-type Si emitters by controlling the interdiffusion between the Al metallic glasses in the electrodes and the Si emitter ρ c,tunneling = | dV dJ | V = 0 = 1 A c k qT A * exp [ 4π ε s m * qh Φ Bn N s ] / Θ , Θ = exp ( π m * q φ B 2 2 ℏ L ) , (1) where k , T , q , A *, ε s , m* , h , Φ Bn , and N s are the Boltzmann constant, temperature, electronic charge, effective Richardson constant (4 πqm*k 2 /h 3 ), relative permittivity (11.68), effective mass of hole or electron, Planck's constant, Schottky barrier height (0.40 eV on a p-type emitter and 0.76 eV on an n-type emitter under the assumption of no surface state on Si), and doping concentration, respectively. First, the interlayer thickness determined the probability of electron tunneling; a thinner L resulted in a higher probability of tunneling and thus a lower ρ c . The interlayer thickness increased as a function of the interdiffusion of Si and Al. 12 The thickness observed with transmission electron microscopy (TEM) was greater than 20 nm, as shown in Fig. 1(a) . Second, the quantity of Al diffused into the n-type Si emitter reduced the effective N s of the n-type dopants. The higher ρ c on the n-type emitter than on the p-type emitter prepared using the Ag/Al MG frit paste is attributed to the fact that diffused Al acts as a p-type dopant, reducing effective N s on the n-type emitter. To minimize the growth of the interlayer and the reduction of the effective N s , Al diffusion into the Si emitter should be minimized. Although there are two more controllable parameters for reducing ρ c in Eq. (1) —the contact area ( A c ) and the tunneling barrier height ( φ B )— A c was not considered in this report because no significant difference in A c was observed, regardless of the addition of Ag acetate into the Ag/Al MG paste. The tunneling barrier height slightly increased by the addition of Ag acetate, which is discussed in detail below. Silver acetate (CH 3 COOAg) was added to the Ag/Al MG paste in order to govern Al–Si interdiffusion. The mechanism of contact formation is shown schematically in Fig. 2 . When the Ag/Al MG frit paste without Ag acetate was fired, it started to exhibit thermoplastic forming behavior in the supercooled liquid region (SCL) [from the glass transition temperature, T g , of 274 °C to the crystallization temperature of 301 °C, Fig. ??? ] and flowed into the gap between the Ag electrode and the native oxide on the Si emitter by capillary force as reported previously. 8,9 As a result of this thermoplastic deformation in the SCL, the contact area between the Ag powder and the Al MG frit dramatically increased. When the temperature increased to 450 °C, the thin native SiO 2 layer was easily reduced by Al ions from the deformed Al MG frit present between the Ag electrode and Si emitter because of the relatively higher potential required for reduction [4/3Al + SiO 2 → 2/3A1 2 O 3 + Si (Δ G = −255 kJ/mol)]. 10 The reduction of the SiO 2 layer provided pathways for the interdiffusion of Si and Al [Fig. ??? ]. The Si that diffused into the interlayer was oxidized during firing and cooling, thus forming an oxide interlayer [Fig. ??? ]. In contrast, when Ag acetate was added into the Ag/Al MG paste, it was randomly distributed in the paste [Fig. ??? ]. The acetate was thermally decomposed at 260 °C and nucleated on the surface of not only the Al MG frit but on the Ag powder and Si emitter as well [Fig. ??? ]. 11 Figure ??? shows the SEM image of this Ag nanocrystallite formation, which prevented the Al MG frit from directly interacting with the Si after the thermoplastic deformation [Fig. ??? ]. Silver nanocrystallites may dissolve into the Al MG frit at temperatures up to 567 °C [Fig. ??? ], which is the Al–Ag eutectic temperature based on the binary phase diagram of Al–Ag. 12 According to the phase diagram, the solubility of Ag in Al increased to about 25 at. % when the temperature was increased up to the eutectic temperature. In addition, α-Al was a major phase of the Al 85 Y 8 Ni 5 Co 2 MG frit after crystallization at 301 °C. 13 The Al–Ag solid solution contacted with Si so that Al–Si interdiffusion was less preferable than the case in which α-Al without the solute of Ag contacted Si. The chemical potential of Al was reduced by forming the solid solution and Ag atoms already occupied vacancies in α-Al into which Si atoms were supposed to diffuse. Both Ag and Si tended to diffuse via a vacancy mechanism because the atomic radius ratios of Si to Al and Ag to Al are both greater than 0.85. 14 Thus, less interdiffusion was attributed to the formation of an Al–Ag solid solution, resulting in a thinner L [Fig. ??? ], less degradation of N s , and finally lower ρ c . As summarized in Table I , the value of ρ c on the n-type Si emitter decreased significantly, whereas that of ρ c on the p-type Si emitter decreased to half. No obvious changes in the electrical resistivity of the electrodes were observed. As shown in Fig. 1 , the thickness of the interlayer between the Ag electrode and the Si emitter decreased from 19 ± 6 nm to 8 ± 3 nm when the Ag electrode was fabricated using the Ag/Al MG paste formulated with Ag acetate. The thinner interlayer was attributed to the reduced interdiffusion, results that are supported by the analysis of Al that remained in the interlayer. The energy dispersive X-ray (EDX) line profiles shown in Fig. 1 and the electron energy loss spectroscopy (EELS) analysis shown in Fig. 3 clearly indicate that Al remained inside the interlayer instead of diffusing into the Si emitter. According to the EELS analysis, Al appeared to be oxidized to Al 2 O 3 during firing followed by cooling. The onset energy of the Al K-edge in pure Al is ∼1550 eV and the onset energy of the Al K-edge obtained in the interface is ∼1560 eV, a value similar to that of Al 2 O 3 obtained from a reference sample. The Al remaining in the interlayer provided clear evidence that Al–Si interdiffusion was impeded by the addition of Ag acetate. The tunneling barrier height and effective N s were further analyzed using Eq. (1) . First, the addition of Ag acetate reduced the contact resistance but the degree of the reduction was lower than the expected value, which was based on the calculation using Eq. (1) . The inconsistency between the measured contact resistance and the calculated one originated from the increase in φ B , which was clearly rationalized by considering the contact resistance on the p-type electrode. Because the contact resistance on the p-type electrode was less affected by interdiffusion, the reduced resistance with the addition of Ag acetate clearly showed the effect of the variation of φ B . If φ B and other constants were fixed, then the value of ρ c calculated in Eq. (1) would be much lower than the measured one, 0.6 mΩ·cm 2 when L decreased from 19 to 8 nm. However, ρ c decreased by only half when Ag acetate was added. It is thought that the increase in φ B is the reason for the limited reduction of ρ c . This conclusion is supported by the fact that no Ag nanocrystallites (NCs) formed within the interlayer region of the sample fabricated without Ag acetate, as shown in Fig. 1(b) , in comparison to the presence of Ag NCs in the sample fabricated with Ag acetate, as shown in Fig. 1(a) . Based on the calculation, the tunneling barrier height increased from 35 meV (with Ag NCs) to 124 meV, thus obtaining a contact resistance of 0.6 mΩ·cm 2 with an L of 8 nm. Silver atoms in the interlayer region may have easily diffused into the neighboring bulk Ag to avoid interface formation, requiring an additional increase in free energy. However, further study is required to clarify the reason for the absence of Ag crystallites in the interlayer. The estimated φ B value of 124 meV was close to the previously calculated value of 140 meV by assuming that tunneling occurred only in that part of the interlayer where no Ag NCs were present. 2 Second, reduction of effective N s by Al–Si interdiffusion was estimated by comparing the contact resistance on the n-type emitter with and without the addition of Ag acetate. According to the calculation using Eq. (1) , excessive Al–Si interdiffusion without the addition of Ag acetate reduced the effective N s from the initially doped value of 10 20 /cm 3 to 3 × l0 18 /cm 3 . As a result, ρ c increased significantly from 0.7 mΩ·cm 2 with acetate to 13 mΩ·cm 2 without acetate. Therefore, the less degradation of N s in the n-type emitter resulted in a ρ c of 0.6–0.7 mΩ·cm 2 on both types of Si emitters, although the advantage of a reduced L was somewhat offset by the increase in φ B . The minimization of Al–Si interdiffusion and the thermoplastic forming behavior of MG allowed us to fabricate IBC cells with an average photovoltaic conversion efficiency of 20.03%. The maximum efficiency was 20.3%. The I–V curve of the IBC solar cell and a summary of the corresponding cell parameters are provided in Fig. 4 . An adhesion test with using 3M tape was passed by all the cells fabricated. The higher efficiency than that in our previous report may be attributed to lower ρ c . 2 The results reported in this paper demonstrate that Ag acetate is a promising additive for the Ag/Al MG paste that serves to impede the Al–Si interdiffusion between the Ag/Al MG electrode and the Si emitter. The impeded Al–Si interdiffusion resulted in a thinner L and less degraded N s and thus, led to a photovoltaic conversion efficiency of 20.3% from IBC cells. We believe that our results will help in improving the cost-effectiveness of solar cells. We would like to acknowledge the support received from the Solar Energy Business Division, Development 1 Group, Samsung SDI Co., Ltd. Table I. Comparisons of both specific contact resistance and electrical resistivity of Ag electrode fabricated with using the Ag/Al MG paste with and without Ag acetate. Ag acetate Specific contact resistance (mΩ·cm 2 ) Electrical resistivity ( μ Ω·cm) n-type p-type Added 0.7 ± 0.3 0.6 ± 0.2 3.0 Not added 13 ± 2 1.2 ± 0.3 3.2 FIG. 1. Cross-sectional TEM images of interlayer between Ag electrode and Si cell formed (a) without and (b) with Ag acetate. Inset shows low-magnification images of cross-sectional view of Ag electrode, and (a) is the high-magnification image of the box area. EDX line profiles were obtained from the dotted lines. Ag acetate in the Ag paste reduced the interlayer thickness from 22 ± 6 nm to 8 ± 3 nm. The EDX profile in (a) shows that Ag was present within the interlayer, as indicated by the contrast difference between the center and the edge of the interlayer. FIG. 2. Schematic diagram showing the mechanism of contact formation [(a-i)–(a-iv)] without Ag acetate and (b-i)–(b-v)] with Ag acetate. Without Ag acetate, (i) Ag paste after printing (ii) thermoplastically formed Al MG frit sandwiched between Ag powders and Si emitter. In the area indicated by a black box in Fig. ??? , this (iii) Al MG frit interdiffused with Ag powder on one side, and with Si emitter on the other side, thus forming Ag/Al and Si/Al solid solutions below the eutectic temperature. (iv) The Ag–A1 or Al–Si eutectic melting was followed by cooling, leading to the formation of a Ag–Al alloy on the surface of the Ag powders and a thick interlayer on Si emitter. With Ag acetate, (i) after printing the Ag paste, Ag acetate was randomly distributed in paste, and (ii) thermal decomposition of Ag acetate formed Ag nanocrystallites. The (iii) thermoplastically formed Al MG frit (iv) interdiffused with both the Ag powders and the Si emitter covered in Ag nanocrystallites below the eutectic temperature. Fig. ??? is an enlarged image of the area indicated by a black box in Fig. ??? . (v) Ag–Al and Al(Ag)–Si eutectic melting was followed by cooling, leading to the formation of a Ag–Al alloy and a thinner interlayer in comparison to the one formed without Ag acetate. (ii-1) SEM image showing Ag nanocrystallites formed from thermally decomposed Ag acetate on the surface of powders and Si emitters. The image was obtained after Ag paste was annealed up to 300 °C. FIG. 3. EELS profiles of Al, Al 2 O 3 , and interlayer formed with Ag acetate showing the presence of Al as Al 2 O 3 in the interlayer. FIG. 4. I–V curve and summary of cell parameters correspond to the interdigitated back contact cell fabricated. A photovoltaic conversion efficiency of 20.3% was achieved at 1-sun concentration and AM 1.5G. The area and thickness of the cell were 154.8 cm 2 and 130 μ m, respectively.
All Science Journal Classification (ASJC) codes
- Physics and Astronomy (miscellaneous)