Optimization of back reflector for high efficiency hydrogenated nanocrystallinesilicon solar cellsGuozhen Yue, Laura Sivec, Jessica M. Owens, Baojie Yan, Jeffrey Yang, and Subhendu Guha Citation: Applied Physics Letters 95, 263501 (2009); doi: 10.1063/1.3279143 View online: http://dx.doi.org/10.1063/1.3279143 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/95/26?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Analyzing nanotextured transparent conductive oxides for efficient light trapping in silicon thin film solar cells Appl. Phys. Lett. 101, 103903 (2012); 10.1063/1.4750242 Efficient light management scheme for thin film silicon solar cells via transparent random nanostructuresfabricated by nanoimprinting Appl. Phys. Lett. 96, 213504 (2010); 10.1063/1.3432739 Plasmonic absorption in textured silver back reflectors of thin film solar cells J. Appl. Phys. 104, 064509 (2008); 10.1063/1.2981194 The effect of front ZnO:Al surface texture and optical transparency on efficient light trapping in silicon thin-filmsolar cells J. Appl. Phys. 101, 074903 (2007); 10.1063/1.2715554 Absorption loss at nanorough silver back reflector of thin-film silicon solar cells J. Appl. Phys. 95, 1427 (2004); 10.1063/1.1633652

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

129.174.21.5 On: Sat, 20 Dec 2014 02:17:21

Optimization of back reflector for high efficiency hydrogenatednanocrystalline silicon solar cells

Guozhen Yue,a� Laura Sivec, Jessica M. Owens, Baojie Yan, Jeffrey Yang, andSubhendu GuhaUnited Solar Ovonic LLC, 1100 West Maple Road, Troy, Michigan 48084, USA

�Received 29 October 2009; accepted 7 December 2009; published online 28 December 2009�

We have studied the effect of texture in Ag/ZnO back reflectors �BRs� on the performance ofhydrogenated nanocrystalline silicon �nc-Si:H� solar cells. While a larger texture provides superiorlight trapping, it also deteriorates the nc-Si:H quality. We have used total and diffused reflection andatomic force microscopy to evaluate the BR texture. A BR with textured Ag and thin ZnO layers hasbeen found to give the best cell performance. Using the optimized BR, we have achieved an initialactive-area efficiency of 10.2% in a nc-Si:H single-junction cell and a stable total-area efficiency of12.5% in a hydrogenated amorphous silicon/nc-Si:H/nc-Si:H triple-junction cell. © 2009 AmericanInstitute of Physics. �doi:10.1063/1.3279143�

Its high short-circuit current density �Jsc� and low light-induced degradation makes hydrogenated nanocrystallinesilicon �nc-Si:H� an ideal candidate for use in the bottom cellof multijunction solar cells. Since nc-Si:H has a low absorp-tion co-efficient compared to hydrogenated amorphous sili-con �a-Si:H�, a relatively thick intrinsic �i� layer is needed toobtain a high Jsc. A thick i layer, however, weakens thebuilt-in electrical field, reduces the carrier collection, andsubsequently decreases the fill factor �FF� of solar cells. Alight trapping technique using a textured back reflector �BR�can achieve a high Jsc without significantly increasing the ilayer thickness, thus playing a critical role in improving theefficiency of nc-Si:H solar cells.1 A commonly used BR con-sists of a Ag layer for reflecting the light that reaches thesubstrate and a ZnO layer, which serves as an optical bufferbetween the metal and the semiconductor. An effective BRmust scatter the light efficiently without suffering reflectionlosses. Intuitively, a flat Ag layer and a highly textured ZnOlayer would be the best combination because the flat Aglayer provides the best reflectance and low parasitic and/orplasmonic losses at the Ag/ZnO interface, and the highlytextured ZnO layer produces randomizing scattering for lighttrapping. Textured ZnO layers can be achieved by optimizingthe ZnO deposition and/or chemical etching.2 However, nc-Si:H materials deposited on textured surfaces usually havehigh defect densities due to crystallite collisions and micro-cracks in the valleys of the textured substrates,3–6 which re-sult in a poor cell performance. An optimized substrate sur-face morphology is needed to minimize this effect. It hasbeen shown that U-shaped6 and W-shaped7 features are bet-ter than the conventional “V-shaped” surface structures forobtaining high quality nc-Si:H.

In this letter, we present results on the effect of Ag andZnO textures on nc-Si:H solar cell performance. We find thata texture that yields superior light trapping without hurtingthe material quality of nc-Si:H is most desirable. A bilayer oftextured Ag and thin ZnO is found to be the best.

Sputtered Ag/ZnO BRs were used for these studies,where the texture was controlled by changing the substrate

temperature and film thickness. The root-mean-square�RMS� of ZnO surface roughness was measured usingatomic force microscopy �AFM�.8 Optical properties of BRswere characterized by measuring the total and diffusive re-flection spectra and angular distribution of the scattered lightintensity.

We deposited nc-Si:H single-junction and a-Si:H/nc-Si:H/nc-Si:H triple-junction solar cells with a modifiedvery high frequency discharge at high rates on various Ag/ZnO BRs. Important technique for improving the nc-Si:Hsolar cell performance include hydrogen dilution profiling9

during the growth of nc-Si:H and optimization of the solarcell structures with buffer and seeding layers.10 The solar cellperformance was characterized using current-density versusvoltage �J-V� measurements under an AM1.5 solar simulatorat 25 °C and quantum efficiency �QE� measurements in thewavelength range between 300 and 1200 nm. The Jsc valueswere calculated by integrating the measured QE curveswith the AM1.5 solar spectrum. Several sets of a-Si:H/nc-Si:H/nc-Si:H triple-junction cells were light-soaked under100 mW /cm2 white light at 50 °C for �1000 hours. Thesolar cell efficiency was also measured at the National Re-newable Energy Laboratory �NREL�.

Although a large number of samples were characterized,we show the total and diffusive reflection spectra of fourrepresentative samples in Fig. 1, where the flat and texturedAg layers were deposited at room and elevated temperatures,respectively, and the thin and thick ZnO layers had thick-nesses of 0.12 and 2.0 �m, respectively. When comparingthe two samples with thin ZnO layers, one notes that the totalreflection from the textured Ag sample is lower than the flatAg sample. The difference could be caused by parasiticlosses associated with the Ag texture. The right-shift of thesmall valley in the short wavelength region is caused by theshift of plasmonic resonance frequency with the increasedAg texture. If we must choose a BR based on total reflectiononly, we would pick flat Ag with a thin ZnO layer. However,the diffused reflection spectra show that this sample does notscatter light efficiently. The two samples with thick ZnOshow similar total reflection spectra with large losses in theshort wavelength region, which could be due to light absorp-tion in the ZnO layer caused by single or multiple passesa�Electronic mail: gyue@uni-solar.com.

APPLIED PHYSICS LETTERS 95, 263501 �2009�

0003-6951/2009/95�26�/263501/3/$25.00 © 2009 American Institute of Physics95, 263501-1 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

129.174.21.5 On: Sat, 20 Dec 2014 02:17:21

within the layer. The diffusive reflection spectra show a largedifference between the two samples with thick ZnO on flatand textured Ag. The sample with textured Ag and thick ZnOhas the best scattering characteristics with a high diffusivereflectance.

From optical characterizations, we may make the follow-ing conclusions. First, the BR with flat Ag and thin ZnO isnot a good choice although the reflectance is the best, be-cause the poor light scattering does not produce enough lighttrapping in nc-Si:H cells. Second, the BRs with flat Ag andtextured Ag do not show much difference in the total reflec-tance when thick ZnO is used, but the diffusive reflectionspectrum is much higher for the sample with textured Ag.Third, although the BR with flat Ag and thick ZnO is ex-pected to have a low loss at the Ag/ZnO interface, the mea-surements do not support this hypothesis because the inter-face loss is probably not dominant when the ZnO layer isthick. Based on these three points and considering the opticalproperties only, one would conclude that the BR with tex-tured Ag and thick ZnO is the best.

As reported in the literature,3–6 nc-Si:H properties areaffected by the surface texture. A high texture may cause ahigh defect density in nc-Si:H solar cells. Table I lists nc-Si:H solar cell J-V characteristics along with the RMS valuesof BRs �from AFM images�, where the four cells were madewith the same recipe but on the four different BRs discussed

above. It is clear that the solar cell on the BR with texturedAg and thin ZnO shows the best performance and the highestJsc. The high Jsc is not expected from the optical character-izations of BRs with textured Ag and thin ZnO. We did fur-ther experiments to understand this observation. Figure 2shows a comparison of AFM images taken from BRs withtextured Ag but with thin and thick ZnO. While the lateralfeature sizes are very similar for the two samples, the verticalfeature sizes are very different. The sample with thick ZnOhas larger vertical features �large RMS� than the one withthin ZnO. The large vertical features produce steeper anglesin the valleys of BRs. nc-Si:H grown on such a surface willhave a high density of defects and microcracks.3–6 Conse-quently, a poor nc-Si:H solar cell performance is obtained onsuch BRs with textured Ag and thick ZnO layers.

In order to confirm the poor cell performance on the BRswith high vertical surface texture, we measured the QEcurves under short-circuit and 2.0 V reverse bias conditions.The difference between the two QE curves reflects the re-combination loss due to poor nc-Si:H quality. Figure 3 showstwo pairs of QE curves measured on two nc-Si:H solar cells.One cell was deposited on the BR with textured Ag and thin

20

40

60

80

100

300 400 500 600 700 800 900 1000 1100 12000

20

40

60

80

(a)

BD

CA

Totalreflection(%)

A (flat Ag, thin ZnO)B (flat Ag, thick ZnO)C (texture Ag, thin ZnO)D (texture Ag, thick ZnO)

(b)

C

D

B

A

Wavelength (nm)

Diffusivereflection(%)

FIG. 1. Total and diffusive reflection spectra of four BRs as described in thetext.

TABLE I. J-V characteristics and BR structure of four nc-Si:H solar cellsdeposited with the same recipe but on different BRs.

NoVoc

�V� FF

Jsc

�mA /cm2�

Efficiency�%� BR

RMS�nm�0 V/�2 V

Difference�%�

1 0.502 0.645 22.59/22.72 0.58 7.31 A 142 0.476 0.528 25.72/26.21 1.91 6.46 B 423 0.482 0.608 26.25/26.50 0.95 7.69 C 334 0.431 0.563 24.59/25.51 3.70 5.97 D 52

Z 350 nm(a) thin ZnO

(b) thick ZnO Z 350 nm

FIG. 2. �Color online� AMF images of two BRs with a textured Ag layer butwith thin and thick ZnO layers.

0.0

0.2

0.4

0.6

0.8

1.0

350 450 550 650 750 850 950

QE

Wavelength (nm)

thin ZnO, 26.28 (0 V)

thin ZnO, 26.54 (-2 V)

thick ZnO, 24.61 (0 V)

thick ZnO, 25.53 (-2 V)

QE (mA/cm2)

FIG. 3. Comparison of QE curves measured under short circuit and �2 Vbias for two nc-Si:H cells deposited on different BRs.

263501-2 Yue et al. Appl. Phys. Lett. 95, 263501 �2009�

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

129.174.21.5 On: Sat, 20 Dec 2014 02:17:21

ZnO and the other on the BR with textured Ag and thickZnO. The nc-Si:H solar cell on the BR with textured Ag andthin ZnO shows very little difference between QE ��2.0 V�and QE �0 V�, indicating a low recombination loss. There arealso significant interference fringes on the QE of this sample,indicating a significant amount of directly reflected light. Onthe other hand, a large difference between QE ��2.0 V� andQE �0 V� is observed with the sample on the BR with tex-tured Ag and thick ZnO, which indicates a significant lossdue to the recombination and thus an inferior nc-Si:H quality.

From the above experiments, we conclude that aproperly textured Ag layer with a thin ZnO layer is the bestchoice for nc-Si:H solar cells. Using these optimized BRsand the improved nc-Si:H material quality obtained byusing a hydrogen dilution profiling technique, we haveachieved Jsc=29.1 mA /cm2 in a nc-Si:H single-junction cellwith open-circuit voltage �Voc�=0.516 V, FF=0.647, andefficiency=9.72%. This is one of the highest Jsc reported fornc-Si:H single-junction solar cells. Using the same BR, wefurther optimized the nc-Si:H growth recipe to obtain higherefficiency. An initial active-area efficiency of 10.2% �Voc=0.540 V, FF=0.680, and Jsc=27.8 mA /cm2� has beenachieved for a nc-Si:H single-junction cell as shown in Fig.4. Further, an a-Si:H/nc-Si:H/nc-Si:H triple-junction cell wasfabricated at a high deposition rate of 10 Å/s. A stable total-area efficiency of 12.5% was obtained as shown in Fig. 5,where the measurement was made at NREL. This efficiencyexceeds the previous world record efficiency of 12.1% mea-sured at NREL from an a-Si:H/a-SiGe:H/a-SiGe:H triple-junction cell.11

In summary, we have systematically studied Ag/ZnOBRs with randomized textures. The results show that a prop-erly textured Ag with thin ZnO is best suited for high effi-ciency nc-Si:H solar cells. Using these optimized BRs andimproved nc-Si:H materials, we achieved an initial active-area efficiency of 10.2% for a nc-Si:H single-junction solarcell and a stabilized total-area efficiency of 12.5% for an

a-Si:H/nc-Si:H/nc-Si:H triple-junction solar cell as con-firmed by NREL.

The authors thank H. Fritzsche for fruitful discussionand E. Chen and T. Palmer for measurement and depositionassistance. This work was supported by U.S. DOE under SAIProgram Contract No. DE-FC36-07 GO 17053.

1J. Yang, B. Yan, G. Yue, and S. Guha, Amorphous and PolycrystallineThin-Film Silicon Science and Technology-2009, MRS Symposia Proceed-ings No. 1153 �Materials Research Society, Pittsburgh, 2009�.

2J. Müller, G. Schöpe, O. Kluth, B. Rech, M. Ruske, J. Trube, B. Szyszka,T. Höing, X. Jiang, and G. Bräuer, Proceedings of 28th IEEE PhotovoltaicSpecialist Conference, Anchorage, AK, 2000, p. 758.

3Y. Nasuno, M. Kondo, and A. Matsuda, Proceedings of the 28th IEEEPhotovoltaic Specialists Conference �IEEE, New York, 2000�, p.142.

4G. Yue, L. Sivec, B. Yan, J. Yang, and S. Guha, Amorphous and Polycrys-talline Thin-Film Silicon Science and Technology-2009, MRS SymposiaProceedings No. 1153, �Materials Research Society, Pittsburgh, 2009�.

5H. Li, R. H. Franken, J. Rath, and R. E. I. Schropp, Sol. Energy Mater.Sol. Cells 93, 338 �2009�.

6M. Python, O. Madani, D. Domine´, F. Meillaud, E. Vallat-Sauvain, and C.Ballif, Sol. Energy Mater. Sol. Cells 93, 1714 �2009�.

7M. Kambe, A. Takahashi, N. Taneda, K. Masumo, T. Oyama, and K. Sato,Proceedings of 33rd IEEE Photovoltaic Specialist Conference, San Diego,2008.

8B. Yan, J. M. Owens, C.-S. Jiang, J. Yang, and S. Guha, Amorphous andNanocrystalline Silicon Science and Technology-2005, MRS SymposiaProceedings No. 862, �Materials Research Society, Pittsburgh, 2005�, p.603.

9B. Yan, G. Yue, J. Yang, S. Guha, D. L. Williamson, D. Han, and C.-S.Jiang, Appl. Phys. Lett. 85, 1955 �2004�.

10G. Yue, B. Yan, C. Teplin, J. Yang, and S. Guha, J. Non-Cryst. Solids 354,2440 �2008�.

11J. Yang, A. Banerjee, and S. Guha, Appl. Phys. Lett. 70, 2975 �1997�.

-30

-25

-20

-15

-10

-5

0

5

-0.2 0 0.2 0.4 0.6

J(mA/cm2 )

Voltage (V)

Jsc=27.80 mA/cm2Voc=0.540 VFF=0.680Eff=10.21 %

RF 18348-22

FIG. 4. J-V characteristics of 10.2% nc-Si:H single-junction cell depositedon the optimized BR.

FIG. 5. J-V characteristics of an a-Si:H/nc-Si:H/nc-Si:H triple-junction cellwith a stable total-area efficiency of 12.5%.

263501-3 Yue et al. Appl. Phys. Lett. 95, 263501 �2009�

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

129.174.21.5 On: Sat, 20 Dec 2014 02:17:21