IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 5, NO. 4, JULY 2015
1217
Status and Potential of CdTe Solar-Cell Efficiency Russell M. Geisthardt, Marko Topiˇc, Senior Member, IEEE, and James R. Sites
Abstract—The status of the highest efficiency CdTe solar cells is presented in the context of comparative loss analysis among the leading technologies for single- and polycrystalline photovoltaic materials. The Shockley–Queisser limit of a single-junction cell, with acknowledgement of variations from standard conditions, is used for reference. The highest CdTe currents achieved are comparable with the best single-crystal cells and superior to other thin-film cells. Voltages match those of multicrystalline Si, but lag behind those of CIGS and crystalline Si, and considerably lag behind crystalline GaAs. The potential for still higher CdTe efficiency will likely require a combination of reduced bulk recombination, smaller back-contact barriers, device structures with advantageous internal fields, and transparent emitters with minimal band offsets. Index Terms—Cadmium compounds, conversion efficiency, solar cell.
I. INTRODUCTION dTe has become a major player in the photovoltaic field, both through successful commercial installations and through high-efficiency solar cells. Several recent record cells have demonstrated substantial increases in the efficiency of CdTe solar cells [1]. CdTe cells, as well as cells based on several other materials, continue to move closer to the theoretical efficiency limit. This paper will explore that efficiency limit, including the limiting efficiency under nonstandard temperature and spectral conditions. The status of existing records will be discussed in comparison with the limiting efficiency. CdTe in particular will be discussed through losses in quantum efficiency (QE). Finally, a discussion of several known loss mechanisms in CdTe will be presented, along with ongoing and potential work, which can reduce these loss mechanisms. This paper complements a previous paper which focused on CIGS [2] by presenting additional perspectives on limiting efficiencies, with specific focus on CdTe cells.
C
II. IDEAL-CELL TARGETS FOR CdTe AND SILICON The thermodynamic limit efficiency of solar cells was originally developed by Shockley and Queisser (SQ) [3] and further explored by several other authors, including [4]–[7]. Since this Manuscript received January 12, 2015; revised March 12, 2015 and April 28, 2015; accepted May 4, 2015. Date of publication June 1, 2015; date of current version June 18, 2015. Colorado State funding was provided by the U.S. Department of Energy’s SunShot program through F-PACE Award DE-EE 0005399. Slovenian support was provided by the Slovenian Research Agency under the Bilateral Project BI-US/13-14/024 and the Research Program P20197. R. Geisthardt and J. R. Sites are with Colorado State University, Fort Collins CO 80523 USA (e-mail:
[email protected]; james.sites@ colostate.edu). M. Topiˇc is with the Faculty of Electrical Engineering, University of Ljubljana, 1000 Ljubljana, Slovenia (e-mail:
[email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JPHOTOV.2015.2434594
Fig. 1. Ideal values for JS C , V O C , fill factor, and efficiency with changes in band gap (under standard test conditions).
limit assumes ideal materials properties, the words ideal and limiting will both be used to describe values calculated under this framework. A. J–V Parameter Dependence on Band Gap In addition to the limiting efficiency, the limiting values for the components short-circuit current density (JSC ), open-circuit voltage (VOC ), and fill factor (FF) can also be calculated as a function of band gap. Limiting JSC is calculated by assuming unity QE for all energies above the band gap. Limiting VOC is calculated by assuming the cell to be a perfect black body, with no nonradiative recombination. Limiting fill factor is calculated by optimizing J∗V where current density J, as a function of voltage, is calculated from the diode equation. Since this is a transcendental equation, it is best calculated graphically or numerically. A numerical approximation was developed by Green [8]. The equations for these calculations were shown in [2], and the results are shown in Fig. 1 for standard test conditions (AM1.5 spectrum, 100-mW/cm2 irradiance, and 25 °C cell temperature [9]). From Fig. 1, it can be seen how these parameters trend with band gap. The limiting JSC is largest at low band gaps, while VOC and FF are largest at high band gap. Since this calculation is not done at absolute zero, the ideal fill factor is less than 100%, and the ideal VOC is below the band gap. The room temperature band gaps of silicon and CdTe are marked on the graph for reference. B. Temperature Dependence of Efficiency The curves shown in Fig. 1 were calculated at standard test conditions. However, cells often operate at nonstandard conditions. Fig. 2 shows how the limiting efficiency changes with
2156-3381 © 2015 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
1218
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 5, NO. 4, JULY 2015
Fig. 2. Limiting efficiency at AM1.5 spectrum with varying cell temperature. Larger band gaps have smaller losses in limiting efficiency at higher temperatures.
terrestrial, and the second is increasing atmospheric absorption, such as midday to sunset. From space to terrestrial (AM0 to AM1), the primary changes are a narrowing of the spectrum and the addition of absorption lines. This has two effects. The narrowing increases the ideal efficiency across most band gaps. The second effect is to add the bumps in the limiting efficiency that are responsible for the flat regions in the limiting JSC curve in Fig. 1. For the increasing atmospheric thickness (AM1 to AM4), there is increasing absorption at the specific spectral regions, as well as increased scattering of high-energy photons. The stronger absorption makes the bumps in the limiting efficiency larger. The increased scattering causes significantly decreased efficiency at the larger band gaps. Note that even though the efficiency may be slightly increased for certain band gaps with increasing air mass, the overall output power is reduced since the integrated irradiance decreases with increasing air mass more quickly than the efficiency increases. In field operation, the performance of cells and PV modules is affected by a number of important effects, many of which can be more significant than the variations in limiting efficiency presented here. Next to cell temperature and spectral changes, PV modules are exposed to a variety of location- and installationspecific conditions [10]. It is not only the band gap of the absorber, where most device physics dependences originate, which matters. Overall technology-related and even module-type dependence can have an impact on the geographical variation in PV performance, as shown for the whole of Europe by Huld et al. [11]. III. CURRENT STATUS OF J–V CURVES
Fig. 3.
Limiting efficiency at 25 °C cell temperature with varying air mass.
varying cell temperature at a fixed AM1.5 spectrum. It can be seen that the limiting efficiency is reduced at higher temperatures; however, the reduction is not uniform with band gap. Smaller band gaps have a greater loss in fill factor with increasing temperature than do larger band gaps, which leads to greater efficiency loss for smaller band gaps at increased temperatures. In particular, comparing the silicon and CdTe band gaps shows that silicon has a 3.0% absolute drop in limiting efficiency from 0 to 60 °C, while CdTe has a loss of only 2.3% absolute. This gives ideal larger band-gap cells such as CdTe an advantage in warmer climates. For less ideal cells, the advantages of the higher band gap should still hold, but the details will be dependent on the specific mechanisms, which reduce the actual efficiency below the limiting efficiency. C. Spectrum Dependence of Efficiency Cells also frequently operate at nonstandard spectral conditions. Fig. 3 shows the limiting efficiency for varying atmospheric thickness at a fixed cell temperature of 25 °C. Two separate features are shown. The first is from space to
In addition to the individual parameters, current–voltage curves can be generated from the SQ approach. It is illustrative to compare these against the record-efficiency curves of an equivalent band gap. The J–V curves are graphed with the x-axis normalized to the limiting open-circuit voltage (VSQ ) and the y-axis normalized to the limiting short-circuit current density (JSQ ) for the respective band gap to allow for direct comparison of performance between cells of different band gaps. Note that under this normalization scheme, the limiting J–V curves will still have different fill factors, since this parameter also varies with band gap. A. Comparison of Crystalline Cells to Shockley–Queisser Limit Fig. 4 compares the record crystalline single-junction cells to the limiting J–V curves under standard test conditions. This includes the comparison of the GaAs record and two recent crystalline silicon records [1] to their respective limiting J–V curves. All of these cells have achieved over 90% of their limiting JSC , although the silicon cells are slightly higher than GaAs in this respect. For VOC , however, GaAs is far ahead with 95% utilization, as compared with the low 80% utilization of silicon. The direct comparison to the limiting J–V curve shows that the
GEISTHARDT et al.: STATUS AND POTENTIAL OF CdTe SOLAR-CELL EFFICIENCY
1219
Fig. 4. Record (full lines) and ideal (dashed lines) J–V curves for singlecrystalline cells.
Fig. 5. Record (full lines) and ideal (dashed lines) J–V curves for multi- and polycrystalline cells.
GaAs cell is approaching the limiting J–V curve in both current and voltage, while the silicon cells are lower in voltage. B. Comparison of Poly- and Multicrystalline Cells to Shockley–Queisser Limit Fig. 5 compares several record poly- and multicrystalline cells with their respective limiting J–V curves at standard test conditions, including CdTe, CIGS, multicrystalline Si (m-Si), and CZTS [1]. The band gap of these cells was calculated from the 35% point of the published QE graphs. These cells have overall lower utilization of the limit than single-crystal cells, as well as larger variations between cells. CdTe has greater than 90% utilization in JSC , whereas the other materials range between 80% and 90% utilization. There is an even larger variation in VOC utilization. In this case, CIGS has leading VOC performance, with utilization comparable with single-crystalline silicon. CdTe and m-Si are comparable with CZTS having lower relative VOC . Both CdTe and CIGS compare favorably with m-Si in overall efficiency. C. Quantum Efficiency Comparisons of CdTe Cells Much of the recent improvement in CdTe record cells has been improvement in JSC . This can be seen through comparisons of the QE curves of several recent record cells, as well
Fig. 6. (a) Current density–voltage and (b) QE curves for CdTe cells, including record cells and cells made at Colorado State University.
as comparisons against more traditional cells made at Colorado State University. Fig. 6 shows the comparison of both J–V and QE curves, including a 13% CSU cell with 100 nm of CdS [12], a CSU cell with an alternative Cd(S,O) window layer [13], the NREL cell which held the CdTe efficiency record for a decade [14], and two recent First Solar records [1], [15]. The impact of thinning or replacing CdS is clear in the shortwavelength performance of these cells. In the case of CSU cells, the replacement of CdS with Cd(S,O) increased the JSC by 5.5 mA/cm2 . The NREL cell used a thin CdS layer paired with alternative Cd2 SnO4 and Zn2 SnO4 TCO/buffer layers, while the First Solar details are unpublished, but clearly have low window layer absorption. A further improvement can be seen by improvements to the transparency of the glass and TCO, as well as improvements in antireflective coatings. This is manifest as a high QE through the mid-range of wavelengths. A sharper knee above the band gap suggests better carrier collection of carriers generated by long-wavelength photons. The larger resulting fill factor is likely due to an improvement in carrier lifetime. A final increase in JSC comes from a modified band gap in the recent First Solar record cell (solid black curve in Fig. 6). All of these increases in QE are manifest in increasing integrated JSC values, as well as increasing JSC values measured from J–V.
1220
IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 5, NO. 4, JULY 2015
TABLE I FILL FACTOR AND V O C UTILIZATION FOR SEVERAL RECORD CELLS
Category
GaAs 28.8% (Alta)
HIT-Si 25.6% (Panasonic)
CIGS 21.7% (ZSW)
CdTe 21.0% (FSLR)
F F I d e a l (% ) FF (%) F F /F F I d e a l (% ) V O C I d e a l (V ) V O C (V ) V O C /V O C I d e a l (% )
89.5 86.5 96.6 1.163 1.122 96.5
87.1 82.7 94.9 0.879 0.740 84.2
87.1 79.3 91.0 0.879 0.746 84.9
89.5 79.4 88.7 1.156 0.876 75.8
Fill factor and V O C ideal and measured values, as well as ratios of the measured to ideal values for several record cells.
Further optical measurements of the individual layers or simulations can provide direct quantification of current losses due to reflection and absorption by each layer in the device. D. CdTe Voltage and Fill-Factor Discussion While losses in JSC can be well explained through QE measurements, losses in fill factor and VOC can be more challenging to explain. Ideal, or limiting, values for fill factor and opencircuit voltage can be calculated based on band gap as in Fig. 1. Table I shows both the ideal and measured fill factor and Vo c values for several record cells, as well as the ratios of the measured to ideal values. From the table, GaAs has excellent utilization of both FF and VOC . Si has good FF, but comparatively poor VOC . CIGS has fair utilization of both FF and VOC , while CdTe has slightly lower FF utilization and significantly lower VOC utilization. Further analysis (not presented here; see [16]) suggests that the vast majority of losses in FF are due to a nonideal value of the open-circuit voltage and ideality factor. Both the CIGS and CdTe record cells also have a nontrivial contribution to FF loss due to parasitic resistances, although this is not the dominant fill factor loss. IV. FUTURE CdTe POTENTIAL Although CdTe solar cells have seen a very significant increase in their record efficiency, partially reflected in more typical cells, there are refinements and additional approaches, which have the potential for further improving the understanding and performance of CdTe cells. The largest improvements in efficiency have resulted from larger photocurrents. Based on the QE curves in Fig. 6(b), the current increases are relatively well understood and have reached about 92% of their fundamental limit, and the potential for future current increases is, therefore, limited. The challenge for the lower current cells is to incorporate the same window-layer, transparent-substrate, and antireflection strategies that have proven successful. The CdTe voltage limitations fall into two major categories. One is the combination of a low recombination lifetime, which is often 1 ns or less, and a low carrier density, which is typically 1014 cm−3 . The record cells have seen some improvement in these values, but still fall significantly short of ideal. One
strategy that may circumvent excess recombination would be to fully deplete the CdTe absorber so that the internal field assists the photocarrier collection. The second major voltage limitation is the nonohmic back contact, which results from either pinning of the CdTe surface or contact materials lacking the work function needed to match the high electron affinity of CdTe. The bending in the valence band limits hole extraction, but more importantly, the corresponding bending in the conduction band significantly enhances the forward current, which lowers the diode turn-on voltage and, hence, the open-circuit voltage. Potential voltage improvement may be possible by mitigating the excess electron flow with a layer before the back contact that has increased conduction-band energy and thus reflects the electrons. Simulation work has shown the potential benefit of such a structure [17], and there has been work to implement an electron reflection with CdMgTe at CSU [18], ZnTe at NREL [19], and a reduced bulk band gap at First Solar [20]. Band-alignment issues can also arise between the emitter and TCO contact layers, especially when exploring window materials that are more transparent. The generally successful Cd(S,O) layer mentioned earlier, for example, has a relatively narrow range of the O/S ratio where the transparency is high and the conduction-band offset is not too large [13]. The potential for increased fill-factor largely follows the same strategy as for voltage. Increased recombination lifetime will benefit both, and some FF increase will result simply by a shift in the J–V curve to higher voltage [8]. In some cases, the FF can also benefit from a reduction in series resistance and shunt conductance.
V. CONCLUSION As solar cell efficiency continues to increase, it becomes more valuable to compare record cells directly with the SQ efficiency limit. Although this limit is often calculated at standard test conditions, evaluation of the limit at varying temperature and spectral conditions can help to predict how future high-efficiency cells will behave under real-world operating conditions. CdTe solar-cell efficiency in particular has shown healthy increases in the past few years, most visibly with the record efficiencies announced by First Solar. Much of the increase at several laboratories has been in higher current through more transparent window layers and reduction in reflection. The highest currents compared with the ideal are similar to those of the best single-crystal cells. Voltage and fill factor of CdTe cells have also increased toward their limiting values as absorber quality has improved. Further increase in these parameters is likely to come through the development of novel cell structures that fully deplete the absorber, reduce the band bending at the back contact, and/or introduce an electron-reflection layer.
ACKNOWLEDGMENT The authors would like to thank W. S. Sampath and M. Joˇst for several fruitful discussions.
GEISTHARDT et al.: STATUS AND POTENTIAL OF CdTe SOLAR-CELL EFFICIENCY
REFERENCES [1] M. A. Green, K. Emery, Y. Hisikawa, W. Warta, and E. D. Dunlop, “Solar cell efficiency tables (Version 45),” Prog. Photovoltaics, vol. 23, pp. 1–9, 2015. [2] M. Topiˇc, R. Geisthardt, and J. R. Sites, “Performance limits and status of single-junction solar cells with emphasis on CIGS,” IEEE J. Photovoltaics, vol. 5, no. 1, pp. 360–365, Jan. 2015. [3] W. Shockley and H. J. Queisser, “Detailed balance limit of efficiency in P-N junction solar cells,” J. Appl. Phys., vol. 32, no. 3, pp. 510–510, 1961. [4] T. Tiedje, E. Yablonovitch, G. D. Cody, and B. G. Brooks, “Limiting efficiency of silicon solar cells,” IEEE Trans. Electron Devices, vol. ED31, no. 5, pp. 711–716, May 1984. [5] O. D. Miller, E. Yablonovitch, and S. R. Kurtz, “Strong internal and external luminescence as solar cells approach the Shockley-Queisser limit,” IEEE J. Photovoltaics, vol. 2, no. 3, pp. 303–311, Jul. 2012. [6] L. C. Hirst and N. J. Ekins-Daukes, “Fundamental losses in solar cells,” Prog. Photovoltaics, vol. 19, no. 3, pp. 286–293, 2011. [7] J. L. Balenzategui and A. Mart, “The losses of efficiency in a solar cell step by step,” in Proc. 14th Eur. Photovoltaic Sol. Energy Conf., 1997, pp. 2374–2377. [8] M. A. Green, “Efficiency limits, losses, and measurement,” in Solar Cells. Englewood Cliffs, NJ, USA: Prentice-Hall, 1982. [9] Standard Tables for Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on 37° Tilted Surface, ATSM G173-03, 2012. [10] J. Kurnik, M. Jankovec, K. Brecl, and M. Topic, “Outdoor testing of PV module temperature and performance under different mounting and operational conditions,” Sol. Energy Mater. Sol. Cells, vol. 95, pp. 373– 376, 2011. [11] T. Huld, R. Gottschalg, H.G. Beyer, and M. Topic, “Mapping the performance of PV modules, effects of module type and data averaging,” Sol. Energy, vol. 84, pp. 324–338, 2010. [12] D. E. Swanson, R. M. Geisthardt, J. T. McGoffin, J. D. Williams, and J. R. Sites, “Improved CdTe solar-cell performance by plasma cleaning the TCO layer,” IEEE J. Photovoltaics, vol. 3, no. 2, pp. 838–842, Apr. 2013. [13] J. Kephart, R. Geisthardt, and W. S. Sampath, “Optimization of CdTe thin-film solar cell efficiency using a sputtered, oxygenated CdS window layer,” Prog. Photovoltaics, 2015, to be published. DOI: 10.1002/pip.2578 [14] X. Wu, J. C. Keane, R. G. Dhere, C. DeHart, D. S. Albin, A. Duda, T. A. Gessert, S. Asher, D. H. Levi, and P. Sheldon, “16.5%-efficient CdS/CdTe polycrystalline thin-film solar cell,” in Proc. 17th Eur. Photovoltaic Sol. Energy Conf., 2001, pp. 995–1000. [15] M. Gloeckler, I. Shankin, and Z. Zhao, “CdTe solar cells at the threshold to 20% efficiency,” IEEE J. Photovoltaics, vol. 3, no. 4, pp. 1389–1393, Oct. 2013. [16] R. Geisthardt, “Device characterization of cadmium telluride photovoltaics,” Ph.D. dissertation, Dept. Phys., Colorado State Univ., Fort Collins, CO, USA, 2014. [17] K.-J. Hsiao and J. R. Sites, “Electron reflector to enhance photovoltaic efficiency: Application to thin-film CdTe solar cells,” Prog. Photovoltaics, vol. 20, no. 4, pp. 486–489, 2012.
1221
[18] P. S. Kobyakov, D. E. Swanson, J. Drayton, J. M. Raguse, K. L. Barth, and W. S. Sampath, “Progress towards a CdS/CdTe solar cell implementing an electron reflector,” in Proc. 39th IEEE Photovoltaic Spec. Conf., 2013, pp. 0386–0391. [19] T. A. Gessert, J. N. Duenow, S. Ward, J. F. Geisz, and B. To, “Analysis of ZnTe:Cu/Ti contacts for crystalline CdTe,” in Proc. 40th IEEE Photovoltaic Spec. Conf., 2014, pp. 2329–2333. [20] M. Gloeckler, “Progress towards practical CdTe solar modules with 17% efficiency,” presented at the E-MRS 2014 Spring Meeting, Lille, France, 2014.
Russell M. Geisthardt received the B.A. degree in physics from Lawrence University, Appleton, WI, USA, in 2008 and the M.S. and Ph.D. degrees in physics from Colorado State University, Fort Collins, CO, USA, in 2011 and 2014, respectively.
Marko Topiˇc (M’96–SM’05) received the Ph.D. degree from the University of Ljubljana, Ljubljana, Slovenia, in 1996. He has been a Full Professor and the Head of the Laboratory of Photovoltaics and Optoelectronics, University of Ljubljana, since 2006 and an Affiliate Professor with the Colorado State University, Fort Collins, CO, USA, since 2011.
James R. Sites received the B.S. degree from Duke University, Durham, NC, USA, in 1965 and the M.S. and Ph.D. degrees from Cornell University, Ithaca, NY, USA, in 1968 and 1969, respectively. Since 1971, he has been with Colorado State University, Fort Collins, CO, USA, where he pursues device physics of photovoltaics.
