Electronic Band Offset Determination of Oxides Grown by Atomic Layer Deposition on Silicon

Minimizing electrical losses at metal/silicon interfaces in high-efficiency single-junction silicon solar cells requires the use of carrier-selective passivating contacts. The electronic barrier heights at the insulator/silicon interface are necessary for calculating the probability of quantum tunneling of charge carriers at these interfaces. Thus, precise knowledge of these parameters is crucial for the development of contact schemes. Using a photoemission-based method, we experimentally determine the electronic band offsets of Al<sub>2</sub>O<sub>3</sub>, HfO<sub>2</sub> and SiO<sub>2</sub> layers grown by atomic layer deposition (ALD) on silicon. For Al<sub>2</sub>O<sub>3</sub>/Si, we determine a valence band offset (Δ<italic>E<sub>V</sub></italic>) and conduction band offset (Δ<italic>E<sub>C</sub></italic>) of 3.29 ± 0.07 eV and 2.24 ± 0.13 eV, respectively. For HfO<sub>2</sub>/Si, Δ<italic>E<sub>V</sub></italic> and Δ<italic>E<sub>C</sub></italic> are determined as 2.67 ± 0.07 eV and 1.81 ± 0.21 eV, while for SiO<sub>2</sub>/Si, Δ<italic>E<sub>V</sub></italic> and Δ<italic>E<sub>C</sub></italic> are 4.87 ± 0.07 eV and 2.61 ± 0.12 eV, respectively. Using technology computer-aided design simulations, we incorporate our experimental results to estimate the contact resistivity that would be attained at various dielectric layer thicknesses. We find that for achieving the 100 mΩ·cm<sup>2</sup> contact resistivity benchmark, Al<sub>2</sub>O<sub>3</sub> layers should be no thicker than 1.65 nm for a <italic>p</italic>-type polysilicon-based hole-selective contact, assuming hole tunneling masses taken from the literature. Correspondingly, for HfO<sub>2</sub> and SiO<sub>2</sub>, an upper limit of 1.4 nm is determined as the thickness threshold in order to utilize these ALD-grown layers for contacts in high-performance silicon photovoltaics.


I. INTRODUCTION
A CCOUNTING for over 95% of commercial production, crystalline silicon solar cells continue to lead in the photovoltaics (PV) landscape and remain a prominent alternative to nonrenewable energy sources [1]. Currently, the bulk of this market is led by architectures that incorporate metal/Si interfaces for electrode formation [2], [3]. Avoiding this type of interface is widely accepted as the strategy toward higher performance silicon solar cells.
Due to the high density of electronically active states that arise from direct contact of metals on silicon, photogenerated charge carriers undergo trap-assisted recombination at such interfaces that limit the electrical device performance. To reach the Shockley-Queisser power conversion efficiency (PCE) limit of 29.4%, passivating and selective contact technologies are being adopted in the silicon PV industry [3], [4], [5]. Generally, passivating contacts are formed by introducing an interlayer or layer stack in between the silicon surface and the metal electrode. This type of technology mitigates the inherent electrical losses in the contacted regions by suppressing charge-carrier recombination as well as maintaining a low enough resistivity [6], [7], [8]. The use of a well-designed passivating contact should increase the carrier collection efficiency and ultimately the PCE. Passivating both electron and hole contacts is needed in order to reach PCEs exceeding 25% [3], [9]. To date, various passivating contact structures have been successfully utilized in solar cell structures, including silicon heterojunctions (HJT) [10], polysilicon on oxide [11], and tunnel oxide passivated contact (TOPCon) [12], [13]. Amongst these contact architectures, a PCE of 26.81% is the best cell performance achieved so far (using HJT) [14], [15].
Typically, SiO 2 -based poly-Si contacts perform better as an electron-selective contact than a hole-selective contact. The valence band offset at the SiO 2 /Si interface is considerably large, resulting in a high barrier for hole tunneling and, hence, a relatively low hole tunneling current [16]. Another reason for the poorer performance as a hole-selective contact is due to the high-temperature anneal step, which causes dopants to diffuse from the poly-Si layer to the Si substrate. As boron diffusion in SiO 2 is not blocked as well as phosphorous, a relatively large boron concentration is found at the SiO 2 /Si interface in this contact architecture, which, in turn, leads to Auger recombination [17]. Despite the key improvements over recent years in passivating contact technologies, the search for a hole-selective contact that can match or even exceed the performance of existing electron-selective contacts continues.
In thin film passivating contacts, the exact mechanism for charge-carrier transport is still under debate in the PV community, but generally, it is understood to be due to quantum tunneling and/or through pinholes [18], [19], [20]. Tunneling refers to the ability of charge carriers to have a wavefunction that can extend through a potential barrier instead of the carrier having to go over the barrier. The probability of carrier tunneling P t through such insulators is described by the Wentzel-Kramers-Brillouin (WKB) approximation [21] where is the reduced Planck's constant, m * is the tunneling charge effective mass, q is the charge of an electron, t is the film thickness, and Δφ b is the potential barrier height at the interface. From the WKB approximation, it follows that the tunneling probability can be controlled by tuning the film thickness and potential barrier heights at the insulator/Si interface. Utilizing thin films for carrier-selective passivating contacts requires fabrication techniques capable of attaining thicknesses with Angstrom (Å) level controllability in a reliable manner. Atomic layer deposition (ALD) offers such benefits through self-limiting surface reactions conducted at relatively low deposition temperatures. It is highly suitable for depositing a variety of materials, including oxides and nitrides, which fit the criteria for contact interlayers. Ultimately, the high level of film and interface control is highly attractive for thin film fabrication for PV applications, particularly for carrier-selective passivating contacts.
In previous work, we report the Si surface passivation quality of various types of ultrathin dielectrics grown via ALD [22]. HfO 2 was identified as the most promising candidate, with 0.9 nm of HfO 2 annealed at 450°C providing a surface recombination velocity (SRV) of 18.6 cm s −1 and 2.2-3.3-nm-thick HfO 2 layers achieving an SRV of ≤2.5 cm s −1 and J 0 of ∼ 14 fA/cm 2 . As the development of an efficient hole-selective contact is a key aim in today's PV industry, we determine the potential for these materials as carrier-selective interlayers. In this article, we report an experimental study that determines the electronic barrier heights at the interface between silicon and ALD-grown Al 2 O 3 , HfO 2 , and SiO 2 . We use our existing ALD growth methods [22], [23] to grow thin films on silicon. We then use X-ray photoelectron spectroscopy (XPS) to probe the electronic core levels (CLs) and valence band maximum (VBM) at the oxide/Si interface. Additionally, we use simulations in Sentaurus technology computer-aided design (TCAD) [24] to estimate the contact resistivity that we would expect in a Si/oxide/poly-Si p-contact formation for each of the investigated oxides. Since the barrier tunneling model in Sentaurus is based on the WKB approximation (1), the resulting contact resistivities highly depend on the oxide thickness t as well as the effective (hole) tunneling mass m h * . Assuming hole tunneling masses from the literature, these simulations indicate that the reasonable contact properties could be achieved and, therefore, provide an incentive for further experimental research on the development of an efficient hole-selective contact based on these oxides. Beyond minimizing contact resistivity, the use of these dielectrics as interlayers in carrier-selective passivating contacts requires the enhancement of the surface passivation properties of these films at ultrathin (sub-3 nm) thicknesses, as explored in [22].

A. Specimen Fabrication
ALD films were grown on p-type (gallium doped) Si (Cz, 5 Ω·cm, <100>, 125 μm thick) substrates that were prepared following a previously reported chemical cleaning and etching procedure [22], [25]. In the first ALD half-cycle reaction, trimethylaluminum, tetrakis(dimethylamido)hafnium, and bis(diethylamido)silane precursors were used to grow Al 2 O 3 , HfO 2 , and SiO 2 films, respectively. An O 2 plasma source was used for the second half-cycle reaction for all three films. All films were grown at a deposition temperature of 200°C. Growth rates per cycle are reported as 1.3Å/cycle (Al 2 O 3 ), 1Å/cycle (HfO 2 ), and 0.6Å/cycle (SiO 2 ) [26], and re-evaluated in [22]. A postdeposition anneal was conducted at 450°C for 30 min for Al 2 O 3 and HfO 2 , and 800°C for 30 min for SiO 2 . All films were grown in a Veeco Fiji G2 plasma-enhanced ALD chamber.

B. X-Ray Photoelectron Spectroscopy
XPS was conducted using a Kratos Axis Ultra delay-line detector (DLD) spectrometer. For XPS, all samples were mounted on a nonmagnetic, stainless-steel bar by using electrically conductive carbon tape. XPS was conducted using a monochromated Al Kα X-ray (1.487 keV) source. The energy resolution of the detector was 0.4 eV. Measurements were conducted at room temperature and at a take-off angle of 90°with respect to the sample surface. The CL spectra and the VBMs were measured using a pass energy of 20 eV, all from an analysis area of 300 μm × 700 μm. To avoid charging effects, a charge neutralizer gun was used for all XPS measurements. Fitting procedures to extract peak positions and relative stoichiometries were performed by using the Casa XPS software. These were fitted and corrected using their corresponding sensitivity factors, taking the mean free path of the photoelectrons and photoionization cross sections of these CLs into account.

C. Electronic Band Offsets Determination
Determination of the valence band offset (ΔE V ) and conduction band offset (ΔE C ) at a semiconductor interface can be done using Kraut's method [27], [28], [29], [30], [31]. These are depicted in the schematic diagram of the band offsets in Fig. 1. This X-ray photoemission-based method uses Poisson's equation to predict the band discontinuities based on the deviations in charge distribution found at the interface relative to the semiconductor bulk. In this approach, the position of the CL at the interface as well as the binding energy difference between the semiconductor CL and the VBM are required to determine the valence band offset In this equation, (E Si CL − E Oxide CL ) Oxide/Si is the energy difference between the CL of the two materials at the interface, namely ΔE CL . Based on the XPS photoelectron sampling depth being under 5 nm in these ALD-grown materials [32], [33], 2-nm and 3-nm-thick films on Si were used to obtain two sets of measurements that probe the interface. To complete the equation for ΔE V , the energy difference between the CL centroids and VBM for Si and all the ALD oxides of interest were obtained from XPS of the respective thick films. For this experiment, the Al 2p, Hf 4f, and Si 2p orbital peaks were used as the CL for Al 2 O 3 , HfO 2 , and SiO 2 , respectively. The elemental silicon region of the Si 2p peak was also used as the CL for Si. For E Si V and E Oxide V determination, linear extrapolation of the leading edge to the baseline of the valence band spectra from the respective thick films was used.
Once ΔE V was determined, ΔE C was calculated following Kraut's method [27] where (ΔE g ) Si/Oxide is the energy difference between the band gap of Si and the respective ALD oxide films. An optical band gap of 1.12 ± 0.01 eV was used for E Si g in these calculations.

A. Electronic Band Offsets
The electronic band offsets at an Al 2 O 3 /Si, HfO 2 /Si, and SiO 2 /Si interface have been reported in the literature. A summary of these calculations from previous reports is presented in Table I.
A large range of band offsets have been presented over the last couple of decades. For example, ΔE V at an Al 2 O 3 /Si interface has been reported between 2.95 and 4.9 eV. The differences are mainly due to dissimilarities in fabrication processes that result in variations in chemical composition and stoichiometry that, in turn, lead to alterations in the band offsets. For example, Alay and Hirose [37] suggest differences in the band offsets for SiO 2 /Si based on whether the SiO 2 layer was fabricated by a dry or wet chemical process as well as the crystal orientation of the underlying Si substrate being (100) or (111). Interfacial effects, such as interfacial dipoles, could also lead to deviations in band offsets. For ΔE CL determination [as a part of (2)], the thickness of the overlayer chosen can also play a role in the band offset calculations. It is generally understood that the thickness must not exceed the photoelectron sampling depth of the overlaying material, but variations in thickness below that limit can cause small shifts in the CL positions. In addition, differences in measurement procedures (e.g., XPS, linear internal photoemission, and synchrotron radiation photoemission) add further uncertainty to the reported band offsets. To accurately determine the probability of carrier tunneling through such interfaces, precise determination of the band offsets is required.
Using XPS, we have identified the CL energy centroids and valence band edges for Si, Al 2 O 3 , HfO 2 , and SiO 2 , as shown in Fig. 2. From Fig. 2(a) and (b), the Si 2p CL energy and leading edge of the valence band spectra (i.e., VBM) for bare (native oxide stripped with Hydrofluoric acid) Si are determined to be 99.71 ± 0.02 eV and 0.69 ± 0.04 eV, respectively. The ± symbol is used to signify the measurement uncertainty. Hence,   Table II.
The CL for Al 2 O 3 (Al 2p) and Si (Si 2p) at the interface from the 2-nm and 3-nm-thick Al 2 O 3 specimens are shown in Fig. 3(a)-(d). From Fig. 3(b) and (d), two peaks are seen for the Si 2p CL at the interface. The CL found at ∼99 eV is detected from elemental silicon and the CL at ∼103 eV is from the presence of SiO 2 . This CL verifies the presence of SiO 2 in these specimens, suggesting the manifestation of a very thin   [45]. For this study, we focus on only using the elemental silicon region for E Si CL in the band offset calculations. From Fig. 3(a) and (c), the Al 2p CL from 2-nm and 3-nmthick Al 2 O 3 on Si is found at binding energies of 74.62 ± 0.02 eV and 74.55 ± 0.02 eV, respectively. Additionally, Fig. 3 Fig. 3(f) and (h) shows a peak at ∼103 eV as well as the Si 2p elemental silicon peak at ∼99 eV, which again demonstrates the presence of an interfacial SiO 2 layer. Taking the elemental silicon contribution into account, the Si 2p CL for 2-nm and 3-nm-thick HfO 2 on Si is found at 99.46 ± 0.02 eV and 99.23 ± 0.02 eV, respectively. Therefore, (E Si CL − E HfO 2 CL ) HfO 2 /Si for 2-nm and 3-nm-thick HfO 2 on Si is determined to be 82.33 ± 0.028 eV and 82.42 ± 0.028 eV, respectively. Fig. 3(i) and (j) shows the Si 2p CL peaks from 2-nm and 3-nm-thick SiO 2 on Si, respectively. Here, we only take the Si 2p into account as we detect contributions from elemental Si and Si-O in the same binding energy region. From Fig. 3 Table III.
From Tables II and III, ΔE V and ΔE C are calculated using (2) and (3), respectively, and are shown in Table IV. To determine ΔE C , we use a bandgap of 6.65 ± 0.11 eV for Al 2 O 3 [46], 5.6  [47], [48], and 8.6 ± 0.1 eV for SiO 2 [49]. (2), an average is taken between the 2 and 3 nm oxide/Si calculations from Table III. From these results, Fig. 4 shows a simplified schematic diagram of the band offsets at the Al 2 O 3 /Si, HfO 2 /Si, and SiO 2 /Si interface.
The ΔE C /ΔE V ratio is a good indication of favorability toward electron/hole transport, where ΔE C /ΔE V > 1 favors hole transport. For Al 2 O 3 /Si, HfO 2 /Si, and SiO 2 /Si, ΔE C /ΔE V can be determined as 0.68, 0.68, and 0.54, respectively. This suggests that all three interfaces would favor electron transport, with SiO 2 being the most favorable toward electrons. SiO 2 is used as the passivating interlayer in the world-leading carrier-selective passivating contact technology, TOPCon, and is known to perform far better as an electron contact than as a hole contact [3]. The band offsets determined indicate that HfO 2 and Al 2 O 3 could offer alternatives for SiO 2 in the hole-selective counterpart. Beyond ΔE C /ΔE V , a direct comparison of the absolute values for ΔE V determined for the three oxides of interest is another good indication toward hole transport favorability [based on (1)]. Evidently, HfO 2 possesses the smallest potential barrier for hole transport but a ΔE V of 2.67 eV is still considerably large.
The band offsets are not the only figure of merit when considering fitting candidates for carrier-selective contacts, so the impact of this must be weighed with their suitability in other important factors, including carrier tunneling mass and SRV. We explored the Si surface passivation quality in previous work [22] and determine the impact of the band offsets and tunneling masses on the contact resistivity using TCAD simulations here.

B. Contact Resistivity Estimation Via TCAD Simulations
For charge-carrier transport through ultrathin dielectrics, one of the theories that is strongly agreed upon is quantum tunneling through the potential barrier created at the Si surface. Based on the WKB approximation (1), the tunneling probability is dependent on the barrier height and film thickness as well as the effective tunneling masses of the charge carriers.
Using the barrier heights determined for the ALD oxides with respect to Si, Sentaurus TCAD is used to estimate the contact resistivity for a range of thicknesses in a typical p-TOPCon format. The contact resistivity is extracted directly from the TCAD simulations for this study. An illustration of the contact structure (i.e., poly-Si/oxide/Si) we are interested in is shown in Fig. 5(a). A p-type Si (1 Ω·cm, 200 μm thick) substrate with the ALD oxides on the front side was devised, with a 50-nmthick p + poly-Si conductive interlayer (with 10 20 cm −3 doping concentration) between the metal contact and the oxide. In the TCAD simulations, an electron effective mass of 0.25 m 0 [46], a hole effective mass of 0.36 m 0 [50], and an optical bandgap of 6.65 eV [46] were used for Al 2 O 3 , where m 0 is the free electron   [52], and an optical bandgap of 5.6 eV [47], [48] were used, as taken from the literature. For SiO 2 , an electron effective mass of 0.4 m 0 [53], [54], a hole effective mass of 0.3 m 0 [55], and an optical bandgap of 8.6 eV were used. Fig. 5(b) shows the tunneling layer thickness versus calculated contact resistivity (ρ c ) for ALD-grown Al 2 O 3 , HfO 2 , and SiO 2 on p-Si as a hole-selective contact.
The calculated ρ c versus dielectric thickness curves show an exponential trend in all three cases. The difference in ρ c between the three ALD-grown dielectrics below 1.1 nm is negligible. From Fig. 5(b), the Al 2 O 3 /(p)Si-based contact outperforms HfO 2 and SiO 2 (i.e., the lowest ρ c ) at most thicknesses. Despite the larger ΔE V measured for Al 2 O 3 /Si in comparison with HfO 2 /Si, the Al 2 O 3 /Si contact outperforms HfO 2 /Si due to the considerably lower hole effective mass. This demonstrates how crucially the contact resistivity depends on the assumed tunneling masses. Based on (1), the tunneling effective mass and interfacial barrier heights are of equal importance and ideally, both parameters should be low enough to ensure a low contact resistivity.
As a standard for carrier-selective passivating contacts, a contact resistivity of 100 mΩ·cm 2 is seen as the upper threshold for high-performance full-area contacts. Achieving a contact resistivity below 100 mΩ·cm 2 provides insignificant improvements to the fill factor, and hence the PCE of such devices. Based on our TCAD simulations with the given effective tunneling masses from the literature, a thickness of 1.65 nm is found as the upper limit for Al 2 O 3 . The upper limit for both HfO 2 and SiO 2 is determined as 1.4 nm.
Incorporating metal oxides, such as Al 2 O 3 with poly-Si for hole-selective passivating contacts, has seen some success in the literature, with the contact resistivity reported at 200 mΩ·cm 2 in multiple findings in the literature [56], [57], [58]. This further illuminates the scope for research on Al 2 O 3 -based hole-selective passivating contacts. For HfO 2 , despite showing a higher contact resistivity than Al 2 O 3 in our TCAD simulations, the surface recombination velocities determined for ultrathin HfO 2 layers outperform Al 2 O 3 , operating just as well as thicker HfO 2 films [59], as found in our previous work [22], [60]. These results were achieved without common postdeposition treatments, such as hydrogenation, which are typically used to enhance the surface passivation properties [22], [57], [61]. However, the impact of growing a p + poly-Si layer on the oxides as well as annealing for crystallizing the poly-Si layer at temperatures exceeding 800°C is yet to be examined. In addition, the use of ALD-grown SiO 2 as a replacement for thermally grown SiO 2 still requires substantial improvements, mainly due to the inferior passivation quality of this type of SiO 2 growth method [13], [22], [57].

IV. CONCLUSION
We explored the electronic band offsets of ALD-grown Al 2 O 3 , HfO 2 , and SiO 2 on silicon using a photoemission-based method. For Al 2 O 3 /Si, we determine ΔE V and ΔE C as 3.29 ± 0.07 eV and 2.24 ± 0.13 eV, respectively. For HfO 2 /Si, ΔE V and ΔE C are determined as 2.67 ± 0.07 eV and 1.81 ± 0.21 eV, while for SiO 2 /Si, ΔE V and ΔE C are 4.87 ± 0.07 eV and 2.61 ± 0.12 eV, respectively. We apply TCAD simulations to predict the contact resistivity at various dielectric thicknesses with a 50 nm p + polycrystalline silicon conductive layer between the metal electrode and thin dielectrics and assuming hole effective tunneling masses taken from the literature. In order to form an efficient hole-selective contact with a p-type polysilicon electrode and not exceed the 100 mΩ·cm 2 contact resistivity benchmark, an upper limit of 1.65 nm in thickness is determined for Al 2 O 3 , while 1.4 nm is calculated as the threshold for HfO 2 and SiO 2 . Experimental demonstration of the contact resistivity for such structures will be a part of future work.

DATA ACCESS STATEMENT
Data underpinning figures in this paper can be downloaded from https://wrap.warwick.ac.uk/75494. Requests for additional data should be made directly to the corresponding author. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.