ALD grown bilayer junction of ZnO:Al and tunnel oxide barrier for SIS solar cell.

Various metal oxides are probed as extrinsic thin tunnel barriers in Semiconductor Insulator Semiconductor solar cells. Namely Al2O3, ZrO2, Y2O3, and La2O3 thin films are in between n-type ZnO:Al (AZO) and p-type Si substrates by means of Atomic Layer Deposition. Low reverse dark current-density as low as 3×10-7 A/cm2, a fill factor up to 71.3%, and open-circuit voltage as high as 527 mV are obtained, achieving conversion efficiency of 8% for the rare earth oxide La2O3. ZrO2 and notably Al2O3 show drawbacks in performance suggesting an adverse reactivity with AZO as also indicated by X-ray Photoelectron Spectroscopy.

Introduction

The increasing demand for global energy has shifted attention to solar energy [1]. Metal Insulator Semiconductor as well as Semiconductor Insulator Semiconductor (SIS) solar cells attracted intensified scientific research interest in the late seventies [2], [3] and also in recent times [4], [5]. Thereby, a Schottky-junction can be easily formed by the deposition of metal or transparent conductive semiconductor films on the absorbing semiconductor substrates. Whereas theoretical predictions claim a possible solar to electricity conversion efficiency η of 21% for Si based SIS solar cell[2], experimental efficiencies of η=12.8% for indium Tin oxide (ITO)/SiO2/p-Si [6], η=13% for ITO/SiO2/n-Si [7], η=14.1% for SnO2/SiO2/n-Si [8], η=8.5% for ZnO/SiO2/n-Si [9], and η=6.8% for Al-doped ZnO (AZO)/SiO2/p-Si [4] have been reported. In all of these studies a thin interfacial SiO2 layer was used, which is well known to offer a high quality in terms of thermal and chemical stability, as well as in terms of interfacial trap density. Notably, reported conversion efficiencies achieved for cells with ZnO or AZO are significantly lower compared to cells with e.g. ITO as Transparent Conductive Oxide (TCO).

Apart from Si-based thin film solar cells, a large potential is provided in the development of emerging alternative solar cell technologies, such as new absorber materials and nanostructures [10], [11], [12], [13]. Concerning the applicability of the SIS concept for alternative absorber films, the use of an extrinsic tunnel barrier oxide is required when no stable intrinsic oxide can be provided. Such oxides can be well grown, for instance, by Atomic Layer Deposition (ALD), which offers a high uniformity, high conformity, and excellent thickness controllability even on complex 3-dimensional structures [14].

ZnO has been extensively studied as TCO in recent years due to a high optical transmission, low resistivity, a high stability in aqueous environment and low material costs [15], [16], [17]. ZnO films can be grown by a variety of methods, including sputtering [18], chemical vapor deposition [19], spray pyrolysis [20], electron cyclotron resonance-assisted molecular-beam epitaxy [21], or pulsed laser deposition (PLD) [22]. In comparison to ZnO, whose resistivity lies in the range of 1–100 Ω cm, values as low as 10−4 Ω cm have been shown for Al-doped ZnO films [23]. AZO also offers a high transmittance and a sufficient large band gap of Eg∼3.3 eV [24]. Such doped ZnO can be also grown by means of ALD by using the trimethylaluminum precursor as Al doping source, which can result in a competitive film resistivity of 4.5×10−3 Ω cm [25].

In this report, we use AZO films deposited by ALD as TCO for metal oxide/p-type Si junctions in SIS solar cells. We show that commonly employed SiO2 can be successfully replaced by ALD grown Y2O3 or La2O3. By using X-ray Photoelectron Spectroscopy (XPS) one focus is set on the chemical stability of Al2O3, ZrO2, Y2O3, and La2O3 tunnel barriers during the ALD of the AZO film.

Experimental

As substrate, p-type monocrystalline (100)-Si with a resistivity of 3–6 Ω cm is used. After etching the native SiO2 in 2% hydrofluoric acid (HF) and subsequent rinsing in DI water and N2 blowing, the substrates are transferred into the ALD chamber (CambridgeNanotech, Savannah 100). The ALD system is capable to deposit homogenous films on 4-in. wafers. Thereby, the ALD-growth proceeds by exposing the substrate surface alternately to different gaseous precursors. One growth cycle consists of (i) a pulse in the ms-range of the metal containing precursor and its chemisorption onto the substrate, followed by an (ii) inert gas purge (20 sccm) to remove the excess precursor. Subsequently, (iii) the oxidizing precursor is pulsed in and reacts on the surface with the bound metal precursor. Finally, another (iv) inert gas purge removes gaseous reaction by-products. The base pressure of the ALD chamber during growth is kept at 1 Torr by using an Alcatel Adixen Pascal 2005 l rotary pump.

Different types of oxides, namely Al2O3 (grown from trimethylaluminum and H2O), ZrO2 (tetrakis-(dimethyl)-zirconium and H2O), Y2O3 (tris(methylcyclopentadienyl)yttrium and H2O), and La2O3 (Tris(N,N′-di-isopropylformamidinate)lanthanum and H2O) are deposited onto the substrates, which are kept at a temperature of 250 °C. The thicknesses of the deposited oxide layers range between 0.8 and 2 nm, whereat growth-rates and optimal pulse and purge times which are needed to settle down in the ALD-window are determined by using spectroscopic ellipsometry (alpha-SE, J.A. Wollam Co.) for film thickness measurements. Subsequently, an AZO layer is grown onto the interfacial oxide. The AZO films are formed from diethylzinc, trimethyaluminum (TMA) and H2O at substrate temperature of 240 °C during ALD. The ZnO doping by TMA is carried out by adding one pulse of TMA every 20th cycle of diethylzinc and H2O pulsing (ratio 20:1). Growth-rates, as well as needed precursor temperatures are depicted in Table 1.

Table 1

Growth-rate of the various oxides and needed precursor temperatures.

Oxide	Growth-rate (Å/cycle)	Precursor temp.
Al2O3	1.1	RT
ZrO2	0.9–1	75 °C
Y2O3	1.4	145 °C
La2O3	0.4–0.5	140 °C
ZnO	1.5	RT

After the deposition of the AZO film, a 1 µm-thick aluminum layer has been sputtered onto the sample and the top contact grid was processed by standard lithography and lift-off techniques.

For high-performing solar cells, it is of great importance to reduce the contact resistance as much as possible. For this aim the thickness of the top metallization and AZO film (250 nm), as well as the geometry of the top metal grid have been optimized as suggested by Bhakta et al. [26].

A shadowing-factor of 10% has been choosen for the fabricated cells. In order to determine the properties of the processed cells, dark I–V and I–V characteristics under illumination (AM1.5 condition) at different light intensities (50 and 100 mW/cm2) are measured. From the measured curves, the open-circuit voltage Voc, fill factor FF, and efficiency η of the cells are obtained. Sizes of the cells are varied in range from 1.175 cm×1.4 cm up to 2 cm×2 cm. In order to detect potential interfacial reactions due to the growth of the AZO layer, XPS measurements are carried out (Specs, Phoibos 150 MCD-9 detector). All obtained spectra are calibrated for the C1s state and peak positions are deduced by using the fitting and analyzing tool Casa XPS from Neal Fairly (VAMAS Processing Software) [27]. The spectra are measured by using at least three scans, a dwell time of 0.15 s, and a step size of 0.01 eV. As anode material Al is used and as entrance slit a size of 0.5×20 mm2 has been chosen. Base pressure of the XPS system during measurement is in range of 1–5×10−9 mbar. The samples have been transferred from the ALD chamber into the XPS system within 2 min.

In order to determine the spectral range of interest of the solar cells the External Quantum Efficiencies (EQE) have been measured (Newport Oriel, Quantum Efficiency measurement kid_QE-PV-Si). The system equipped with a Si reference detector and a 500 W arc Xe ozone free lamp (Newport Oriel) has been calibrated and optimized by using a crystalline-Si reference cell. The normalized EQE has been measured for wavelengths in the range of 310–1100 nm.

Additional Raman spectroscopy has been performed to assess potential interfacial strain. For this purpose a system (Witec, alpha300) equipped with an achromatic objective (Nikon, E Plan) has been used in back-scattering geometry. The samples are illuminated with a frequency-doubled Nd:YAG laser at low intensities of about 13 kW/cm² to avoid heating of the sample. In the Stokes Raman spectrum of crystalline silicon, the Brillouin zone-center first-order Raman active mode can be monitored to investigate strain in the silicon layer [28].

Results

In a first step, the transmittance of the grown AZO films is assessed by reflectometry. The measured transmittance of 100 nm-thick AZO layers is higher than 90% in the spectral range of interest. Afterwards, as a function of the film thickness the resistivity of the AZO films is measured and the results are depicted in Fig. 1.

Fig. 1

Resistivity versus the thickness of the AZO film. The inset shows reported values for ALD grown AZO and ZnO at different temperatures taken from Otto et al. [29].

The lowest value of 1.4 mΩ is obtained for a 100 nm thick AZO film, which agrees well with reported values for ALD grown films with similar thickness (see Fig. 1 inset) [29].

In Table 2, the reverse dark current Js, open-circuit voltage Voc, fill-Factor FF, and effective conversion efficiency η of the fabricated SIS solar cells are shown as a function of the oxide type. The oxide thickness is about 1 nm in all cases.

Table 2

Cell-parameters measured in dark (Js) and under illumination (AM1.5 condition) at 100 mW/cm2.

Oxide	Js [A/cm2]	Voc [mV]	FF [%]	η [%]
ZrO2	2·10−6	250	40	1.6
Y2O3	3·10−7	350	67	5.75
La2O3	3.3·10−7	527	71.3	8.05

The processed cells with Al2O3 as tunnel barrier exhibit very high reverse dark current Js in the mA-region independent of the barrier thickness. By using a 1.1 nm thick ZrO2 barrier, reverse dark currents are lowered down to Js=2×10−6 A/cm2. A low open-circuit voltage of Voc=250 mV and an adverse fill factor of about 40% lead to a low efficiency of η≈1.6%. Much better results are achieved by applying 8 ALD cycles of Y2O3 (1.1 nm) as thin oxide barrier in between the Si absorber and 250 nm AZO. The reverse dark current Js=3·10−7 A/cm2 is about one order lower, the fill factor of FF=67% is relatively high and open-circuit voltage is increased to Voc=350 mV. These values result in an effective efficiency of η=5.75%.

The best results, however, have been achieved by using a thin La2O3 barrier (0.9 nm). In this case, a reverse dark current as low as 3.3×10−7 A/cm2, a low contact resistance (Rs=1.78 Ω), a high shunt resistance (RH=192 kΩ), an open-circuit voltage as high as 527 mV, and a fill factor of FF=71.3% are obtained.

From these curves shown in Fig. 2, an effective conversion efficiency of η=8% is obtained for a cell-area of 1.4 cm2. The short-circuit current is Isc=30.02 mA for this cell.

Fig. 2

Plot of the current–density versus voltage versus voltage of the AZO/La2O3/Si solar cell, measured in dark, at 50 mW/cm2, and 100 mW/cm2 (left y-axis) and the corresponding output power characteristic (right y-axis). The inlet shows the current versus voltage characteristic (left y-axis) and derive dV/dI (right y-axis) measured in dark. The points are modeled data.

The unsealed cells show degradations over time: the solar cell with La2O3 tunnel barrier shows a drop in RH resulting in a Fill-Factor below 60%, 6 months1 after fabrication. Also Isc is lowered but notably Voc remains almost stable.

XPS analysis

From a theoretical point of view, Al2O3 should offer the highest conduction-band offset to Si with ϕB=2.8 eV compared to the other oxides ZrO2 (ϕB=1.5), L2O3 (ϕB=2.3 eV), and Y2O3 (ϕB=2.3 eV) [30], [31]. As the reserve current of the cells depends as Js∼exp[−qϕB/kT], the lowest Js is expected when using Al2O3 as tunnel barrier. This is, however, in contrast to our results. In order to clarify the impact of the AZO deposition on the chemical stability of the interfacial oxide, XPS measurements have been carried out.

For this purpose, onto the various oxides with a thickness of ∼1 nm grown on Si, a 3 nm thick AZO layer is deposited. These samples are analyzed then by XPS and referenced to the corresponding pure oxides deposited on Si with the same thickness.

In Fig. 3, the metal peaks of the oxides, namely the Al 2p state of Al2O3 at binding energy of 73.72 eV (a), the Zr 3d state of ZrO2 at 181.68 eV (b), the Y 3p state of Y2O3 at 300.49 eV (c), and the La 3d state of La2O3 at 835.02 eV (d) are shown, respectively with and without the AZO layer. In all cases a significant shift of the metal-oxide peak toward higher binding energies is visible when 3 nm thick AZO is deposited onto the oxides. Strongest shifts are indicated for Al2O3 (0.82 eV) and ZrO2 (0.88 eV) followed by La2O3 (0.57 eV) and Y2O3 (0.54 eV). Although the correct interpretation of these shifts is not trivial, it is clear that the shifts are correlated to an interaction between the oxides and AZO. On the out-diffusion of Zn species from ZnO into an Al2O3 layer during low temperature annealing with the creation of deep-level defects has been reported by Wang et al [32]. Nevertheless, no significant shift of the Zn 2p state can be found making the formation of ZnMO compounds unlikely. However, the complete substitution of metal oxide atoms located within the tunnel barrier by diffused Zn is quite possible. The observed shift of the metal oxide peaks toward higher binding energies suggest rather the formation of non-stoichiometric MO (M=metal) in a higher oxidation state (as O-atoms offer a high electron negativity) or the insertion of exceeding OH-groups as reported for Y2O3 during ALD leading to Y-OH bonds [33]. This implies that during the growth of the AZO film, oxygen or OH-groups are provided in excess leading to a non-stoichiometric metal oxide. On the doping of ZnO by Y, Zr, and La was reported [34], [35], [36]. For instance in carefully grown AZO films, only stoichiometric Al2O3 can be found, indicating the complete substitution of Zn by Al [37], [38]. In our case, the potential metal dopant species are located in excess within the tunnel barrier. Hence, a possible doping of ZnO by diffused metal atoms from the tunnel barrier may result additionally in non-stoichiometric MO compounds.

Fig. 3

XPS spectrum for (a) the Al2p state of 1.2 nm thick Al2O3 deposited on Si with and without an AZO layer (3 nm), (b) the Zr3d state of ZrO2 (1.2 nm) w and w/o AZO (3 nm), (c) the Y3p state of Y2O3 (1.1 nm) w and w/o AZO (3 nm), and (d) the La3d state with satellite peak on the left side of La2O3 (0.9 nm). The magnitudes of the peak-shifts are indicated.

Fig. 4 shows the relative EQE for the solar cell with La2O3 as tunnel barrier. A relatively good EQE is obtained for wavelengths at around 400 nm and in the range between 500 nm and 700 nm.

Fig. 4

Plot of the relative external quantum efficiency of the AZO/La2O3/Si cell.

The observed decrease of the EQE at wavelengths between 900 nm and 1100 nm and below 400 nm is typical for the band gap cut-off in the Si absorber material.

It is noted that no interfacial strain was detected by Raman spectroscopy as a Raman shift of 521 rel. 1/cm was observed which is typical for unstrained silicon.

Conclusions

Induced by the deposition of the AZO layer, a chemical instability of the metal oxide lowers the effective barrier-height by the insertion of defect states into the oxide band-gap [39]. As deduced by temperature dependent I–V measurements, effective barrier heights ϕB (not shown) of the thin Al2O3, ZrO2, Y2O3 and La2O3 layers are indeed significantly lower compared to data from literature for pure oxides on Si [40], whereat influence of the enhanced tunnel probability for electrons must be considered. Non-stoichiometry of the metal oxide, out-diffusion of metal species from the tunnel barrier, as well as a possible substitution of metal atoms by Zn within the barrier induce amounts of interface traps at the AZO/tunnel barrier and the oxide/Si interface, which affects the cell performance adversely. On this basis, we can conclude that the reactivity of the AZO layer lead to unfavorable characteristics of the cell, which most likely can explain the main part of the gap between experimentally determined effective efficiencies and theoretical predictions [2]. It is said that overall efficiencies of the fabricated cells should be improvable by applying e.g. suitable anti-reflection coatings or a higher dopant level of the substrate. Nevertheless, obtained efficiencies for La2O3 are higher compared to reported values for AZO/SiO2 [4] and competitive to ZnO/SiO2 junctions [9]. Generally, we attribute the lower efficiencies reported for AZO/oxide junctions in SIS cells to a higher reactivity of ZnO compared to, for instance, ITO. In order to reduce the reactivity of ZnO a lowering of the deposition temperature or changing the dopant species [41] may be helpful.

In summary, we used ALD grown Al2O3, ZrO2, Y2O3 and La2O3 as tunnel barrier in SIS solar cells with n-type AZO and p-type Si as absorber. While unexpectedly the samples with Al2O3 and ZrO2 show either high dark reverse currents or low open-circuit voltages, the rare earth oxides Y2O3 and La2O3 show promising results. For 1.1 nm thick Y2O3 we obtained a conversion efficiency of η=5.75% and an open-circuit voltage of Voc=350 mV. A high conversion efficiency of η=8%, low reverse-current Js=3.3·10−7 A/cm2 and a good Voc=527 mV are achieved for AZO/La2O3/p-Si samples with 1.4 cm2 cell-area. XPS measurements reveal the transformation of the former stoichiometric tunnel barrier to a non-stoichiometric oxygen- or OH-rich barrier during the deposition of the AZO film at 240 °C. Fewest changes in binding energy of the metal oxide states are observed for La2O3 and Y2O3 indicating an enhanced chemical stability toward AZO. These rare earth oxides are therefore well suited for AZO/insulator/p-absorber solar cells, in which an extrinsic oxide as tunnel or passivation layer is necessary.