1. Introduction

Solar cells composed of group Ⅴ and group Ⅲ elements have achieved the highest efficiency of 35.1% under AM0 spectrum^[1], which is obtained from a multi-junction solar cell (MJSC). Solar cells of this type have been widely used for space and terrestrial purposes. These devices are typically manufactured by metal organic chemical vapor deposition (MOCVD), and tunnel junctions (TJ) are used to interconnect the sub-cells for the convenience of carrier transport. A tunnel junction that can be applied in a cell device needs to be optimized for the best electrical and optical characteristics.

With the increasing bandgap of solar cells, the demand for broader bandgap TJ is becoming increasingly urgent. AlGaInP and Al$_{0.2}$GaAs were chosen as materials for the 1st and 2nd junction cell in 5-junction solar cell, thus the traditional Ga$_{0.52}$InP/Al$_{0.4}$GaAs TJ cannot fulfill the requirement of optical transparence due to its narrow bandgap. One promising TJ is AlGaInP/Al$_{0.8}$GaAs, which has a broader bandgap than Al$_{0.2}$GaAs subcell and which makes the TJ layer more optically transparent to the underlying middle cell than a Ga$_{0.52}$InP/Al$_{0.4}$GaAs TJ. However, a high doping concentration becomes more difficult to achieve in a broader bandgap n-type material.Tellurium is a promising n-type dopant for tunnel junction between top and middle cells in broadband solar cells^[2]. Material doped with tellurium allows higher degeneracy of the electron states in the conduction band, which provides large tunneling current densities. Another benefit of tellurium doping is the relatively low diffusion of Te in epi-layer due to its large atomic size^[3]. However, tellurium may produce a large compressive strain in GaInP or AlGaInP due to its large atomic size. Moreover, tellurium, as a surfactant, can affect the layer composition by slightly unbalancing the incorporation of gallium and indium into the crystal lattice^[4]. All of the above phenomena make the lattice-matched epi-layer lattice mismatched to the substrate. In addition, when incorporated into MJSCs, a sharp doping profile with high carrier concentration is required in a thin epitaxial layer (15-50 nm) in the TJ^[5]. The aim of this paper is to study how to acquire an excellent tellurium doped ultrabroad band TJ that can be applied to multi-junction solar cells. To simplify the experimental procedure, parameter optimization was carried out in Ga$_{0.52}$InP/Al$_{0.4}$GaAs TJ because its properties are similar to that of AlGaInP/Al$_{0.8}$GaAs. The optimized growth conditions were then transplanted to AlGaInP/Al$_{0.8}$GaAs TJ.

2. Experiment

Tellurium doped GaInP and AlGaInP samples were grown in a horizontal, low-pressure, AIX-2600 reactor at 650 ℃, n type GaAs oriented (100) 15$^\circ$ off towards the nearest (011) plane was used as substrates. The source materials were trimethylgallium (TMGa), trimethylindium (TMIn), purified phosphine (PH$_{3})$, purified arsine (AsH$_{3})$ and diethyltelluride (DETe). Pd-purified hydrogen was used as a carrier gas. The growth parameters varied in these experiments, including growth temperature, DETe flow, and TMGa flow, while keeping the flow of TMIn constant. To achieve a lattice-matched epi-layer, high resolution double crystal X-ray diffraction (HRXRD) rocking curves and photoluminescence (PL) measurement were used to characterize material composition. Electrochemical capacitance voltage (ECV) profiling was used to measure the doping concentration of the TJ, and an $I$-$V$ test was applied to confirm the tunneling effect.

3. Results and discussion

When tellurium is incorporated on the phosphorus site as an n type dopant in GaInP and AlGaInP layers, it produces a compressive strain on the lattice matched epi-layers because its atom size is much bigger than that of phosphorus. In addition, Te will act as a surfactant when the surface coverage of Te reaches a critical value, it tends to control group Ⅲ adatom incorporation and create a positive feedback mechanism with indium^[6]. Therefore, the affinity of tellurium and indium manifests and the latter begins to be preferentially incorporated into the alloy. A higher indium content in the GaInP will make possible a higher Te incorporation. High concentration of tellurium required for tunnel junctions causes a lattice mismatch between the epitaxial layer and the substrate. The HRXRD rocking curve in Fig. 1 illustrates the variation of the lattice. The TMGa flow is 11.2 sccm in undoped GaInP, which is lattice matched to substrate. However, the epi-layer became lattice mismatched when doped with tellurium, as shown in sample 129. To reduce the compressive strain, we intentionally increased the TMGa flow during the growth of Te doping GaInP single layer. Sample 130-132 shows a decreasing lattice mismatch, the inset in Fig. 1 describe this tendency. Sample 132 is nearly lattice matched (as confirmed by the PL measurement), whose peak located in 675nm has a gallium mole fraction of 0.53. Sample 133 in Fig. 1 illustrates the XRD curve of GaInP: Te/AlGaAs: C TJ, peak A is probably due to the tensile strain produced by carbon doping in the AlGaAs layer. To acquire a high quality lattice matched epi-layer, we adopt the strain balance method. In detail, the tensile produced by carbon doped AlGaAs was used to balance the compression produced by Te doped GaInP. With the help of gallium mole variation, the epi-layer again became lattice-matched to the substrate.

Tellurium acts as a surfactant in the MOCVD growth epilayer^[7]. It accumulated in the surface of the newly grown layer and was slowly incorporated into the next grown layer, even when the DETe valve was turned off. The remaining tellurium would neutralize the carbon dopant in the subsequent AlGaAs layer and decrease the tunneling effect of the TJ. To desorb the residual tellurium, the wafer was heated up by approximately 50 ℃ in phosphine and then the temperature was adjusted for growth of the AlGaAs layer, the corresponding temperature variation and valve switching of sample 142 are shown in Fig. 3. This measure caused an obvious increasing carbon doping concentration in the AlGaAs layer, and resulted in the improvement of the peak current of the corresponding TJ.

To obtain the designed doping concentration, sample 135 was grown with gradient Te doping GaInP, as illustrated in Fig. 4. The corresponding numerical result is listed in Table 1.

The inset in Fig. 4 shows that the doping concentration is linear-related to DETe flow, thus the DETe flow needed for TJ could be calculated by extrapolation and then applied to the fabrication of broadband AlGaInP/AlGaAs TJ. The doping profile of an optimized TJ (sample 142) is shown in Fig. 5, the calculated flow of DETe is 4.2 $\times$ 10$^{-6}$ sccm, and both the p-and n-type doping fulfill the requirement of TJ, theoretically.

By further optimizing the growth parameter, GaInP:Te/Al$_{0.4}$GaAs:C broadband TJ was fabricated in a 0.18 $\times$ 0.176 cm$^{2}$ wafer, whose peak current was $\frac{2.07}{(0.18\times 0.176)}$ $=$ 65.3 A/cm$^2$. The corresponding Ⅰ–Ⅴ characteristic is shown in Fig. 6.

There is more difficulty in achieving a high doping concentration in ultrabroad band AlGaInP/AlGaAs TJ because its band gap is broader than that of GaInP/AlGaAs TJ. The problems of lattice mismatching and turn on/off effect of DETe also exist in ultrabroad band TJ. Therefore, the doping characteristics of AlGaAs and AlGaInP were studied separately. In detail, GaAs: Si/Al$_{0.8}$GaAs TJ and AlGaInP:Te/GaAs:C TJ were studied to achieve the optimized parameters of Al$_{0.8}$GaAs and AlGaInP involved in ultrabroad band TJ.

The method used in optimizing GaInP/Al$_{0.4}$GaAs broadband TJ was applied to the growth of AlGaInP/Al$_{0.8}$GaAs TJ, and a peak current density of $\frac{0.75}{0.35\times 0.35}$ $=$ 6.1 A/cm$^2$ was obtained in a 0.35 $\times$ 0.35 cm$^{2}$ wafer; the corresponding Ⅰ–Ⅴ characteristic is shown in Fig. 7.

4. Conclusion

Broad and ultrabroad band tunnel junctions using DETe as an n-type dopant were fabricated, the growth temperature, valve switching and flow of DETe parameters were studied. Our experiment demonstrated that tellurium acts as a surfactant during MOCVD growth and it introduced positive feedback with indium, which resulted in lattice mismatching between substrate and epi-layer. Predoping and desorption of DETe before and after growth were used to weaken the effect of turn-on and off. A high peak current needs an appropriate flow of DETe, consequently, by all of the the measures mentioned above, a GaInP/AlGaAs broadband TJ with a peak current of 65.3 A/cm$^2$ and an AlGaInP/AlGaAs ultrabroad band TJ with peak current of 6.1 A/cm$^2$ were fabricated.