Titanium dioxide hole-blocking layer in ultra-thin-film crystalline silicon solar cells

One of the remaining obstacles to approaching the theoretical efficiency limit of crystalline silicon (c-Si) solar cells is the exceedingly high interface recombination loss for minority carriers at the Ohmic contacts. In ultra-thin-film c-Si solar cells, this contact recombination loss is far more severe than for traditional thick cells due to the smaller volume and higher minority carrier concentration of the former. This paper presents a novel design of an electron passing (Ohmic) contact to n-type Si that is hole-blocking with significantly reduced hole recombination. This contact is formed by depositing a thin titanium dioxide (TiO2) layer to form a silicon metal-insulator-semiconductor (MIS) contact. A 2 {\mu}m thick Si cell with this TiO2 MIS contact achieved an open circuit voltage (Voc) of 645 mV, which is 10 mV higher than that of an ultra-thin cell with a metal contact. This MIS contact demonstrates a new path for ultra-thin-film c-Si solar cells to achieve high efficiencies as high as traditional thick cells, and enables the fabrication of high-efficiency c-Si solar cells at a lower cost.

the metal Ohmic contacts, becomes the main obstacle for high performance. The key to solve this issue is forming carrier-selective contacts in c-Si solar cells, which can provide both lower minority carrier recombination velocity and more efficient majority carrier transport. Traditional c-Si solar cells use a diffused emitter and back surface field (BSF) to create carrier-selective layers. However, it has been shown that recombination at the metal/silicon interface still causes more than 40% of the total recombination losses [5,6] . To address the challenge of further controlling the contact recombination loss, various advanced cell designs have been demonstrated. One example is the heterojunction solar cell with an intrinsic larger bandgap thin-film (HIT) [7][8][9] . However, the a-Si:H layer in HIT cells has high parasitic absorption and high defect concentrations that results in a loss in the short circuit current (J sc ) [10,11] . Another method is to deposit a thin tunneling silicon dioxide (SiO 2 ) layer as a carrier selective contact [12][13][14][15] . However, to achieve low contact resistance, the thickness of the tunneling SiO 2 layer has to be very precisely controlled, and challenging for large-scale manufacturing. Recently, c-Si solar cells with Si/organic and/or Si/metal-oxide heterojunctions [16,17] have been demonstrated, providing an alternative solution of carrier-selective layer. However, those designs suffer the quality of surface passivation and the low V oc .
In this paper, we demonstrate a carrier-selective contact formed by depositing a thin layer of TiO 2 between the metal and n-type Si to form a metal-insulator-semiconductor (MIS) contact, which can effectivity block holes without compromising the conductivity for electrons. TiO 2 has a large valance band (VB) offset (DE v ), resulting in holes being blocked.
Meanwhile, good passivation can be provided by TiO 2 on an n-type Si surface, achieving a surface recombination velocity (SRV) as low as 100 cm-sec -1 [18][19][20] . TiO 2 can thus prevent holes from recombining at this interface or diffusing into the metal. For the conduction band (CB), TiO 2 can unpin the Fermi level at the Si surface, eliminate the Schottky barrier, and pin the Fermi level of the metal very close to the CB edge of Si [21,22] . This MIS contact thus has a small CB offset (DE c ), allowing electrons to transport freely through the hole blocking layer. With the aligned CBs, this MIS contact also has good tolerance for thickness variations of the TiO 2 layer. We experimentally demonstrate TiO 2 carrierselective contact on an ultra-thin-film c-Si junction solar cell, which can reduce the SRV at the contacts and suppress the recombination current. A 2 µm thick c-Si solar cell with the TiO 2 MIS hole-blocking contact achieves a V oc of 645mV, which is significantly higher than recently published c-Si cells of similar thickness [1][2][3][4] .
To demonstrate the benefits of carrier-selective contacts in solar cells, we simulated the effects of applying carrier-selective contacts to various Si solar cells using Synopsys Technology Computer-Aided Design (TCAD). These simulations consider both intrinsic recombination (Auger and radiative recombination) and extrinsic recombination (Shockley-Read-Hall (SRH), surface and contact recombination). The simulated cell structure is illustrated in Figure 1a, which consists of a lightly p-doped bulk absorber with a doping concentration of 10 16 cm -3 , and an n + emitter and a p + back surface field (BSF) layer, each with a doping concentration of 10 20 cm -3 and thickness of 50 nm. The cell thickness varies from 1 µm to 100 µm and the minority carrier lifetime is assumed to be 1 ms. In order to demonstrate the effects of carrier-selective contacts, the minority carrier recombination velocity at the contacts is set at 10 2 cm-sec -1 for both electron and hole selective contacts and 10 7 cm-sec -1 for metal contacts. The surface recombination velocity (SRV) at the passivated surfaces is assumed to be 10 cm-sec -1 . The recombination currents are analyzed under different mechanisms, including bulk, BSF, emitter, and contact recombination. The bulk recombination current accounts for the SRH, Auger and radiative recombination in the bulk region. The front/rear recombination current includes all the recombination at the emitter/BSF and at the front/rear interfaces. The contact recombination current includes the recombination at both front and rear contacts.
The V oc of cells with and without carrier-selective contacts are plotted versus the cell thickness in Figure 1b. It can be clearly observed from this plot that the carrier-selective contact can improve the V oc of cells with thickness from 1 µm to 100 µm, with the enhancement far more prominent in the thinner cells. In the 1 µm cell, the carrier-selective contact improves the V oc by 60 mV, while the carrier-selective contact improves the V oc by 12 mV in the 100 µm thick cell. Such a difference in V oc enhancement for different cell thicknesses can be explained by analyzing the components of the recombination current density J 0 , as shown in Figure 1c. In a 1 µm thick cell with metal contacts, the contact recombination current (J 0,contact ) is significantly larger than the other components, contributing to ~90% of J 0 . Applying carrier-selective contacts effectively suppresses this J 0,contact and reduces J 0 from 12 fA-cm -2 to 1.2 fA-cm -2 , resulting in a V oc improvement of 60 mV in the 1 µm thick cell. In the 100 µm thick cell with metal contacts, bulk and contact recombination current each contribute to ~40% of the total recombination current. Carrierselective contacts can therefore only reduce J 0 from 24 fA-cm -2 to 15 fA-cm -2 and improve V oc by 12 mV, which is less significant than the thin cells.
On the other hand, applying carrier-selective contacts enhances the increase in V oc achieved through decreasing the cell thickness. In the metal contact cell, the V oc is increased by only 20 mV as the cell thickness decreases from 100 µm to 1 µm, while that change is enhanced to 60 mV in the carrier-selective contact cells. Especially, the J 0, contact in the metal contact cells becomes the majority term in J 0 when the cell is thinned down below 10 µm.  The key to the TiO 2 MIS contact is the property of interface passivation at the TiO 2 /n-Si interface, which can be improved by post deposition annealing 26 . Samples with TiO 2 MIS contacts were annealed in forming gas at different temperatures ranging from 300 o C to 500 o C for 90 sec. In Figure 3, it can be seen that as the annealing temperature increases from 300 o C to 450 o C, J sc increases significantly by 13% and J 0 drops by 43%, which is equivalent to 16 mV increases in V oc . These results show that annealing improves the passivation effect of TiO 2 as well as the EQE at short wavelengths. As the annealing temperature further increases to 500 o C, J 0 increases by 38% and J sc drops by 3% compared to the optimized condition. These are probably due to degradation in the passivation effect, which is caused by a phase change of TiO 2 from anatase to rutile at high temperature 26,27 .
Annealing at 450 o C produces the best device with the lowest J 0 of 4.6 pA-cm -2 , highest J sc of 16.7 mA-cm -2 and best efficiency of 8.9%. The performance of this device is detailed in the first row of Table 1.
In summary, we have demonstrated a carrier-selective contact with a TiO 2 MIS structure.
With this carrier-selective contact, a 2 µm thick Si solar cell can achieve a V oc of 645 mV, which is 10 mV higher than that of a comparable cell with metal contact. We use TCAD simulations to analyze the recombination loss in thin cells and reveal the origin of the V oc improvement by the MIS contact. Our results show contact recombination is the major recombination mechanism in thin cells and eliminating this source is essential to the design of high-efficiency c-Si cells. We also studied the post deposition annealing of TiO 2 MIS contact cells and point out a method to improve the effect of this carrier-selective contact.
Over all, this work demonstrates a new design for carrier-selective contacts and its application in ultra-thin-film c-Si solar cells, and provides a path to high-efficiency ultrathin-film c-Si solar cells.

Experimental Section
Our ultra-thin-film c-Si solar cells are fabricated on a silicon-on-insulator (SOI) wafer to precisely control the cell thickness. The buried oxide blocks any carriers generated in the To optimize for best performance, cells are annealed under forming gas at different temperatures in the range from 300 o C to 500 o C for 90 seconds. Finally, the thermal oxide is thinned down to ~80 nm by dry etch to function as the single layer anti-reflective coating.
Control samples without the TiO 2 layer are fabricated with an identical process except that the TiO 2 deposition step is skipped. Solar cell efficiency is measured under AM 1.5G normal illumination (1000 W/m 2 , 1 sun) at room temperature. A standard solar simulator is used as the light source, with its intensity monitored by a certified photodetector. For the EQE measurement, a mechanically chopped monochromatic light beam is used as the light source, and the photocurrent is measured using a lock-in amplifier. The light intensity for the EQE measurement is calibrated with an amplified, calibrated photodetector. The saturation current is measured using the quasisteady-state open circuit voltage (QSSV oc ) method with a Sinton Instrument WCT-120.