WEBVTT 00:00:00.880 --> 00:00:05.360 This is the loss analysis of that champion Topcon cell. 00:00:06.960 --> 00:00:11.016 On the right hand side, you can see a pretty busy ... 00:00:11.016 --> 00:00:13.720 graph of the different losses. 00:00:13.720 --> 00:00:15.655 I don't expect you to read them all, 00:00:15.655 --> 00:00:18.926 but I'm just sort of pointing out here that we're able to do a ... 00:00:18.926 --> 00:00:22.397 pretty detailed understanding of where the different losses ... 00:00:22.397 --> 00:00:23.598 occur in this device. 00:00:24.400 --> 00:00:27.113 The brown ones are optical losses, 00:00:27.113 --> 00:00:30.294 the blue ones are recombination losses, 00:00:30.294 --> 00:00:34.598 and then the green and pinky ones are resistive losses. 00:00:36.200 --> 00:00:38.740 In order to determine the optical losses, 00:00:38.740 --> 00:00:41.715 we analyze the quantum efficiency and apply a ray ... 00:00:41.715 --> 00:00:44.763 tracing model using SunSolve from PV Lighthouse, 00:00:44.763 --> 00:00:47.158 which is a very nice tool for doing that. 00:00:47.480 --> 00:00:50.449 We found in this case we have a path length ... 00:00:50.449 --> 00:00:54.466 enhancement factor of 32, which is rather typical for a ... 00:00:54.466 --> 00:00:56.038 random pyramid cell. 00:00:56.360 --> 00:00:59.928 But for those of you who saw Eli Yablonovich's talk ... 00:00:59.928 --> 00:01:02.476 yesterday, it's a fair bit below the ... 00:01:02.476 --> 00:01:05.449 Lambertian limit of the 4n squared limit, 00:01:05.449 --> 00:01:07.318 which is about z equals 50. 00:01:10.840 --> 00:01:13.344 So for the recombination and resistive losses, 00:01:13.344 --> 00:01:16.599 we did that through three dimensional simulations in Quokka. 00:01:17.080 --> 00:01:20.518 We took measurements and best guesses for values of the ... 00:01:20.518 --> 00:01:23.134 J0 contact resistance, line resistance, 00:01:23.134 --> 00:01:27.319 all of those things you need to know in order to build a Quokka model. 00:01:28.440 --> 00:01:31.992 This cell ended up having an area average J0 of two and a ... 00:01:31.992 --> 00:01:34.439 half femtoamps, which is pretty low. 00:01:35.960 --> 00:01:40.122 And what we found actually in this modeling is that you don't ... 00:01:40.122 --> 00:01:45.199 necessarily need to model all of the different J0 values geometrically. 00:01:45.200 --> 00:01:48.208 You can just take this area average value and that will ... 00:01:48.208 --> 00:01:50.079 predict the Voc very accurately. 00:01:50.400 --> 00:01:52.906 Because the diffusion length is so long in these devices ... 00:01:52.906 --> 00:01:55.732 that all the different contacts are in electrical contact with ... 00:01:55.732 --> 00:01:56.638 one another anyway. 00:01:57.040 --> 00:02:00.825 So we can think of this area average J0 as a really good ... 00:02:00.825 --> 00:02:03.319 proxy for surface recombination. 00:02:03.320 --> 00:02:07.481 For the total device, we found that the simulation ... 00:02:07.481 --> 00:02:09.759 matched the measurements. 00:02:09.760 --> 00:02:11.260 Well, that's not surprising, 00:02:11.260 --> 00:02:12.760 I guess, because if it didn't, 00:02:12.760 --> 00:02:15.520 I would have asked our postdoc to go back and do it again. 00:02:16.960 --> 00:02:21.514 And what I'd like to point out here is the importance of the ... 00:02:21.514 --> 00:02:25.119 Auger recombination, which you can see here. 00:02:25.120 --> 00:02:27.301 This is the sort of medium blue, 00:02:27.301 --> 00:02:28.559 0.62 milliwatts. 00:02:29.280 --> 00:02:31.120 That's an intrinsic loss, of course. 00:02:31.600 --> 00:02:35.624 And if you're wondering why the total efficiency is up ... 00:02:35.624 --> 00:02:38.396 around 30%, you have to subtract that ... 00:02:38.396 --> 00:02:41.704 auger loss to get you back down around 29.5%, 00:02:41.704 --> 00:02:46.264 which is what we're familiar with as the efficiency limit for ... 00:02:46.264 --> 00:02:48.678 a single junction silicon cell. 00:02:52.840 --> 00:02:54.624 Okay, so let's work our way through ... 00:02:54.624 --> 00:02:56.039 some of the big losses here. 00:02:56.040 --> 00:02:58.140 So in terms of the optical losses, 00:02:58.140 --> 00:03:00.840 the biggest single one is infrared escape. 00:03:01.320 --> 00:03:03.868 And that is because our light trapping is not at the ... 00:03:03.868 --> 00:03:04.839 Lambertian limit. 00:03:05.560 --> 00:03:08.840 We also have a bit of absorption in the polysilicon layer. 00:03:08.840 --> 00:03:10.960 That's the parasitic absorption component. 00:03:10.960 --> 00:03:13.615 It's only small, but it's something we can do, 00:03:13.615 --> 00:03:16.759 something we can work on and actually hope to improve. 00:03:18.200 --> 00:03:21.960 The front finger reflection is still a large loss, 00:03:21.960 --> 00:03:23.840 that's 0.44 milliwatts. 00:03:23.840 --> 00:03:26.600 And of course we can go for thinner fingers to help reduce that. 00:03:27.800 --> 00:03:29.634 In terms of recombination, as I said, 00:03:29.634 --> 00:03:32.119 most of the recombination in this device is auger. 00:03:32.920 --> 00:03:35.222 Surface recombination is still present, 00:03:35.222 --> 00:03:37.919 but it's pretty small compared to the other ones. 00:03:37.920 --> 00:03:39.923 We can do a little better on the front surface, 00:03:39.923 --> 00:03:41.399 but there's not a lot to gain here. 00:03:42.600 --> 00:03:46.452 And resistive losses, we do have some significant ... 00:03:46.452 --> 00:03:47.919 transport losses. 00:03:47.920 --> 00:03:50.980 That's the two purple colored ones here, 00:03:50.980 --> 00:03:54.040 hole transport and electron transport. 00:03:54.600 --> 00:03:57.739 That is the resistive losses caused by carriers having to ... 00:03:57.739 --> 00:03:59.439 make their way to the contacts. 00:03:59.880 --> 00:04:03.292 And it's not surprising when you think that the contacts are ... 00:04:03.292 --> 00:04:07.319 separated by more than five times the cell thickness in these devices. 00:04:07.320 --> 00:04:10.155 So they've got a long way to travel and they're not getting a ... 00:04:10.155 --> 00:04:12.875 lot of help from the films on the surfaces or the diffused ... 00:04:12.875 --> 00:04:15.479 layers on the surface because they're lightly doped. 00:04:17.160 --> 00:04:19.720 So we need to go to closer contacts to solve that problem. 00:04:21.320 --> 00:04:23.430 Okay, so we've got a good model of ... 00:04:23.430 --> 00:04:24.759 that champion cell. 00:04:25.280 --> 00:04:27.970 Let's pick out some of the low hanging fruit and make ... 00:04:27.970 --> 00:04:30.911 realistic assumptions for how much better we could make ... 00:04:30.911 --> 00:04:33.038 those things and see where that takes us. 00:04:34.000 --> 00:04:36.640 We've improved the front surface passivation a bit. 00:04:36.800 --> 00:04:40.560 Got it down to an area average J0 of two femtoamps. 00:04:40.720 --> 00:04:42.400 Gives us a little bump in the voltage. 00:04:43.360 --> 00:04:46.086 We've gone to a high resisitivty wafer because as ... 00:04:46.086 --> 00:04:48.945 you'll see in a moment, that gives us the best chance ... 00:04:48.945 --> 00:04:52.070 to get a high voltage and a high fill factor and that gives ... 00:04:52.070 --> 00:04:53.399 us some benefits there. 00:04:53.960 --> 00:04:56.787 We've reduced the front finger width by about 30%, 00:04:56.787 --> 00:04:59.479 which we think is achievable in the near future. 00:05:00.840 --> 00:05:04.020 We've thinned the poly a bit on the rear to reduce some of ... 00:05:04.020 --> 00:05:07.408 those parasitic losses and we've reoptimised the contact ... 00:05:07.408 --> 00:05:10.519 geometry and with that we can get a little bit over 27%. 00:05:10.840 --> 00:05:15.320 And I think that's quite plausible within the next 12 to 18 months.