42.3% Efficient InGaP/GaAs/InGaAs concentrators using bifacial epigrowth

42.3% Efficient InGaP/GaAs/InGaAs Concentrators using Bifacial Epigrowth P. Chiu1, S. Wojtczuk1, X. Zhang1, C. Harris1, D. Pulver1, and M. Timmons2 1 Spire Semiconductor, 25 Sagamore Park Road, Hudson, NH 03051 2 Independent Consultant, Durham, NC, 27703 ABSTRACT

processing steps could potentially lead to low yields and higher costs.

Spire Semiconductor has demonstrated a new bi-facial epigrowth manufacturing process for InGaP/GaAs/InGaAs N/P tandem concentrator cells. A lattice-mismatched 0.94 eV InGaAs cell is epitaxially grown on the backside of a lightly doped, N-type GaAs wafer, the epiwafer is flipped, rinsed, and 1.42 eV GaAs and 1.89 eV InGaP cells are grown lattice matched on the opposite wafer surface. Cells are then made using only standard III-V process steps. NREL verified that 1 cm cells achieve 42.3% efficiency at 406 suns, AM1.5D, 25qC. This establishes a new world record, exceeding the previous record efficiency of 41.6% set by an InGaP/InGaAs/Ge cell. The bi-facial tandems exhibit superior performance relative to Ge based triple junction cells due to improved current matching. In addition they are simpler to fabricate than the inverted metamorphic (IMM) process because epitaxial liftoff and wafer bonding are not required.

Spire Semiconductor’s bifacial approach utilizes a 1eV InGaAs bottom subjunction to improve current matching with the top junctions, but avoids potential yield issues associated with the IMM process (wafer bonding and epitaxial liftoff). The InGaAs bottom cell is grown on the backside of the GaAs wafer in the bifacial process. The wafer is flipped, rinsed, and the top two junctions are grown on the frontside of the wafer. The thick GaAs substrate isolates dislocations generated in the InGaAs subjunction and prevents those defects from affecting the top junctions. Previously, Spire Semiconductor reported near record performance of 41.4% at 334 suns using this approach [4,5]. In this paper we discuss additional improvements made on the 41.4% cell to reach a 42.3% efficiency at 406 suns.

INTRODUCTION

The bifacial tandem cells were grown on N-type 4 inch (001) GaAs wafers, with the epi-ready surface miscut 10 degrees off toward the (111)A. The substrate needs to be transparent to infrared light past the cutoff of GaAs at 870 nm. To limit free carrier absorption, the main absorption past the cutoff of GaAs, N-type GaAs wafers were 17 -3 purchased with a low doping of ~10 cm . For a doping 17 cm-3 N-type GaAs, the free carrier absorption 10 -1 coefficient is ~0.6 cm between 1-2 Pm [6]. This results in less than 4% QE loss through a 650 Pm thick wafer.

The concentrator solar cells most widely adopted by module manufacturers utilize a triple junction design where lattice matched InGaP and InGaAs top junctions are grown on a Ge substrate. Spectrolab used this latticematched design to achieve an efficiency of 41.6% at 364 suns [1]. However, the Ge junction is poorly current matched, absorbing approximately two times more photons than the top two junctions. One way to improve the InGaP/InGaAs/Ge tandem is to grow metamorphic top and middle cells to achieve better current matching and higher Jsc. The improved current matching comes at the expense of minority carrier lifetime lowering dislocation defects. Due to these defects, the In0.65Ga0.35P/In0.17Ga0.83As/Ge metamorphic cell actually has a 0.5% lower efficiency than the lattice matched InGaP/InGaAs.Ge cell [2]. An alternative approach to achieve higher efficiency is to grow lattice matched InGaP and GaAs cells on GaAs substrates, and replace the Ge junction with a lattice mismatched InGaAs cell with a ~1 eV bandgap. The inverted metamorphic (IMM) process can maintain lattice match for the top and middle cells by growing an inverted cell stack with the 2% lattice mismatched InGaAs cell grown last. The top and middle cells remain defect free because dislocations from the InGaAs cell thread upward during growth. The IMM cell achieved 40.8% at 326 suns [3]. The disadvantage of the IMM cell is that a carrier wafer must be bonded on top of the InGaAs cell, followed by an epitaxial liftoff of the original GaAs substrate. These

978-1-4244-9965-6/11/$26.00 ©2011 IEEE

EXPERIMENTAL PROCEDURE

The wafers were epitaxially grown in a Veeco E450, low pressure metal organic chemical vapor deposition (MOCVD) reactor with a capacity of 13x4” wafers. For these growths the hydrogen mainflow was 120 slpm and pressure was 50 torr. Metal-organics used were TMIn, TMGa, TMAl and hydrides were 100% arsine and phosphine. Dopant sources were a DMZn, DiSi, CBr4, and DeTe. The InAlGaAs grading layer is grown first on the (111)B side (non epi-ready side) on the GaAs wafer followed by the lattice mismatched 0.95eV N/P InGaAs cell. After bottom cell growth, the epiwafer is then simply flipped in the MOCVD glove box, and a tunnel junction, a 1.42eV GaAs middle cell, a tunnel junction, and the final 1.89eV InGaP cell are grown lattice matched on the (111)A (epi-ready) GaAs wafer surface. The cell process after epigrowth is relatively simple and the main steps are outlined here. The backside 400 nm thick InGaAs cap is wet etched to remove a layer of InGaAs “damaged” by phosphine from the final InGaP growth. CrNiAu is evaporated over the entire wafer backside and acts as a back contact for the cell and

000771

ISOTYPE CELL RESULTS

0.8 0.7 0.6

EQE

optical mirror. New positive resist is applied and openings are made for the gridlines in an amine image reversal process for clean metal liftoff. The evaporated gridlines of the high efficiency cells reported here are 4 Pm of TiPtAu. A second photolithography step is used to define areas over the busbars where the AR coating will be removed by liftoff. The GaAs cap is then stripped in a citric-based etch selective against the InAlP window. A double layer MgF2/ZnS antireflection coating is evaporated and the AR film over the busbars is lifted off in acetone. The wafer fronts are protected with resist, mounted on tape, and then diced. The dicing is the last fabrication step and defines the cell junction area.

0.5 0.4 0.3

constant base data graded base data constant base fit grade base fit

0.2 0.1 0.0 1000

In order to improve the previous 41.4% bifacial cell, we focused on improvements in the voltage and current of filtered InGaAs isotype, filtered GaAs, and InGaP isotype cells. For the InGaAs bottom cell, we changed from a base with a constant doping of 6x1016 cm-3 (L829-990-1) to a 16 -3 17 -3 linear doping grade from 6x10 cm to 4x10 cm (L829-991-1). The purpose of the base doping grade is to create an electric field that effectively enhances the base diffusion length [7]. Equation 1 describes the diffusion length in the presence of doping electric field.

Ldh

Lb ( E

qLb  1) kT

1200

1400

Wavelength (nm)

Fig. 1: Experimental QE curves for InGaAs isotype cells with constant (L829-990-1) and graded (L829-9911) base doping. Theoretical curves (solid lines) generated by QE model are also plotted for comparison. Table 1: Summary of illuminated IV data at 1x and 500x for InGaAs isotype cells with constant (L829-9901) and graded (L829-991-1) base doping.

(1)

where Ldh is the base diffusion length in a cell with a graded base, Lb is the base diffusion length in a cell with constant base doping, and E is the electric field. Using the doping conditions from above, we calculate an E of 225 V/cm. We have fit the measured QE curve for the InGaAs cell with a constant base doping using a standard QE model [8], and determined an Lb of approximately 3 Pm. Substituting the values of Lb and E, we find that Ldh is 11 Pm, which is more than three times greater than Lb. Using the same QE model we can generate a theoretical QE curve for a cell with a graded base and Ldh equal to 11 Pm. Integration of the model QE curves of InGaAs cells with a constant and graded base doping shown in Fig. 1, indicates an increase in the Jsc of 0.5 mA/cm2. The measured QE curve for the cell with the graded base is shown in Fig. 1. Integration of the experimental QE curves 2 indicates an increase in the Jsc of 0.52 mA/cm , that agrees well with that predicted by the QE model. Moreover, illuminated IV measurements summarized in Table 1 also indicate an improvement in Voc of the cell by 22 and 15 mV at 1x and 500x respectively with the use of a graded base. The increase in voltage is attributed to a reduction in surface and bulk recombination. It should be noted that doping grades do always enhance cell performance because it can be difficult to increase the doping required for the grade without reducing the lifetime as well. In the case of the InGaAs cell, the minority carrier lifetime is insensitive to the increase in base doping because is it dominated by SRH dislocation defects.

978-1-4244-9965-6/11/$26.00 ©2011 IEEE

1x Voc

500x Voc

(mA/cm )

(V)

(V)

Constant

13.59

0.372

0.607

Graded

14.11

0.394

0.621

Base doping

1x Jsc 2

In the case of the filtered GaAs, we improved the current by adding a pseudomorphic InGaAs layer between the base and the BSF. The purpose of the pseudomorphic In0.01GaAs layer was to recover some of the current lost due to substrate absorption. In the bifacial cell, the substrate is between the middle and bottom cells. In addition to the previously discussed free carrier absorption, the substrate also absorbs light near the band edge, pushing the effective cutoff of the substrate from 872 to 890 nm. The light between 872 to 890 nm cannot be collected by a lattice matched GaAs middle cell or the InGaAs bottom cell [4,5]. Increasing the cutoff of the middle cell with the pseudomorphic layer allows us to recover some of the wasted current without affecting the current of the bottom cell. Choosing the composition and thickness of the InGaAs layer is critical to cell performance. To minimize the effect of any heterojunction barriers on minority carrier transport, the In composition was kept at 1%. Discontinuities in the valence and conduction band are