E.H.A. Granneman and V.I. Kuznetsov. Levitech BV, Versterkerstraat 10, 1322AP Almere, The Netherlands

Transcription

1 Spatial ALD, Deposition of Al 2 O 3 Films at Throughputs Exceeding 3 Wafers per Hour E.H.A. Granneman and V.I. Kuznetsov Levitech BV, Versterkerstraat 1, 1322AP Almere, The Netherlands Al 2 O 3 is the first ALD film being introduced in the solar cell (PV) industry. The primary driver for implementation is two-fold: excellent surface passivation and low cost-of-ownership. Spatial ALD systems satisfy both criteria. On Fz and Cz-type wafers minority carrier lifetimes in the range of 5-8 and.5-1ms, respectively, are obtained. With this type of film, PERC-type cells with efficiencies > 2% were realized. Further, such cells are processed with throughputs > 3 wafers/hr. The thermal stability of the film is improved by selecting thin films (< 6nm), and by carrying out a post-deposition anneal at 6 C. This anneal drives out excess hydrogen that is incorporated during deposition, thereby avoiding the formation of blisters upon exposure to subsequent high-temperature ( firing ) steps. Introduction During the last few years various types of ALD films have successfully been introduced in high-volume manufacturing in a number of industries. Particularly in the semiconductor (IC) industry the implementation is wide spread (high-k gate stacks, capacitor dielectrics, diffusion barriers, etc.). More recently, a specific type of ALD film, Al2O3, was widely researched in the PV industry as a potential surface passivation layer for solar cell surfaces (1-5). Thin (< 1nm) Al 2 O 3 films are able to passivate p-type surfaces very effectively. On the one hand intrinsic negative charges in the dielectric film repel charge carriers, while on the other hand hydrogen that is present in the ALD films passivates dangling bonds at the Si/SiO 2 / Al 2 O 3 interface. The combination of both effects reduces the charge recombination losses considerably. The overall result is an increase in cell efficiency of.5-1% (absolute) (6). Following extensive testing of these films, the expectation is that large scale implementation in PV manufacturing will take place in the near future. In this paper we describe the film and deposition requirements, that are specific for the PV industry, and we make a comparison with the IC Industry. This is followed by a brief description of the Spatial ALD process in general and of the Levitrack system in particular. Cost aspects of this approach relative to system design and precursors are discussed as well. After that, the basic material and electrical properties of blanket ALD Al 2 O 3 films are discussed. Special attention is paid to the formation of defects. The socalled blisters are a consequence of the accumulation of hydrogen at the Si/ Al 2 O 3 interface and are formed upon exposure to high temperatures. Finally, the performance of spatial ALD films in solar cells is summarized.

2 Application of ALD films in the industry Al 2 O 3 is currently the only type of ALD film being introduced in manufacturing of solar cells. The main reason is the excellent surface passivation capabilities. A secondary reason is that this is probably the only ALD film that can be deposited at a cost that is acceptable for the PV industry. To make the cost issue more clear, table I shows a comparison between film and equipment requirements in the IC and PV industries. TABLE I. Comparison typical film and equipment requirements Semiconductor and PV industries. Paramater Semiconductor Industry PV Industry # Process Steps per device Film Uniformity (1sigma %) < 1 < 4 Particle generation Very Important Irrelevant Metal (Fe) contamination (cm-2) < 1E1 < 1E12 CoO deposition process ($/wfr) Equipment Capital Cost (M$) Equipment Throughput (wfrs/hr) Equipment Uptime (%) > 95% >95% Wafer Breakage 1:5, 1:1 From table I it is clear that film specs such as defects, contamination and uniformity are much less stringent in the PV industry compared with the Semiconductor Industry. In contrast with that, equipment productivity and film cost are much more critical in PV. There are several approaches to deposit Al 2 O 3 films with high throughput: ALD, PECVD, APCVD, and PVD. Of these, the first two techniques are being evaluated most. The advantage of PECVD is that the technology is well-known in PV as this technique has been used for several decades to deposit SiNx films (anti-reflection coatings). Consequently, it is a mature technology with a large installed base. ALD is completely new; it is practiced in two different types of systems, batch ALD and Spatial ALD. In the first approach 5-1 wafers are loaded back-to-back in a horizontal or vertical furnace tube, with three or four tubes operating in parallel. One of the main drawbacks of the temporal batch ALD approach is the long turn-around time, which is critical in an industry relying heavily on in-line, short-turn-around, production equipment. The remainder of the paper deals with Spatial ALD. Spatial ALD In spatial ALD, the role of time and place are inverted relative to conventional (temporal) ALD. In temporal ALD the substrate is kept at a fixed position while the precursor and inert gas supplies are pulsed. In that way, the substrate is exposed to a specific sequence of precursor exposures (precursor 1, inert gas, precursor 2, inert gas, etc.). In spatial ALD the precursors are flowing continuously but injected at spatially different positions. The substrate is exposed to the same sequence of precursors by moving it through space from one precursor injection area to the next. This technique is applied in the so-called Levitrack system (7-9). A schematic cross section of the Levitrack reactor is, shown in figure 1. Wafers are loaded in a track in

3 which zones with TMA and H 2 O injection zones are spatially separated by zones of inert gas. The distance between the wafers and the track wall is.15mm. When the gas flow is Gas Flow r = 78mm 15 µm TMA H 2 O TMA H 2 O Figure 1. Cross section of the Levitrack Spatial ALD reactor. The TMA (Tri-Methyl- Aluminum) and H 2 O vapors are mixed with N2 to a concentration of a few percent. above a well-defined level, the wafers will float, kept in position by strong forces above and below the wafer. The reactor is installed under an angle, which causes the floating wafers to move forward through gravity. The wafers pass ALD cells with fixed sequences of TMA and H 2 O, adding a thin layer of Al 2 O 3 with each passage through a cell. The system is designed such that the different precursors do not mix, independent of the presence of wafers. This implies that deposition only takes place on objects that pass the subsequent precursor zones, i.e. only on the wafers (on one side of the wafer) and not on reactor walls. As there is no belt carrying the wafers, no deposition takes place on wafer carriers. The system length defines the number of ALD cells. At a growth rate at 2 C of.12nm/cycle, 1nm Al 2 O 3 is deposited in each meter of active length. Figure 2. Layout of the Levitrack system. The wafers travel with a typical velocity of.2m/s, and each ALD cell has a length of 12cm. When wafers are injected with a pitch of 25cm, the throughput is 3 wafers/hr. In addition to a high throughput, this approach has several other advantages: as the process takes place under atmospheric pressure, no pumps or valves are required. Further, no moving parts are present in the deposition part of the track resulting in excellent reliability and reduced maintenance. A schematic layout of the system is shown in figure 2. It is made up of segments of 1m that can be switched on/off individually. A common precursor evaporation module supplies vapor to all segments.

4 It is particularly important to be able to stop the introduction of new wafers when anomalies such as wafer breakage, or unstable transport occurs. This is done as follows: P High pressure Low pressure in between wafers P H.Figure 3. Tracking of wafer position by means of an array of pressure sensors. Pressure sensors are located in various positions along the track, see figure 3. When a wafer is present underneath a pressure sensor, the gas flowing from the point of injection to the exhaust (see cross section left in figure 1) is slightly obstructed, resulting in an increased pressure. Consequently, whenever a wafer passes a sensor, a pressure peak is observed. This results in plots as shown in figure 4. It is worth mentioning that in a track with large gas flows entering above and below the wafer, wafers tend to repel each other. This is caused by the fact that the pressure in between wafers increases as wafers approach each other. Consequently, wafers try to move away from each other as far as possible. This is clearly illustrated by the peaks on the right end of the pressure traces: the last wafers are lagging behind more and more moving from ALD segment 1 to 6. Heat-up segment ALD segment #1 ALD #2 Individual wafers Cool-down Time ALD #3 ALD #4 #5 #6 Figure 4. Pressure/time plot registered by sensors present in each of the ALD segments of a system with six ALD segments, during transport of 1 wafers. One pressure sensor per ALD segment is shown. Each peak corresponds with a specific wafer. The shift in time between peaks in adjacent ALD segments corresponds with a wafer velocity of ~.2m/s. The pressure sensor read-outs are used to monitor the progress of wafer transport during processing of batches of wafers. When something irregular happens, the operator is alerted immediately. An example of incorrect wafer transport is shown in figure 5. To demonstrate production-worthiness several large scale marathons were done. The current status is that throughputs of 3 wafers/hr are realized routinely, and with some more complex automation 36 wafers/hr can be reached.

5 Irregular wafer transport ALD #6 Cool-down Time Wafer stops Control system stops wafer input ALD #1 ALD #2 ALD #3 ALD #4 ALD #5 Wafers stagnate underneath sensor Figure 5. Example of pressure/time plots in case of intentionally obstructed wafer transport. A continuous high pressure is indicative for a stagnant wafer underneath a sensor, while a continuous low pressure indicates an area where no more wafers pass. At the onset of stagnation, the sensors show erratic wafer motion with some wafers approaching and some others repelling each other. Cost ALD precursors To give some insight in the relative contribution of the cost of TMA in the overall Cost-of-Ownership, the following comments can be made: the cost of TMA depends strongly on its purity. In e.g. the IC Industry the highest purity (electronic grade, %) is used which costs ~7$/gram. As this is too expensive for PV applications a less pure TMA is being used. This so-called Solar-grade material which is typically 99% pure (1), and costs < 1$/gram. It turns out that the latter quality material appears to result in good sufficient (11). Of course, the cost of TMA should be seen in relation with the overall costs. When we consider a state of the art, 6MWp, PV production line with an output of 2M solar cells in which one 6nm, 2M$, 24 wafers/hr, ALD system is operational that is depreciated over 7 years, we arrive at a contribution of the depreciation to the total CoO of.143 $/wafer. This is to be compared with a TMA cost of.31 $/wafer, i.e. roughly 2% of the depreciation of the system, and 1% of the total Cost-of-Ownership. Material characterization ALD Al 2 O 3 film In figure 6 a uniformity plot is shown for a wafer that was processed in one pass through a 6nm system (i.e. 6 ALD segments). The uniformity on the deposition side of the wafer is 2.5% (1σ). The structure that can be observed in the plot is indicative for the lines of injection points through which the precursor are introduced. The plot on the right demonstrates that there is no deposition on the non-deposition side of the wafer. Upon a closer examination some wrap-around of deposition onto the backside can be observed at the wafer edge (~ 1mm). It is worthwhile to make a comment on the importance of uniformity. It is wellknown that ALD is able to deposit films with very good uniformity. Typical uniformities realized in the IC Industry is < 1%, see table I. One way to get a good uniformity is to over saturate the precursor supply. In this way, non-uniformities resulting from local

6 Figure 6. Al 2 O 3 film thickness on the deposition side and on the non-deposition side of the wafer (left and right picture, respectively). The thin native SiO 2 film that is always present on the wafer ( 1nm) cannot be distinguished from the Al 2 O 3 film. The measurement is done with an ellipsometer on polished wafers, with 16 points per wafer. depletion of precursors are reduced. However, this requires the injection of higher precursor flows. In the PV industry a uniformity of ~4% is sufficient to obtain a good passivation of the surface. Consequently, in order to avoid waist of costly material, the precursor concentration is reduced to the point where an acceptable uniformity of ~ 4% is combined with a minimum amount of precursor usage. Ideally, the tool operates at the onset of depletion. Other film characteristics are summarized in table II: TABLE II. Characteristics Spatial ALD ALD Al2O3 film Parameter Number Uniformity, within-wafer (1σ) < 3% Uniformity, wafer-to-wafer (1σ) Refractive Index < 1% 1.65 Density (g/cm3), RBS/XRR 2.6 / 2.85 Al/O ratio.76 C content (At.%) < 1 % Cl content (At.%) <.2 % H content (At.%), as deposited 1 ± 3 H content (At.%), 6 C anneal 3 ± 1 Note: Deposition at 2 C; all concentrations measured with SIMS Electrical characterization ALD Al 2 O 3 film Lifetime. The lifetime of minority carriers in solar cells is a combination of bulk and surface recombination lifetimes. In order to isolate the surface passivation capabilities of Al 2 O 3, usually high-resistivity Fz Si wafers are used, having a near perfect bulk lifetime. Consequently, when Al 2 O 3 is deposited on both sides of such a wafer, one basically measures the surface recombination lifetime. An example of a sample processed in the Levitrack is shown in figure 7 (12). The carrier lifetime is in the range of 7-8ms, a high value, that should be compared with values <.1ms of uncoated samples. One of the main issues with Al 2 O 3 films is the high-temperature stability. When samples are heated at temperatures higher than the optimum PDA at 4-45 C, the minority carrier lifetime deteriorates. In almost all solar cell process flows, one of the last steps is the firing of metal contacts at temperatures > 8 C.

7 Figure 7. The minority carrier lifetime of double-side Al 2 O 3 -coated, high-resistivity Fz wafers as a function of carrier density. A 4 C Post-Deposition Anneal (PDA) was applied. Effective lifetime (µs) min anneal fired nm (ref. TU/e) Film thickness (nm) Figure 8. Minority carrier lifetime of double-side coated Fz samples as a function of ALD Al 2 O 3 film thickness for samples that were either annealed at the optimum temperature of 4 C, or in an industrial firing furnace set at 85 (13) As can be seen in figure 8, this causes a dramatic reduction in the surface recombination lifetime. Although values of.5ms are currently still acceptable, a lot of research was done to investigate this phenomenon and to improve the high-temperature stability. As yet, it is not fully understood what the physical/chemical mechanisms are that cause this reduction in lifetime (see also next paragraph). One observation that can be made is that the differences in lifetime that are observed initially, all basically disappear upon firing, for thicknesses in the range 5-15nm. Defects (blisters). Although this has not been completely resolved yet, one of the reasons for the deterioration of lifetime upon exposure of the film to high temperatures is the occurrence of macroscopic defects, commonly referred to as blisters. At higher temperatures hemispherical macroscopic features show up that are filled with gas (14), see figure 9. These features are found preferentially, but not exclusively, at grain boundaries. Usually they look like almost perfect hemispherical domes with, depending

8 on the thickness and process history of the films, diameters in the range 1-25µm, i.e. much larger than the film thickness. When the Al2O3 films are coated with a PECVD SiNx film the formation of blisters is more pronounced. Figure 9. Upon exposure of the Al2O3 films, hemispherical features (blisters) show up. Hennen et all (15) carried out extensive experiments on the formation of these defects, resulting in the following observations: The diameter of the blisters increases with Al2O3 film thickness, see figure 1 The number of blisters increases at higher annealing temperatures 6 PDA 8 C, 3s Blister density (cm-2) Al2O3 film thickness (nm) Figure 1. The appearance of blisters as a function of Al2O3 film thickness. No SiNx film is present. All post-deposition anneals were executed at 8 C, 3s. Blisters show up predominantly at crystal grain boundaries and at areas where wafers are damaged (i.e. scratches). Overall, the formation of blisters depends strongly on the surface topology and structure For Al2O3 films thinner than ~1nm, not covered with a SiNx film, the density of blisters is very small, even at high annealing temperatures, see figure 1. In case of films coated with SiNx, blisters are observed even at thinner Al2O3 films Upon exposure to identical (high-temperature) thermal budgets, the combined volume of all blisters is more or less constant, irrespective of their dimensions. Upon annealing above ~5 C, the concentration of hydrogen in the films decreases from an initial value of ~ 1 to ~2-3 At%, see figure 11.

9 Figure 11. SIMS measurement of the hydrogen concentration in 24nm ALD Al 2 O 3 films, upon annealing at various temperatures. The narrow peak at the surface and the broad one at depth ~1. are SIMS artifacts. These observations result in the following physical picture (15): At the deposition temperature of 2 C, a large concentration of hydrogen is incorporated in the Al 2 O 3 film, probably mostly in the form of OH bonds Diffusion of hydrogen from the film outwards does take place, but is very slow. Out-diffusion is more effective when films are thin. Al 2 O 3 is basically a good diffusion barrier for such out-diffusion. When a SiN x film is present on top of the Al 2 O 3 film, out-diffusion is (further) hindered The hydrogen is extremely mobile in the very thin SiO 2 film that is always present in between the Si substrate and the Al 2 O 3 film. The hydrogen accumulates at blister nucleation sites such as grain boundaries and scratches. When the Al 2 O 3 films are thicker, blisters have more time to coalesce, resulting in fewer but larger blisters. As the total amount of hydrogen in the film is more or less constant, and outdiffusion limited, the combined blister volume is also constant. From these observations the following strategy to reduce blisters was derived: at annealing temperatures in the range C hydrogen out-diffuses at a reasonable rate, while blister formation is still limited. In thinner Al 2 O 3 films out-diffusion of hydrogen competes favorably relative to blister formation. Therefore, thin films ( 6nm) are further optimized for PV surface passivation applications. Figure 12 shows results from this approach. Samples with 6nm Al 2 O 3 films on both sides are coated with a 1nm SiN x film; in half of the cases a 6 C, 2min. anneal is carried out prior to SiN x deposition. In the other half a SiN x film is deposited on as-deposited Al 2 O 3 samples. Both types of samples are subjected to firing anneals at 83, 85 and 87 C. These graphs confirm the observations mentioned above: the electrical parameters (lifetime, implied open-circuit voltage Voc) degrade upon exposure to higher firing temperatures. However, it is also clear that there is a substantial improvement when a PDA is applied. Consequently, a PDA is now used in all solar-cell related activities. To avoid this additional process step, customers often incorporate the PDA in the subsequent PECVD SiN x system as a pretreatment prior to the deposition of the SiN x film.

10 Lifetime (µs) no PDA PDA Implied V oc (mv) no PDA PDA 83 C 85 C 87 C C 85 C 87 C Firing Temperature Firing Temperature Figure 12. Lifetime and implied Voc for 6nm Al 2 O 3 films fired at various temperatures. All Al 2 O 3 samples were coated with a 1nm SiN x film. Half of the samples received a 6 C, 2min. post-deposition anneal prior to SiN x deposition. Performance ALD Al 2 O 3 films in Solar Cells Currently, so-called PERC cells with Al 2 O 3 back-side surface passivation films are being introduced in the PV industry (16). A typical layout of a PERC cell is shown in figure 13. The backside of the PERC cell is prepared as follows: after texturization of front and back side of the wafers, dopant diffusion and SiN x Anti-Reflective coating deposition are carried out, followed by a back-side polishing of the back-side. Subsequently the Al 2 O 3 and SiN x capping layer are deposited on the back side. With a fast pulsed laser the Al 2 O 3 /SiN x stack is removed locally (17). After that, Al metallization paste is screenprinted and fired. The Al paste penetrates through the laser-ablated holes in the dielectric stack, and reacts with the Si substrate to form the local Al-BSF contact (Back-Surface Field). As it turns out, the step coverage and the surface-passivation capabilities of the ALD Al 2 O 3 film are so good that the back-side polishing step can be omitted. Ag line SiO 2 / SiN x stack Texturized surface n + P-type Si p + Al ALD Al 2 O 3 PECVD SiN x capping Al Contact Hole or line Figure 13. Application of Al 2 O 3 films in back-side passivated PERC cells This approach results in a considerable cost saving, and is currently being looked at seriously. The implementation of Al 2 O 3 films in this type of device has been extensively tested with positive results, and is currently transferred to high-volume manufacturing As is to be expected in the development of advanced devices, a considerable part of the implementation of new technology is consumed by the integration of the various

11 types of processes and films in specific device sub-modules. In the case of back-side surface passivation of Al 2 O 3 this involves the pre-cleaning step prior to the ALD process, the ALD process itself, the SiN x capping layer deposition step, the subsequent laser ablation step, and the paste firing process. In figure 14 the optimization of the SiN x deposition process is shown. The result of this optimization step is both a gradual increase of the efficiency of the cell (left picture) and a narrower efficiency distribution. In this particular example, the efficiency that was realized is > 2%. One of the main contributors of the efficiency increase is the gradual improvement of the short-circuit current J sc (right figure). The current status for industrial mono-crystalline Si wafers is an efficiency of 2.2% average. For multi-crystalline wafers this number is 18.8%. Optimization cycles ALD Al 2 O 3 PECVD SiN x Optimization cycles ALD Al 2 O 3 PECVD SiN x Figure 14. The efficiency N cell and short-circuit current J sc of p-type mono-crystalline PERC cells in which Levitrack Al 2 O 3 films are used. The SiN x deposition conditions are optimized. Cell Efficiency (%) # Backside metal lines Figure 15. Efficiency as a function of the number of metal lines on the wafer backside. A higher number of metal lines reduces the series resistance of the device. Another example of PERC process optimization is shown in figure 15. In this case the laser opening step and number of metal contacts are optimized. Again, efficiencies above 2% are demonstrated.

12 Conclusions The following conclusions can be drawn: It was demonstrated that ALD processes with very high throughputs can be realized through a fundamental change in concept: spatial ALD instead of the conventional temporal ALD. Al 2 O 3 Films with a thickness 6nm are sufficient for a good surface passivation The Levitrack is a reliable spatial ALD system with excellent repeatability, and many features to track the transport of wafers, and monitor system performance. Throughputs exceeding 3 wafers/hr are realized in combination of a low Costof-Ownership < 5$ct /wafer are feasible. Within-wafer and wafer-to-wafer uniformities are typically 2.5 and 1% (1σ), respectively Al 2 O 3 high-temperature stability can be improved considerably through a postdeposition anneal (prior to SiNx deposition). P-type PERC cell efficiencies exceeding 2.2% (average) are realized in highvolume manufacturing Acknowledgments The project is executed with subsidy from the Ministery of Economic Affairs, Agriculture and Innovation, programs EOS: Long-Term and TKI (Eco-PV), both executed by Agentschap NL References 1. B. Hoex, J.J.H. Gielis, M.C.M. van de Sanden and W.M.M. Kessels, J.Appl. Phys 14, (28) 2. J.Schmidt, A. Merkle, R. Brendl, B. Hoex, M.C.M. van de Sande and W.M.M. Kessels, Prog. Photovoltaics 16, 461 (28) 3. B. Vermang, H. Goverle, L. Tous, A. Lorenz, P. Choulat, J. Horzel, J. John, J. Poortmans and R. Mertens, Prog. Photovolt 2, 269 (212) 4. A. Richter, J. Benick and M. Hermle, IEEE J. PhotoVoltaics 3 (1), 236 (213) 5. G. Dingemans, R. Seguin, P. Engelhart, F. Einsele, B. Hoex, M.C.M. van de Sanden, W. M. M. Kessels, J. Appl. Phys 16, (29) 6. Efficiency > 2%, paper 5 Z. Xia, J. Dong, Y. Gao, X. Li, H. Qiao, R. Sidhu, L. Tao, K. Teng, Z. Yang, B. Zhang, G. Xing, Proc. Proc. 28th Eur. PVSEC, 3 September - 4 October 213, Paris, p E.H.A. Granneman, P. Vermont, V. Kuznetsov, M. Coolen, K. Vanormelingen, Proc. 25 th Eur. PVSEC, Valencia, September 6-9, 21, p P.G. Vermont, V. Kuznetsov, E.H.A. Granneman, Proc. 26 th Eur. PVSEC, Hamburg, September 5-8, 211, p B. Vermang, H. Goverde, A. Rothschild, J. John, J. Poortmans, R. Mertens, P. Vermont, K. Vanormelingen and E.H.A. Granneman, Proc. 26 th Eur. PVSEC, Hamburg, September 5-8, 211, p See e.g. precursor brochures Air Liquide ( and AkzoNobel ( 11. G. Dingemans and W. M. M. Kessels, Proc. 25th Eur. PVSEC, September 6-1, 21, p. 183.

13 12. Courtesy PhotoVoltec, Belgium 13. Courtesy G. Dingemans and W. M. M. Kessels Technical University Eindhoven 14. B. Vermang, H. Goverde, A.Lorenz, A.Uruena, J. Das, P. Choulat, E. Cornagliotti, A. Rothschild, J. John, J. Poortmans and R. Mertens, Proc. 26 th Eur. PVSEC, Hamburg, September 5-8, 211, p L. Hennen, E.H.A. Granneman and W.M.M. Kessels, Proc. 38 th IEEE-PVSC Austin, June 3-8, S.Wang, T. Xu, F. Ye, Z. Li, Y. Yang, W. Deng, C. Zhang, Z. Feng, P. Verlinden, Q. Huang, Proc. 27 th Eur. PVSEC, Hamburg, September 24-28, 212, p P. Jeffrenou, A. Uruena, J. Das, J. Penaud, M. Moors, A. Rothschild, B. Lombardet, J. Szlufcik. Proc. 26 th Eur. PVSEC, Hamburg, September 5-8, 211, p 218