Schottky contacts to In₂O₃

n-type binary compound semiconductors such as InN, InAs, or In₂O₃ are especial because the branch-point energy or charge neutrality level lies within the conduction band. Their tendency to form a surface electron accumulation layer prevents the formation of rectifying Schottky contacts. Utilizing a reactive sputtering process in an oxygen-containing atmosphere, we demonstrate Schottky barrier diodes on indium oxide thin films with rectifying properties being sufficient for space charge layer spectroscopy. Conventional non-reactive sputtering resulted in ohmic contacts. We compare the rectification of Pt, Pd, and Au Schottky contacts on In₂O₃ and discuss temperature-dependent current-voltage characteristics of Pt/In₂O₃ in detail. The results substantiate the picture of oxygen vacancies being the source of electrons accumulating at the surface, however, the position of the charge neutrality level and/or the prediction of Schottky barrier heights from it are questioned.

Semiconducting oxides (SO) find application in transparent electrodes for thin film solar cells, as rectifiers, as transistors, as UV-photodetectors, and as chemical, gas and pH-sensors. In order to realize devices with optimized performance, the bulk as well as the surface properties of the SO-layers must be controllable and adjustable. In₂O₃ thin films with high room temperature (RT) mobilities (up to 230 cm²/V s^1,2) and controllable conductivities ranging from semi-insulating (by Mg-doping³) to highly conductive (by Sn-doping⁴) have favorable bulk transport properties for transistor applications. However, especially the surface electron accumulation layer (SEAL) of SOs is difficult to modify and in some cases like for In₂O₃ not sufficiently tunable to allow the formation of Schottky contacts. King et al. explained the surface electron accumulation as well as the ease of extrinsic n-type doping of In₂O₃ by the position of the branch point energy E_bp (sometimes referred to as charge neutrality level) within the conduction band and about 0.4 eV above the conduction band minimum (CBM).⁵ Recently, a value of 0.35 eV above CBM was derived by quasi-particle calculations in close agreement with the experimental value.⁶ The microscopic origin of surface electrons is likely the formation of oxygen vacancies at the topmost surface layer(s).^7,8 A remote oxygen plasma treatment was capable of removing the electron accumulation at the surface of ZnO^9,10 and allowed the fabrication of Schottky contacts with acceptable rectification thereon. Qualitatively the same results have been obtained with SnO₂ films by an oxygen plasma treatment of the surface.^11,12 For In₂O₃, Bierwagen and co-workers investigated the change of the surface band bending due to the same oxygen plasma treatment by x-ray photoemission spectroscopy.¹³ As-grown samples showed a downward band bending (surface Fermi level is above conduction band minimum) independent of the growth conditions. The oxygen plasma treatment reversed the band bending at the surface such that an upward bending was observed after the treatment, however, current-voltage (IV) characteristics of mercury (Hg) contacts remained poor. Here we present a proof of principle study demonstrating that a reactive sputtering process of the Schottky contact metal in an oxygen-containing ambient addresses the obstacles for fabrication of Schottky barrier diodes on In₂O₃ and show its suitability for the fabrication of Schottky diodes on In₂O₃(001) thin films grown by molecular beam epitaxy (MBE). The rectification achieved is sufficient for space charge spectroscopic characterization methods.

Three approximately 500 nm-thick In₂O₃(001) films were grown by plasma-assisted molecular beam epitaxy on (001)-oriented yttria-stabilized zirconia (YSZ) substrates. These films comprise an unintentionally doped (UID) film, a lightly Mg-doped film with Mg concentration of N_Mg = 1.3 × 10¹⁷ cm⁻³, and a lightly Sn-doped film with Sn concentration of N_Sn = 10¹⁸ cm⁻³. Details on Mg and Sn-doping are described in Refs. 3 and 4, respectively. Differences in the post-growth cool-down conditions in the MBE chamber (oxygen plasma or vacuum), and their respective consequences on oxygen-related defects,^1,3,4 were equalized by annealing the samples for 1 min at 750 °C in air prior to contact deposition. Standard photolithography was used to pattern ohmic and circular Schottky contacts. For ohmic contacts gold was dc-sputtered under Ar atmosphere. RT Hall effect measurements were performed at a magnetic field of 0.43 T. Here, ohmic contacts were placed at the corners of square-shaped 5 × 5 mm² samples. For the creation of Schottky contacts sputtering in Ar was not successful independent of the contact metal used. Similar to former reports for ZnO we applied reactive dc-sputtering of Au, Pd, and Pt.^14–16 The gas flow of 50 vol.% argon and 50 vol.% oxygen was 100  standard cubic centimeter per minute resulting in a total pressure of 0.025 mbar during the sputtering process. The target diameter is 30 mm, the target-to-substrate distance was 4 cm and a sputtering power of 30 W was used. The properties of the different contact metals are compared by means of RT IV measurements conducted in a SUESS probe station P200 connected to an Agilent Semiconductor Parameter Analyzer 4155C. For each metal an ensemble of 30 contacts or more was investigated per sample. For selected diodes the IV measurements were recorded in a temperature range from 293 K to 443 K.

Hall effect measurements on the annealed samples yielded free electron concentration n and the Hall mobility μ_H. The RT values are n = 1.9 × 10¹⁷ cm⁻³ and μ_H = 201 cm²/V s for the UID, n = 1.6 × 10¹⁷ cm⁻³ and μ_H = 147 cm²/V s for the Mg-doped and n = 8.5 × 10¹⁷ cm⁻³ and μ_H = 140 cm²/V s for the Sn-doped sample. As expected, the free electron concentration is highest for the Sn-doped and lowest for the Mg-doped sample; Hall mobility is highest for the UID sample and lowest for the Sn-doped sample.

Similar to evaporated metal contacts to SnO₂ described already in the literature¹¹ we could not observe a rectifying behavior for non-reactively dc-sputtered Au, Pd, and Pt on In₂O₃. In Fig. 1(a) the characteristic of a non-reactively sputtered Pt contact on the UID sample is shown exemplarily (see grey line in the figure) for all metal contacts that were realized that way on each sample type. None of these contacts showed rectification. Rectifying contacts were obtained by reactive sputtering of the respective metal as depicted in Fig. 1(a) for the case of Pt. A detailed microscopic understanding of Schottky contact formation does not exist so far. We speculate that, similar to the oxygen plasma treatment reported by Bierwagen and co-workers,¹³ negative oxygen ions impinge the sample surface at the very beginning of the sputtering process. These oxygen ions might fill surface and surface-near oxygen vacancies, by that remove the origin of the SEAL, and move the surface Fermi level below the CBM. Positively charged argon ions simultaneously ablate the metal target such that the cleaned In₂O₃ surface is covered right away by the Schottky contact metal. Best room temperature current-voltage characteristics of reactively sputtered Au, Pd, and Pt contacts are depicted in Fig. 1 for the UID, the Mg-, and the Sn-doped sample with corresponding fits. For the fit we assumed that current transport across the barrier is due to thermionic emission, only (see the supplementary material for more details¹⁹).

Temperature-dependent current-voltage characteristics were measured for upward and downward temperature ramping for Pt/In₂O₃ in air between RT and 150 °C and are shown and discussed in the supplementary material.¹⁹ At a temperature of 120 °C (393 K) a strong increase of the diode current is observed. At this temperature the barrier properties change. The series resistance of the diode decreased such that the current under flat-band conditions is now similar to that of Au Schottky barrier diodes. Further, the RT barrier height and the ideality factor are lower after the temperature cycle. For both the upward and the downward temperature ramping we determined the barrier height and ideality factor which are depicted in Fig. 2 in dependence on the inverse temperature. From the

$ΦB eff (T)$ vs. 1/T plot the homogeneous or mean barrier height and the standard deviation of the Gaussian broadened barrier potential are obtained.^17,18 During the heat cycle the homogeneous barrier height

$\Phi _\mathrm{B}^\mathrm{hom}$

$ΦB hom$ decreased from about 1.13 eV to 0.99 eV, whereas the standard deviation σ remained with 0.17 and 0.16 eV about constant. This is also reflected by the fact that the voltage dependence of the standard deviation is with about −0.046 and -0.049 eV essentially the same prior and after heating. The voltage dependence ρ₂ of the homogeneous barrier height changes consequently from 0.21 to 0.19. We emphasize that the diode undergoes irreversible changes at a temperature of 120 °C after these changes the diode is stable and fully functional up to at least 150 °C.

Now, we can compare the experimentally derived homogeneous barrier height to values predicted within the metal-induced gap state (MIGS) and electronegativity concept:²⁰ 

where

$Φ bp n$ is the n-type branch-point energy (difference between energy of the conduction band minimum E_c and the branch-point energy E_bp,

$\Phi _\mathrm{bp}^n\,=\, E_\mathrm{c}-E_\mathrm{bp}$

$Φ bp n=Ec−E bp$⁠), S_X is the slope parameter and depends on the high-frequency dielectric constant

$S_X\,=\, \Phi _\mathrm{B}^\mathrm{eff}A_X/(1+0.1(\varepsilon _\infty -1)^2)$

$SX=ΦB eff AX/(1+0.1(ɛ∞−1)2)$⁠, X_M and

$X_\mathrm{In_2O_3}$

$XIn2O3$ are the electronegativity of the contact metal and indium oxide in Miedema units.

$X_\mathrm{In_2O_3}\,=$

$XIn2O3=$ 5.73 eV was obtained converting its Pauling scale value of 2.52 eV²¹ to Miedema scale. A_X = 0.86  is a proportionality factor linking metal work function and the electronegativity of the metal in Miedema unit. Using the high-frequency dielectric constant ɛ_∞ = 4.43 derived from ɛ_∞ × m_opt = 1.24 m₀ and m_opt = 0.28 m₀²² and the experimentally determined branch-point energy of King et al.⁵ we find for Pt/In₂O₃:

$\Phi _\mathrm{B,Pt}^\mathrm{hom}$

$ΦB,Pthom$ = −0.4 eV + 0.395 × (5.65 eV − 5.73 eV) = −0.43 eV. This actually states that a Schottky barrier should not form between Pt and In₂O₃ being in conflict with the above data. The large discrepancy between predicted and experimentally determined homogeneous barrier height is illustrated in Fig. 2(c). The dashed green area indicates values calculated using Eq. (1) for 0.1 < S_X < 0.4 eV and 0 < E_bp − E_c < 0.5 eV. The green rectangle represents the range of homogeneous barrier heights (including experimental errors) deduced for Pt/In₂O₃ from the temperature-dependent IV data. According to the MIGS and electronegativity concept a Schottky barrier should not form between Au, Pd, Pt, and In₂O₃, respectively. For the II-VI semiconductor ZnO a similarly large discrepancy between predicted and experimentally determined barrier heights was reported²³ and suggests revisiting the predictive power of the MIGS and electronegativity concept and/or branch-point energies. Of course, more experimental work on Schottky barrier heights on In₂O₃ is required to establish experimentally chemical trends of the barrier height for various contact metals.

For Pt/In₂O₃(UID) diode space charge spectroscopic investigations are briefly discussed. From capacitance-voltage measurements at RT a mean net doping density of 1.5 × 10¹⁷ cm⁻³ was determined, a value in excellent agreement with the Hall effect data. However, the net doping density is not constant but varies along the growth direction. This hinders an unambiguous determination of the homogeneous barrier height. Thermal admittance spectroscopy was measured in a temperature range from about 20 K to 330 K at zero bias. The temperature dependence of the diode capacitance is depicted for selected probing frequencies ω in Fig. 3. In the temperature range from 20 K to 60 K freeze-out of carriers occurs. For temperatures higher than 200 K a deep-level defect contributes to the capacitance. For the deep-level defect standard maximum evaluation of G(ω)/ω was applied to construct the Arrhenius plot depicted in the inset of Fig. 3. The thermal activation energy and apparent capture cross-section are determined from linear regression of the Arrhenius straight line to be E_t = 290 meV and σ_app = 6 × 10⁻¹⁵ cm². For evaluation of carrier freeze-out Pautrat et al.²⁴ proposed a model that requires the temperature dependence of the carrier mobility in the temperature range in which freeze-out occurs as input.

Temperature-dependent mobility data do not exist, hence we consider ionized impurity (IIS) and grain boundary scattering (GBS) typically limiting low temperature mobility of oxide semiconducting thin films.^25,26 For IIS an unrealistically small value E_t = 5.9 meV is obtained and so GBS is considered. Here, the height of the potential barrier between adjacent grains enters evaluation. We have constructed an Arrhenius plot for a typical range of V_b and obtain the following E_t(V_b): E_t(5 meV) = 18 meV, E_t(10 meV) = 23 meV, and E_t(20 meV) = 33 meV. We assume the activation energy between about 25 and 30 meV.

In summary we have realized Schottky barrier contacts on UID, lightly Mg-doped and lightly Sn-Doped MBE-grown In₂O₃(001) thin film samples by a reactive sputtering process in an O₂/Ar environment. The diode current can be well described by thermionic emission across a laterally inhomogeneous barrier. Temperature-dependent measurements reveal that these barrier fluctuations can be well modeled by a Gaussian distribution. Best contacts were obtained with Pt on Mg-doped samples and showed rectification of more than three orders of magnitude, the averaged RT effective barrier height of Au, Pd, and Pt on the UID sample is 0.50, 0.49, and 0.57 eV, respectively. For Pt contacts on the Mg-doped sample changes of the contact properties were observed at 120 °C. After that, the diode was stable up to at least 150 °C and its mean barrier height was determined to be about 1 eV. The realization of first rectifying contacts to In₂O₃ enables systematic characterization of In₂O₃ by depletion region based characterization techniques and the use in unipolar field-effect devices. During the reactive sputtering process oxygen ions likely removed surface oxygen vacancies that were attributed to cause the SEAL. We note, however, that removal of the SEAL contrasts the charge neutrality picture that predicts donor type defects (the surface itself or defects caused by the O-ion damage) to move the Fermi level into the conduction band. Further insights into the effect of the oxygen ions on surface accumulation and a refined theoretical concept for predicting Schottky barrier heights on SOs like In₂O₃ or ZnO and/or additional investigations of their branch point energy are demanded.