Excellent Responsivity and Low Dark Current Obtained With Metal-Assisted Chemical Etched Si Photodiode

Metal-assisted chemical etched (MACE; also known as MacEtch or MCCE) nanostructures are utilized widely in the solar cell industry due to their excellent optical properties combined with a simple and cost-efficient fabrication process. The photodetection community, on the other hand, has not shown much interest toward MACE due to its drawbacks, including insufficient surface passivation, increased junction recombination, and possible metal contamination, which are especially detrimental to p-n photodiodes. Here, we aim to change this by demonstrating how to fabricate high-performance MACE p-n photodiodes with above 90% external quantum efficiency (EQE) without external bias voltage at 200–1000 nm and dark current less than 3 nA/cm2 at −5 V using industrially applicable methods. The key is to utilize an induced junction created by an atomic layer deposited (ALD) highly charged Al2O3 thin film that simultaneously provides efficient field-effect passivation and full conformality over the MACE nanostructures. Achieving close to ideal performance demonstrates the vast potential of MACE nanostructures in the fabrication of high-performance low-cost p-n photodiodes.


I. INTRODUCTION
S ILICON-BASED p-n photodiodes that convert incident light to an electrical signal are extensively used in a wide range of commercial applications ranging from telecommunication to medical sensing and imaging and from consumer wearables to self-driving vehicles-just to name a few [1], [2], [3], [4]. To achieve high performance, namely, high sensitivity, from a p-n photodiode, minimizing reflection and recombination losses is of vital importance. In that sense, high-performance p-n photodiodes resemble high-efficiency Si solar cells [5], [6], although there are obviously some differences in, e.g., operating voltages and signal levels.
In state-of-the-art silicon solar cells, the reflectance losses have been mitigated by utilizing surface nanostructures [i.e., black silicon (b-Si)], which enable a drastic decrease in total reflectance (R) over a wide wavelength range (200-1000 nm) [7], [8], [9]. While there are different methods for the fabrication of surface nanostructures, in photovoltaics, the metal-assisted chemical etching (MACE; also known as MacEtch or MCCE) technique has turned out to be the most attractive, as it is a simple, cost-efficient, and crystal-damagefree process that has been shown to be easy to scale-up industrially [10], [11], [12], [13], [14], [15]. MACE should, therefore, offer an attractive means to reduce the reflectance losses also in industrial photodiodes.
The second key factor for achieving a high-sensitivity p-n photodiode is the reduction of carrier recombination. This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ Recombination in the Si bulk can be minimized by using high-quality substrates and avoiding contamination during semiconductor processing. Surface recombination can be suppressed by utilizing the so-called field-effect passivation by a highly charged atomic layer deposited (ALD) Al 2 O 3 film [16], [17]. This method has been introduced already in 2006 [18] and is nowadays commercially exploited, e.g., in Si solar cells. An additional benefit of ALD is that it forms a highly conformal and stable coating [19], [20], [21], and thus, the method works well also on nanostructured surfaces [16]. More specifically, for MACE nanostructures, we have previously reported [22] a high effective carrier lifetime of ≈400 µs (at an injection level of 10 15 cm −3 ) using ALD Al 2 O 3 passivation while keeping the surface reflectivity as low as 3.9% in the wavelength range of 300-1000 nm. After further development, even millisecond-range lifetimes were reached, indicating that MACE causes no contamination to Si bulk. Such high carrier lifetimes should fulfill the strict requirement for low dark current in photodiodes.
An additional source of recombination is the chargecollecting p-n junction that is traditionally fabricated via external doping, such as ion implantation or dopant diffusion. A high concentration of dopants results in increased Auger recombination inside the junction, and implantation may cause crystal damage that is not fully recovered during subsequent annealing. In p-n photodiodes, the dopantrelated recombination source can be mitigated by replacing the external doping with a built-in inversion layer, which forms underneath the same charged ALD Al 2 O 3 film that is used for surface passivation [23]. In solar cells, such an inversion layer would cause too high resistive losses due to a much higher current density, but in photodiodes, the inversion layer can boost the performance especially at short wavelengths, which are absorbed inside the junction [24], [25]. So far, the integration of such an ALD-based inversion layer with MACE nanostructures has not yet been reported, but it could be the key to raising the performance of MACE photodiodes to a completely new level.
While MACE has previously been applied to Schottky photodiodes [26], [27], it is well known that they always suffer from high dark current and, consequently, have limited sensitivity. For instance, Zhang et al. [26] have recently reported a high-responsivity MACE Si Schottky photodiode; however, the achieved dark current density is as high as 130 nA cm −2 at −1-V bias, which cannot compete with commercial Si photodiodes (typically <10 nA cm −2 ). Unlike Schottky diodes, p-n diodes are known for their low dark currents. However, there are only limited investigations made on MACE p-n photodiodes (e.g., by Zhong et al. [28]), and the obtained results have left room for improvement, falling clearly below the commercial counterparts [24], [29]. Thus, we believe that the full potential of MACE in p-n photodiode manufacturing has not yet been demonstrated.
In this article, our aim is to demonstrate a high-performance Si p-n photodiode by utilizing industrially feasible MACE nanostructures together with ALD Al 2 O 3 that reduces surface recombination and simultaneously induces an inversion layer forming a charge collecting junction. We characterize the spectral responsivity (R λ ) of the photodiodes in a wide wavelength range and compare the results to the existing literature. We analyze the results considering separately the optical and recombination losses along with other typical figures of merit, such as dark current and detectivity. Finally, the potential of using MACE in fabrication of high-performance commercial photodiodes is discussed. Fig. 1(a) presents a photograph of the fabricated MACE b-Si p-n photodiode. Several MACE b-Si and reference planar devices with a circular active area with a radius of 2.5 mm were fabricated on a 380-µm-thick 4-in n-type float-zone Si wafer with a resistivity of >9000 cm and a (111) orientation. The fabrication of the MACE b-Si devices is described next, while the planar reference devices follow the exact same steps excluding the actual MACE process. First, the wafer was treated at 1000 • C under an O 2 atmosphere for 84 min to grow ≈500-nm SiO 2 on the surfaces by wet oxidation. The oxide layer was patterned by photolithography and used as a mask material for the following ion implantation. Boron (B) and phosphorus (P) implantations were used to form highly doped areas allowing ohmic contact between the Si bulk and the contact metals on the front and rear sides of the wafer, respectively. The front opening areas were implanted with a B dose of 1 × 10 15 cm −2 and an ion energy of 33 keV, while the whole rear surface was implanted with a P dose of 2 × 10 15 cm −2 and an ion energy of 50 keV. The implantation was followed by a SiO 2 removal and drive-in for the dopants at 1000 • C for 50 min. A new SiO 2 layer was grown during the drive-in step, which served as an etching mask during the MACE process; i.e., the SiO 2 layer was removed locally only from the b-Si (or planar) active device areas. Next, a twostep room temperature MACE process was used to form b-Si nanostructures. The etching parameters were chosen based on our earlier work [22] to optimize the electrical and optical properties of the final devices. First, the wafer was dipped in a solution of HF (50 wt%) 80 mL, AgNO 3 (0.005 mol/L) 400 mL, and H 2 O 400 mL for 20 s to deposit Ag nanoparticles on the surface. Second, the wafer was etched in a solution of HF (50 wt%) 150 mL, H 2 O 2 (30 wt%) 100 mL, and H 2 O 2700 mL for 3 min to form surface nanostructures. Afterward, the Ag nanoparticles were removed in 69-wt% HNO 3 for 15 min. Fig. 1(b) shows a scanning electron microscope (SEM) image of the resulting nanostructures after Ag removal. The average diameter and height of the nanostructures measured by SEM were ≈0.25 and ≈1.5 µm, respectively.

II. EXPERIMENTAL SECTION
A 50-nm Al 2 O 3 layer was deposited with Beneq TFS-500 ALD reactor on the nanostructures at a temperature of 200 • C with trimethylaluminum and water as precursors. The negatively charged Al 2 O 3 film passivates the nanostructured surface and generates the p + inversion layer into the lowly doped n-type substrate, as demonstrated in Fig. 1(c). The Al 2 O 3 film was then patterned with photolithography to make the contact openings for the front metal contacts. Subsequently, 300-and 500-nm-thick Al layers were deposited by the MRC 903 sputtering system on the front-and reardoped surface regions, respectively. This was followed again with photolithography to pattern the front contacts. Finally, the wafers were annealed at 425 • C for 20 min in forming gas to activate the Al 2 O 3 passivation and to sinter the Al contacts. The cross section of the final MACE device is presented in Fig. 1(d). The cross section of the planar reference device is alike apart from the MACE b-Si structures.
For determining the external quantum efficiency (EQE), the photodiodes were attached to a printed circuit board and placed in a dark box sealed from room lighting. The used light source was Bentham ILD-D2-QH-24, which was directly coupled to the entrance port of a Bentham TMC300_0060 monochromator. The monochromator was used to select the wavelength (200-1100 nm) with 10-nm bandwidth, and the resulting beam was focused on the sample surface using lensbased relay optics. The light beam was first aligned perpendicular to the diode front surface, and the light-generated output current of the photodiodes was measured as a function of wavelength. Subsequently, the output current was transformed into EQE by calibrating the values against a b-Si detector, which had been calibrated using standards traceable to the Physikalisch-Technische Bundesanstalt (PTB) [25]. The corresponding spectral responsivity was calculated with R λ = EQE λ × (qλ/ hc), where q is the elementary charge, h the Planck constant, and λ and c the wavelength and speed of light, respectively.
The total reflectance was characterized between 250 and 1100 nm with a Cary 5000 UV-Vis-NIR spectrophotometer equipped with an integrating sphere. In addition, the homogeneity of the reflectance was characterized at the wavelengths of 656 and 984 nm using a PV2000A metrology tool (SEMI-LAB SDI) with a spot size of 100 µm. The internal quantum efficiency (IQE) was determined from the EQE and total reflectance measurements. Dark I -V curves were measured by a Hewlett-Packard Model 4145A semiconductor parameter analyzer. Finally, the specific detectivity (D * ) was calculated, assuming that the dark current is the dominating noise source using the equation [30], [31], D * = R λ ((A/2q I D )) 1/2 , where A is the active area of the detector, and I D is the dark current. Fig. 2(a) and (b) presents the spectral responsivity and the EQE of our MACE and planar reference photodiodes measured at zero bias voltage. These parameters illustrate how well the photodiode detects the light signal: the former describes the amount of output current per incident power at a given wavelength, and the latter describes the probability at which the incoming photon will be transferred to a current signal. As shown in Fig. 2, our MACE photodiode exhibits excellent responsivity (and EQE) for a wide wavelength range of 200-1000 nm, and the values are well above those reported previously for MACE photodiodes. Note also that when the wavelength of the incoming photon is 200 nm, the responsivity exceeds even the theoretical one electron per one photon limit raising the EQE to ≈102%. The phenomenon related to the planar Si photodiodes fabricated in this work measured at zero bias, and as a comparison, a state-of-the-art MACE p-n photodiode measured at -12-V bias [28].

III. RESULTS AND DISCUSSION
above unity EQE has been previously correlated with impact ionization; i.e., the UV photons have a high enough energy to excite multiple electron-hole pairs without the presence of an external bias voltage [32]. This phenomenon can be observed only in devices that have low enough recombination rates. The planar counterpart shows rather high responsivity as well, especially around 400 nm, but is still clearly below the performance of the MACE photodiode. This is because a single layer of Al 2 O 3 (50-nm thick) cannot provide similar antireflection performance to the MACE surface.
As a further comparison, the performance of a state-of-theart MACE p-n photodiode [28] is also included in Fig. 2. It is apparent that this photodiode performs best at near-infrared (NIR) wavelengths. However, it can reach a similar level to our MACE photodiode only in a rather narrow wavelength range (1000-1100 nm). Note that our and the reference MACE photodiodes have been measured at different reverse bias voltages (0 versus -12 V). We would expect an increase, albeit a minor one, in the responsivity of our device when measured at higher biases due to more efficient charge collection.
In order to explain the much higher performance of our MACE photodiode compared with the previous work, we analyze the optical and recombination properties as well as the dark current of the diodes. We start by examining the optical properties, namely, the total reflectance [ Fig. 3(a)]. It can be seen that the reflectance in our MACE nanostructured surface is rather constant (≈5%) in the whole wavelength range (250-1100 nm). Fig. 3(a) (inset) confirms that the reflectance of our MACE photodiode is also highly uniform spatially. The uniformity at both 656 and 984 nm is approximately ±1.2% absolute (note that the small difference between the absolute values of spectral and spatial reflectance is due to two different characterization tools). As expected, in the planar photodiode, the reflectance is much higher as compared with MACE surfaces and reaches its minimum value (≈13%) near 400 nm correlating with the antireflection effect produced by a 50-nm-thick Al 2 O 3 on planar Si surface. The reflectance curve of the reference MACE photodiode [28] is relatively flat and in the same range as our MACE photodiode. Therefore, the reflectance cannot explain the difference in responsivity between the two MACE photodiodes.
As mentioned in Section I, it is important to also analyze the possible recombination losses in the photodiode. This is best done via IQE, which disregards the reflection losses and, thus, describes how many electrons can be collected to the contacts per the number of absorbed photons. In other words, the IQE is a direct measure of recombination activity. Fig. 3(b) presents the IQE of our MACE and planar photodiodes as well as the reference MACE photodiode. The IQE of our MACE photodiode is close to ideal (100%) in a wide wavelength region confirming that it does not considerably suffer from either surface or bulk recombination. This is because we have used the following: 1) a high bulk lifetime substrate; 2) ALD Al 2 O 3 surface passivation; and 3) a low-recombination inversion layer as the charge collecting junction. It is worth highlighting that the high IQE in a wide wavelength range also proves that MACE is not a significant source of (metal) contamination to the wafers. In addition, it is worth noting that our planar photodiode also exhibits near perfect IQE, and in the UV (250-400 nm), the planar device even outperforms the MACE photodiode. This means that in MACE, the surface recombination is higher than in the planar surface, which is a reasonable result considering the higher surface area related to the nanostructures.
The nanostructures of the reference MACE photodiode were not passivated, and the p-n junction was formed by dopant diffusion. Both issues are likely to cause severe recombination losses near the front surface, and thus, these factors explain the resulting negligible IQE in UV. In the longer wavelength region (800-1000 nm), both our and reference MACE photodiodes result in relatively high IQE. Since these longer wavelengths are absorbed mostly deep in the Si bulk, the IQE results confirm that bulk or the rear surface recombinations are not significantly limiting the device performance in either MACE device. In addition to the spectral responsivity, the final device performance is also affected by dark current [ Fig. 4(a)]. These two together determine the sensitivity of the device. In our MACE photodiode, the dark current density measured at 1-V reverse bias is 1.3 nA cm −2 , which is comparable to commercial Si photodiodes (≈0.1-10 nA cm −2 ) [33], [34]. The dark current remains low with increasing reverse bias; e.g., at −5 V, the dark current density of the MACE device is around 2.1 nA cm −2 . Zhong et al. [28] did not report the dark current for their MACE photodiode, and therefore, it is not shown here. Interestingly, the dark current of our MACE device is on the same level as the planar reference photodiode. The low dark current means that there cannot be a significant amount of thermal generation centers (defects), and since all generation centers are also recombination centers [35], [36], this result further proves that MACE does not introduce significant metal contamination or other impurities into silicon. This has been a constant debate earlier when the reported leakage currents have been much higher in MACE diodes.
Another important parameter of the photodiode is the sensitivity, which is best described with the so-called specific detectivity that reflects the smallest detectable light intensity and allows the comparison between photodiodes of different sizes. Our MACE photodiode shows the detectivities of 8.0 × 10 12 and 3.6 × 10 13 Jones (Jones = cm Hz 1/2 W −1 ) at the wavelengths of 200 and 1000 nm under a reverse bias voltage of 1 V, respectively [ Fig. 4(b)]. Due to its higher responsivity, the MACE photodiode shows higher detectivity and, consequently, higher sensitivity to small light intensities compared with the planar one.
While the MACE photodiode presented here is outperforming the one reported earlier, it is worth noting that there are previous works on b-Si p-n photodiodes textured with more complicated and expensive methods, namely, using plasma etching [24], [25]. Such photodiodes exhibit even higher responsivity (closely following the ideal curve all the way from UV to NIR). However, plasma texturization cannot compete with the simplicity, up-scalability, and cost provided by MACE. Nevertheless, the comparison indicates that there is still room for improvement in the performance of MACE photodiodes. Indeed, as reported in the literature [27], even lower reflectance should be achievable by MACE. Likewise, as seen earlier in Fig. 3(a), the IQE of MACE could be slightly improved by reducing further the surface recombination. Therefore, by proper MACE and ALD process tuning, p-n photodiodes with low-cost and simple b-Si fabrication are realistic without compromising the high device performance.

IV. CONCLUSION
We have fabricated an MACE p-n photodiode with a recordhigh responsivity from 200 nm (164.5 mA W −1 ) to 1100 nm (456.0 mA W −1 ) with a peak at 1000 nm (731.3 mA W −1 ). Effective mitigation of reflectance by MACE nanostructures as well as junction and surface recombination by ALD Al 2 O 3induced junction and effective surface passivation, respectively, allowed reaching near ideal performance (EQE close to 100%) over the whole wavelength range. Furthermore, our photodiode demonstrates the possibility to reach low dark currents even with MACE nanotexturing, which, consequently, is seen as the detectivities of 8.0 × 10 12 and 3.6 × 10 13 Jones at the wavelengths of 200 and 1000 nm at a bias voltage of −1 V, respectively. The high sensitivity also proves the compatibility of the MACE process with conventional p-n photodiode fabrication without introducing contamination issues. Consequently, similar to solar cells, MACE has the potential to reform the photodetector industry by significantly improving device performance with low fabrication costs.