Characterization of graded CIGS solar cells

Available online at www.sciencedirect.com

Energy Procedia

Energy Procedia 00 (2009) 000–000

Energy Procedia 2 (2010) 49–54

www.elsevier.com/locate/procedia

www.elsevier.com/locate/procedia

E-MRS Spring Meeting 2009, symposium B

Characterization of graded CIGS solar cells Koen Decocka,*, Johan Lauwaerta,b, Marc Burgelmana a

Department of Electronics and Information Systems (ELIS), University of Gent, St-Pietersnieuwstraat 41, B-9000 Gent, Belgium b Department of Solid State Sciences, University of Gent, Krijgslaan 281-S1, B-9000 Gent, Belgium.

Received 1 June 2009; received in revised form 1 December 2009; accepted 20 December 2009 Abstract The CIGS material system has an adjustable bandgap. A ‘grading’ of the bandgap through the thickness of absorber can be used to improve cell performance. Some modern CIGS-solar cells already have a graded bandgap profile; for this work, such cells were provided by AVANCIS. We have set up a numerical model with the SCAPS program, based on a variety of cell measurements. This model then was applied to analyze the true benefit of bandgap grading. Grading appears to have only a limited effect on cell performance. Not only the effectively graded parameters, but also non graded parameters have an influence on the net benefit of grading. © B.V. c 2009

2010Published PublishedbybyElsevier Elsevier Ltd Keywords: CIGS; graded bandgap; SCAPS.

1. Introduction The CIGS material system has a bandgap which is dependent on the exact composition of the material. This property can be used to engineer the bandgap of the absorber layer throughout a solar cell and in this way to make a cell with a ‘graded’ bandgap. Two effects then come into play. First of all, locally increasing the bandgap reduces recombination in that part of the cell, but it also deteriorates the absorption. Secondly by changing the conduction band level, additional electric fields can be built in in the cell. The in-depth variation of the bandgap can have several benefits; a good overview is given by Lundberg et al. [1]. Two different approaches can be considered: changing the bandgap towards the buffer ((i), ‘front grading’) and changing the bandgap towards the back of the absorber ((ii), ‘back grading’); of course, both can be combined (‘double grading’). (i) For modern CIGS devices the dominating part of the recombination occurs in the space charge region (SCR). The main goal of front grading is trying to reduce recombination here by locally increasing the bandgap and this way trying to increase the Voc. Unfortunately this technique introduces a counteracting electric field (if the bandgap shift occurs in the conduction band (CB)) and shifts the absorption somewhat deeper in the cell, further away from the SCR, where the collection probability is lower. (ii) Alternatively, or additionally, one can increase the bandgap towards the back of the

* Corresponding author. Tel.: +32 9 264 8953; fax: +32 9 264 3594. E-mail address: [email protected].

c 2010 Published by Elsevier Ltd 1876-6102 doi:10.1016/j.egypro.2010.07.009

50

K. Decock et al. / Energy Procedia 2 (2010) 49–54 Author name / Energy Procedia 00 (2010) 000–000

absorber. The main goal here is to introduce a field which pushes the minority carriers towards the SCR. This way, carrier collection is increased, and the influence of the back contact, which has a rather high recombination velocity, is reduced. Several CIGS-based solar cells with a graded bandgap have already been produced. Some modern production methods automatically lead to a graded bandgap profile [2]. The net effect of grading however is difficult to identify and dependent on a lot of non-graded properties of the cell as well ([3], [4], [5]). Therefore we have investigated the influence of several parameters starting from a consistent model which is able to reproduce diverse measurements on a modern graded CIGS-solar cell. 2. Model We will analyze the effects of bandgap grading on a modern CIGS solar cell. We therefore performed various measurements on an AVANCIS sample which is produced with the laboratory line process. The AVANCIS solar cell contains both front S-grading and back Ga-grading [2]. C-V, C-f and I-V measurements were performed at different temperatures, next to spectral response and DLTS-measurements. A consistent model for this cell was set up in SCAPS, a software tool of the University of Gent available to the PV research community ([6], [7]). Version SCAPS 2.8 can handle graded cell structures [8]. The model consists of a 2.3 μm CIGS absorber layer, together with a 0.1 μm thick CdS buffer layer and a 0.2 μm thick ZnO window layer. The absorber has a gallium grading towards the back which is more or less step-like, and a sulphur grading towards the buffer which can be approximated by an exponential. Both incorporating extra gallium or sulphur leads to a bandgap widening. Incorporation of Ga leads to a raise in the conduction band (CB)-level without changing the valence band (VB)-level (common anion rule). Incorporation of S affects both the CB and VB, where the VB change is somewhat larger than the CB-change. The exact distribution between CB and VB effect is however complicated, not exactly known and moreover dependent on the Ga concentration [9]. Therefore we assumed in our model that the S-grading only results in a VB-lowering (common cation rule). The resulting band structure of the absorber is shown in figure 1. 1.1μm

0.7μm

0.5μm

Back

CIGSe bac k CISSe

Front

CISSe Graded

Buffer layer

Back contact

d grad

Figure 1: Schematic view of the band diagram of the absorber in the SCAPS model, based on SIMS data reported in [2]. Back Ga-grading is incorporated in the CIGSe back layer and is modelled as a conduction band step. Front S-grading is incorporated in the CISSe graded layer as an exponential decrease of the valence band. The CISSe graded layer is split in two parts (front and back) for the modelling of the defects.

In the model we assumed the background composition of the CISSe layer to be CuIn(S0.1Se0.9)2. The 0.5 μm wide front part of this layer has an exponential S-grading, that is: the S/(S+Se) ratio decays from 0.3 at the buffer to the background composition 0.1, with a characteristic decay length of 0.1μm (Lchar see formula 6 in [8]). The bandgap of CuIn(SxSe1-x)2 is calculated as Eg(eV) = 1.04(1-x)+1.53x-0.14x(1-x). At the back side of the cell there is a CIGSe layer with composition Cu(In0.6Ga0.4)Se2 with Eg = 1.26 eV. These bandgap data are consistent with the formula reported in [10] combining literature data of several authors. One of the objectives for using a graded structure is to rearrange the recombination throughout the cell. So it is important that our model is realistic for the defect distributions in the cell. We achieved this by setting up a model

51

K. Decock et al. / Energy Procedia 2 (2010) 49–54 Author name / Energy Procedia 00 (2010) 000–000

which can reproduce the apparent doping density for a wide temperature span (240 - 360 K) (figure 2). It could be emphasized that the defects included in this model have been observed by means of DLTS, not shown here. Additionally, the model is also able to reproduce the I-V measurements in dark (T = 240 – 360K) and in light conditions (figure 3), as well as the spectral response measurement. To be able to do this, we had to split the graded CISSe layer (figure 1) in two parts (‘front’ and ‘back’). An overview of the defect and doping properties is given in table 1. The effective electron diffusion lengths are respectively 0.56, 1.17 and 1.25 μm for the CISSe-layer and the back- and front-CISSe graded layer. 80 240K measurement 300K measurement 360K measurement 240K simulation 300K simulation 360K simulation

T=300K T=360K

60 40 2

J[mA/cm ]

10

-3

Apparent doping density [cm ]

17

16

10

20 dark

0 -20

T=240K

illuminated 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

-40 0.0

0.1

Width [μm]

0.2

0.3

0.4

0.5

0.6

V[V]

Figure 3: Comparison between modelling and measurement for the I-V Figure 2: Comparison between modelling and measurement for the measurements in dark (T = 240 – 360 K) and under illumination. apparent doping profile at different temperatures. The good agreement supports validity of the defect and shallow doping distribution used in the model.

Table 1: SCAPS parameters used to define doping and defect distribution in the front part of the absorber. Defect 2 has an exponential distribution in the back part of the CISSe graded layer. The doping and defect densities result from the CV-measurement shown in figure 3. Defect 1 is included in the absorber to avoid regions without recombination. The energy levels and capture cross sections of defect 2 and 3 result from DLTS-measurements with subsequent fine-tuning. More details about the parameters used are reported in [10].

Cell

CISSe 5 1015

CISSe graded back 5 1014

CISSe graded front 1016

Shallow doping concentration (acceptor) [cm-3] Defect 1(neutral) Capture cross section electrons [cm2] Energy level Concentration [cm-3] Diffusion length electrons [μm] Defect 2(acceptor) Capture cross section electrons [cm2] Energy level Concentration at the left side [cm-3] Concentration at the right side [cm-3] Characteristic length [μm] Diffusion length electrons [μm] Defect 3(acceptor) Capture cross section electrons [cm2] Energy level Concentration [cm-3] Diffusion length electrons [μm]

10-15 Ei 1015 5.6

10-15 Ei 1015 5.6

10-15 Ei 1015 5.6

10-14 Ev+0.4eV 1016 1016 # 0.56

10-15 Ev+0.4eV 2 1016 5 1015 0.05 1.2

10-15 Ev+0.4eV 5 1015 5 1015 # 2.5

# # # #

# # # #

10-15 Ev+0.55eV 1.3 1016 1.5

52

K. Decock et al. / Energy Procedia 2 (2010) 49–54 Author name / Energy Procedia 00 (2010) 000–000

3. Numerical investigation with SCAPS We performed several simulations with SCAPS in order to investigate the benefit of the grading on cell performance and to optimize the grading structure. 3.1. Front grading We investigated the influence of the magnitude, shape and depth of the front grading. In order to do this we varied respectively the S/(S+Se) content at the absorber-buffer interface, the characteristic length Lchar of the exponential describing the front grading, and the grading depth dgrad (see figure 1). Changing these parameters influences the average bandgap of the absorber layer. This acts upon the cell performance next to the net influence of the grading. As a consequence, if we want to identify the real influence of the grading, we should always compare with a cell layout which has a uniform (front part of the) absorber with the bandgap equal to the mean bandgap of the graded structure. In our analysis we compared with a uniform reference where the bandgap was averaged out over the front layers (CISSe + CISSe graded), leaving the back grading (CIGSe back-layer) unchanged. (i) In a first simulation we varied the amplitude of the grading together with its shape, keeping the dgrad constant at 0.5 μm. The amplitude was varied by changing S/(S+Se) at the buffer interface from 0.1 to 0.5 and the shape by changing the characteristic length from 0 to 0.3 μm. The net grading benefit is shown in figure 4 (top left). At the left side of the figure there is a uniform layer, to the right there is strong grading. From the bottom to the top of the figure the grading changes from a very steep frontgrading to a more or less triangular shaped one. The effect of grading seems to be rather disadvantageous, deteriorating efficiency up to 1.35 % absolute for triangular grading. There is a slight Voc-gain, but this is totally annealed by Jsc-loss and especially by a considerable FF-loss (up to 6% absolute). There is however a small benefit of grading (up to 0.05 % absolute), which appears for a rather moderate grading (S/(S+Se) = 0.2, Lchar = 0.1 μm), but the game doesn’t seem worth the candle. (ii) In a second simulation we varied the amplitude together with the depth of the grading, keeping the characteristic length of the grading constant at 0.1 μm. The amplitude was changed similarly as in the first simulation. The depth is changed by increasing dgrad and simultaneously decreasing the thickness of the CISSe layer, keeping the total thickness of both layers constant at 1.2 μm. The ratio of the thicknesses of the back and front CISSe graded layer are kept constant. Because changing the layer thicknesses also influences the defect composition, we performed 3 simulations varying the effective electron diffusion length (Ln) of the CISSe layer. In simulation A, Ln in the CISSe layer is smaller than in the CISSe graded layer (this is the same situation as in the first simulation setup). In simulation B, it is more or less the same and in simulation C, Ln is larger in the CISSe layer. The results are shown in figure 4. The grading benefit is in all three cases rather moderate or even negative. The net effect of grading is strongly dependent on the defect distribution. Whether or not grading is successful seems to depend on the following rule of thumb: “Grading reshapes the generation profile. If it can reshape the profile in such a way that more charge carriers are generated in regions with less recombination probability, grading is beneficent.” Again it seems to be the fill factor which determines the net efficiency gain. The Voc-gain and Jsc-loss cancel each other out. The largest grading benefits occur for moderate grading (S/(S+Se) § 0.25). The ideal grading depth depends then on the defect distribution. Adding a front grading to the absorber should thus be performed in a very cautious way, there are a lot of parameters influencing the final result, and often grading can even be harmful.

53

K. Decock et al. / Energy Procedia 2 (2010) 49–54 Author name / Energy Procedia 00 (2010) 000–000

'K[%] simulation A

'K[%]

0.30

0.70 -0.80

0.65

0.25

-0.60

0.20 -0.10

0.10

0

-1.00 -1.20

dgrad [Pm]

0.55

-0.20

0.15

0

0.05

-0.10 -0.20

0.50 -0.30

0.45

-0.40

0.40

0.03 0.05

0.05

0.10

0.60

-0.40

Lchar [Pm]

0.15

0.35 -0.50

0.30 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S/(S+Se) near buffer

S/(S+Se) near buffer

'K[%] simulation B

'K[%] simulation C

0.70

0.70 0.20

0.65

0.15

0

0.10

0.60

0.60

0.05 -0.05 -0.10

0.45

-0.15 0.15

0.40

0.50 0.45 0.40

0.10

0.35

0.35

0.05

0.30 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S/(S+Se) near buffer

0.05 0.10

0.55

0

0.50

dgrad [Pm]

0.55

dgrad [Pm]

0.65

0.05

0.15 0.25

0.10 0.15

0.20

0.20 0.30

0.30 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50

S/(S+Se) near buffer

Figure 4: Net grading contribution to the efficiency. For every simulation point the difference between the graded and corresponding uniform efficiency is given. This uniform efficiency is the efficiency of a cell where the CISSe-part of the absorber layer is uniform and has the same composition as the mean composition of the graded variant of the cell. (i) The top left figure gives the results for the analysis of the shape and the magnitude of the grading. From left to right there is a change from no to maximum grading. From the bottom to the top there is a change from very steep to almost triangular grading. (ii) The other three figures give the result for a variation in depth and magnitude of grading. From left to right again there is a change from no to maximum grading. From bottom to top, a change from shallow to thick grading. In simulation A, Ln in the CISSe is smaller than in the CISSe graded layer. In simulation B, Ln is comparable in both layers and in simulation C, Ln is larger in the CISSe layer.

3.2. Back grading We investigated the influence of the back grading on the cell performance, which special attention to the back surface recombination and possible beneficial fields. With respect to the back surface recombination, the back grading seems to be redundant. Even for rather large changes of the back grading, no effect on the cell-performance is noticed. Changing the thickness of the back grading for example, over a distance of 0.8μm only leads to an efficiency change of 0.06%. This is because the absorber layer is rather thick (2.3 μm) and thus almost all the generation takes place far enough from the back surface anyway. The shape and magnitude of the back grading has only a negligible effect on cell performance. Changing the grading towards the back only has some influence on the quantum efficiency of lower energy photons (as the cell is rather thick). This effect however is only dependent on the average width of the bandgap at the back, and not on the shape of the bandgap. So there is no electric field manifestly improving cell performance. The modelling concludes thus that the back grading in the investigated sample doesn’t improve cell performance significantly, but it isn’t harmful either.

54

K. Decock et al. / Energy Procedia 2 (2010) 49–54 Author name / Energy Procedia 00 (2010) 000–000

3.3. Concluding remarks Even though grading doesn’t seem to improve cell performance significantly, it does not mean grading cannot do this at all. Several authors ([4], [5]) mention efficiency gain up to 3%. This is because the effect of grading is dependent on a plethora of other parameters as well. 4. Conclusion The influence of a graded bandgap structure in a modern CIGS solar cell has been investigated. Therefore a realistic numerical model which is able to mimic several different measurements on a real device, under various circumstances, has been set-up. By means of this accurate model it could be shown that front grading can have some small (ǻȘ < 1%) beneficial effect. Unfortunately, it depends on a lot of parameters, the most important being the grading depth, the material composition at the buffer and the recombination properties in the entire absorber. Moreover, the parameters governing the effect of the grading are often not directly linked to the grading. As a guideline one could say front grading can improve the efficiency when it reshapes the generation profile in such a way that regions with smaller recombination are favoured. For the back grading it is concluded that no significant influences on the solar cell performance could be expected for cells with relatively thick absorber layers (i.e. thick enough so that most of the generation takes place far enough from the back contact).

Acknowledgement: We acknowledge the support of the Research Foundation – Flanders (K. D., FWO Research Assistant); of the European Integrated project ‘Athlet’. We thank AVANCIS (München, Germany) for providing CIGS cell samples.

References: O. Lundberg, M. Edoff, L. Stolt, The effect of Ga-grading in CIGS thin film solar cells, Thin Solid Films 480-481 (2005) 520. J. Palm, V. Probst, F.H. Karg, Second genereation CIS solar modules, Solar Energy 77 (2004) 757. M. Topiþ, F. Smole, J.Furlan, Band-gap engineering in CdS/Cu(In,Ga)Se2 solar cells, J. Appl. Phys. 79 (1996) 8537. M. Gloeckler, Device physics of Cu(In,Ga)Se2 thin-film solar cells, Ph.D. thesis, Colorado State University, Fort Collins, 2005. M.A. Contreras, J. Tuttle, A. Gabor, A. Tennant, K. Ramanathan, S. Asher, A. Franz, J. Keane, L. Wang, R. Noufi, High efficiency graded bandgap thin-film polycrystalline Cu(In,Ga)See-based solar cells, Sol. Energy Mater. Sol. Cells, 41-42 (1996) 231. [6] M. Burgelman, P. Nollet and S. Degrave, Modelling polycrystalline semiconductor solar cells, Thin Solid Films 361-362 (2000) 527. [7] M. Burgelman, J. Verschraegen, S. Degrave and P. Nollet, Modeling thin film PV devices, Progress in Photovoltaics 12 (2004) 143. [8] M. Burgelman, J. Marlein, Analysis of graded band gap solar cells with SCAPS, in: Proceedings of the 23rd European Photovoltaic Conference, Valencia, Spain, 2008, pp. 2151-2155. [9] M. Turcu, I.M. Kötschau, U. Rau, Composition dependence of defect energies and band alignments in the Cu(In 1-xGax)(Se1-ySy)2 alloy system, J. Appl. Phys. 91 3 (2002) 1391. [10] K. Decock, J. Lauwaert and M. Burgelman, Modelling thin film solar cells with graded band gap, Proceedings of the 45th international conference on microelectronics, devices and materials (MIDEM), Postojna, 2009. [1] [2] [3] [4] [5]