An approach to downconversion solar cells

An approach to downconversion solar cells

Solar Energy Materials & Solar Cells 108 (2013) 241–245 Contents lists available at SciVerse ScienceDirect Solar Energy Materials & Solar Cells jour...

547KB Sizes 5 Downloads 57 Views

Solar Energy Materials & Solar Cells 108 (2013) 241–245

Contents lists available at SciVerse ScienceDirect

Solar Energy Materials & Solar Cells journal homepage:

An approach to downconversion solar cells Mark B. Spitzer b,n, Hans P. Jenssen a, Arlete Cassanho a a b

AC Materials, Inc., Tarpon Springs, FL 34689, United States Photonic Glass Corporation, 661 Pleasant Street, Norwood, MA 02062, United States

a r t i c l e i n f o


Available online 10 October 2012

We describe an approach to solar downconversion in which a semiconductor is used for absorption, and substitutional rare earth ions are used for emission. The semiconductor would provide broad-band absorption, and the rare earth ions would provide emission of multiple photons by a cross relaxation process in a narrow wavelength band. The emitted photons are then absorbed by a solar cell formed from a low band gap semiconductor such as silicon. Er and Yb are suggested owing to strong emission at 980 nm which are useful when the down-converter is paired with silicon solar cells. The use of InGaN is proposed for the absorbing host semiconductor. & 2012 Elsevier B.V. All rights reserved.

Keywords: Downconversion Solar cell Cross relaxation Quantum cutting KY3F10 BaY2F8

1. Introduction The last few years have seen renewed interest in methods to break the Shockley–Queisser limit [1] to the energy conversion efficiency of single-junction solar cells [2,3]. The Shockley– Queisser limit is a consequence of detailed balance which assumes that for a single absorber of energy band gap EG, each photon with energy greater than EG generates not more than one photon, and that photons with energy less than EG generate no photocurrent. Although current practice in space solar cells is to use multijunction (tandem) solar cells to overcome this limit, there may still be advantages to single-junction cells if the efficiency can be raised beyond the Shockley–Queisser limit. Possible advantages may be reduced cost, reduced weight, elimination of currentmatching issues and improved radiation tolerance. Since such cells have not yet been made and these advantages are speculative; it is nevertheless interesting to examine whether such a cell could be fabricated. In this paper we consider how such cell might be designed. The keys to developing a single-junction cell with conversion efficiency greater than the Shockley–Queisser limit are: (i) to combine sub-EG photons into one photon than can be absorbed (upconversion) or (ii) to split high energy photons into multiple photons that each have energy greater than EG (downconversion). An excellent review may be found in Ref. 4. Upconversion has been the subject of a number of investigations that have shown small gains in efficiency by placing an upconverter behind a bi-facial Si solar cell [2,3]. When placed behind


Corresponding author. Tel.: þ1 781 492 9200; fax: þ1 781 551 9801. E-mail address: [email protected] (M.B. Spitzer).

0927-0248/$ - see front matter & 2012 Elsevier B.V. All rights reserved.

the cell, the up-converter does not interfere with absorption of photons with energy greater than EG. A down-converter must be placed in front of the cell and for this reason, great care must be taken to maintain the cell’s short circuit current (Jsc). We are not aware of an integrated solar cell and down-converter that has yielded improved Jsc but as we will discuss, we believe that such a cell is possible. The approach discussed here is different than the intermediate band solar cell [5,6] which has shown promise for upconversion via intermediate states in the energy band gap of the absorbing semiconductor within the solar cell. In such cells, recombination via these intermediate states has the potential to interfere with minority carrier transport and may lead to increased dark saturation current. Our approach is similar to Shalav’s ([2,3] because the up or downconversion process is separated from the photovoltaic process and consequently does not interfere with photogenerated carrier collection. In this paper we will be concerned primarily with downconversion. We make a distinction here between down-shifting and downconversion. By down-shifting, we mean a process in which a single photon is absorbed at a short wavelength followed by emission of a single photon at a longer wavelength. While down-shifting may improve cell efficiency if the solar cell quantum efficiency is weak at short wavelength, it cannot break the Shockley–Queisser limit because down-shifting does not change detailed balance. By downconversion we mean a process comprising the absorption of one photon followed by the emission of more than one photon, and therefore downconversion changes detailed balance. If both photons emitted in downconversion are to be absorbed by the solar cell, the original photon must have energy greater than twice the band gap of the absorbing semiconductor. For Si cells, the wavelength range of interest is 300 nm–560 nm.


M.B. Spitzer et al. / Solar Energy Materials & Solar Cells 108 (2013) 241–245

2. Er downconversion materials Adding and splitting of photons can be attained by using cooperative nonradiative resonant energy transfer between nearby rare earth ions. These processes obtain owing to the unfilled states in the inner 4 f shell [6]. The lanthanides provide many inner shell states with transitions at optical wavelengths [7]. Er is a good candidate for downconversion. When Er substitutes for Y in fluoride crystals such as NaYF4, transitions can occur among inner shell states at energies in the desired optical range. Nonradiative recombination is low because such transitions require a phonon, and in the yttrium fluoride based crystals, the phonon spectra does not provide sufficient energy to support the larger transitions (such as 4I13/2-4I15/2), therefore except for small transitions, the decay is largely radiative and the lifetime in the excited state is on the order of 1 ms. In Er-doped materials, the absorption and re-emission process begin with absorption of a photon by the excitation of an electron from the 4I15/2 ground state to an excited state such as for example the 4F7/2 state. The absorption may also proceed by excitation to a higher energy state such as the 4F5/2 followed by nonradiative decay to a nearby lower state, or in some cases excitation to a nearby higher state through phonon interactions, followed in some cases by radiative decay to the ground state. Downconversion occurs when energy is exchanged non-radiatively between nearby Er ions. Fig. 1a shows a simplified energy diagram in which some 4 f states have been omitted for clarity. One possible downconversion process begins by excitation of an electron to the Er 4F7/2 state. The excited electron may then relax to the 4I11/2 state via non-radiative energy exchange, by contributing energy to a nearby electron in an Er 4I15/2 ground state, thereby exciting this nearby electron from the Er 4I15/2 ground state to the Er 4I11/2 state. In this cross relaxation process, nonradiative energy transfer results in conversion of the energy of one photon to the energy of two excited electrons in 4I11/2 states. These electrons return to the ground state by emitting two photons with wavelength of approximately 980 nm. The density of states associated with the 980 nm emission can be increased by adding Yb, which has only one level (2F5/2) and the energy of this level is approximately the same as the Er 4I11/2 state [7]. Cross relaxation between Yb and other lanthanides is well-known [6]; the addition of Yb enhances the probability of a transition that emits at 980 nm. Fig. 2b shows how Er and Yb may cooperate in downconversion. In this case the excited electron

cross relaxes to the Er 4I11/2 state by nonradiative energy transfer to an electron in the Yb ground state, thus exciting the electron to the Yb 2F5/2 level. The two electrons decay to their ground states by emitting two photons at 980 nm.

3. Downconversion results Er-doped KY3F10 and BaY2F8 samples were grown by the Czochralski method, and were sliced and polished. The sample thickness is 1 mm. The transmission of these samples was measured and data for KY3F10 are shown in Fig. 2. The absorption lines can be associated with known Er 4 f levels [9]. The transmission data reveal several problems with the use of such crystals for downconversion. Note first that reasonably thick ( 41 mm) samples are needed to absorb fully the wavelengths associated with these levels, meaning that thin Er-doped films may not be useful for absorption. A second problem with the use of Er-doped KY3F10 or similar crystals for either up or downconversion that is evident in Fig. 2 is that Er-doped KY3F10 will only absorb in specific narrow bands, whereas worthwhile downconversion requires that all photons in the desired band be absorbed and converted. A further problem with the use of Er-doped KY3F10, particularly on the front of a solar cell, is the parasitic absorption at 650 nm and 800 nm. We have also examined the photoluminescence (Fig. 3) of KY3F10 and BaY2F8 by pumping with an Ar laser (488 nm). The strong signature of the Er 4I11/2 state is present suggesting downconversion. We also observe upconversion emission at about 550 nm when we pump with 980 nm [see Ref. 8,which also provides some initial lower bounds on quantum efficiency of the processes discussed here]. While it could be argued that we are observing down-shifting and not downconversion, the fact that we observe upconversion is a strong indicator of the presence of the two-electron cross relaxation processes we described in Fig. 1. Nevertheless, proof that we observe conversion of one absorbed photon into two emitted photons requires a definitive measurement of quantum efficiency. Quantitative measurements of photoluminescence after absorption of 980 nm photons by the Er 4I11/2 state have only shown a quantum efficiency of approximately 0.5 [8], meaning that quantitative proof has not yet been attained by us. We used a second PL system to probe below 1000 nm and the result is shown in Fig. 4. These data show the signature of the Er

Fig. 1. Downconversion based on Er. (a) Two Er ions exchange energy non-radiatively. (b) Energy is exchanged between Er and Yb.

M.B. Spitzer et al. / Solar Energy Materials & Solar Cells 108 (2013) 241–245


Fig. 2. Transmission of Er-doped KY3F10 as a function of Er concentration.

Fig. 3. Photoluminescence of KY3F10 for three Er concentrations. The pump wavelength is 488 nm. The data indicate downconversion to 980 nm. Fig. 4. Photoluminescence of KY3F10 and BaY2F8. The pump wavelength is 488 nm. 4

I13/2 level (1550 nm). These data indicate downconversion via both the 4I11/2 and the 4I13/2 levels. The 4I13/2 level is useful for upconversion but since 1550 nm is not absorbed by silicon, the 4 I13/2 level is parasitic for downconversion. However, we believe that transitions via 4I13/2 can be reduced by increasing the density of states associated with the 4I11/2 transition. We can increase the effective density of states of the levels that emit at 980 nm by adding Yb, which has only one state (2F5/2). In this way the branching to output states that emit at 980 nm can be increased, and the output at 1550 nm correspondingly decreased.

4. Discussion 4.1. Separation of absorption and emission Narrow band emission from a rare earth such as Er or Yb is very well suited to efficient absorption in a silicon solar cell. The central problem in realizing up and downconversion solar cells is the attainment of broad-band absorption. As was shown in Fig. 2,

Er absorbs in narrow bands, and in the YF4-based crystals, requires substantial thickness ( 41 mm). In order to overcome this problem, the absorption and emission processes must be separated [7]. If the absorption occurs in a direct wide band gap semiconductor, the short wavelength portion of the solar spectrum could be fully absorbed in a thin film. If the absorber is doped with rare earth ions, and the recombination occurs radiatively through cross relaxation processes, then downconversion can occur. In this way, the absorption is strong and broad-band, and the emission is narrow. Such a process requires a host semiconductor with a high diffusion length (much longer than the film thickness) and high quality surface passivation. If the recombination in the semiconductor is primarily radiative, then the quantum efficiency of the process can be high. 4.2. Er doping of wide band gap semiconductors Er doping of wide band gap semiconductors is beginning to receive attention. Steckl et al. have reported on the fabrication of


M.B. Spitzer et al. / Solar Energy Materials & Solar Cells 108 (2013) 241–245

Fig. 5. Integrated down-converter and solar cell structure.

GaN-based display materials doped with various high concentrations of rare earth ions and have fabricated electroluminescent devices by molecular beam epitaxy (MBE) [10]. Steckl also reports photoluminescence that we presume occurs when photons are absorbed by the GaN host to create electron–hole pairs. These pairs are presumed by us to diffuse to the rare earth ions where they recombine. This observation supports the notion that semiconductors can exchange electrons and holes with tri-valent rare earth ions, and that the rare earths can radiate when excited in this way. Although the alignment of the Er levels and the semiconductor energy bands is not well known, in principal, downconversion could occur in GaN-based materials. If a wide band gap material such as InxGa1  xN can host a sufficient density of rare earth ions in tri-valent substitutional lattice positions, then it may be possible to select a composition with the conduction band edge tuned to one of the Er levels between 2 and 3 eV, and the valence band edge tuned to the Er ground state, so that electrons and holes are efficiently exchanged between the conduction and valence bands and the Er levels. In this hypothetical device, the InGaN would absorb all of the energy above EG, and transfer the excited carriers to the Er3 þ where the electrons would recombine via downconversion. This approach overcomes both the problem of narrow band absorption and the problem of weak absorption in Er-doped thin films. Downconversion via the undesired 4I13/2 state could be suppressed by adding Yb as has previously been discussed. If the Er ions are surrounded by Yb, then the process shown in Fig. 1b will be favored; this process has no emission at 1550 nm. Fig. 5 shows how a solar cell might be formed with an integral down-converter. The high band gap absorber layer comprises InGaN with a thickness sufficient to absorb most of the energy above EG. A layer of InGaN that is highly doped with Er and Yb for downconversion could be placed at the back of the absorber layer. The purpose of confining the Er and Yb is to attain a high concentration and short distance between the Er and Yb, and, by placing a pure layer of InGaN in front, to favor absorption in the pure InGaN. Since the InGaN is in optical contact with the solar cell, emitted photons easily couple to the cell. The emission process is isotropic, meaning that one half of the emitted photons will be radiated toward the front of the converter. However, most of these photons will be trapped by total internal reflection at the front surface, and will eventually be absorbed by the solar cell. In the InGaN/GaN system, a heteroface can be formed between InGaN and GaN. In this structure, the GaN serves both to reduce surface recombination velocity at the boundary of the

InGaN, and also as a buffer layer for InGaN heteroepitaxy. Growth of GaN has been demonstrated on both Si and sapphire[9,11]. If carrier transfer between the Er ions and the bands is efficient and the InGaN quality is high, it may be straightforward to develop Er-doped InGaN down-converters on a low cost solar cell or substrate. If grown on sapphire, the assembly might someday be used as a solar cell cover that boosts efficiency by downconversion.

5. Conclusions We have described a new approach to downconversion in which broad-band absorption is proposed in a semiconductor, followed by emission via rare earth ions. Er-doped materials such as InGaN seem best suited for InGaAs or Si low band gap solar cells. If Er-doped films can be grown on Si, the approach may provide a way to form a single-junction cell that breaks the Shockley–Queisser limit.

Acknowledgments Part of this work was supported by the National Science Foundation under Grant no. IIP-1013378. The authors are grateful to B. Goldberg and A. Kitt for helpful measurements. References [1] W. Shockley, H. Queisser, Detailed balance limit of efficiency of p-n junction solar cells, Journal of Applied Physics 32 (1961) 510–519. [2] Shalav, A., Richards, B.S., Kramer, K.,Gudel, H., Improvements of an upconversion NaYF4:Er3 þ phosphor/silicon solar cell system for an Enhanced response in the near-infrared, in: Record of the IEEE Photovoltaic Specialists Conference, Institute of Electrical and Electronics Engineers (IEEE), 2005, p. 114. [3] Shalav, A., Richards, B.S., Kramer, K., Conibeer, G., and Green, M., Two-colour excitation up-conversion efficiency enhancement for a silicon photovoltaic device using Er3 þ doped phosphors, in: Record of the IEEE 4th World Conference on Photovoltaic Energy Conversion, Institute of Electrical and Electronics Engineers (IEEE), 2006, p. 45. [4] B.M. van der Ende, L. Aarts, A. Meijerink, Lanthanide ions as spectral converters for solar cells, Physical Chemistry Chemical Physics 11 (2009) 11081–11095. [5] Antolı´n, E., Martı´, A., Linares, P.G., Ramiro, I., Herna´ndez, E., Farmer, C.D., C.R. Stanley, and Luque, A., Advances in quantum dot intermediate band solar cells, in: Record of the 35th IEEE Photovoltaic Specialists Conference, Institute of Electrical and Electronics Engineers (IEEE), 2010, pp. 65–70. [6] Hubbard, S.M., Plourde, C., Bittner, Z., Bailey, C.G., Harris, M., Bald, T., Bennett, M., Forbes, D.V., and Raffaelle, R. InAs quantum dot enhancement of GaAs solar cells, in: Record of the 35th IEEE Photovoltaic Specialists Conference, 2010 ,pp. 1217–1222. [7] G.H. Dieke, H.M. Crosswhite, Applied Optics 2 (1963) 675–686.

M.B. Spitzer et al. / Solar Energy Materials & Solar Cells 108 (2013) 241–245

[8] Spitzer, M.B., Jenssen, H., Cassanho, A., Goldberg, B., and Kitt, A., Er-doped fluoride crystals for up and down conversion in single junction solar cells, in: Record of the 37th IEEE Photovoltaic Specialists Conference, Seattle WA, Institute of Electrical and Electronics Engineers (IEEE), 2011. [9] Spitzer, M.B., New up- and down-conversion concepts for high efficiency photovoltaics, in: Record of the 21st NASA Space Photovoltaic Research and Technology Conference, The United States National Aeronautics and Space Administration (NASA), Cleveland OH, 2009.


[10] A.J. Steckl, J. Heikenfeld, D.-S. Lee, M.J. Garter, C.C. Baker, Y. Wang, R. Jones, IEEE Journal of Selected Topics in Quantum Electronics 8 (2002) 749. [11] For example: B. Jampana, T. Xu, A. Melton, M. Jamil, R. Opila, C. Honsberg, I. Ferguson. Realization of InGaN solar cells on (111) silicon substrate, in: Record of the 35th IEEE Photovoltaic Specialists Conference, Institute of Electrical and Electronics Engineers (IEEE), (2010, p. 457.