Hydrogenation of the wide-gap oxide semiconductor as a room-temperature and 3D-compatible electron doping technique

Hydrogen underlies the operations of various solid-state devices, passivating defects in silicon transistors and solar cells,¹ causing leakage current in ceramic capacitors and piezoelectric devices,² mediating charge transfer in fuel cells and atomic switches,^3,4 etc. Hydrogen can also behave as donor impurity in various oxide semiconductors generating one electron carrier and increasing electrical conductivity.^5,6 When considered as donor impurity, hydrogen has a unique character of extremely high permeability in solids,⁷ inducing various intriguing phenomena at room temperature.^8–11 By taking advantage of this characteristics, hydrogen may facilitate efficient electron doping for a range of oxide devices, which is applicable at room temperature in a 3D-compatible way.

Generally speaking, the technique of electron doping is essential for the study of device functionalities and physical phenomena in oxides.^12,13 So far, electron doping to oxides has often used in-situ doping during the growth or thermal treatment in the reactive environment.^14,15 However, these techniques lack the microscopic controllability of dopant distribution in the device structure, in contrast to the well established ion implantation for conventional semiconductors.¹⁶ For this issue, an electrochemical hydrogenation combined with photolithography could provide an alternative pathway, which could control the dopant distribution with high spacial resolution. The exploitation of electrochemistry is reasonable considering a rich variety of electrochemical phenomena in oxides such as anodic oxidization,^17,18 electrochromism,^19,20 resistance switching,^21,22 and so on. While the electrochemical hydrogenation has been used for nanotubes and nanoparticles to optimize the optical, catalytic, and photovoltaic properties,^19,23–25 it has rarely been applied to oxide semiconductor thin films in combination with photolithography. In this study, the electrochemical hydrogenation is successfully demonstrated at room temperature in the archetypical wide-gap oxide semiconductor TiO₂ thin film with photo-resist mask on the surface [Fig. 1(a)]. The unmasked region of the oxide thin film works as the cathode, inducing the cell reaction with the anode, namely, the external or surface metal electrode. This technique is inherently 3D-compatible with a reasonably high spatial resolution of several tens nanometer, showing the potential to be used for various oxide devices.

The electrochemical hydrogenation is demonstrated in polycrystalline anatase TiO₂ thin films on SiO₂/p:Si substrates. Polycrystalline anatase TiO₂ thin films with the thickness of 30-40 nm are fabricated by amorphous deposition at room temperature and crystallization during the thermal treatment.²⁶ Amorphous deposition is performed on the SiO₂(100-500 nm)/Si substrate by pulsed laser deposition (KrF) at the 1 Pa oxygen pressure. Crystallization is induced by thermal treatment in O₂ at 500-600 ^oC for 30-60 min. Scanning electron microscopy (SEM) in Fig. 1(b) shows anisotropic polycrystalline grain with a micrometer length scale in the in-plane direction. Transmission electron microscopy (TEM) in Fig. 1(c) shows a well-defined plate-like anatase grain, which helps us elucidate the role of grain boundary in the hydrogenation process as shown later.

Electrochemical hydrogenation is performed with the three-terminal potentiostat. The working electrode is the TiO₂ thin film with a small gold pad on the surface for the electrical contact. The counter electrode is platinum and the reference electrode is Ag|AgCl in the saturated KCl solution. The working electrode and the counter electrode are immersed in the potassium phosphate buffer solution with pH = 6.3, which is bridged to the reference electrode via the KCl salt bridge. After keeping the target potential for 1 min, the working electrode is rinsed by water, and the electrical measurement is performed. Here, the electrical potential is defined by the external voltage applied on the working electrode with respect to the reference electrode, from which the electrode potential of the reference electrode (0.199 V vs. SHE) is also subtracted.

As shown in Fig. 2(a), the sheet resistance of the dried TiO₂ thin film decreases from ∼10⁹ Ω (insulating) at -0.6 V vs. standard hydrogen electrode (SHE), and saturates at ∼10⁴ Ω (conductive) below -0.8 V vs. SHE (filled diamonds). This decrease in the sheet resistance indicates the electrochemical reduction of TiO₂, in other words, the electron doping. The saturation of the sheet resistance implies the reduction reaction of TiO₂ is impeded by the side reaction such as hydrogen generation. This reduced TiO₂ thin film is further exposed to various electrical potentials subsequently, and the sheet resistance is found to be modulated within one order reversibly as shown by open diamonds and square in Fig. 2(a). However, the sheet resistance never recovers to the initial value ∼10⁹ Ω, possibly because the positive electrical potential on TiO₂ forms a blocking contact (depletion region) between TiO₂ and electrolyte.

In spite of the large decrease in sheet resistance, the TiO₂ thin film shows negligible change in its crystalline structure by electrochemical hydrogenation. X-ray diffraction before (“Initial”) and after (“Doped”) hydrogenation shows no difference neither in the peak position, the peak intensity, nor the peak width [Figs. 2(b) and 2(c)], indicating negligible change in the lattice constant and crystallinity. Here, X-ray diffraction is performed in the plane by using grazing incident X-ray in order to maximize the diffraction intensity. The angle between the X-ray source and the sample surface is 0.3° (below the total reflection angle), and the detector is parallel to the sample surface (2θ = ω = 0.3°). In the plane, the angle between the X-ray source and the detector (2θ_χ) is kept two times larger than the angle of in-plane sample rotation (Φ). All the diffraction peaks in Fig. 2(b) correspond to the anatase phase of TiO₂. The absence of structural change in TiO₂ is also confirmed by Raman spectroscopy (not shown), indicating a great advantage of this technique for damage-less electron doping.

In order to elucidate the mechanism of hydrogenation and also the ability of restricting the hydrogenated region, the electrochemical process is performed through the line-and-space photo resist mask on the TiO₂ surface. In this case, because the sheet resistance of the TiO₂ thin film saturates below -0.8 V vs. SHE as shown in Fig. 2(a), the same electrochemical reduction occurs simply by putting aluminum anode (-1.66 V vs. SHE) on the TiO₂ surface as shown in Fig. 1(a), and immersing the sample in diluted hydrochloric acid or diluted tetramethylammonium hydroxide solution for 1 min without external voltage. Then, the photo resist mask is removed by acetone, and the conductive area on the TiO₂ thin film is measured by conductive atomic force microscopy (cAFM). Here, the AFM and cAFM is performed by using Cypher S of Oxford Instruments. The iridium coated silicon tip is grounded, and the hydrogenated TiO₂ thin film is electrically biased by 3 V during the measurement. The tip curvature radius of 28 nm enables the spatial resolution of a few tens nm.

The cAFM image clearly shows the line-and-space pattern corresponding to the area which is uncovered by the photo-resist mask [Fig. 3(a)]. This line-and-space conductivity pattern is in contrast to the simultaneously measured TiO₂ surface morphology [Fig. 3(b)], which only shows a flat surface with a root mean square of ∼0.3 nm. In this electrochemical process, there are three possible mechanisms of TiO₂ reduction: the reduction simply through the top surface, the reduction mediated by the grain boundary, and the reduction via the underlying SiO₂ layer.^27,28 However, the last possibility can be denied by the experimentally observed conductivity pattern because the reduction from the underneath would not be affected by the surface mask and lead to a homogeneous conductivity pattern. In order to further clarify the reduction mechanism, the magnified scan is performed for the square region in Figs. 3(a) and 3(b). The surface morphology in this region shows several grain boundaries as shown by arrows in Figure 3(d). On the other hand, the corresponding cAFM image in Figure 3(c) shows nontrivial conductivity even in the area far from the grain boundaries, indicating the reduction occurs relatively homogeneously inside the TiO₂ grain. Besides, the edge line of the doped region is straight and almost unaffected by the position of each grain boundary. This negligible influence of the grain boundary indicates the reduction reaction is not mediated by the grain boundary but mainly occurs through the grain surface. This result is also consistent with the ability of electrochemical reduction even in the anatase single crystal, which has no grain boundary.¹⁰ It should be noted the apparent high conductivity along the grain boundary in Fig. 3(c) does not necessarily mean the higher conductive at the boundary, but includes crosstalks from the surface morphology signal.

The specific reduction reaction in this electrochemical process would be the hydrogenation according to the past studies.^9,10,25 The hydrogenation of TiO₂ is actually confirmed by secondary ion mass spectroscopy (SIMS). The SIMS is measured by NANO SCIENCE CORPORATION with 2 keV cesium ion beam in ADEPT-1010 (ULVAC-PHI, Inc.). The backside SIMS (shown later) is measured by Toray Research Center, Inc. with 3 keV cesium ion beam in ADEPT-1010 (ULVAC-PHI, Inc.). When SIMS is applied to hydrogen, it should be distinguished from the hydrogen inside the surface adsorbed water. To address this issue, the TiO₂ thin film is electrochemically reduced in the electrolyte using D₂O, and is subsequently rinsed in H₂O in order to exchange the adsorbed water. In the experiment, this “Doped” sample is prepared by immersing the TiO₂ thin film with a small aluminum pad in 0.5 mol L^-1 deuterium chloride solution for 1 min, and then by rinsing it in H₂O for 30 sec. The “Non-doped” sample (for reference) is prepared by immersing the TiO₂ thin film without an aluminum pad in 0.5 mol L^-1 deuterium chloride solution for 1 min, and then by rinsing it in H₂O for 30 sec.

The SIMS spectra in Fig. 4(a) show a large amount of deuterium in this “Doped” sample while showing almost no deuterium in the reference “Non-doped” sample. The deuterium concentration integrated throughout the spectrum (from the SiO₂ layer to the TiO₂ surface) is around 1 × 10¹⁴ cm^-2, approximately consistent with the sheet electron density from the Hall measurement (N_2D = 6 × 10¹³ cm^-2). It should be noted Al and Cl are also measured (using 2 keV oxygen ion beam only for Al), but the obtained concentration is below ∼3 × 10¹⁶ cm^-3 for Al, and below ∼3 × 10¹⁷ cm^-3 for Cl.

Importantly, the deuterium profile in Fig. 4(a) is strongly inhomogeneous in the plane-normal direction, and shows a larger concentration on the side of the TiO₂/SiO₂ bottom interface. This inhomogeneous deuterium profile is not attributed to the deuterium diffusion during the measurement because the back-side SIMS, which is measured in the opposite direction from the SiO₂ side, also shows the similar inhomogeneous profile (“Doped, back” in Fig. 4(a)). In order to further investigate the influence of the interface, the N_2D in the hydrogenated TiO₂ thin films is plotted as a function of the TiO₂ film thicknesses in Fig. 4(d). A nontrivial amount of N_2D remains at the zero thickness limit (∼4 × 10¹³ cm^-2), indicating a considerable amount of donor hydrogen is located at the interface rather than inside the TiO₂ thin film. The existence of the interfacial hydrogen is also consistent with the large interfacial deuterium concentration for the “Doped” sample in Fig. 4(a). This interfacial deuterium concentration is unaccountable only by the matrix effect when it is compared with the SIMS spectra of hydrogen (H) as shown in Figs. 4(b) and 4(c). In both cases, the interfacial concentration of D is much larger than that of H although the matrix effects for H and D have approximately similar trend to each other.²⁹ It should be noted the interfacial deuterium concentration becomes smaller in the back-side SIMS because the thermal process during the preparation process for this measurement (60 °C for 6 hours) may partially release deuterium.

One of the concerns about the electrochemical hydrogenation is the low stability of doped hydrogen, which is previously reported.⁹ Indeed, most electrochemically doped hydrogen in bulk TiO₂ does not form -OH bonds, and hence, diffuses out only in a few hours. The formation of rigid -OH bonds requires the thermal treatment in the hydrogen atmosphere. Surprisingly, the electrochemically doped hydrogen in the TiO₂ thin film at room temperature is stable as indicated by the time dependence of N_2D [Fig. 5(a)]. Although N_2D decreases by 20 % in the first day, the decay becomes slower, and N_2D almost saturates after several tens days. Fig. 5(b) shows the log-log plot of N_2D as a function of time, where the simple extrapolation of the experimental data implies N_2D can be maintained above 1 × 10¹³ cm^-2 for as long as 10⁶ days (∼2,700 years). Generally, the hydrogen can be stabilized by dangling bonds and interstitial sites, for example, the dangling bonds in amorphous silicon,³⁰ and the interstitial sites in LnNi₅-type (Ln: lanthanoid) hydrogen storage alloy.³¹ Considering a lot of dangling bonds and interstitial sites exist at the interface, they may account for the experimentally observed high stability of hydrogen. It should be noted that the strong interaction of the interface with hydrogen has also been recognized in other materials systems such as the SiO₂/Si interface in silicon transistors,³² the Pd/oxide interface in hydrogen gas sensors,³³ and the metal/ceramics interface in the mechanical component.³⁴ 

One of the unique characteristics of this electrochemical hydrogenation with respect to other doping technique is the inhomogeneous dopant distribution perpendicular to the interface; the hydrogen concentration is largest at the interface and becomes smaller far from the interface [Fig. 4(a)]. Obviously, this distribution has no problem when we just impart electrical conductivity to the TiO₂ thin films or nanoparticles. In such cases, the inhomogeneous distribution would be rather advantageous due to the high hydrogen stability as shown in Fig. 5. On the other hand, the inhomogeneous distribution may be problematic for fundamental research, which often needs to control the absolute dopant concentration with perfect spatial homogeneity. In such cases, the thermal treatment in hydrogen atmosphere is more preferable because it may achieve homogeneous hydrogen distribution by stabilizing hydrogen in the bulk part of TiO₂.⁹ Thus, the inhomogeneity of electrochemical hydrogenation makes this technique more suitable for simply imparting electrical conductivity to the TiO₂ thin films and nanoparticles than for precisely controlling hydrogen dopant concentration.

In conclusion, we demonstrate the electrochemical hydrogenation of the polycrystalline anatase TiO₂ thin film at room temperature, transforming the insulating TiO₂ to the highly conductive n-type semiconductor. The hydrogenation occurs almost homogeneously in the plane and little influenced by the grain boundary. As a result, the hydrogenated area can be precisely controlled by photo-resist mask on the surface, providing a powerful electron doping technique for oxide devices. In the plane-normal direction, it is indicated a considerable amount of hydrogen exists around the bottom interface in the hydrogenated TiO₂ thin film. This interfacial hydrogen reservoir will affect electrochemical phenomena in TiO₂, as well as facilitate novel devices such as ion-electron transducers and neuromorphic devices.^35,36

This work was supported by MEXT/JSPS KAKENHI Grant Number JP17K18869.