Ultra‐High Performance Amorphous Ga2O3 Photodetector Arrays for Solar‐Blind Imaging

Abstract The growing demand for scalable solar‐blind image sensors with remarkable photosensitive properties has stimulated the research on more advanced solar‐blind photodetector (SBPD) arrays. In this work, the authors demonstrate ultrahigh‐performance metal‐semiconductor‐metal (MSM) SBPDs based on amorphous (a‐) Ga2O3 via a post‐annealing process. The post‐annealed MSM a‐Ga2O3 SBPDs exhibit superhigh sensitivity of 733 A/W and high response speed of 18 ms, giving a high gain‐bandwidth product over 104 at 5 V. The SBPDs also show ultrahigh photo‐to‐dark current ratio of 3.9 × 107. Additionally, the PDs demonstrate super‐high specific detectivity of 3.9 × 1016 Jones owing to the extremely low noise down to 3.5 fW Hz−1/2, suggesting high signal‐to‐noise ratio. Underlying mechanism for such superior photoelectric properties is revealed by Kelvin probe force microscopy and first principles calculation. Furthermore, for the first time, a large‐scale, high‐uniformity 32 × 32 image sensor array based on the post‐annealed a‐Ga2O3 SBPDs is fabricated. Clear image of target object with high contrast can be obtained thanks to the high sensitivity and uniformity of the array. These results demonstrate the feasibility and practicality of the Ga2O3 PDs for applications in solar‐blind imaging, environmental monitoring, artificial intelligence and machine vision.


S2
The supporting information file includes: Figure S1. Schematic diagram of the MSM a-Ga2O3 SBPD.        Figures S10/S11/S12. DFT calculation analysis on Ga2O3.          This result also can be verified by the percentage of Ga2O species of 49.3% in the asdeposited a-Ga2O3 film, suggesting that annealing process results in the increase in the concentration of VO. As is shown in Figure S4c, the broad defect peak centered around 567 nm indicates the existence of VO in the as-deposited a-Ga2O3 film. However, the CL spectrum intensity of the as-deposited a-Ga2O3 film is lower than that of annealed a-Ga2O3 film, further confirming that annealed a-Ga2O3 film has higher concentration of VO. According to SCLC model, the electrical properties of annealed a-Ga2O3 film is clearly better with lower concentration of trap density and higher electron mobility, which is significantly better for photosensor application. As shown in Figure S6b, it can be found that the current increases a lot when the device is under 254 nm light illumination, and the photocurrent increases with the light intensity. However, dark current of the as-fabricated MSM a-Ga2O3 SBPD is much larger than that of the post-annealed (PA) MSM a-Ga2O3 SBPD, whereas the photocurrent is lower. According to the time-dependent photoresponse characteristics, the as-fabricated MSM a-Ga2O3 SBPD exhibits longer recovery time with τd1/τd2=22/383 ms. Based on the noise spectrum of the device, the specific detectivity (D * ) was calculated to be 3.9×10 12 Jones at 5 V bias voltage at 1kHz. As is presented in Figure S6f, the as-fabricated MSM a-Ga2O3 PD also exhibits solar-blind photodetection characteristics, but the rejection ratio is much smaller than that of the PA MSM a-Ga2O3 SBPD. Hence, the PA MSM a-Ga2O3 SBPD shows much higher performance than the as-fabricated MSM a-Ga2O3 SBPD.      Since the Ga2O3 film was annealed in N2 atmosphere, it is necessary to investigate the status of N in the Ga2O3 film. According to our DFT calculation, a deep defect level above the VBM was introduced when N was doped in Ga2O3. As shown in Figure S13b, a b S18 the defect level is very flat, indicating the effective mass is very large, and the mobility for the trapped electron is very low. This is one of the reasons that the PA MSM a-Ga2O3 SBPD has lower dark current.   The total scanning area is 10 µm×10 µm with width consisting of 5 µm wide a-Ga2O3 channel and 2.5 µm wide Ti/Au electrodes on both sides. Figure S7b shows surface a b c S22 potential variation process along the red arrow line in Figure S7c. It can be found that the CPD of as-fabricated MSM a-Ga2O3 SBPD under 254 nm light illumination also increases overall. However, surface CPD in the as-fabricated a-Ga2O3 channel region shows less increment compared to that in the PA MSM a-Ga2O3 SBPD, which indicates more carriers are photo-generated and trapped in the PA a-Ga2O3 film. Besides, surface CPD between Ti/Au metal stacks and as-fabricated a-Ga2O3 film became bigger after the device was illuminated with 254 nm light. This may well explain why the photocurrent of as-fabricated MSM a-Ga2O3 SBPD is much smaller than that of the PA MSM a-Ga2O3 SBPD by the following statements.
As confirmed in the XPS and TEM measurement, the PA a-Ga2O3 film has higher concentration of oxygen vacancy in the body area and at the interface of electrodes and a-Ga2O3 film. The neutral VO are tend to be ionized to VO 2+ states by the photoexcitation, donating two electrons to the conduction band, which makes contribution to the measured photocurrent. [18,19] Therefore, more photo-induced carriers will be generated in the PA a-Ga2O3 film. Additionally, the Schottky barrier lowering effect is enhanced in the PA a-Ga2O3 SBPD. Thus, increased electrons inject to Ga2O3 film from metal contact, producing large internal gain in PA MSM a-Ga2O3 SBPD.
It is worth noting that the potential distribution curve of the PA MSM a-Ga2O3 SBPD is bend up in Figure 4h, whereas the as-fabricated MSM a-Ga2O3 SBPD is the opposite.
This can be explained by work function difference of the a-Ga2O3 film. [20]  Besides, the sub-gap defect states were also obtained by CL and PL measurements.
Based on these measurements, the band information of the as-deposited and PA a-Ga2O3 film are presented in Figures S8 c and d. As was stated in the DFT calculation, VO 2+ are more easily to be produced in the PA a-Ga2O3 film, which may result in lower work function of the PA a-Ga2O3 film.

Supplementary Note 3. DFT calculations on Ga 2 O 3 and the corresponding analysis
First of all, the crystalline β-Ga2O3 model was used to analyze the formation energies of different oxygen vacancies and their effects on the electronic structure. And the amorphous model was used to simulate the annealing process. In all calculations, the AM05 functional was adopted, [21] which in general yields accurate lattice parameters.
For electronic structure calculations, the shGGA−1/2 method based on a GGA+U S25 ground state was employed, [22,23] which yields a 4.94 eV bandgap. The calculated bandgap value is in excellent agreement with the experimental value of 4.98 eV. The crystalline supercell structure of Ga2O3 used in our DFT calculations is shown in Figure   S10g. We have calculated the formation energies of three different kinds of oxygen vacancies under both O-rich and O-poor conditions, using the method as described in the work of Matsunaga et al. [24] Here, O1 is fourfold coordinated, while O2 and O3 are threefold coordinated. The formation energy results (shown in Figure S10b The formation of VO will introduce defect levels near the mid-gap, as shown in Figure   S10c (red dashed line). The introduced defect levels are located at 2.766 eV, 2.695 eV and 2.030 eV above the VBM for O1, O2 and O3, respectively. These values are consistent with the charge transfer levels (ε (+2/0)) for all inequivalent oxygen sites

S26
(2.569 eV, 2.316 eV and 2.094 eV, respectively). Therefore, ε (+2/0) reflects the gap between the VBM and the defect levels introduced by the oxygen vacancies. And the closer ε (+2/0) is to CBM, the easier it is for the formation of VO 2+ . In these defect states, electrons are found to be strongly localized with strong and narrow DOS as illustrated in Figure S10d. [19,25] Photoexcited holes are likely to be captured by these acceptor-like traps. In the region of metal electrodes, this will cause the lowering of barrier height, and thus contributing to the high internal gain of the PA MSM a-Ga2O3 SBPD. [26][27][28] Similarly, the energy band diagrams and the DOS of the Ga2O3 film with ionized VO 2+ are shown in Figure S10e  To better recover the realistic experimental situation, we also calculated the formation energy of oxygen vacancies in amorphous Ga2O3. To generate the a-Ga2O3

S27
in this work, DFT-based ab initio molecular dynamics (AIMD) simulations were performed. [29,30] The temperature and time step were respectively set to 3000 K and 5 ns during AIMD simulations, and structure relaxation was carried out to reduce the internal stress. On that basis, the formation energies of oxygen vacancies in a-Ga2O3 were calculated and the PA of a-Ga2O3 with oxygen vacancies was simulated. The temperature and time step were respectively set to 675 K and 5 ns for the PA simulation.
The calculation results are similar to the results obtained in single crystal β-Ga2O3.
The supercell structure of amorphous Ga2O3 is shown in Figure S11a. indicating that VO 2+ is created more easily after annealing. It is known that VO can be ionized to VO 2+ under photoexcitation, which acts as shallow donor level resonant with the conduction band. [19,31,32] The oxygen vacancies also play a vital role in accelerating S28 the electron-hole recombination, which will be discussed in detail in Supplementary Note 4. Therefore, the PA MSM a-Ga2O3 SBPD can achieve higher photocurrent and faster recovery speed.
Another effect of annealing for Ga2O3 is structural reconstruction and diffusion. As shown in Figure S11d, the atomic structure relaxation proceeds after the formation of VO on O1, O2 or O3 sites. Red dashed circles denote the oxygen vacancies, and the virtualized and materialized atoms indicate the atomic locations before and after the annealing process, respectively. We observe that the surrounding oxygen atoms tend to move to the vacancy sites, annihilating the original vacancy sites. On the other hand, Ga atoms tend to move away from oxygen vacancies, which is enhanced after annealing.
As listed in Table S2, the distance variations between the oxygen vacancies and neighboring Ga atoms are calculated. For O1, the displacement of coordinated Ga2 atom away from VO after annealing is 0.67 Å, much larger than that before annealing (0.11 Å). According to previous analysis, such outward displacement strengthens the Ga-O bond, raising the defect level to CBM. This is the reason why VO 2+ is more energetically favorable after annealing. Here we shall discuss the VO-assisted carrier recombination model using firstprinciples calculation. In the CL and PL spectra, the blue emission peaks were attributed to donor-acceptor pairs recombination. Through DFT calculation, the VO introduced defect levels were found to reside in the mid-gap, where electrons are found to be highly localized. The holes are easily trapped by these localized states under photoexcitation.

Supplementary Note 4. Discussion of recombination model
When the VO traps a hole, it becomes VO + , and the defect level moves to ~1.5 eV below the CBM as shown in Figure S12c. respectively. According to the differential charge diagrams ( Figure S14 c and d), after the formation of VGa the surrounding oxygen atoms will get more electrons from other Ga atoms, and the bonding will be strengthened. The electron is localized near the oxygen atom (in the opposite direction of the VGa). After VGa is formed near N atoms, the surrounding oxygen atoms still get more electrons from other Ga atoms, while electrons around nitrogen atoms gather toward VGa, forming defect levels. These defect states are not good recombination centers and do not match the CL and PL results.

Supplementary Note 6. Computational details of DFT calculation
The DFT calculations were carried out using the plane-wave-based Vienna Ab initio Simulation Package (VASP 5.4.4). [33,34] All calculations were based on a 1×4×2 β-Ga2O3 supercell that includes 64 Ga atoms and 96 O atoms. The generalized gradient approximation (GGA) functional with the Armiento-Mattsson 2005 (AM05) flavor was employed to account for the exchange-correlation energy, [21] since it is known to predict accurate lattice constants. The valence electron configurations were 4s and 4p for Ga; 2s and 2p for O. Projector augmented-wave pseudopotentials were used to replace the S32 core electrons. The plane wave kinetic energy cutoff was fixed to be 500 eV. For all structural relaxations, the convergence criterion in electronic self-consistent runs was set to 10 −6 eV, and structural optimization was obtained until the Hellmann-Feynman force acting on any atom was less than 0.02 eV/Å in each direction. The Brillouin zones were sampled by a 3×3×3 equal-spacing k-point mesh. The electronic structure calculation for β-Ga2O3 is quite challenging in that it involves closed shell 3d electrons that cannot be described well by DFT-GGA. Hence, a standard 8 eV on-site Hubbard correction was applied to the 3d orbitals of Ga. Subsequently, we employed an efficient self-energy correction scheme (DFT−1/2) originally proposed by Ferreira et al., and later improved by Xue et al. (shDFT−1/2). [22,23] The ab initio molecular dynamics (AIMD) technique was utilized to establish the amorphous Ga2O3 supercell, which was used to analyze the defect formation energy in amorphous Ga2O3. For the simulation of the annealing process, the annealing temperature was set to 675 K, and heat-up time was set to 5 ns.

Supplementary Note 7. Evaluation of uniformity test of the image sensor array
As shown in Figures S16 a and  Under 180 μW/cm 2 254 nm light illumination, photocurrent of hundreds of μA was measured in all pixels, demonstrating a high current contrast of ~10 4 . As presented in Figure S16d, the photocurrent shows large fluctuation, as high as 20% along line 3 and 4 in Figure S16c, which indicates the uniformity of photocurrent response of the image sensor array should be further improved in the future. Figure Figure S16a. Additionally, the output images in Figure 5d show unsharp shape of the light beam. This is probably because of cross-talk issue due to existence of current sneak path in the array. One possible way to solve this problem is to adopt one transistor-one photodetector (1T−1PD) or one diode-one photodetector (1D−1PD) structure. [35,36]