PROGRESS IN PHOTOVOLTAICS: RESEARCH AND APPLICATIONS Prog. Photovolt: Res. Appl. (2014) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/pip.2468

RESEARCH ARTICLE

Two-terminal metal-inter-connected multijunction III–V solar cells Chieh-Ting Lin1*, William E. McMahon2, James S. Ward2, John F. Geisz2, Mark W. Wanlass2, Jeffrey J. Carapella2, Waldo Olavarria2, Emmett E. Perl1, Michelle Young2, Myles A. Steiner2, Ryan M. France2, Alan E. Kibbler2, Anna Duda2, Tom E. Moriarty2, Daniel J. Friedman2 and John E. Bowers1 1 2

Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, CA, USA National Renewable Energy Laboratory, Golden, CO, USA

ABSTRACT A novel bonding approach with an interface consisting of a metal and dielectric is developed, and a “pillar-array” metal topology is proposed for minimal optical and electrical loss at the interface. This enables a fully lattice-matched twoterminal, four-junction device that consists of an inverted top two-junction (2J) cell with 1.85 eV GaInP/1.42 eV GaAs, and an upright lower 2J cell with ~1 eV GaInAsP/0.74 eV GaInAs aimed for concentrator applications. The fabrication process and simulation of the metal topology are discussed along with the results of GaAs/GaInAs 2J and (GaInP + GaAs)/ GaInAs three-junction bonded cells. Bonding-related issues are also addressed along with optical coupling across the bonding interface. Copyright © 2014 John Wiley & Sons, Ltd. KEYWORDS III–V semiconductor; photovoltaic cells; multijunction; concentrator photovoltaic; device bonding; thermal compression bond *Correspondence Chieh-Ting Lin, Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, CA, USA. E-mail: [email protected] Received 5 February 2013; Revised 26 October 2013; Accepted 10 December 2013

1. INTRODUCTION Multijunction photovoltaics have made significant progress recently because of an increase in the number of junctions and advancements in epitaxial growth technology. Researchers are now able to grow three-junction (3J) cells with low defect density and high efficiency with methods such as inverted metamorphic growth, which allows integration of lattice-mismatched materials with improved bandgap selection. State-of-the-art 3J cells have achieved efficiencies greater than 40% using epitaxial methods [1–3], and further increases in efficiency are expected with the use of additional lattice-mismatched junctions. Lattice-mismatched growth and the utilization of novel materials allow for better bandgap selection, but maintaining low defect densities due to lattice constraints can still be an issue as additional junctions are added to achieve higher efficiencies. Wafer bonding offers a method for combining materials with dissimilar lattice constants, with each subcell grown on an appropriate lattice-matched substrate. Therefore, a wafer-bonded four-junction (4J) cell could be fabricated Copyright © 2014 John Wiley & Sons, Ltd.

from nearly defect-free subcells and thereby achieve higher efficiencies than other approaches. Soitec recently reported a direct-bonded 4J cell that achieved 44.7% efficiency at 297× concentration, which is a world record at the time of this writing [4]. Furthermore, wafer bonding widens the design space for future devices and would allow for new concepts such as the integration of InGaN-based wide bandgap material for 5+ junction devices. Simulations for this type of device show that these devices can attain efficiencies greater than 50% at 1000× concentration under the ASTM G173 direct spectrum if losses can be kept to levels typical of current commercial multijunction cell structures [5–8].

2. BACKGROUND A wafer bonding process applicable to photovoltaic devices must simultaneously satisfy strict requirements for electrical, optical, thermal, and mechanical coupling. Furthermore, the bonding process cannot severely degrade the device performance of either tandem. While it is possible

C.-T. Lin et al.

Two-terminal metal-interconnected multijunction III–V solar cells

for this process to have an electrically separated configuration with three or more terminals, a two-terminal configuration is preferred because of simplicity. Direct semiconductor–semiconductor bonding and nondirect bonding have been utilized in various applications, such as the hybrid silicon platform that integrates photonic devices based on III–V materials onto silicon [9]. Direct semiconductor bonding has been applied to photovoltaic devices, and provides great mechanical, thermal, and optical coupling. However, because of the nature of the high temperatures required in the process (400–600 °C) and extended bonding times necessary to ensure good mechanical stability, there is a high possibility of degradation in device performance [10]. Typical contact resistances for direct-bonded devices are on the order of 0.1 Ω cm2, which is suitable for 1-sun applications where typical triple-junction devices generate current densities of 14 mA/cm2 at 1 sun [10–12]. However, under high concentration (1000×), a bonded device with contact resistance of 0.1 Ω cm2 will suffer an open circuit voltage drop of roughly 1.4 V, which is a significant portion of the voltage generated [13,14]. Nondirect wafer bonding with a carbon-nanotube mixture assisting layer has also been attempted, in which the two semiconductor interfaces do not come into direct contact and no covalent bond is formed. This method makes it possible to reduce the annealing temperature and time compared with the direct bonding approach. Therefore, the devices are subject to less thermal damage. Surface roughness requirements are also reduced because of the soft bonding layer. However, there is some optical absorption at the bonding interface, and the contact resistance is too large for operation at high concentrations [14]. Here, we propose and demonstrate an alternative hybrid bond for which the bonding interface is formed by a matrix of metal and filler material (Figure 1). Metal bonding is widely used in electronics packaging and microelectromechanical systems (MEMS) packaging because of its strong mechanical integrity and high electrical conductivity. The metal can reduce series resistance across the interface to an acceptable level for high concentration. The filler material between the metal features can provide for optical and thermal coupling, as well as some mechanical support. The filler can consist of an index-matched material or an optical coating designed to transmit and/or reflect certain portions of the spectrum to the relevant subcells. Heavily doped semiconductor layers adjacent to the bonding interface are necessary to carry current laterally until collected by the interfacial metal. This metal allows for vertical conduction across the junctions with low resistance to current flow. Because the cells are currentmatched, there is no need for additional terminals; therefore, a conventional metal grid pattern at the bonding interface is not necessary. Instead, an array of pillars can be used to minimize shadowing loss for the bottom tandem. Simulations were performed to compare the performance of 5-μm grid fingers to 20-μm × 20-μm pillars using various pitch lengths. Even considering process tolerance, the pillar array outperformed the grid topology. Details of this simulation can be found elsewhere [8].

Figure 1. Schematic cross-sectional view of the bonding process showing the interfacial metal contacts and optical filler. The optical filler can be either a dielectric or an epitaxially grown semiconductor. The interfacial contacts are deposited on both bonding surfaces and then bonded using thermocompression. The top substrate is then removed, and fabrication is completed using standard cell processing techniques.

3. EXPERIMENTAL DETAILS Prior to bonding, a 1-μm-thick optically transparent filler is first deposited or epitaxially grown on one of the samples. This film is patterned and wet etched to form trenches or pits to a highly doped contact layer. A metal stack is then deposited on both the InP-based and GaAs-based subcells to allow a gold–gold (Au–Au) bond to be formed at the interface. The metal stack consists of a 5- to 30-nm-thick titanium (Ti) adhesion layer, a 30- to 100-nm platinum (Pt) layer used as a Au-diffusion barrier, and a 370- to 465-nm gold bonding layer. The combined thickness of the metal stacks from both samples must be slightly thicker than the filler to ensure proper contact and bonding of the Au layers. The filler material occupies close to 95% of the interfacial volume and is crucial for the optical coupling between the devices. The material must be optically transparent, mechanically stable enough to support the top cells, and compatible with the fabrication process. Several materials could be used for the filler; a dielectric filler (SiO2) and two semiconductor fillers (GaInP2 and AlGaAs) were explored. When using a dielectric as the filler, a 1-μm SiO2 layer is deposited on the InP-based subcells using plasmaenhanced chemical vapor deposition. The film is then patterned and etched to allow metal contact deposition; this cross section is shown in Figure 2(a). Because of the large optical index mismatch between SiO2 (n = 1.5) and the adjacent III–V layers (n > 3), reflection losses are significant. Therefore, an optimal design would also need to include optical coatings to maximize light transmission to the lower cells as shown in Figure 2(b). In addition to minimizing reflection of bottom-cell light, these coatings could also Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

C.-T. Lin et al.

Two-terminal metal-interconnected multijunction III–V solar cells

Figure 2. (a) Cross-sectional view of the bonding interface with SiO2 as a filler material. The SiO2 layer is deposited on the bottom sample because it remains attached to a (rigid) substrate that is not noticeably affected by stresses between the III–V material and the SiO2 film. (b) Bonding interface with optical coatings added to the SiO2 filler AR, anti-reflection. (c) Bonding interface using a GaInP2 filler (the GaInP2 is replaced with an AlGaAs filler in the current device design.).

be tailored for high reflectance of top-cell light and/or unconvertible infrared light. Respectively, this could enhance the top-cell efficiency and reduce cell heating. When using a semiconductor as the filler material, a 1-μ m layer of GaInP2 or AlGaAs is grown epitaxially on the inverted top cell. This layer is then patterned and etched for metal contact deposition as shown in Figure 2(c). A major advantage of this approach is that the index of refraction of the semiconductor filler and the adjacent epitaxial layers are nearly identical, and therefore, negligible reflection at the interface is expected. Once the matrix of metal and filler material is formed, the samples are ready for bonding. The bonding process starts with a Au/Zn/Au back-contact deposition on the bottom of the InP substrate using thermal evaporation and electroplating. The metal stack is annealed at 380–400°C for 1 min to allow for the formation of ohmic contacts prior to bonding. To remove remaining contaminants at the bonding interface, a thorough solvent clean, a 30-s O2 plasma ash, and a 20-min ultraviolet–ozone clean are performed. Samples are then loaded onto a Finetech flipchip bonder (FCB) (Berlin, Germany) for an aligned bonding. It is important to note that the pattern at the bonding interface must have reflection symmetry so that the regions of filler and metal can be aligned after flipping one of the samples prior to bonding. The bond is performed at 30 N/cm2 for 2 min in ambient air at 320°C. Samples are then transferred to a custom-designed high purity graphite fixture that applies >150 N/cm2 force while the sample is annealed in an ambient air oven at 300°C for 1 h. This pressure is maintained during a 10-min cool down. The bonded sample is then processed into devices with standard III–V wet-etch processing. The sample first undergoes a substrate removal process where the GaAs Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

substrate is removed with a 1:1 solution of NH4OH : H2O2, and the GaInP2 stop etch is removed with concentrated HCl. An infrared (IR) contact aligner is used to align the top metal contacts with metal features at the bonding interface. The top-cell mesa pattern is etched with H3PO4 : H2O2 : H2O (3:5:1) for the GaAs-based layers and concentrated HCl for the InP-based layers. The lower-cell mesa is slightly larger than the top cell to allow additional masking to protect the top-cell perimeter during the lower-cell mesa etch [15].

4. RESULTS A series of samples were fabricated for process development and for various testing purposes. This section will discuss the initial and current devices; the following section will discuss the associated process development steps. As an initial demonstration of the mechanical stability of the pillar-array bonding configuration, a GaAs cell (with 1mm2 active area) is bonded to a GaAs wafer. The resulting device is shown in Figure 3. The current–voltage characteristic of this device is compared with a conventional GaAs device in Figure 4. This initial device has a wide Au border that increases the dark/light area ratio to 1.16:1 and explains the majority of the observed 30-mV drop in Voc. The quality of the bonded cell is similar to that of a cloned cell bonded to a handle wafer using epoxy, indicating that bonding-related damage to the top cell is negligible. The most complicated device fabricated so far is a 3J GaInP (1.85 eV)/GaAs (1.42 eV) tandem bonded to a GaInAs (0.74 eV) cell. Figure 5 shows the current–voltage characteristics of this device. The device parameters for this device are Voc = 2.70 V, Jsc = 12.7 mA/cm2, fill factor (FF) = 83.0%, and efficiency = 28.4% under the 1-sun ASTM G173 direct spectrum.

C.-T. Lin et al.

Two-terminal metal-interconnected multijunction III–V solar cells

GaInP GaAs GaInAs

VOC = 2.7034 V 2 JSC = 12.659 mA/cm Fill Factor = 83.0% Efficiency = 28.39% 2

Figure 3. Infrared image of a GaAs cell (with a 1-mm active area) bonded to a GaAs wafer with an array of metal pillars. The bright features are metal. The wide borders around the cell and the four lines are the top contact grid of the device. The dots overlaying the grid lines are the pillar structures at the bonding interface. Figure 5. Official measurement result for a bonded GaInP/ GaAs/GaInAs triple-junction device (with AR coatings) under the 1-sun ASTM G173 direct spectrum.

Bonded GaAs cell

Current (mA/cm2)

0

coupling. In order to gauge the effect of optical losses at the bonding interface, a 3J device with a new bandgap combination that requires current matching is being fabricated and tested.

GaAs junction GaAs handle

Cloned cell, without bonding

-10

5. PROCESS REFINEMENT

GaAs junction

-20

0.0

0.2

0.4

0.6

0.8

1.0

Voltage (V) Figure 4. J–V measurement comparison between a GaAs cell bonded to a GaAs wafer and a cloned cell without bonding. The J–V curve is nearly identical, indicating minimal processrelated damage to the GaAs cell.

The efficiency is calculated based on the lower mesa area, which is larger than the top mesa because of processing constraints. The efficiency is also affected by a suboptimal top grid, which is not designed for 1-sun operation and could be redesigned to further reduce shadowing losses. Furthermore, this device uses a three-terminal structure for testing purposes, and the additional shadow loss caused by the interfacial metal feature is approximately 2.9% of the overall device area (11.7% if trench width is considered). Because this bandgap combination provides the third junction with excess photocurrent, this device is a good test of series resistance but is not a good test of optical

To progress from the initial one-junction device to the current 3J device, several refinements were made based upon intermediate results. The three most critical will be discussed here: (i) the optical filler material was changed from SiO2 to GaInP2, and then to AlGaAs; (ii) bondingrelated damage to the lower cell was mitigated; and (iii) an interfacial air gap was investigated and reduced. The first devices were bonded using a dielectric filler, but this was changed to GaInP2 to simplify the fabrication process and improve optical coupling and thermal conductivity. However, it was subsequently found that the GaInP2 filler layer suffers from a significant lateral etch, which widens the trench required for a 10-μm grid finger deposition to ~26 μm (Figure 6(a)). The resulting interfacial air gap adjacent to the metal reflects a significant amount of light (~50%), which increases the effective lower-cell shadowing loss from ~8% (10-μm shadow loss) to ~14% (10 μm shadow loss + 16 μm shadow loss at 50%). To mitigate this problem, the GaInP2 optical filler was replaced with AlGaAs, which etches isotropically and is less prone to lateral etching based on experimental observation. As shown in Figure 6(b), this reduced the trench width to ~17 μm, which is reasonable considering the flip-chip bonder alignment tolerance is approximately ±2–3 μm, and the grid finger width is 10 μm. Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

C.-T. Lin et al.

Figure 6. Scanning electron microscope image of a trench etched into the interfacial optical filler. In a subsequent step, a 10-μm-wide metal finger is deposited in this trench. Ideally, the trench width should be a few microns wider than the finger to allow for Au deformation during thermocompression bonding. Two optical filler materials are shown: (a) for the GaInP2 filler, lateral etching creates a (25.8 10)/2 = 7.9-μm-wide gap to either side of the subsequently deposited 10-μm-wide metal finger. (b) For the AlGaAs filler, this dimension is reduced to (17.6 10)/2 = 3.8 μm, which reduces the associated optical shadowing of the lower cells to an acceptable level.

Process-related damage motivated additional process changes. Initial attempts to fabricate GaAs (1.42 eV)/GaInAs (0.74 eV) two-junction (2J) test devices resulted in cells with Voc in the range of 1.15 V, which is significantly lower than the expected value (1.3 V). Testing of nonbonded GaInAs cells showed that the O2 plasma ashing step was responsible for this drop in the Voc. This is because the GaInAs junction was only 50 nm beneath the surface for these initial devices. Because some O2 plasma cleaning is necessary for surface treatment before bonding, the recipe was adjusted to reduce the plasma ashing time, and an ultraviolet–ozone treatment was added to ensure proper surface treatment.

Two-terminal metal-interconnected multijunction III–V solar cells

We also found a link between the bonding temperature and the device performance. The samples were initially annealed at 400°C for 1 h. However, it was found that under such conditions, the Voc and FF suffer a dramatic drop. The annealing temperature and time was reduced to 300°C for 1 h. In addition, the FCB bonding temperature was also reduced to 320°C [15]. Possible contributions to the thermal-related damages are speculated to be as follows: (i) thermal stability of the Ti/Pt/Au contact and (ii) window layer damage during the ambient anneal [16,17]. Further details regarding thermal-related damage that steered process modification decisions are discussed elsewhere [15]. This reduction in bonding temperature combined with a reduction in the O2 plasma treatment time increased the Voc of the GaAs (1.42 eV)/GaInAs (0.74 eV) 2J test devices from 1.15 to 1.27 V. To further protect the GaInAs cell from bonding-related damage, a 1-μm-thick buffer layer of heavily doped InP was added to the surface of the GaInAs cell. The heavy doping pushes the InP optical absorption edge from 1.35 to above 1.42 eV, which provides the necessary optical transparency for the lower cells. The Voc further improved from 1.27 to 1.29 V with this design change. An air gap at the bonding interface can also have a significant impact on the performance of the overall device. In a current-matched configuration, optical losses at the bonding interface will lead to an overall reduction in performance due to an undersupply of photons to the current limiting lower cells. Figure 7 shows a cross-sectional scanning electron microscope image of a bonded GaAs/GaInAs device with no air gap at the bonding interface. This demonstrates the possibility of intimate contact between semiconductor layers; a specular reflectivity scan suggests that low optical loss can be achieved. This is shown in Figure 8. Furthermore, a bonded GaAs/GaInAs tandem cell was fabricated and measured under a flash simulator to examine the overall series resistance (RS) of the device. The J–V curve at a concentration of 200 suns and Jsc versus Voc curve is shown in Figure 9. The slope of the measured J–V curve under 200 suns concentration closely matches that of a modeled device with a RS of 0.05 Ω cm2. The bonding interface only represents a part of this value because the RS of a device is the sum of series

Figure 7. Cross-sectional scanning electron microscope image of a well-bonded device showing no air gap across the interface between the (GaInP2) optical filler and the lower cells. The bright central features are the interfacial grid finger and the top grid finger. The void next to the interfacial grid is the trench etched for contact deposition. The width of this void can be reduced changing the optical filler from GaInP2 to AlGaAs. Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

C.-T. Lin et al.

Two-terminal metal-interconnected multijunction III–V solar cells

likely contributing significantly to our measured RS. The series resistance related to the top grid including spreading, finger, and contact resistance is estimated to be 0.033 Ω cm2 based on the method discussed by Moore [19].

6. CONCLUSION

Figure 8. Specular reflectivity scan comparison between three triple-junction devices (InGaP/GaAs/InGaAs) without AR coating: one monolithically grown inverted structure (MJ106n8), one poorly bonded (MK373s1n1), and one well bonded (MK373s2n1). The reflectivity of the well-bonded and monolithically grown devices is similar. The poorly bonded device has a significantly higher reflectivity at longer wavelengths that are vulnerable to an air gap at the bonding interface.

A hybrid bonding approach utilizing an array of metal– metal bonds and an optically transparent filler material has been developed. Results were presented for an initial GaAs device bonded to a GaAs conducting wafer and for a more recent GaInP/GaAs on GaInAs bonded 3J device. Process-related damage was investigated using 2J GaAs on GaInAs bonded test devices. By adjusting the process to mitigate plasma and thermal damage, the Voc of these devices was increased from 1.15 to 1.29 V. Optical coupling issues related to air gaps at the interface were also addressed by switching to an AlGaAs optical filler material. The specular reflectivity of a well-bonded device is comparable with a monolithic device, showing that good optical coupling at the interface is possible. Implementation of these process refinements has produced a bonded (GaInP + GaAs)/ GaInAs 3J device with a Voc of 2.70 V, a Jsc of 12.66 mA/cm2, an FF of 83.0%, and an efficiency of 28.39% under the 1-sun ASTM G173 direct spectrum. This illustrates the potential for the fabrication of high-performance multijunction solar cells from latticemismatched materials using a cell bonding approach. 4J bonded devices based on these refinements are promising for higher efficiency.

ACKNOWLEDGEMENTS

Figure 9. Illuminated J–V curve of a bonded GaAs/InGaAs twojunction device under 200 suns concentration is shown in the red curve. Modeled J–V curves with different series resistances are also shown to estimate the series resistance of the actual device. The curves indicate that the device has a series resistance that is closely matched to a model with a total series resis2 tance 0.05 Ω cm . This sets an upper bound for the series resistance of the bonding interface.

This material is based upon work supported as part of the Center for Energy Efficient Materials (CEEM), an Energy Frontier Research Center (EFRC) funded by the US Department of Energy, Office of Science, and Office of Basic Energy Sciences under award number DE-SC0001009. Part of this work is performed in the University of California, Santa Barbara Nanofabrication Facility, supported by the National Science Foundation and the National Nanofabrication Infrastructure Network (NNIN). E. E. Perl is supported by the National Science Foundation Graduate Research Fellowship under grant no. DGE-1144085.

REFERENCES resistances from the top grid, semiconductor stack, bonding interface, and back contact. Therefore, this measurement can be used to set an upper bound on the series resistance of the bonding interface at 0.05 Ω cm2. This value is lower than the previous reports of series resistance at a direct bonded interface [11–14,18]. Also note that the top grid is not optimized for high concentration and is

1. Geisz JF, Friedman DJ, Ward JS Duda, A, Olavarria WJ, Moriarty TE, Kiehl JT, Romero MJ, Norman AG, Jones KM. 40.8% efficient inverted triple-junction solar cell with two independently metamorphic junctions. Applied Physics Letters 2008; 93: 123505. 2. Guter W, Schone J, Philipps SP, Steiner M, Siefer G, Wekkeli A, Welser E, Oliva E, Bett AW, Dimroth F. Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

C.-T. Lin et al.

3.

4.

5.

6.

7.

8.

9.

10.

11.

Current-matched triple-junction solar cell reaching 41.1% conversion efficiency under concentrated sunlight. Applied Physics Letters 2009; 94: 223504. King RR, Law DC, Edmondson KM, Fetzer CM, Kinsey GS, Yoon H, Sherif RA, Karam NH. 40% efficient metamorphic GaInP/GaInAs/Ge multi-junction solar cells. Applied Physics Letters 2007; 90: 183516. Kazmerski LL. “Best research-cell efficiencies”, Available: http://www.nrel.gov/ncpv/ images/efficiency_chart. jpg, revision 08/2013, accessed 09/06/2013. G173-03 Standard tables for reference solar spectral irradiances: direct normal and hemispherical on 37° tilted surface. Book of Standards Volume: 14.04, American Society for Testing and Materials, West Conshohocken, PA, 2004. Friedman DJ. Progress and challenges for nextgeneration high-efficiency multijunction solar cells. Current Opinion in Solid State and Materials Science 2010; 14: 131. Friedman DJ, Olson JM, Kurtz SR. High-Efficiency III-V Multijunction Solar Cells. Handbook of Photovoltaic Science and Engineering, 2nd Edition. Luque A, Hegedus S (eds). Wiley: Sussex UK, 2010. McMahon WE, Lin C-T, Ward JS, Geisz JF, Wanlass MW, Carapella JJ, Olavarria W, Young M, Steiner MA, Frances RM, Kibbler AE, Duda A, Olson JM, Perl EE, Friedman DJ, Bowers JE. Metal pillar interconnection topology for bonded two-terminal multijunction III-V solar cells. Journal of Photovoltaics 2013; 3(2). Alexander F, Park H, Cohen O, Jones R, Paniccia M, Bowers J. “Electrically pumped hybrid AlGaInAs-silicon evanescent laser.” Optics Express 2006; 14(20): 9203–9210. Black A, Hawkins AR, Margalit NM, Babic DI, Holmes AL, Chang Y-L Jr., Abraham P, Bowers JE, Hu EL. Wafer fusion: material issues and device results. IEEE JSTQE 1997; 3(3): 943–951. Tanabe K, Fontcuberta I, Morral A, Atwater HA, Aiken DJ, Wanlass MW. Direct-bonded GaAs/InGaAs

Prog. Photovolt: Res. Appl. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/pip

Two-terminal metal-interconnected multijunction III–V solar cells

12.

13.

14.

15.

16.

17.

18.

19.

tandem solar cell. Applied Physics Letters 2006; 89: 102106. Tanabe K, Watanabe K, Arakawa Y. III-V/Si hybrid photonic devices by direct fusion bonding. Scientific Reports 2012; 2: 349. Law DC, Bhusari DM, Mesropian S, Boisvert JC, Hong WD, Boca A, Larrabee DC, Fetzer CM, King RR, Karam NH. Semiconductor-bonded III-V multijunction space solar cells. Conference Proceeding of 35th IEEE PVSC, 002237, (2009). Boca A, Boisvert JC, Law DC, Mesropian S, Karam NH, Hong WD, Woo RL, Bhusari DM, Turevskaya E, Mack P and Glatkowski P. Carbon nanotubecomposite wafer bonding for ultra-high efficiency III-V multijunction solar cells. Conference Proceeding of 36th IEEE PVSC, 003310, (2010). Lin C-T, McMahon WE, Ward JS, Geisz JF, Wanlass MW, Carapella JJ, Olavarria W, Young M, Steiner MA, France RM, Kibbler AE, Duda A, Olson JM, Perl EE, Friedman DJ, Bowers JE. Fabrication of twoterminal III-V multijunction solar cells. Conference Proceeding of 38th IEEE PVSC, pp. 000944–000948, (2012). Sehgal BK, Bhattacharya B, Vinayake S, Gulati R. Thermal reliability of n-GaAs/Ti/Pt/Au Shocttky contacts with thin Ti films for reduced gate resistance. Thin Solid Films 1998; 330: 146–149. Riesen van S, Bett AW. Degradation study of III-V solar cells for concentrator applications. Progress in Photovoltaics: Research and Applications 2005; 13: 369–380. Chiu PT, Law DC, Woo RL, Singer SB, Bhusari D, Hong WD, Zakaria A, Boisvert J, Mesropian S, King RR, Karam NH. Direct semiconductor bonded 5J cell for space and terrestrial applications. Conference Proceeding of 39th IEEE PVSC, (2013). Moore AR. An optimized grid design for a sunconcentrator solar cell. RCA Review 1979; 40: 140.