IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 5, MAY 2002 949

V and a maximum breakdown voltage of 1250 V. This is the higheston/off ratio reported to date for a SiC Schottky diode.

ACKNOWLEDGMENT

The authors would like to thank L. Coulbeck and P. Taylor of DynexSemiconductor, Lincoln, U.K., for gel encapsulating the diodes forbreakdown measurements.

REFERENCES

[1] (2001, Feb.) Infineon Technologies Produces World’s First PowerSemiconductors in Silicon Carbide. Infineon Ltd.. [Online]. Available:http://www.infineon.com/news/press/102-019e.htm

[2] M. Bhatnagar, P. K. McLarty, and B. J. Baliga, “Silicon-carbide high-voltage (400 V) Schottky barrier diodes,”IEEE Electron Device Lett.,vol. 13, pp. 501–503, Oct. 1992.

[3] W. Wright, J. Carter, P. Alexandrov, M. Pan, M. Weiner, and J. H. Zhao,“Comparison of Si and SiC diodes during operation in three-phase in-verter driving ac induction motor,”Electron. Lett., vol. 37, pp. 787–788,Dec. 2001.

[4] B. J. Baliga,Power Semiconductor Devices. Boston, MA: PWS, 1996.[5] D. Defives, O. Noblanc, C. Dua, C. Brylinski, M. Barthula, V. Aubry-

Fortuna, and F. Meyer, “Barrier inhomogeneities and electrical char-acteristics of Ti/4H-SiC schottky rectifiers,”IEEE Trans. Electron De-vices, vol. 46, pp. 449–455, Mar. 1999.

[6] V. Saxena, J. N. Su, and A. J. Steckl, “High-voltage Ni- and Pt-SiCSchottky diodes utilizing metal field plate termination,”IEEE Trans.Electron Devices, vol. 46, pp. 456–464, Mar. 1999.

Highly Stable Hydrogenated Amorphous SiliconGermanium Solar Cells

Aad Gordijn, Raúl Jimenez Zambrano, Jatindra Kumar Rath, andRuud E. I. Schropp

Abstract—This article shows an optimized a-SiGe:H material that be-haves highly stable in solar cells. The a-SiGe:H material is deposited byPECVD with high hydrogen dilution, near the microcrystalline depositionregime. We made various a-SiGe:H single solar cells to optimize the devicedesign. The band gap in the central part of the cell is 1.53 eV. The hydrogenbonding configuration in the a-SiGe:H material suggests the presence ofvoids, however, the material has no noticeable sign of crystallinity. Lightsoaking experiments showed that the present single junction a-SiGe:H solarcells are highly stable. After one hour of light soaking, a slight improvementin fill factor is observed and an improvement in carrier collection in the redregion is evident from spectral response. The stable a-SiGe:H material isincorporated as the bottom cell of a-Si:H/a-SiGe:H tandem solar cells. Un-like the single junction cell, this tandem cell slightly degrades under lightsoaking. This is solely the result of degradation of the a-Si:H top layer.

Index Terms—Amorphous silicon, photovoltaics, solar cells.

I. INTRODUCTION

Hydrogenated amorphous silicon (a-Si:H)-based solar cells are the-oretically more efficient in a dual band gap tandem cell configuration

Manuscript received September 24, 2001; revised February 11, 2002. Thework presented here was supported by the Dutch organization “NederlandseOrganizatie voor Energie en Milieu (NOVEM).” The review of this brief wasarranged by Editor P. Panayotatos.

The authors are with the Debye Institute, Surfaces, Interfaces, andDevices, Utrecht University, 3508 TA Utrecht, The Netherlands (e-mail:j.k.rath@phys.uu.nl).

Publisher Item Identifier S 0018-9383(02)04341-1.

compared to a single junction, because a wider part of the solar spec-trum can be used to generate charge carriers. The incorporation of ger-manium results in a low band gap material (suitable as the bottom cellof a tandem cell), but it has drawbacks such as the deteriorated elec-trical properties [1] and the high costs. Moreover, though a-SiGe:H israther stable as a material, solar cells incorporating this material showdegradation under light soaking [2]. Carlsonet al. [3] reported thatafter long illumination times, a-SiGe cells degraded more than a-Sicells. This behavior bears serious implications for the long term sta-bility of a-Si/a-SiGe tandem modules. The degradation property of thecell, manifested as a decrease in the efficiency, is caused by an in-creasing number of mid-gap defect statesNd. This is known as theStaebler–Wronski effect (SWE) [4]. During light soaking, the defectstates are created when weak bonds are broken due to electron–holepair recombination. Hydrogen atoms bound in void-like configurationsgive rise to faster metastable defect creation. On the other hand, thepresence of microcrystals seems to retard the degradation behavior,even when the grain boundaries are passivated with hydrogen. An ex-ample of this is PECVD heterogeneous a-Si:H material [5]. Further-more, it is known that in a-Si:H solar cells, the SWE is affected by thehydrogen content [6] and by the order in the network represented bythe Urbach energy and microstructure parameter [7].

In recent years, several methods have been applied to reduce thedegradation in a-Si:H. Materials that have a polymorphous structure[8], a structure containing microcrystals [5] or an improved structuralorder [9], have been made. In these cases, the materials are depositedat the onset of the crystalline deposition regime by applying high dilu-tions or in a regime of powder formation.

In this study, we present an optimized a-SiGe:H material that is madefrom highly hydrogen diluted source gas and we correlate the stabilityof the resulting solar cells to some aspects of the microstructure. Lightsoaking experiments have been performed on these a-SiGe:H singlejunction solar cells and on a-Si:H/a-SiGe:H tandem cells to study theirlong-term efficiency and the impact of the SWE. The fill factor (FF),derived from the diode current–voltage (I–V ) characteristics of thesolar cells, is used as the main quality monitoring parameter for thedevices because it is considered to be correlated with the defect den-sity [10].

II. EXPERIMENTAL

The samples were deposited by the plasma enhanced chemicalvapor deposition (PECVD) technique [11]. Hydrogenated amorphoussilicon is deposited by dissociating silane(SiH4) in an rf electricfield, whereas a-SiGe:H was obtained from a gas mixture of SiH4,GeH4 and H2. The band gap of a-SiGe:H materials can be tunedby adjusting the GeH4=SiH4 ratio. P-i-n solar cells were depositedon transparent conducting oxide (TCO) coated glass (Asahi U-typeSnO2:F) and supplied with an evaporated silver back contact,resulting in a configuration given by SnO2:F/p-a-SiC:H/a-Si:H(buffer)/graded layer/i-a-SiGe:H/graded layer/n-a-Si:H/Ag. The bandgap graded layer, which is of staircase type, is made by stepwiseincreasing and decreasing the germanium concentration. This gradingprocess results in an overall (U-shape) profile as shown in Fig. 1.A series of a-SiGe:H solar cells was made with subtle variants ofthe band gap grading profile. In a tandem configuration two p-i-nstructures are deposited on top of each other in the configura-tion SnO2/p-a-SiC:H/buffer/i-a-Si:H/n-�c-Si:H/p-a-SiC:H/a-Si:H(buffer)/grading/i-a-SiGe:H/grading/ n-a-Si:H/Ag. None of the solarcells comprises any special optical back reflector (as for exampleZnO/Ag [12]).

0018-9383/02$17.00 © 2002 IEEE

950 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 5, MAY 2002

Fig. 1. Band gap diagram of the grading used in the a-SiGe:Hi-layers.

All tandem cells are designed in such a way that the top cell is cur-rent limiting (as verified by determining the current density of the topand bottom cells using spectral response measurements). Clearly, thetop cell for the a-Si:H (1.85 eV)/a-Si:H (1.78 eV) tandem cell structureis different from a top cell for a a-Si:H (1.78 eV)/a-SiGe:H (1.53 eV)tandem. The totali-layer thickness (including the graded layers) of thea-SiGe:H subcell is 150 nm. The samei-layer thickness is used in thea-SiGe:H subcells of tandems, in which thei-layer thickness of the topcell is 100 nm. The solar cells were encapsulated to perform accuratelight soaking experiments. The encapsulation serves to prevent deteri-oration of the metal back contact due to repeated measurements and/oroxidation.

Both a-Si:H/a-SiGe:H tandem solar cells and a-SiGe:H single junc-tion solar cells were exposed to light. An a-Si:H/a-Si:H tandem celland a number of a-Si:H single junction cells (150, 215, and 500-nmthick i-layers with different hydrogen dilutions) served as references.The cells were continuously illuminated at a temperature of 50�C usinglight with an intensity of around 90 mW/cm2, close to the AM1.5 spec-trum. Although some cells were illuminated with unfiltered light, mostof the a-SiGe:H cells and the 500 nm thick a-Si:H reference cell wereilluminated through an a-Si:H film. This layer simulates the opticaltransmission of the a-Si:H top cell to study the case of the bottom cell ofa tandem cell. Before and during the light soaking the properties of thedevices were monitored byI–V and spectral response measurements.The measurements were done at logarithmically increasing time inter-vals.

I–V diode characteristics were measured under illumination usinga Wacom dual beam solar simulator. Spectral response measurementswere performed under the AM1.5 bias light condition for singlejunction cells. To estimate the separate responses of the top and thebottom cell of a tandem cell, the device was illuminated by red andblue bias light respectively. Material properties were characterized byFourier transform infrared spectroscopy (FTIR), the steady state pho-tocarrier grating technique (SSPG), the constant photocurrent method(CPM), Raman spectroscopy, selective area electron diffraction pattern(SADP), Rutherford back scattering (RBS) and photoconductivitymeasurements on individual layers.

III. RESULTS AND DISCUSSION

For the optimized intrinsic amorphous SiGe material, we used ahigh hydrogen flow rate just below the flow at which the material be-comes microcrystalline. The important deposition parameters and ma-terial properties of the intrinsic low band gap a-SiGe:H material arelisted in Table I.

With respect to the structure, the obtained material shows a consider-able microstructure manifested as the presence of the 2100 cm�1 peak

TABLE IDEPOSITION PARAMETERS AND MATERIAL

PROPERTIES OF THE A-SIGE:H FILMS

Fig. 2. Deconvoluted FTIR spectrum showing peaks of Ge–H, Si–H, andSi–H modes of vibration.

Fig. 3. SADP image of the a-SiGe:H film.

in addition to Si–H bonds at 2000 cm�1 (see Fig. 2). The 2100 cm�1

peak is caused by dihydrides and/or hydrogen at the surfaces of voids.The selective area electron diffraction pattern in Fig. 3 shows ringswhich are slightly sharper than the pure amorphous structure but thisdoes not imply the presence of nanocrystals in the matrix (nanocrys-tals are not evident from Raman spectroscopy). We speculate that thesecharacteristics of the SADP pattern are related to an improvement ofthe medium range order (MRO). The MRO may have a pronounced ef-fect on the stability of the material.

The initial cell parameters of the single junction a-SiGe:H andtandem a-Si:H/a-SiGe:H cells are shown in Table II. As mentionedin the experimental section, three reference a-Si:H solar cells areused. An optimized 500 nm thick solar cell with an initial efficiencyof 9.7% (initialFF = 0:674) is used to compare with the optimizeda-SiGe:H cells under similar illumination (illuminated through an

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 5, MAY 2002 951

TABLE IICELL PARAMETERS OF A-SIGE:H-BASED SINGLE JUNCTION AND

A-SI:H/A-SIGE:H TANDEM CELLS

Fig. 4. Relative changes in the fill factor with light soaking time for threesingle junction a-Si:H cells: 215 nm thick (degraded with full light), 500 nmthick, and 150 nm thick (with high hydrogen dilution) and for two a-SiGe:Hcells, which are 150 nm thick, light soaked with filtered light and with unfiltered(full) light. The initial fill factors are normalized to 1.

a-Si:H film). To compare an a-SiGe:H cell with an a-Si:H cell withsimilar efficiency, a 215 nm thick a-Si:H cell with an initial efficiencyof 7.5% (initialFF = 0:669) is used (both illuminated in full light).Moreover, the results are compared with a 150 nm thick a-Si:H cellwith a high hydrogen dilution (initialFF = 0:607). Details of theresults of the degradation of high hydrogen diluted a-Si:H cells will bepublished separately [13]. In Fig. 4, we see that the standard 500 nmthick a-Si:H cell is degraded about 18%.

Unlike the amorphous silicon single junction cells with differentthicknesses made using different hydrogen dilutions, the a-SiGe:H de-vices show a stable behavior, independent of the illumination condition(see also Fig. 4). Variations in the fill factor with time are in the order of4%. We do not see significant differences in the degradation behaviorfor the different a-SiGe:H cells that slightly differ from each other withrespect to their bandgap grading design. The stable behavior is intrinsicto the type of material used in the cell.

A more detailed look at the degradation behavior of the FF showslittle degradation during the first hour, after which a small improvementappears. This behavior is consistent for all the a-SiGe:H cells degradedin this study. An explanation for this inverse Staebler–Wronski effect isthat in the first short periods between the measurements only the lightinduced effects play a role while the temperature dependent effects donot yet have an effect. The improvement after 1 h can be explainedby a temperature induced rearrangement toward a hydrogen mediatedthermal equilibrium of defects that may have been “frozen in” duringthe cooling after deposition.

The improvement in the fill factor (after 1 h) is confirmed by spec-tral response measurements which show a consistent increase of thecollection in the red part of the spectrum (Fig. 5). Fig. 6 shows that theimprovement in red is more pronounced at forward bias.

The changes in fill factor during light soaking of the tandem cells areplotted in Fig. 7. A comparison of the degradation rates of the top celllimited a-Si:H/a-SiGe:H and a-Si:H/a-Si:H (initial efficiency= 9.9%,initial FF = 0:730) tandem cells to those of the a-SiGe:H and a-Si:H

Fig. 5. Variation of the spectral response of an a-SiGe:H cell with light soakingtime (white bias light).

Fig. 6. Relative changes of spectral response of an a-SiGe:H cell at positive(+0.3 V) bias compared to that at short circuit case at different light soakingtimes.

Fig. 7. Changes in fill factor (initial FF normalized to 1) with degradationtime for the two a-Si:H/a-SiGe:H tandem cells (averaged) and the a-Si:H/a-Si:Htandem cell.

Fig. 8. Spectral response of the a-Si:H/a-SiGe:H tandem cell after differentlight soaking times.

single junctions infers that the tandem cells show degradation('10%),in contrast to single junction a-SiGe:H cells which showed virtually nodegradation. Moreover, a decreasing integrated short circuit current inthe top cell ('8% for the top cell, compared to 0.5% for the a-SiGe:Hbottom cell) is measured by spectral response (Fig. 8). This impliesthat the degradation of the a-Si:H/a-SiGe:H tandem cell is mainly dueto the a-Si:H top cell degradation.

952 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 49, NO. 5, MAY 2002

The stability of these a-SiGe:H cells, in spite of the presence of a no-ticeable 2100 cm�1 mode in the IR spectrum, suggests that the struc-ture (we speculate the medium range order) plays a crucial role in thedegradation behavior. This adds a new dimension to the understandingso far on the a-SiGe:H solar cells’ degradation behavior which is at-tributed to microstructure/voids [2].

IV. CONCLUSION

After material and device design optimization, we depositeda-SiGe:H single junction cells with an initial efficiency of� = 7:4%,and a-Si:H/a-SiGe:H tandem cells with� = 9:1% without the useof any enhanced optical back reflector like ZnO. The optimizeda-SiGe:H material with a Tauc band gap ofEg = 1:53 eV was madewith a high hydrogen dilution, so that the material is near the edge ofmicrocrystallinity. The a-SiGe:H solar cells show a stable behaviorduring light soaking, independent of their initial efficiency values.A-Si:H cells (optimizedi-layer of undiluted SiH4) degrade 18% underthe same conditions.

The subtle differences in device design do not affect the degradationrate but rather the initial (and in equal extent the stabilized) efficiency.The stable behavior is intrinsic to our a-SiGe:H material. The a-Si:H/a-SiGe:H tandem cells degrade around 10%, which is solely due to thea-Si:H top cell degradation. The a-Si:H/a:Si:H tandem cell degradesstronger and has a lower stabilized efficiency.

We managed to reduce the degradation of a-SiGe:H solar cells by op-timizing the material such that, in spite of a significant microstructure(which usually gives rise to stronger degradation), the material behavesstable in a solar cell. In the material, no microcrystals (that usually pre-vent degradation) are present, but the medium range order is improved.

ACKNOWLEDGMENT

The authors would like to thank K. van der Werf, who deposited thesamples.

REFERENCES

[1] J. K. Rath, A. R. Middya, and S. Ray, “Influence of deposition parame-ters on defect formation in a-SiGe:H alloys,”Phil. Mag. B, vol. 71, no.5, pp. 821–839, 1996.

[2] X. Xu, J. Yang, and S. Guha, “On the lack of correlation between filmproperties and solar cell performance of amorphous silicon–germaniumalloys,” Appl. Phys. Lett., vol. 62, pp. 1399–1401, 1993.

[3] D. E. Carlson, L. F. Chen, G. Ganguly, G. Lin, A. R. Midday, R. S.Crandall, and R. Reedy, “A comparison of the degradation and annealingkinetics in amorphous silicon and amorphous silicon–germanium solarcells,” Mat. Res. Soc. Proc., vol. 557, pp. 395–400, 1999.

[4] D. L. Staebler and C. R. Wronski, “Reversible conductivity changes indischarged-produces amorphous silicon,”Appl. Phys. Lett., vol. 31, pp.292–295, 1977.

[5] D. V. Tsu, B. S. Chao, S. R. Ovshinsky, S. Guha, and J. Yang, “Effect ofhydrogen dilution on the structure of amorphous silicon alloys,”Appl.Phys. Lett., vol. 71, pp. 1317–1319, 1997.

[6] C. Godet, P. Morin, and P. Roca i Cabarocas, “Influence of the dilute-phase SiH bond concentration on steady-state defect density in a-Si:H,”J. Non-Cryst. Solids, vol. 198–200, pp. 449–452, 1996.

[7] E. Bhatacharya and A. H. Mahan, “Microstructure and the light-inducedmetastability in hydrogenated amorphous silicon,”Appl. Phys. Lett., vol.52, pp. 1587–1589, 1988.

[8] P. Roca i Cabarocas, “Plasma deposition of silicon clusters: A way toproduce silicon thin films with medium range order,”Proc. Mat. Res.Soc. Symp., vol. 507, pp. 855–865, 1998.

[9] B. P. Nelson, E. Iwaniczko, A. H. Mahan, Q. Wang, Y. Xu, R. S. Cran-dall, and H. M. Branz, “High deposition rate, a-Si:H,n–i–p solar cellsgrown by HWCVD,”Thin Solid Films, vol. 395, pp. 292–297, 2001.

[10] W. Frammelsberger, H. Rubel, P. Lechner, R. Geyer, and N. Kniffler,“Defect characterization in amorphous silicon based solar cells bysub-band gap spectroscopy with constant photocurrent measurements,”Appl. Phys. Lett., vol. 58, pp. 2660–2662, 1991.

[11] R. E. I. Schropp and M. Zeman,Amorphous and Microcrystalline SiliconSolar Cells: Modeling, Materials and Device Technology. Norwell,MA: Kluwer, 1998.

[12] J. Fölsch, D. Lundszien, and H. Wagner, “Stability investigation ofa-Si:H/a-SiGe:H tandem solar cells,” in14th Eur. Photovoltaic SolarEnergy Conf., 1997, pp. 601–604.

[13] G. Munyeme, private communication.