Sub-bandgap near-infrared photovoltaic response in Au/Al2O3/n-Si metal–insulator–semiconductor structure by plasmon-enhanced internal photoemission

Silicon sub-bandgap near-infrared (NIR) (λ > 1100 nm) photovoltaic (PV) response by plasmon-enhanced internal photoemission was investigated. The Si sub-bandgap NIR PV response, which remains unexploited in Schottky junction-like solar cell device, was examined using nanometer sized Au/Al2O3/n-Si junction arrays. This kind of metal–insulator–semiconductor structure was similar in functionality to Schottky junction in NIR absorption, photo-induced charge separation and collection. It showed that NIR absorption increased steadily with increasing volume of Au nanoparticles (NPs) till a saturation was reached. Simulation results indicated the formation of localized surface plasmon on the surfaces of Au NPs, which was correlated well with the observed NIR absorption. On the other hand, the NIR PV response was found sensitive to the amount and size of Au NPs and thickness of Al2O3. Chemical and field-effect passivation of n-Si by using Al2O3 and SiO2 were used to optimize the NIR PV response. In the current configuration, the best PV conversion efficiency was 0.034% at λ = 1319 nm under illumination power of 0.1 W/cm2. Supplementary Information The online version contains supplementary material available at 10.1186/s11671-023-03818-4.


Introduction
Today's low-carbon economy demands more efficient renewable energy sources. At present, Si solar cells dominate the photovoltaic (PV) market [1,2]. Significant advances have been made in promoting the photoelectric conversion efficiency of Si solar cell in the past decade [3,4]. Typical examples include perovskite/Si tandem solar cells with large open-circuit voltage and TOPCon Si ones with large short-circuit current density and fill factor [5][6][7][8]. Although the photoelectric conversion efficiency of Si solar cell is approaching the Shockley-Queisser limit of 30% for single junction Si solar cell [9], the current methodology to improve the cell performance is still limited to the reduction of electrical loss as well as optical loss mainly in the wavelength range of < 1100 nm, corresponding to photon energy higher than Si bandgap of 1.1 eV. The solar radiation in the Si sub-bandgap near infrared (NIR) region has not been fully utilized for PV applications. How to extend the PV response to the Si sub-bandgap infrared region is therefore well worth investigation to further improve the Si PV performance and overcome the Shockley-Queisser limit. So far, various efforts have been made to up-convert NIR photons into visible ones [10][11][12][13][14]. Nonetheless, the intrinsic low efficiency of non-linear process of up-conversion has made it unpromising for utilization of NIR in photovoltaics [12,13]. On the other hand, it has been found that the structure of metal semiconductor contact, or Schottky junction, has great potential in expanding the responsive spectra range of solar cells, because the Schottky barrier is usually lower than the semiconductor bandgap [15,16]. By forming Schottky junction, the hot carriers, generated in metallic nanostructure through non-radiatively decay of excited surface plasmons [16,17], can be extracted by semiconductor via plasmon-enhanced internal photoemission (IPE) before thermal relaxation [18][19][20][21]. However, relevant researches mainly focus on TiO 2 cells (bandgap of 3.2 eV), reporting power conversion efficiencies of 0.02-0.03% working in visible wavelength range [22,23]. Although these reported efficiencies are seemingly low, they demonstrated a breakthrough in material bandgap limitation and provided beneficial inspiration to study the Si counterpart. In fact, Schottky junctions as well as metal-insulator-semiconductor (MIS) ones have been used in NIR photo-detection (PD) [19,[24][25][26][27][28]. Considering that PD and PV devices are similar in working process and device structure, in this work, we investigated the Si sub-bandgap NIR PV response of nanometersized Au/Al 2 O 3 /n-Si MIS junction arrays. It is found that the nano-MIS junction arrays not only absorb the NIR light, but also offer built-in electric field to separate photo-induced charges, via a process of IPE, which is enhanced by localized surface plasmons (LSPs) on Au nanoparticles (NPs).

3
was then deposited on Al 2 O 3 at a rate of 5 nm/min by pulsed laser deposition with a frequency-doubled Q-switched Nd:YAG laser at a base pressure lower than 2 × 10 −4 Pa. The laser worked at a constant repetition rate of 10 Hz and the laser fluence on Au target was about 1.0 J·cm −2 . Subsequently, the deposited Au thin film was annealed in a forming gas of hydrogen and nitrogen (H 2 :N 2 = 5%:95% in volume) at 450 °C for 30 min to form Au NPs. An 80.0-nm-thick indium-tinoxide (ITO) layer and ~ 1.0 μm thick Ag grid were deposited as the front electrode. A 20.0-nm thick SiO 2 layer for surface passivation was deposited onto the backside of the textured Si by electron beam evaporation, and a 1.0 μm thick Al layer was grown by resistance heating as the rear electrode. Finally, thermal annealing was conducted at 450 °C in nitrogen for 5 min to form Ohmic contacts. The area of solar cell device was 1 × 1 cm.

Characterization
The surface morphology and components of the Schottky junction were measured with scanning electron microscopy (SEM, Philips, XL30) and energy dispersive X-ray spectroscopy (EDS, Aztec X-MaxN 80, Oxford Instruments), respectively. The absorption spectra were obtained by an NIR spectrometer (Ideaoptics, NIR2500) with an integrating sphere. External quantum efficiency (EQE) of the solar cell was measured by a QE/IPCE system of Oriel/Newport. The current density-voltage characteristics of solar cells were measured under NIR light illumination using a source meter (Keithley, SMU2400). The NIR light source was a 1319 nm laser diode (CNI laser, MIL-H-1319). Figure 1 shows a schematic of the prepared nano-MIS NIR PV device. The front side of the cell faced a 1319 nm light beam with illumination power of 0.1 W/cm 2 . The textured Si enhanced the light absorption and helped the formation of nano-MIS junctions for NIR absorption at the meantime. The SiO 2 and Al 2 O 3 were passivation layers that saturated the dangling bonds on the Si surface so that trapping of photo-induced charges by these defects could be diminished [29][30][31][32][33]. Meanwhile, they also provided electric fields at the interface of oxide and Si, which could facilitate charge separation of photo-induced carriers [34].

Results and discussion
The top-and side-view SEM images of the textured Si are given in Figs. 2a, b, respectively. It featured micrometer sized pyramid-like structures. The average width of the Si pyramid base was 6.8 ± 1.5 μm, and the average height was 5.0 ± 0.9 μm. Figures 2c, d show the corresponding top-and side-view images for textured Si with Au NPs formed by Au film annealing. The Au deposition time was 5 min, which was equivalent to an apparent thickness of 25 nm. The inset shows an enlarged SEM image, and the average size of Au NPs was 36.2 ± 4.8 nm. Figure 2e shows the 25 nm Au film before annealing. The Au film was flat and no Au NPs had been formed. Different sizes and distributions of Au NPs exhibited in Figs. 2f, g were formed by annealing Au films with apparent thicknesses of 10 and 50 nm, respectively. The corresponding average diameters of Au NPs were 24.1 ± 1.8 and 53.6 ± 4.3 nm, respectively. The average size of Au NPs increased with the growing thickness of the deposited Au film, which is consistent with the relevant studies [35]. For case of 50 nm Au, a small amount of obviously larger Au islands existed on the ridges of Si pyramids that could be more favorite for nucleation of Au NPs. Figure 2h gives the EDS spectrum of Si/Au sample in Fig. 2c. Signals of Si and Au were observed as expected. Figure 3a shows the absorption spectra for planar and textured Si and for different nano-MIS junctions. The NIR absorption of planar Si could be ascribed to the defect states induced by n-type doping and free carrier absorption [36]. After fabrication of textured Si, the NIR absorption arising from dopant energy levels was improved considerably, indicating the availability of light trapping by the pyramid micro-structures [37][38][39]. With the deposition of Au and the formation of nano-MIS junctions, the NIR absorption was enhanced considerably due to IPE mechanism [19,[40][41][42]. From 10 to 50 nm apparent thicknesses of deposited Au, the absorption increased steadily with the growing thickness of Au and reached saturation. The NIR absorption enhancement for larger Au deposition thickness could be attributed to the increasing area of nano-MIS junctions, as shown in Figs. 2c, f and g. On the other hand, localized surface plasmons (LSPs) induced by the incident NIR light on Au NPs would also contribute to the NIR absorption [22,[43][44][45]. To demonstrate the role of incorporated Au NPs in NIR absorption, Fig. 3b gives the absorption difference spectra between the nano-MIS structures with Au NPs and the structures with corresponding Au film before annealing. A clear trend of further NIR absorption enhancement by formation of Au NPs and LSPs was thus seen. The absorption difference spectra in Fig. 3b showed resonant absorption peaks in 1100-1600 nm NIR wavelength range, which were induced by the excitation of LSPs [46,47]. Identical Au nanoparticles periodically distributed on a silicon substrate would show resonance absorption peaks  The absorption difference spectra between the nano-MIS structures with Au NPs formed by annealing and the structures with the same apparent thickness Au without annealing in NIR wavelength, while in a fabricated sample, nanoparticles with various sizes would result in broadband absorption enhancement [48]. By subtracting the absorption of corresponding Au film structure without LSPs from the broadband absorption in Fig. 3a, the resulting Fig. 3b could be regarded as a representative or effective resonance absorption peak for fabricated samples with Au NPs, which exhibited the strength and wavelength of LSP qualitatively [24,48]. The resonance peak wavelength experienced a red-shift from 1156 to 1262 nm as the Au NP size increases from 24.1 nm (corresponding to 10 nm Au) to larger values formed by 50 nm Au, similar to the red-shift in the literature [42,48]. It is seen that with the increasing deposition of Au, the strength of LSP became weaker, indicating that more Au thin film were grown.
To elucidate the LSP-enhanced inner photoemission, electric field distribution was calculated by finite-difference time-domain (FDTD) simulation, as shown in Fig. 4. A simulated distribution of LSP electric field around a half-sphere Au NP surrounded by 1 nm Al 2 O 3 , Si and 80-nm thick ITO film was displayed, under the illumination of 1319 nm incident light. The electric field intensity (E) is re-scaled by the incident field intensity (E 0 ). For simplicity, the radius of Au NP was set at 20 nm, and the NP was placed on the surface of Al 2 O 3 and planar Si. It is seen that bright spots existed mainly at the bottom of the Au NP near the boundary between Au NP and Si, which showed enhancement of electric field. The LSP electric field increased the kinetic energy of hot electrons [45], and helped to overcome the MIS barrier, thus increasing the photo-induced current. Figure 5a shows the PV response of current density-voltage curves of the MIS PV devices at 1319 nm with illumination power of 0.1 W/cm 2 . Table 1 lists the measured PV parameters of short-circuit current density (J SC ), open-circuit voltage (V OC ), fill factor (FF) and conversion efficiency (η). Results for the devices with Au film not annealed were also exhibited, referred to as 25 nm Au film and 50 nm Au film. Such devices had explicitly lower J SC and V OC and worse performance, compared with devices with Au NPs, demonstrating the importance of LSP in sub-bandgap NIR photovoltaic response [49,50]. Moreover, for ITO/Al 2 O 3 /n-Si structure device without Au NPs, almost no photovoltaic response was found, indicating that Au NPs were the main absorber and determinant factor for photoresponse of 1319 nm light (Fig. S1).
The performance of the devices of nano-MIS junctions with Au NPs did not evolve in a similar way as the NIR absorption shown in Fig. 3a. The energy conversion efficiency was the largest for the case of 25 nm Au. This is because that the strength of LSP decreased with the increasing Au deposition time, hence the enhancement of inner photoemission by LSP was weaker, as a result, a compromise of conversion efficiency appeared for the case of 25 nm. Another reason could be due to the difference in the so-called mean free path of hot electron at Au NPs. It is noticed that the average radii of Au NPs formed at 10, 25 and 50 nm apparent thicknesses were 12.0, 18.1 and 26.8 nm, respectively. The mean free path of hot electron at Au NP was ~ 23 nm in length [51][52][53], which defines a mean distance that a photo-induced hot electron could move freely on the Au NP surface before being thermally dissipated. The average size of Au NPs formed by 50 nm apparent thickness was larger than the mean free path, while that for 25 nm was within the range. Since that the NIR light was incident directly onto the nano-MIS junctions as shown in Fig. 1, hot electrons appeared at the Au NP surface region away from the MIS junctions. Therefore, hot electrons were more difficult to reach and tunnel through the junctions for the case of 50 nm Au compared to the 25 nm one. Since the NIR absorption of nano-MIS structure for 10 nm Au was evidently lower as shown in Fig. 3a, the J SC of corresponding device had lower value than the one of 25 nm and 50 nm Au. When the apparent thickness of Au was raised from 10 to 50 nm, the V OC increased gradually from 0.20 to 0.22 V. This could be explained by the increasing area of nano-MIS junction [54]. For an ideal n-type Si/Au Schottky solar cell under AM0 illumination, the V OC was ~ 0.3 V [55], therefore, it is acceptable that the V OC of the nano-MIS solar cell here was ~ 0.2 V under NIR illumination. Figure 5b displays the EQE of devices with different sizes of Au NPs in sub-bandgap NIR wavelength range. With the increasing size of Au NPs, the variation of EQE was consistent with the changing trend of J SC tested under 1319 nm illumination. Figure 6 shows the schematic energy diagram of the nano-MIS NIR PV device. Since fixed positive and negative charges existed at the interfaces between SiO 2 /Si and Al 2 O 3 /Si, respectively [30,32], they strengthened the energy band bending of the MIS junction region, and facilitated the photo-induced charge transport. According to the IPE mechanism, hot electrons were excited from Au NPs and entered the conduction band of Si by tunneling through the Al 2 O 3 layer  [15,19]. They were then pushed toward the rear electrode of Al by the built-in electric field. At the meantime, the holes generated in Au NPs were pushed toward the front electrode of ITO and Ag grid [42].
To examine the effect of Al 2 O 3 layer on the PV parameters, Figs. 7a-d plot the short-circuit current density, opencircuit voltage, fill factor and conversion efficiency as functions of the layer thickness, respectively. From Fig. 7a, it is seen that the J SC decreased all the way with the increasing thickness. To fulfill the separation of photo-induced charges, hot electrons should tunnel through the MIS barrier as indicated in Fig. 6. Therefore, the photo-induced current would depend on the barrier width in a decay manner. Figure 7a suggested this trend. When Al 2 O 3 thickness was within 1 nm, the passivation of interface defects helps to improve the J SC , thus slowing down the decaying trend. The J SC followed an exponential dependence law with barrier thickness for cases beyond 1 nm, which is in agreement with the mechanism of Fowler-Nordheim field-assisted transport [22]. For the evolution of V OC , it increased at first till a maximum, and then decreased. It is known that fixed negative charges existed at the interface between Al 2 O 3 and Si [32]. This kind of fieldeffect passivation strengthened the band bending at the MIS junction, which tended to increase the V OC . However, when the Al 2 O 3 layer thickness further increased, the number of fixed negative charge approached saturation, but the J SC decreased significantly as shown in Fig. 7a, leading to decrease in the voltage [56]. Another explanation for the improved V OC by proper Al 2 O 3 thickness could be associated with the increment of Schottky barrier height ( b ). The Schottky barrier height was 0.77 eV and 0.81 eV for Au/n-Si and nano-MIS structure with 1 nm Al 2 O 3 , respectively, which was obtained by fitting of square root of EQE versus photon energy in 1200-1400 nm wavelength range [57]. For Schottky diode, V OC is linearly proportional to the Schottky barrier height [58]. Within 1 nm Al 2 O 3 thickness, the V OC of device increased in accordance with the barrier height (Fig. S2). The b value increased by 0.04 eV, consistent with the average improvement (~ 0.04 V) between the two devices. From I-V curves of two devices under dark condition (Fig. S3), it could be noticed

Research
Discover Nano (2023) 18:33 | https://doi.org/10.1186/s11671-023-03818-4 1 3 that the saturation current at reverse bias was evidently lowered after introducing 1 nm Al 2 O 3 , which verified the role of Al 2 O 3 to raise the Schottky barrier height. Figure 7c gives the dependence of fill factor on Al 2 O 3 thickness. The FF is mainly related to the sheet resistance of these devices [22]. Since the sufficient passivation of dangling bond defects at Au/Si interface could decrease sheet resistance, FF reached maximum when Al 2 O 3 thickness was 1 nm. For larger values of thickness, FF evidently decreased due to increasing sheet resistance of thicker Al 2 O 3 film. From Fig. 7d, it is seen that an optimal thickness of Al 2 O 3 existed in this nano-MIS device in terms of conversion efficiency. The average conversion efficiency of 1319 nm light was highest for 1 nm Al 2 O 3 , and the best efficiency of 0.034% has been experimentally found with PV parameters shown in Table 1. Besides, the role of SiO 2 layer on the rear surface of device was mentioned in Fig. S4. Figure 8a gives the EQE curve of photovoltaic response between 1200 and 1400 nm for the best nano-MIS PV device. The EQE at 1319 nm was 0.06%. For comparison, the EQE of a commercial PN junction crystalline Si solar cell was also drawn. In 1200-1250 nm wavelength range, the EQE of the PN junction solar cell dropped sharply almost to zero, as it could hardly absorb the sub-bandgap NIR light. However, the EQE of the nano-MIS structure was > 0.1%, and decreased more slowly beyond 1250 nm. The internal photoemission yield can be described by simple Fowler equation [59] where hυ is energy of incident photon, b is the barrier height in units of energy, and E F is the Fermi energy of the metal emitter. The EQE of device can then be approximated by A(hυ) ⋅ Y(hυ) , where A is absorption of incident photon. Theoretical calculation result of EQE was also plotted in Fig. 8a. It can be noted that experimentally tested EQE was in good agreement with calculated EQE, both in absolute value and spectral line shape. Additionally, the extra J SC gained from utilization of Si sub-bandgap NIR can be estimated by integral equation where q is electron charge, b S ( ) is photon flux density of solar spectrum. The result of extra J SC calculated from tested EQE was ~ 0.01 mA/cm 2 . The photoresponse characteristic of nano-MIS device at zero bias was investigated in Fig. 8(b). The observed responsive current to 1319 nm light illumination was 282 μA, corresponding to the J SC of device. Considering the dark current of 20 nA, it was proposed that our device could also function as a high-performance NIR photodetector which could operate at zero bias voltage, with responsivity (R) of 2.82 mA/W and detectivity (D * ) at 3.5 × 10 10 cm × Hz 1/2 /W calculated as follows.
(1) Compared to performance of NIR solar cell made by other materials such as PbS quantum dot working in the same wavelength range [60], the nano-MIS junction Si solar cell developed here had similar PV parameters of V OC and FF, only the J SC was one order of magnitude lower than their value, leading to our efficiency lower than theirs. Nonetheless, our device efficiency has reached the same level as the performance of TiO 2 cells working in visible range. According to the theoretical analysis of efficiency limit by IPE mechanism [59,61], the efficiency we obtained here was also reasonable, although it left room of promotion by, for example, improving the built-in electric field and further optimizing the generation, transportation and extraction of hot carriers [17]. Therefore, conversion efficiency close to 1% would be expected according to their simulation and this work could lay foundation for future researches on Si sub-bandgap NIR photovoltaic response.

Conclusions
In summary, we investigated a sub-bandgap NIR (λ > 1100 nm) Si solar cell with nano-Au/Al 2 O 3 /n-Si MIS junctions, demonstrating the Si sub-bandgap NIR PV response that has not been exploited by Schottky junction solid-state device. It employed LSP-enhanced inner photoemission mechanism, and light trapping mechanism by textured surface. Under the illumination of 1319 nm light with intensity of 0.1 W/cm 2 , a best conversion efficiency achieved here was 0.034%, with open-circuit voltage of 0.21 V, short-circuit current density of 0.282 mA/cm 2 and fill factor of 0.582, with the assistance of SiO 2 and Al 2 O 3 passivation. Our work is the first experimental demonstration of sub-bandgap NIR Si solar cell based on the MIS structure at room temperature. Such a Si-based plasmon-enhanced MIS photovoltaic device will be a promising strategy for application of Si sub-bandgap NIR photovoltaic response after further interfacial engineering.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
All authors gave their consent for publication.

Competing Interests
The authors declare that they have no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.