Introduction

Transition metal oxides possess a wide range of work functions, spanning from 3.5 eV for defective ZrO2 to 7.0 eV for stoichiometric V2O5 [1,2,3,4,5,6]. Among them, MoOX is one of the most extensively studied materials for applications in optoelectronic devices [7,8,9] due to its high transparency, nontoxicity and moderate evaporation temperature [10, 11]. MoOX is reported to have a large work function of ~ 6.7 eV and is being widely used as hole extraction layers in photovoltaic devices [12], light emitting devices [13], sensors [14, 15] and memories [16]. For photoelectric devices involving MoOX hole extraction layers, the device performance is strongly dependent on both the optical and electronic properties of the MoOX thin films. In the photovoltaic field, MoOX thin films were initially applied in organic devices [17,18,19]. In recent years, a lot of research has been done on the application of MoOX films to crystalline silicon (c-Si) solar cells [9, 20,21,22]. The ionization energy of c-Si is about 5.17 eV, which is the lower limit for the work function of hole selective contact materials [23]. The high work function of MoOX will induce a large band bending at the c-Si/MoOX interface and lead to the accumulation of holes in p-type silicon (p-Si) or the depletion of electrons in n-type silicon (n-Si), thus favoring the holes transport [24]. By substituting the p-type amorphous silicon layer with MoOX film in the classical silicon heterojunction solar cell, an power conversion efficiency (PCE) of 23.5% has been achieved [25]. Compared to MoOX contacts made to n-type wafers, those made to p-type wafers (without amorphous Si layer) show better performance in terms of surface passivation and contact resistivity [24]. The feasibility of MoOX films as hole-selective contacts on p-Si solar cells has been demonstrated in our previous work [26], and an efficiency of 20.0% was achieved based on p-Si/SiOX/MoOX/V2OX/ITO/Ag rear contact [27].

MoOX (X ≤ 3) has a large work function because of the closed shell character in its bulk electronic structure and the dipoles created by its internal layer structure [28]. The presence of oxygen vacancy defects will decrease the work function of MoOX [4] and result in an n-type material [29]. Numerical simulations indicated that higher work function of MoOX induced a favorable Schottky barrier height as well as an inversion at the MoOX/intrinsic a-Si:H/n-type c-Si (n-Si) interface, stimulating the path of least resistance for holes [30]. Therefore, tuning the electronic structure and work function of MoOX is of great significance for passivating contact c-Si solar cells.

MoOX films can be deposited by atomic layer deposition [30,31,32,33,34], reactive sputtering [12], pulsed laser deposition [35], thermal evaporation [24, 36] and spin coating [37]. In most of the solar cell researches based on Si/MoOX contact, MoOX films are prepared by thermal evaporation at room temperature [8]. Because the controllability of the properties of MoOX films by thermal evaporation is limited, various methods of post-treatments were studied to tune the work function of thermally evaporated MoOX. UV-ozone exposure could increase the work function of evaporated MoOX films on gold substrates from 5.7 eV to 6.6 eV [8]. Irfan et al. performed air annealing of MoOX films on gold substrates at 300 °C for 20 h and found that the long-time annealing does not assist in reducing the oxygen vacancies due to the diffusion of gold from substrate toward the MoOX film [38]. The work function of MoOX films on p-type c-Si (p-Si) was found to decrease after in situ vacuum annealing in the temperature range from 300 to 900 K [39].

In this work, p-Si solar cells with MoOX passivating contacts on rear sides are configured. The optical and electronic properties as well as the influence of the post-annealed MoOX films on p-Si/MoOX solar cells are systematically investigated through experiments and energy band simulations. A linear relationship between the work function and the O/Mo atomic ratio is found. It is interesting that compared with the intrinsic sample, the 100 °C-annealed sample with a higher work function exhibits a lower contact resistivity in spite of its thicker SiOX interlayer. According to the energy band simulation, the variation of MoOX’s work function has a little effect on the band bending of p-Si, while the band bending of MoOX increases significantly as its work function increases. Therefore, it is suggested that higher work functions are vital for effective hole transport from p-Si to MoOX where the interfacial SiOX layer is in a moderate thickness range. Our results provide valuable details of the interface characteristics of the p-Si/MoOX in view of high-performance heterojunction solar cells with oxide-based carrier selective contacts.

Methods

Film Deposition, Post-Annealing Process and Solar Cell Fabrication

Solar cells are fabricated on p-type < 100 > CZ wafers with a resistivity of ~ 2 Ω·cm and wafer thickness of 170 μm. The silicon wafers are precleaned by mixed solution of NaOH and H2O2 and then textured by NaOH solution. The wafers are then washed by deionized water (DI water) following 1 min’s dip in dilute hydrofluoric acid (HF). Heavily doped n+ front surface (ND ≈ 4 × 1021 cm−3) is achieved by diffusing phosphorus from a POCl3 source in a quartz furnace. A double-layered SiNX:H passivation and antireflection coating is then deposited by plasma-enhanced chemical vapor deposition (PECVD). The silver paste is screen-printed on the solar cells with a selective emitter [40]. Subsequently, a fire-through process is conducted at 850 °C for ~ 1 min, after which Ohmic contacts with low resistivity result [41]. The rear surface of each sample is rinsed with dilute HF before MoOX deposition. MoOX films are thermally evaporated at the rear side with a deposition rate of ~ 0.2 Å/s under 8 × 10–4 Pa [26]. Post-annealing treatments of the room-temperature-deposited MoOX films are carried out in a rapid thermal processor in air. The setting temperature was reached in 10 s and held for 5 min. MoOX films with different annealing temperatures are applied to p-Si solar cells with full rear MoOX/Ag contacts.

Measurements

The transmittance spectra of the MoOX films deposited on 1.2-mm-thick silica glasses are measured using a UV–Vis spectrometer with an integrating sphere. Surface morphology and roughness of the films are measured by atomic force microscope (AFM). The optical properties of the MoOX films are analyzed using spectroscopic ellipsometry (J.A. Woollam Co., Inc., M2000U ellipsometer), and the measured results are fitted using the native oxide model. High-resolution X-ray photoelectron spectroscopy (XPS) of Mo 3d and Si 2p are measured employing monochromate Al Kα X-rays with a photon energy of 1486.7 eV. The ultraviolet photoemission spectroscopy (UPS) spectra are recorded by using unfiltered He I 21.22 eV excitation with the sample biased at − 10 eV. Before XPS and UPS detecting, the surfaces of the samples were precleaned by argon ions.

The contact resistivity at p-Si/MoOX interface is extracted by the Cox and Stack method [42], which involves a series of resistance measurements on a probe station with different diameter front Ag contacts. The passivation qualities of MoOX films with different thicknesses are determined from effective lifetime measurements via quasi-steady-state photo conductance (QSSPC) method. The samples for QSSPC test are asymmetric as the front sides are textured, n+ doped and passivated by means of a double-layered SiNX:H films [43], while the rear sides are covered with the MoOX films [26]. The current density–voltage characteristics of the solar cells (3.12 × 3.12 cm2) are measured under standard one sun conditions (100 mW·cm−2, AM1.5G spectrum, 25 °C) as the luminous intensity is calibrated with a certified Fraunhofer CalLab reference cell.

Simulations

Numerical simulation of the band structure of the p-Si/MoOX contacts is done with AFORS-HET, which is based on solving the one-dimensional Poisson and two carrier continuity equations [44]. The key parameters are listed in Table 1. The front and back contact boundary is set as fixing metal work function to flat band. The interface between p-Si and MoOX is set as “thermionic-emission” (one of the numerical models). Tunneling properties of thin SiO2 film are commonly set by changing the interface parameters under the “thermionic-emission” model only for metal/semiconductor Schottky contact. Therefore, the actually existed tunneling SiOX at the Si/MoOX interface is omitted. For p-Si, electroneutral defects at the central energy with total trap density is set as 1 × 1014 cm−3. For MoOX, donor-type conduction tail defects with total concentration are set as 1 × 1014 cm−3.

Table 1 Parameters used for AFORS-HET simulation
Full size table

Results and discussion

Figure 1a represents the photographs of the 10-nm-thick MoOX films on silica glass annealed in air for 5 min at different temperatures (100 °C, 200 °C and 300 °C). All of the samples are visually colorless and transparent. From the corresponding optical transmittance spectra in Fig. 1b, one can see that the transmittance spectrum of the 100 °C-annealed MoOX film almost overlaps with that of the unannealed film. Higher annealing temperatures result in a lower transmittance at 600–1100 nm range, which could be assigned to free carrier absorption induced by oxygen vacancies [46]. Thicker MoOX films (20 nm) are deposited onto polished Si wafers to measure the refractive index n and extinction coefficient k more accurately. The refractive index in Fig. 1c lies in the 1.8–2.5 range, which is consistent with other studies [31, 32]. The n curves as well as the k curves (Fig. 1d) have a little difference among the four samples. The n at 633 nm of the 20-nm-thick films decreases slightly, which is summarized in Table 2.

Fig. 1
figure 1

a Photographs and b transmittance spectra of the 10-nm-thick MoOX films on silica glass annealed in air for 5 min at different temperatures. c Refractive indices n and d extinction coefficient k curves of the 20-nm-thick MoOX films on polished silicon wafers

Full size image
Table 2 Root-mean-square roughness (unit: nm) of 10 nm/20 nm post-annealed MoOX films on SiO2 wafers and refractive index n at 633 nm of the 20-nm films
Full size table

The surface morphologies are then characterized by AFM as shown in Additional file 1: Figure S1. The corresponding root-mean-square (RMS) roughness is listed in Table 2. The as-deposited 10-nm-thick MoOX thin film (Additional file 1: Figure S1a) has an RMS roughness of 4.116 nm, which is in accordance with the wave-like surface morphology. As the annealing temperature goes higher (Additional file 1: Figure S1b–d), the surface undulation of the MoOX film becomes larger, while the featured structures become smaller and much denser probably due to the dewetting process [47]. After annealing at 300 °C, the RMS roughness reaches 12.913 nm. The 20-nm-thick films are less rough with the RMS around 1 nm (Table 2). The dewetting process is also suppressed as indicated by the RMS measurements as a function of annealing treatments. The above morphology evolution does not fully reflect the changes in the oxide film in the device level, where the MoOX films are deposited on Si and capped with Ag electrodes, but the morphology evolution can do give us the intrinsic properties of MoOX on SiO2 surface.

MoOX has a natural tendency to form oxygen vacancy defects [48], which may impact on the molecular structure. In order to identify such vacancy-related molecular structure variations, Raman spectroscopy measurements are taken on MoOX(20 nm)/Si(< 100 >). There are no characteristic peaks of MoOX in the Raman spectra under green light (532 nm) excitation (Additional file 1: Figure S2), which is independent to the thermal treatment. When the excitation is changed to ultraviolet light of 325 nm, characteristic bands of MoOX appear, which generally locate at 600–1000 cm−1 (Fig. 2). The sharp peak of 515 cm−1 in all samples corresponds to Si–Si bond. For the intrinsic and 100 °C-annealed MoOX films, Raman bands are present at 695, 850 and 965 cm−1, which are from [Mo7O24]6−, [Mo8O26]4− anions, and (O =)2Mo(–O–Si)2 dioxo species, respectively [49]. When the film is annealed at 200 °C, the 965 cm−1 band shifts to 970 cm−1, which is assigned to Mo(= 16O)2 dioxo species [50]. The Raman spectrum of the 300 °C-annealed MoOX film exhibits bands at 695, 810 and 980 cm−1. The band at 810 cm−1 is from Si–O–Si bond, while the (O =)2Mo(–O–Si)2 contributes the band at 980 cm−1. The results indicate that annealing at different temperatures will affect the chemical composition of MoOX film, which may indicate the difference of oxygen vacancy concentration of each sample.

Fig. 2
figure 2

The UV Raman (325 nm) spectra of post-annealed 20-nm-thick MoOX films on polished silicon wafers

Full size image

XPS is conducted on MoOX films (10 nm) to quantify the relative content of each oxidation state and the oxygen to molybdenum (O/Mo) atomic ratios. After Shirley background subtraction and fitting by Gaussian–Lorentzian curves, a multi-peak deconvolution of the XPS spectra is conducted. The Mo 3d core level is decomposed into two doublet peaks with a doublet spin–orbit splitting ΔBE 3.1 eV and a peak area ratio of 3:2 [11]. As shown in Fig. 3, the peak of Mo6+ 3d5/2 core level centers at ~ 233.3 eV binding energy. For all of the samples, a second doublet at ~ 232.0 eV, which is denoted as Mo5+, is required to obtain a good fit to the experimental data [8]. The O/Mo ratio is calculated by the following formula [51]:

$$X = \frac{1}{2} \cdot \frac{{\mathop \sum \nolimits_{n} n \cdot I({\text{Mo}}^{n + } )}}{{\mathop \sum \nolimits_{n} I({\text{Mo}}^{n + } )}}$$

where I(Mon+) is the individual component intensities from the Mo 3d spectra. n relates to the valence state of Mo ion, i.e., 5 for Mo5+ and 6 for Mo6+. The factor 1/2 is due to that each oxygen atom is shared by two molybdenum atoms.

The O/Mo ratios of all samples as listed in Table 3 are below 3. Oxygen loss and oxidation state transitions have been reported during transition metal oxides deposition [1]. Since the XPS measurements are ex-situ, the air exposure to the thermally evaporated MoO3 films at room temperature could also increase the oxygen vacancies [18, 52]. The O/Mo ratio of the unannealed MoOX film is 2.958, while post-annealing at 100 °C increases the value to 2.964. Higher annealing temperatures then reduce the O/Mo ratio gradually. The highest O/Mo ratio of the 100 °C-annealed sample might be explained by the thermally activated oxygen injected from air to the MoOX film [38]. Additional file 1: Figure S3 compares the Si 2p XPS spectra of the 10-nm-thick annealed MoOX films. The Si 2p XPS spectrum of the unannealed sample shows dual peaks of silicon elements and Si4+ peak. A Si2+ peak appears when annealed at 100 °C. When annealed at 200 and 300 °C, peaks of Si4+, Si3+ and Si2+ exist simultaneously. In addition, the calculated X in SiOX for the four samples are 2, 1.715, 1.672 and 1.815, respectively. The oxygen atoms in SiOX are from MoOX; therefore, the O/Mo ratio depends on the balance between SiOx taking oxygen and air injecting oxygen. By the way, as the annealing temperature goes higher, the signal of Si element becomes weaker, indicating thicker SiOX interlayers [26].

Fig. 3
figure 3

Mo 3d core-level XPS spectra of the 10-nm-thick MoOX films on silicon wafers a without post-annealing, with post-annealing at b 100 °C, c 200 °C and d 300 °C

Full size image
Table 3 O/Mo ratio and work function of the post-annealed 10-nm-thick MoOX films on silicon wafers. Effective minority carrier lifetime of silicon wafers covered by the post-annealed MoOX films
Full size table

Reducing the cation oxidation state of an oxide tends to decrease its work function [1]. UPS is utilized to calculate the work function of MoOX films as a function of thermal treatment. Figure 4a shows the secondary electron cutoff region of the UPS spectra, from which a minor vibration of work function can be seen. From Fig. 4b we can see, after post air annealing, the defect peaks in the valence band area [37] become weaker. Table 3 lists the O/Mo ratio evaluated from XPS fitting and corresponding work function evaluated from UPS secondary electron cutoff for samples on polished silicon wafers. The results of the work function and the stoichiometry of MoOX are also depicted in Fig. 4c, where a strong positive correlation is disclosed. An increase of the O/Mo ratio from 2.942 to 2.964 leads to an increase of the work function by roughly 0.06 eV.

Fig. 4
figure 4

a The secondary electron cutoff region and b valence band from the UPS spectra of the post-annealed MoOX films on silicon wafers. c Work function plotted against the stoichiometry (O/Mo ratio)

Full size image

Before applying the MoOX films as passivating contacts on p-Si wafers, one-dimensional energy band simulations are conducted using AFORS-HET [44] to get a clear image of the p-Si/MoOX heterocontacts. The thicknesses of p-Si and MoOX film are set as 1 μm and 10 nm, respectively. The acceptor concentration of p-Si is 1 × 1016 cm−3, resulting in a work function of 4.97 eV. Since MoOX is an n-type material [53], oxygen vacancies concentration variation is simulated by changing the donor concentration at the range of 1 × 1016 cm−3 to 1 × 1020 cm−3. Figure 5a shows that the work function and donor concentration of MoOX are exponentially correlated. Figure 5c, d depicts the simulated band structure as the donor concentration (ND) of MoOX is 1 × 1016 and 1 × 1020 cm−3, respectively. Both the bands of p-Si and MoOX are bent due to the work function difference and Fermi energy equilibrium. Efficient carrier extraction requires that photogenerated holes in the valence band of p-Si recombine with electrons presented in the MoOX conduction band that are injected from the adjacent metal electrode [7, 54]. The band bending in p-Si, MoOX and the total band bending are shown in Fig. 5b. As the work function of MoOX (WFMO) changes, there is no obvious change in the band feature of p-Si. In contrast, the band bending in MoOX, which represents a favorable built-in electric field for electron injection, increases as its work function increases. We can conclude that the increase in the MoOX work function will raise the total band bending of p-Si/MoOX contact, most of which lies in the MoOX part. Therefore, a high work function of MoOX is desired from the aspect of electron injection at the p-Si/MoOX interface.

Fig. 5
figure 5

Simulated energy band results of the p-Si/MoOX contact. a The relationship between the work function and ND of MoOX (ND-MO). b The p-Si, MoOX and the total band bending for p-Si/MoOX contact. The acceptor concentration of p-Si is 1 × 1016 cm−3. Simulated band diagrams of p-Si/MoOX contact as the ND-MO is c 1 × 1016 cm−3 and d 1 × 1020 cm−3, respectively

Full size image

Figure 6 depicts the dark I–V characteristics of the p-Si/MoOX contacts using Cox and Strack method (see Additional file 1: Figure S4 for the schematic illustration) [42]. The slope of the I–V curve increases with the increase of the diameter of dot electrode. The I-V curves of the unannealed and 100 °C-annealed samples are linear, with the specific contact resistivity (ρc) fitted as 0.32 and 0.24 Ω‧cm2, respectively. Although annealing at 100 °C would make the SiOX layer at the p-Si/MoOX interface thicker, the WFMO is higher than that of the unannealed MoOX film, so the corresponding sample shows the best hole transport characteristic. The I-V curves of the samples annealed at 200 and 300 °C become nonlinear at small dot diameter and could not be considered as Ohmic contact. Compared with the samples annealed at 100 °C, samples annealed at higher annealing temperatures possess lower currents. As the small drop of work function, the main reason would be that higher annealing temperature causes thicker SiOX layer at the p-Si/MoOX interface, making it more difficult for carriers to tunnel through the oxide barrier.

Fig. 6
figure 6

Contact resistance measurements of the 10-nm-thick MoOX films on polished silicon wafers a without post-annealing, with post-annealing at b 100 °C, c 200 °C and d 300 °C

Full size image

The passivation qualities of the MoOX(10 nm)/p-Si heterojunctions as a function of thermal treatment are characterized in terms of effective minority carrier lifetime (τeff). The injection-level-dependent τeffs is shown in Additional file 1: Figure S5, where the τeffs at an injection level of 1 × 1015 cm−3 are listed in Table 3. The unannealed MoOX film shows the best passivation ability. Higher treating temperature leads to lower τeff, which is the combined result of the chemical passivation of the interfacial SiOX and the field effect passivation of MoOX, as larger X in SiOX means fewer dangling bonds of silicon and larger X in MoOX means larger built-in electric field intensity.

The MoOX films are then adopted into the p-Si/MoOX(10 nm)/Ag configuration (Fig. 7a) to investigate the influence of MoOX’s electronic properties on the device performance. The light current density versus voltage (J–V) curves are shown in Fig. 7b. The average J–V characteristics are shown in Fig. 7c–f. The lower VOCs after annealing are in line with the lower τeff. All cells, except for the ones with MoOX annealed at 300 °C, share similar JSC (~ 38.8 mA/cm2), which means the minor difference in optical index of MoOX and variation in the thickness of the interfacial SiOX have little influence in the effective optical absorption of bulk silicon at long wavelength range. The best PCE of solar cells with unannealed MoOX films is 18.99%, which is similar to our previous report [26]. A PCE of 19.19% is achieved when 100 °C annealing is applied. The PCE improvement mainly comes from the elevated fill factor (FF) with reduced series resistance, which is consistent with the low contact resistance in Fig. 6b. Inefficient transport of holes leads to the decrease of FF, which is prominent on the devices with 300 °C annealing. Higher annealing temperatures lead to PCEs drop that is originated from reduced VOC (degraded field effect passivation of MoOX) and FF (thicker SiOX interlayer reduces the carrier tunneling probability). As the MoOX thin films are capped with Ag electrodes, the performance degradation could be mainly originated from the high-temperature induced elemental diffusion at the MoOX/Ag interface as demonstrated in the previous report [26]. The diffusion of Ag atoms into MoOX will decrease MoOX’s work function, as the Fermi levels align at equilibrium by the transfer of electrons from metals to MoOX [19, 55, 56].

Fig. 7
figure 7

a Cross-sectional schematic, b J–V curves and c–f average J–V parameters of the p-Si/MoOX/Ag solar cells with MoOX films annealed at different temperatures

Full size image

Overall, the performance of the p-Si/MoOX heterojunction solar cell is affected by the passivation quality, work function and band-to-band tunneling [34] properties of the hole-selective MoOX film. The passivation performance of the present structure is still poor, leading to relatively lower VOC. Therefore, efficient surface passivation will be a research focus for nondoped carrier selective contacts.

Conclusions

In summary, MoOX films with different oxygen vacancy concentrations were obtained by post-annealing at different temperatures. The O/Mo atomic ratio of MoOX films is linearly related to their work function. Compared with the intrinsic MoOX film, the one annealed at 100 °C obtained less oxygen vacancy and higher work function. Energy band simulation shows that the band bending of p-Si in the p-Si/MoOX contact is basically the same when the work function of MoOX varies from 6.20 eV to 6.44 eV. Nevertheless, a larger work function yields increased band bending in MoOX film. Experimental results indicate that the moderately improved work function of MoOX annealed at 100 °C is favorable for hole selectivity. The corresponding solar cell with optimized full rear p-Si/MoOX/Ag contact achieved a PCE of 19.19%.