Fabrication and characterization of Schottky barrier diodes on rutile TiO2

Schottky barrier diodes (SBDs) were fabricated by depositing Pd, Pt or Ni on single crystal, conductive n-type rutile TiO2 using e-beam evaporation. As-grown and nominally undoped rutile TiO2 single crystals are semi-insulating, and were heat-treated in forming gas flow, N2 flow or H2 gas to obtain conductive n-type crystals displaying electrical conductivities in the range of ( 0.5 − 8 ) × 10 − 2 Ω − 1 cm − 1 . Additionally, SBDs were deposited on Nb-doped conductive n-type rutile TiO2 with a conductivity of around 0.25 Ω − 1 cm − 1 . Generally, SBDs displaying a rectification of up to eight orders of magnitude were obtained, when comparing the current under reverse and forward bias. The extracted ideality factors were in the range of 1.1 − 4.0 . From Capacitance-Voltage measurements, the built-in voltage was derived to be around 1.2 V – 1.9 V , depending on the doping concentration of the specific TiO2 single crystal. Series resistances as low as 19 Ω were achieved. A considerable variation in the electrical characteristics of different SBDs deposited on the same crystal was found, regardless of the metal or doping strategy used. Moreover, the SBD characteristics change over time, particularly seen as a degradation in rectification, mainly related to an increase in the current under reverse bias. Additional surface treatments such as boiling in H2O2 and etching in HF do not have a significant effect on the quality of the SBDs. Clear indications for poor adhesion between TiO2 and Pd are shown. In conclusion, we demonstrate the fabrication of SBDs which are suitable for studying the fundamental properties of metal/TiO2 junctions and the characteristics of electrically-active defects in TiO2 using space-charge spectroscopy.

-, depending on the doping concentration of the specific TiO 2 single crystal. Series resistances as low as W 19 were achieved. A considerable variation in the electrical characteristics of different SBDs deposited on the same crystal was found, regardless of the metal or doping strategy used. Moreover, the SBD characteristics change over time, particularly seen as a degradation in rectification, mainly related to an increase in the current under reverse bias. Additional surface treatments such as boiling in H 2 O 2 and etching in HF do not have a significant effect on the quality of the SBDs. Clear indications for poor adhesion between TiO 2 and Pd are shown. In conclusion, we demonstrate the fabrication of SBDs which are suitable for studying the fundamental properties of metal/TiO 2 junctions and the characteristics of electricallyactive defects in TiO 2 using space-charge spectroscopy.
In applications for photo-catalysis, systems consisting of TiO 2 and noble metals, such as Pd, Pt, Au or Ag, display a better performance compared to just TiO 2 [29][30][31][32][33]. This is believed to be due to an improved charge Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. transfer at the interface between the metal and TiO 2 due to the formation of a rectifying junction [34]. Besides photo-catalytic applications, rectifying metal/TiO 2 junctions are also of interest for UV photo-detectors [35], hydrogen sensors [36] and switching devices [37][38][39].
In order to improve metal/TiO 2 junctions for all applications mentioned above, the charge transfer across the interface needs to be understood better. This requires knowledge regarding the electrical characteristics of rectifying metal/TiO 2 junctions, where simplified model systems are often a superior starting point for such investigations. Besides studying the properties of the metal/TiO 2 junctions themselves, the junctions can also be used to probe the properties of defects in TiO 2 using space-charge spectroscopy [16,40,41]. Thus, the fabrication of Schottky barrier diodes (SBDs) involving conductive TiO 2 single crystals obtained by employing different doping strategies paves the way for systematic defect studies in such material as well as enables investigations of the fundamental properties of the corresponding metal/TiO 2 junctions.
SBDs have been achieved on TiO 2 using Pt [42][43][44], Pd [45][46][47], Ni [48] and Au [19,[49][50][51]. The characteristics of these SBDs show a varying degree of non-ideal behaviour. Although there is a considerable amount of literature on SBDs involving TiO 2 , only a few reports exist comparing SBDs on differently doped TiO 2 crystals fabricated in the same laboratory. Additionally, there is a general concern about the long-term stability of metal/TiO 2 junctions [42,46,52], but the change in electrical characteristics over time has not been addressed in depth. In hydrogenated TiO 2 , the incorporated H might affect the SBDs. It has been shown that the current in junctions formed between different metals and n-type-type single crystalline TiO 2 is sensitive to hydrogen in the ambient [49], and a reaction between H and Pd may cause SBDs to degrade [46,47,49].
In the present work, we have systematically studied the electrical behaviour of junctions formed between Pd, Pt, Ni as well as Al and n-type TiO 2 single crystals doped with Nb or doped by heat-treatments in hydrogenating or reducing atmospheres. The concentration of interstitial hydrogen (H i ) in the different types of TiO 2 crystals was measured. The variation in SBD characteristics for diodes deposited on the same TiO 2 crystal was investigated. First, SBD characteristics using different metals are compared. SBDs were obtained for Pd/TiO 2 , Pt/TiO 2 and Ni/TiO 2 with up to eight orders of magnitude in rectification. Following this, a comparison of Pd/TiO 2 SBDs using TiO 2 with different dopants and different H i concentration is presented. The diode characteristics showed a pronounced change over time, related to a decrease in overall rectification. Detailed measurements concerning the degradation are presented. The fabricated Pd/TiO 2 junctions are found to exhibit suitable electrical characteristics for studying fundamental junction properties and the characteristics of electrically-active defects in TiO 2 using space-charge spectroscopy.

Experimental
Rutile TiO 2 single crystals with a thickness of 0.5 mm exhibiting different surface orientations grown by the float-zone (FZ) method or the Verneuil (V) method were purchased from MTI Corporation [53] and Shinkosha [54], respectively. In the following, a TiO 2 crystal with surface orientation (X) is denoted as TiO 2 (X). Nominally undoped FZ-and V-grown single crystals were obtained, while V-grown single crystals doped with 0.05wt% of Nb were also acquired. Nominally undoped TiO 2 single crystals were transparent and semi-insulating with a conductivity of s < W ---10 cm 7 1 1 . The as-received wafers were cut into pieces measuring approximately 5×5 mm 2 using a laser cutter. After cutting, all crystals were cleaned in an ultrasonic bath with acetone, isopropanol and de-ionized water for 5 min each (standard cleaning procedure).
Afterwards, the undoped samples were exposed to one of the following heat treatments (annealings): 1. Annealing in forming gas (FG) flow (N 2 + H 2 with H N 2 2 [ ] [ ] ≈1/9) at temperatures between  500 C and  600 C. The annealing duration was 35 90 min -. After annealing, the samples were cooled-down rapidly in the gas flux.
2. Annealing in closed ampoules filled with approximately 0.5 bar of H 2 gas. The ampoules were evacuated before filling with H 2 . These crystals were annealed at temperatures between  400 C and  600 C for 10 90 min -. After annealing, the crystals stayed in the ampoules during cool-down.
3. Annealing in N 2 flow at temperatures between  980 C and  1200 C. The typical annealing duration was in the order of hours (1.25-25 h). After annealing, the samples were taken out of the furnace and the cooldown occurred in N 2 flow.
A tube furnace was used for annealing. All TiO 2 crystals were put into the furnace after the desired annealing temperature was reached. In the following, TiO 2 crystals which have been annealed in FG flow, H 2 gas or N 2 gas are denoted as FG-TiO 2 , H 2 -TiO 2 and N 2 -TiO 2 , respectively. Nb-doped crystals are denoted as Nb-TiO 2 .
Ni, Pd, Pt and Al were deposited onto TiO 2 single crystals subjected to the heat-treatments described above, and Pd was deposited onto Nb-TiO 2 crystals (thickness ∼150 nm). Circular metal contacts displaying areas of -´-0.67 6. 36 10 cm 3 2 ( ) were obtained using a shadow mask. Either eutectic InGa was applied to or a stack of Ti/Al was deposited onto the back side of these crystals as Ohmic contacts. All metal depositions were performed using e-beam evaporation. In between depositing or applying metals, the crystals were cleaned using the standard cleaning procedure described above. Typically, between 15 to 20 SBDs are obtained on a piece of TiO 2 crystal (see figure 1 for an overview of the resulting sample layout). Some crystals were subjected to surface treatments prior to the deposition of the circular Ni, Pd, Pt or Al contacts. These crystals were exposed to boiling H 2 O 2 for 1 3 min -, and some of them were additionally exposed to HF for 5 min. Crystals that were exposed to HF in addition to H 2 O 2 will be denoted as TiO 2 (X)-HF in the following.
Fourier-Transform infrared spectroscopy (FT-IR) was used to determine the concentration of H i denoted as [H i ] in TiO 2 single crystals. The investigated TiO 2 single crystals had a surface orientation of (001). Infrared (IR) transmittance spectra were measured using an evacuated Bruker IFS 125HR spectrometer equipped with a globar light source, a KBr beamsplitter, and a liquid-nitrogen-cooled InSb detector. The IR beam was kept at normal incidence (±3°) with respect to the (001) surface of the TiO 2 single crystals. Measurements were performed at room temperature using a spectral resolution of -1 cm 1 . [H i ] was determined from the integrated area of the optical absorption associated with a local vibrational mode of H i [55] using a calibration factor determined by Johnson et al [56].
The electrical conductivity, σ, and resistivity, ρ, of TiO 2 single crystals was determined using a four-point probe measurement according to the van-der-Pauw method [57,58]. The measurement utilized a Keithley 7001 switching system, a Keithley 2182A nano-volt-meter and a Keithley 6221 current source. Eutectic InGa pads were used as Ohmic contacts in the corners of the TiO 2 single crystals.
A sample for Scanning Transmission Electron Microscopy (STEM) was prepared by focused ion beam (FIB) using a Ga-ion JEOL JIB-4500. Before milling, a 700 nm W film was deposited in order to protect the underlying sample against Ga damage.
Current-Voltage (IV) measurements were carried out on the metal/TiO 2 junctions under dark conditions at variable temperatures in the range between 25 K and 330 K using a Keithley 6487 unit. Measurements below room temperature were performed in vacuum and within a closed-cycle He cryostat. Capacitance-Voltage (CV) measurements were carried out under dark conditions using an Agilent 4284A LCR meter at six different probing frequencies f meas between 1 kHz and 1 MHz. According to the Mott-Schottky theory of metalsemiconductor junctions, the barrier height, F B , depends on the metal work function, F m , and the semiconductor electron affinity, c ; for TiO 2 was assumed [60]. From IV measurements, the series resistance, R s , shunt resistance, R sh , ideality factor, η, and average rectification was deduced. η incorporates all the effects that make the device non-ideal [61]. Here, R s represents the transport of charge carriers through the bulk (on-state resistance) and through the Ohmic contact. r IV was calculated from R s , extracted from IV measurements on SBDs, and sample dimensions; r = R A l IV s . m = l 500 m is the crystal thickness, and A is the area of the different SBDs. R sh determines the amount of reverse leakage current. The average rectification of the diodes was calculated by computing the ratio between the average current under reverse and forwards bias. For averaging, the voltage ranges between 2 V and 4 V for V was chosen. Diodes that showed a rectification lower than one order of magnitude after contact deposition were disregarded. From CV measurements, the donor concentration, N d , and the built-in voltage, V 0 , were deduced, using the depletion approximation [59]. In order to calculate N d , knowledge of the static relative dielectric constant for TiO 2 ,  TiO 2 , is required. Here, a value of =  160  ) and a dark bluish color. It is generally observed that conductive n-type TiO 2 single crystals display a bluish color [18,25,26,[65][66][67][68]. The relation between the electrical conductivity and optical absorption of the heat-treated TiO 2 single crystals is investigated further in [69]. Figure 2 displays infrared absorption spectra recorded on as-received TiO 2 single crystals (nominally undoped as well as Nb-doped) and heat-treated TiO 2 single crystals. Data are shown for the wavenumber region where optical absorption associated with a local vibrational mode of H i can be seen [55].  [55,70,71], and is likely to be responsible for the conductivity observed in samples heat-treated in FG flow or H 2 gas [69]. Notably, no change in the [H i ] of the samples was observed over the course of a few months when storing the crystals at around - 20 C [69]. σ, however, decreased slightly with time. σ was measured on FG-TiO 2 crystals after annealing, and  remeasured 40 days later. These measurements showed a reduction of~5% in σ after 40 days of storage, which was larger than the relative standard deviation of the measurements (~1%).

Metal/TiO 2 junctions
3.2.1. Influence of the choice of metal Figure 3 shows IV curves recorded on junctions between FZ-grown FG-TiO 2 (001) and various metals. Rectifying junctions are formed for Pd/TiO 2 , Pt/TiO 2 and Ni/TiO 2 , while Ohmic contacts are obtained for Al/TiO 2 , as anticipated from the expected barrier heights stated in table 1. The highest rectification observed was eight orders of magnitude and was observed for Pd/TiO 2 junctions. Usually, independent of the exact annealing condition, rectifications between 2−5 order of magnitude are achieved for Pd, Pt as well as Ni contacts on TiO 2 . Series resistances in the order of W 100 to W 1 M were obtained. Typically, ideality factors between 1.1 and 4.0 are found. η larger than 2 can, for example, be related to the presence of a highly-compensated interfacial layer or to recombination in the depletion region [19,59,72].

Influence of surface orientation and treatment
In figure 3, one can observe a significant spread in electrical characteristics of the SBDs deposited on the same TiO 2 single crystal (see figure 1). There is a wide spread in the values for the current under reverse as well as forward bias. The differences in electrical characteristics seen for SBDs deposited on the same TiO 2 crystal originate from a spread in values of R R , s sh and η, suggesting lateral inhomogeneities over distances of around 1 mm (see figure 1). Particularly, the differences seen for η suggest significant surface inhomogeneities, e.g., due to lateral inhomogeneities in the density of interface states or the thickness of a highly-compensated interfacial layer [57,72,73]. The spread observed for R s and R sh could also be related to surface inhomogeneities. R s will here mainly comprise of contributions from the bulk resistivity and the contact resistance of the Ohmic contact, and especially the resistance of the Ohmic contact could be influenced by surface inhomogeneities [57]. R sh is related to leakage currents whose magnitude might, for example, be sensitive to the density of defects within the space-charge region [74]. In figure 4, the average rectification determined for metal/TiO 2 junctions is shown, fabricated using FZ-grown FG-TiO 2 (001), FG-TiO 2 (110) or FG-TiO 2 (100) crystals, all treated in boiling H 2 O 2 . Results are also shown for samples for which the FG-TiO 2 crystals were treated in HF prior to metal deposition (FG-TiO 2 (X)-HF). The displayed uncertainty is the standard deviation determined from the averaging of the rectifications of SBDs on the same FG-TiO 2 crystal. Regardless of surface orientation, surface treatment or choice of metal, a large spread in rectification can be seen. This suggests that either the spread in rectification is related to a universal surface characteristic of TiO 2 or it is related to a bulk property. Importantly, the spread found for η and R sh is most likely related to surface properties, while the spread in R s can be related to bulk and surface properties. It has previously been found that hydroxyl layers can form on the TiO 2 surface and influence the electrical characteristics of metal/TiO 2 junctions [19]. One would, however, expect HF treatments to affect the electrical characteristics in this case. It was also found that the treatment in H 2 O 2 did not have a significant effect on the quality of the SBDs (not shown). The experiments summarized in figure 4 seem to demonstrate that deposition of Pd yields the best SBDs. Therefore, Pd diodes were selected for further characterization.

Comparison between resistivity and SBD series resistance
Values for ρ of FZ-grown H 2 -TiO 2 crystals annealed at different temperatures between  400 C and  600 C for durations of 10 90 min -were measured prior to metal contact deposition. In figure 5, the resistivity estimated from IV measurements, r IV , for several SBDs on each H 2 -TiO 2 crystal, is plotted as a function of resistivity, ρ. As shown in the figure, the least resistive crystals have values for r IV very close to ρ. For more resistive crystals, r IV are larger than ρ. This result indicates that for conductive crystals, the bulk resistivity is limiting the series resistance, but for more resistive crystals, contact resistance becomes important. Moreover, this result suggests that the contact resistance is dependent on the doping concentration of the crystal. The relative deviation in r IV between different diodes is similar for all crystals, as shown in the inset of figure 5.

Influence of doping
Doping levels suitable for electrical measurements were achieved by annealing in forming gas at  600 C for 1 h, in H 2 at  450 C for 30 60 min -, or in N 2 at  1100 C for~90 min. Figure 6 shows IV curves recorded on Pd/TiO 2 junctions comprising FZ-grown H 2 -TiO 2 (001), V-grown FG-TiO 2 (001), V-grown N 2 -TiO 2 (001) and V-grown Nb-TiO 2 (001). Rectifying junctions are obtained regardless of the doping procedure used. The resulting junctions exhibit ideality factors from 1.1 to 4, and values for the series resistance in the order of W 10 to W 100 M . A large spread in electrical characteristics is also seen for Pd/TiO 2 junctions comprising TiO 2 crystals doped using different strategies, both between the differently treated crystals, and between SBDs on the same crystal (see figure 1). In figure 6, one can observe a significant spread in the values of R R , s sh and η of the SBDs deposited on the same TiO 2 single crystal, resulting in a wide range of values for the current under reverse as well as forward bias, similar to what is seen in figure 3. Notably, Nb-doped TiO 2 displays the same spread in electrical characteristics. Thus, one can conclude that the inhomogeneity causing the large spread in electrical characteristics is not caused by heat-treating the samples in reducing or hydrogenating atmospheres. Moreover, H is not likely to play a role in the observed inhomogeneity taking into account that the Nb-doped TiO 2 crystals were shown to contain significantly less H i than the hydrogenated or reduced TiO 2 crystals (see figure 2).

Schottky barrier height and Built-in voltage
The determination of barrier heights from IV characteristics is only reliable if the current is determined by thermionic emission [75]. For the SBDs studied here, this is not the case, since η is larger than 1.1 at room temperature [75]. The SBDs that exhibited the lowest values of η were measured also at lower temperatures T. IV-T curves from a Pd/H 2 -TiO 2 junction are plotted in figure 7(a). h T ( )displays an abnormal decrease with Figure 5. r IV , extracted from IV measurements, as a function of ρ for a series of FZ-grown H 2 -TiO 2 crystals annealed at temperatures from  400 C to  600 C for 10 90 min -. The resistivity of each crystal was measured before Pd contacts were deposited. Each data point corresponds to a measurement on a single SBD.  1). IV curves of the diode with the highest rectification, the lowest rectification and a medium rectification are plotted in green, red and blue, respectively.
increasing T, as shown in the inset. Such a temperature-dependence leads to a non-linear shape of the Richardson plot I T ln 0 2 ( )versus T 1 , where I 0 is the reverse saturation current [75], and the value for the effective Richardson constant A* and F B could not be extracted. A similar temperature dependence of η was found regardless of the crystal growth method, choice of metal or the doping procedure used. These measurements are in line with previously reported temperature-dependent IV data measured on Schottky junctions involving TiO 2 [44,48,76]. Recently, F B has been deduced from IV-T data measured on junctions involving TiO 2 that display similar non-ideal behaviour [48,76]. In [76] and [48], F B was extracted assuming barrier height inhomogeneities (BHI) and using a model proposed by Werner and Güttler [77]. For the junctions studied here, however, which were measured at lower temperatures, this model can only explain the non-ideal behaviour of IV measurements recorded at T larger than~100 K 150 K -. For < T 100 K, the requirement of a linear relationship between η and F B , which confirms the existence of BHI [78], is not fulfilled. In addition, the modified Richardson plot provides reasonable values for A* and F B , but only in the temperature range above 100 K. Therefore, it is possible that BHI affect the current transport, but at low temperatures, other mechanisms are also important.
The built-in voltage for Pd/TiO 2 junctions was determined from CV measurements. Figure 7(b) shows results of CV measurements recorded on Pd/TiO 2 junctions comprising FZ-grown FG-TiO 2 (001), V-grown N 2 -TiO 2 (001) and V-grown Nb-TiO 2 (001) crystals. The representation of CV measurements as C 1 2 versus V are straight lines, suggesting a homogeneous donor concentration in the depletion region. Notably, CV measurements at different f meas display a slight dispersion. V 0 was determined to be 1.2 V 1.3 V -for Pd/FG-TiO 2 (001), 1.4 V 1.6 V -for Pd/N 2 -TiO 2 (001) and 1.8 V 1.9 V -for Pd/Nb-TiO 2 (001) junctions, depending on f meas . Notably, a N d of´-1.4 10 cm 19 3 was determined for the Nb-TiO 2 (001) crystal, being close to the Nb concentration of´-1.38 10 cm 19 3 expected in these crystals according to the wt% of Nb specified by the supplier [54]. Thus, the determination of N d from CV measurements is corroborated. The V 0 deduced is larger than expected, considering the theoretically calculated SBD heights given in table 1.
The results shown in figure 7(b) show that the fabricated Pd/TiO 2 junctions are suitable for space-charge spectroscopy techniques, such as deep-level transient spectroscopy [79] or thermal admittance spectroscopy [80], and hence the junctions can be used to study electrically-active defects in TiO 2 [41]. It is, however, important to verify the suitability of a specific junction for space-charge spectroscopy, i.e., junctions suitable for space-charge spectroscopy should display low series resistance as well as ideality factors close to 1 [57].

Time-evolution of electrical characteristics
Pd/TiO 2 junctions were stored at - 20 C for several months, and IV curves were recorded every few days. Most Pd/TiO 2 junctions exhibit a decrease in rectification over the course of a few weeks regardless of the doping used for the TiO 2 crystal. It has been reported previously that the electrical characteristics of rectifying metal/TiO 2 junctions change over time, and a reaction between H and Pd has been suggested to be the cause of the degradation [46,47,49]. Figure 8(a) displays IV curves from a Pd/TiO 2 junction recorded at different times after initial contact deposition. The junction comprises a FZ-grown FG-TiO 2 (001) crystal. The series resistance and ideality factor of the Pd/TiO 2 junction do not change significantly over the course of several weeks, while the current under reverse bias notably increases, and thus the overall rectification decreases. Such a degradation was typically observed for the SBDs comprising FG-TiO 2 (001) crystals. Pd/TiO 2 junctions comprising TiO 2 crystals doped using heat-treatments in H 2 or N 2 also display a similar degradation behaviour. Thus, the observed degradation is related to a decrease in shunt resistance of the Pd/TiO 2 junctions. Figure 8(b) shows the current under reverse bias as a function of time. Initially, there is very little change in the magnitude of the reverse current. After a few weeks, a pronounced increase in the current under reverse bias can be seen. We fitted this increase with a first-order kinetics expression: for -4 V; and t = 21 days for both -2 V and -4 V.
In FG-TiO 2 , H i is expected to be the main donor, and thus the observation regarding the series resistance is in accordance with the observation that [H i ] does not change in similarly treated samples. Interestingly, this is contrary to expectations taking the large diffusion constant for H i in TiO 2 ( m D 0.2 m hour 2 at room temperature) into account [81].
In order to study the observed degradation further, IV curves of the Pd/TiO 2 junctions comprising a FZgrown H 2 -TiO 2 (001) crystal were recorded every few days until the Pd contacts were removed after 42 days. During contact removal, some of the surface was polished off, leaving a rougher surface compared to as-received crystals. Fresh Pd contacts were deposited on the same crystal piece. The average rectification after first contact deposition was initially 5.2. Just before the contacts were removed, the average rectification had dropped to 1.8. Finally, after re-deposition of contacts, the average value increased to 3.0. This experiment demonstrates that a new annealing procedure is not required to improve the quality of the SBDs, and that the degradation is related to the interface. Interestingly, only R sh changes over time, suggesting that the change at the Pd/TiO 2 interface related to the observed degradation does not affect the barrier itself. This observation may indicate that the degradation is caused by enhanced tunnelling of charge carriers through the space-charge region. The origin of such an enhanced tunnelling is, however, unclear. Studying the defects present at the Pd/TiO 2 interface using space-charge spectroscopy could help shedding light on the role of defects in TiO 2 regarding the observed leakage current [41].

Adhesion of Pd on TiO 2
In an attempt to study the interface between Pd and TiO 2 , STEM was performed on a FIB sample fabricated from the edge of a Pd/TiO 2 junction, covering both the junction and plain TiO 2 . Figure 9 displays a STEM micrograph recorded on a Pd/TiO 2 junction after preparing a lamella using FIB. A gap between the Pd contact and TiO 2 crystal is clearly visible. A 100 nm thick Pd layer was confirmed present by energy-dispersive x-ray spectroscopy and high-angle annular dark-field imaging. However, it was found that the adhesion of Pd on TiO 2 crystals is poor. During FIB milling, local stress relaxation or heating induced by the beam at the Pd-TiO 2 interface may have caused the Pd-film to detach. Poor Pd-adhesion was also found during fabrication of the samples: Some Pd contacts detached from the TiO 2 during cleaning in acetone and isopropanol in an ultrasonic bath.

Conclusion
We have studied the formation of metal/TiO 2 junctions comprising Pd, Pt, Ni and Al. For this purpose, different strategies for obtaining conductive n-type TiO 2 single crystals were employed. Conductive n-type TiO 2 with conductivities ranging from -´W - ) was also utilized to form metal/TiO 2 junctions. Rectifying junctions are obtained for Pd/TiO 2 , Pt/TiO 2 and Ni/TiO 2 regardless of the doping procedure or the surface orientation of the TiO 2 crystals, while Al forms an Ohmic contact with TiO 2 . From CV measurements, the built-in voltage was deduced to be around 1.2 V 1.9 V -, depending on the doping concentration of the specific TiO 2 single crystal. A large spread in electrical characteristics is found for SBDs fabricated on the same TiO 2 crystal regardless of the doping strategy or the surface orientation. Ideality factors extracted from IV measurements were in the range of -1.1 4.0 at room temperature, whereas values for the series resistance range from~W W 10 100 M -. Most Pd/TiO 2 junctions exhibit a degradation over time. The degradation is related to an increase in the current under reverse bias voltage, whereas the series resistance and ideality factor stay approximately constant. The degradation can be described by first-order kinetics with a time constant of 21 days at room temperature. Adhesion between Pd and TiO 2 appears to be poor, which could be one of the reasons for the large spread in electrical characteristics of SBDs. Importantly, we demonstrated the fabrication of Pd/TiO 2 junctions which can be used to study fundamental properties of metal/TiO 2 junctions and electrically-active defects in TiO 2 using space-charge spectroscopy.