Oxygen Plasma Treatment to Enable Indium Oxide MESFET Devices

Metal‐semiconductor field‐effect transistor (MESFET) devices based on pulsed laser deposition (PLD) grown In2O3 thin films with on–off ratios exceeding 6 orders of magnitude and low sub‐threshold swing values close to the thermodynamic limit are reported. Oxygen plasma treatment and compensation doping with Mg are utilized to suppress the accumulation of electrons at the surface of In2O3, which is a major obstacle for its use as an active material in electronic devices. The influence of both methods is investigated on the electrical properties of thin films as determined by Hall effect measurements on samples of varying film thickness. Using the performance of vertical Schottky barrier diodes as a benchmark, fundamental plasma parameters such as input power and background gas pressure are optimized.


Introduction
Wide-bandgap semiconductors are the basis for fully transparent electronics, including existing and emerging technologies, such as UV-active photovoltaic devices, displays, and virtual and augmented reality applications. [1]A promising material is indium oxide In 2 O 3 .It is an n-type semiconducting metal oxide with a wide optical bandgap of 3.7 eV and thus, is transparent in the visible spectrum. [2]The most stable modification of In 2 O 3 is its cubic bixbyte structure having a high electron mobility values of up to 160 cm 2 V −1 s −1 . [3]These properties make it suitable for applications within transparent active electronics.However, the most common application today is indium tin oxide as a material for transparent electrodes. [4]In the past, In 2 O 3 has been proven to form an electron accumulation layer at its surface (SEAL). [5,6]his makes it difficult to fabricate electrically rectifying contacts with In 2 O 3 , required for all active devices such as transistors or photodetectors.It is known from other metal oxides, such as zinc oxide ZnO, that the key for fabricating high-quality, highly

Experimental Section
All In 2 O 3 films in this work were grown by PLD on r-plane corundum (Al 2 O 3 ) substrates with 10 × 10 mm 2 dimension.The PLD system uses a LPXpro 305 KrF excimer laser by Coherent operating at a wavelength of 248 nm and an energy of 600 mJ per pulse as described in detail in Ref. [15].In 2 O 3 films were deposited at a temperature of ≈ 500°C in an oxygen ambient at a pressure of 0.02 mbar.Homogeneous In 2 O 3 targets with different amount of MgO (0 wt %, 0.5 wt %, and 1 wt %, corresponds to 0 at %, 3.25 at %, and 6.5 at % MgO) were used to deposit films ranging from few 10 nm thickness up to 1 m.The deposition processes were performed with a 1-h preheating period and resulted in growth rates of about 0.25 Å per pulse.
For plasma treatment of the samples, a PlasmaPro 80 RIE unit (Oxford Instruments) was used.The system allows controlling multiple process parameters, in particular substrate temperature, pressure, and composition of the process atmosphere, and electrical power used to drive the two plasma sources, radio frequency (RF), and inductive coupled plasma (ICP).The default process parameters are t = 1 min, pressure p = 0.027 mbar, gas flow rate V = 50 sccm, plasma powers P HF = 100 , and P ICP = 300 W.
The different sample structures investigated in this study are visualized in Figure 1: Thick In 2 O 3 films were used for initial plasma treatment experiments (a).For fabrication of vertical The patterning of active In 2 O 3 devices is not possible with conventional wet chemical etching due to the low solubility of In 2 O 3 in chemicals.In-plane structuring via UV-photolithography was therefore used only to define metal contacts (top-contact design).In Figure 2, a flow diagram of the FET fabrication process is shown.After the initial film deposition (step 1), Ohmic contacts (Au, step 2) are deposited by DC magnetron sputtering with an inert background gas (Ar) at room temperature.For the realization of rectifying contacts, a reactive sputter deposition is used, that is, under a mixed Ar/O 2 atmosphere (50 sccm gas flow each, step 4).This process introduces additional oxygen at the metalsemiconductor interface, which saturates oxygen vacancies at the In 2 O 3 surface turning an initial downward band bending into an upward bending.It is a prerequisite for rectifying contacts as found in this and previous studies. [11,12]n order to specifically investigate the role of plasma treatment and to ensure that neither the use of solvents nor the photo resist itself will subsequently alter the properties of already plasmatreated indium oxide surfaces, two points in the manufacturing process have been identified where plasma treatment is necessary: On the one hand, it can take place directly before the deposition of the contact material (step 3).At this stage the photo resist was already applied and developed, and the solvent for the lift-off step cannot interact with the treated interface since it is already covered by the deposited metal.Therefore, only the area under the rectifying contacts is plasma treated.The other time for effective plasma treatment was after fully completing the device preparation (step 5).This effectively results in subjecting the exposed areas between already applied metal contacts to plasma treatment.
Layer thickness measurements were performed using a profilometer (DektakXT by Bruker) and via spectroscopic ellipsometry measurements (M-2000 by J.A. Woollam).Composition of deposited thin films were analyzed via energy-dispersive X-ray spectroscopy (EDX) measurements using a Nova NanoLab 200 by FEI company.Hall effect measurements were performed at a magnetic field of 0.43 T using a Keithley current source, switch system, and multimeter.For the Au contacts Ohmic behavior was confirmed.Static current-voltage measurements were done using the 4155C/4156C Precision Semiconductor Parameter Analyzers (by Agilent).

Results
Electrical properties of PLD-desposited In 2 O 3 films with various film thickness were investigated in an as-grown state as well as after an oxygen plasma treatment under default parameters.We used Ar plasma as a chemically inert reference process with the same parameters.EDX measurements quantified magnesium cation contents to (1.2 ± 0.1) at % and (2.3 ± 0.2) at % in our In 2 O 3 films prepared with targets containing 0.5 wt % and 1 wt % MgO, respectively.For nominally undoped films, the Mg concentration was within error margins of the instrument, (0.08 ± 0.10) at%.
Results from our resistivity and Hall effect measurements confirm the role of Mg as a compensating acceptor in In 2 O 3 , rendering those films semi-insulating. [13,16,17]At 1.2 at % Mg doping concentration, the resistivity  increases by two to three orders of magnitude, at 2.3 at % by an approximate factor of 10 5 (cf. Figure 3).Nearly all of those films showed ambiguous Hall coefficients in the as-grown state, so that their charge carrier density n and mobility  could not be determined.We see that the trend of increasing  with rising Mg content is due to a decrease in both n and  for the (thinnest) sample where comparison between doped and undoped films is possible.This effect explains the purpose of Mg doping In 2 O 3 in previous publications: [11,16,18] The shift of the surface Fermi level deeper into the bandgap restrains the formation of a surface electron accumulation layer (SEAL) and, thus, enables the formation of a depletion region, a requirement for rectifying contacts to In 2 O 3 .
The electrical transport properties of the samples significantly change after they have been subjected to plasma treatments, especially regarding the involved ion species.For the undoped In 2 O 3 films, the conductivity of Ar-plasma-treated samples increases by about onefinally order of magnitude.This is mainly due to an increasing carrier concentration n, whereas the car-rier mobility  varies only slightly.O 2 plasma treatment of undoped samples, however, does not significantly change their initial state (regarding all three quantities, , n, and ).For In 2 O 3 films with 1.2 at % Mg content, a strong decrease of resistivity after plasma treatment can be observed.This effect is much larger for Ar plasma than for O 2 plasma, that is,  decreases by four and two orders of magnitude, respectively.As stated above, for those Mg-doped films, only plasma treated samples can be evaluated regarding carrier concentration and mobility.Therefore, it cannot be finally confirmed whether the decrease in resistivity was mainly caused by an increase in charge carrier concentration n or mobility .However, a change over several orders of magnitude is much more likely to be expected for n than for . [19]he investigation on films of different thickness d was aimed at directly proving the existence of a SEAL in our samples.When charge transport is dominated by conduction through the bulk material, the charge carrier density n is independent of the layer thickness (n ∝ d 0 ).However, the formation of a conductive layer dominating the current flow should have a constant surface charge carrier density n □ = n/d.In other words, the charge carrier density n calculated from Hall effect measurements would then follow a 1/d dependence.
Both cases could be found in our measurement results, evaluated by fitting the data to a power law A • d B .The Mg-doped In 2 O 3 films show a pronounced n ∝ d −1 behavior after plasma treatment, regardless of the ion species involved.This leads to the assumption that a SEAL is still present after the treatment in those samples.Due to similar power law fits for resistivity, the existence of a degenerate layer in the as-grown state is probable, presumably at the In 2 O 3 -substrate interface.For the undoped films, however, the observed behavior approaches a layer thicknessindependent charge carrier density n.The power law exponent increases significantly to more than −1, that is, −0.70 and −0.38 for the samples treated with Ar plasma and O 2 plasma, respectively.The untreated films (n ∝ d −0.17 ) displayed the behavior closest to bulk-dominated charge transport (n ∝ d 0 ).Oxygen plasma treatment causes only a small change in the film thickness-dependent trend as well as in the absolute values of n, while Ar plasma both significantly enhances the trend and increases the calculated carrier densities by about an order of magnitude.The observation suggests that the Ar plasma process is significantly more effective in generation of surface-near free charge carriers than O 2 plasma, even though the same parameters were used.Since, as mentioned, oxygen deficiency plays a major role in charge transport in In 2 O 3 , this effect is probably superimposed by the saturation of vacancies in the oxygen process.Although reduced by the generally very high bulk conductivity of undoped In 2 O 3 , all three states show a significant trend of decreasing carrier density with increasing sample thickness.This indicates that the existence of a SEAL in these samples is very probable.
Another prominent point is that after Ar plasma treatment, undoped and Mg-doped samples have almost identical electrical properties.For instance, all values for the charge carrier density are between 10 20 and 10 21 cm −3 and show the same layer thickness dependent trend.Studies have been performed showing that intentionally damaged In 2 O 3 films display similar electrical properties with increasing damage regardless of their initial state (undoped, donor-doped, or acceptor-doped). [19]This is justified by the fact that the Fermi level in these samples is eventually raised into the conduction band by introduced intrinsic defects and defect complexes.Especially in the case of compensation doped samples (with Mg, as in this work as well) this leads to a drastic increase of the charge carrier density and, thus, decrease of the resistivity.
To provide a rationale for the Hall effect measurement results in general, we propose the following explanation.Ar plasma treatment seems to structurally damage our In 2 O 3 sample surfaces and introduces a shift of the Fermi level above already existing surface donor states such that they release their electrons into the conduction band.In contrast to Ref. [19], this damage only occurs locally at the thin film surface and, therefore, induces a SEAL.In nominally undoped samples, the conductivity of the SEAL increases so that it exceeds the otherwise dominant bulk conductivity.In Mg-doped thin films, whose as-deposited conductivity is too small for Hall effect measurements, measurable conductivity is achieved by inducing this conductive layer via plasma treatment.A similar statement can be made about the oxygen plasma processes, although the effect here is at least an order of magnitude weaker compared to the Ar plasma treatment.In Mg-doped samples, the O 2 process provides significant overall conductivity in the first place.However, due to a constant surface charge carrier density, it can be assumed to be an induced SEAL as well.
In general, comparison of films grown under identical conditions shows significant differences in electrical properties after plasma treatment.With the exception of the oxygen plasma treated undoped In 2 O 3 , the conductivity of all other samples increases by several orders of magnitude after the plasma treatment.This behavior, upon initial observation, is in disagreement with other studies that report on the depletion of the SEAL in In 2 O 3 after oxygen plasma treatment. [12,14]An explanation could be that saturating the oxygen vacancies in the plasma process is counteracted by inducing significant structural damage due to a too high plasma power.This hypothesis is supported by the fact that plasma power in Ref. [12] was a factor of 3 lower than our processes (P ICP = 100 W, P HF = 50 W, similar pressure of 0.025 mbar).To verify, we varied the plasma process parameters in the next sample series.
We investigated the effect of oxygen plasma treatment on the performance of Pt-In 2 O 3 Schottky barrier diodes (SBDs) by exposing the films directly to oxygen plasma prior to reactive sputtering of the metal contacts as in Ref. [12] (cf.pre-gate plasma treatment, step 3 in Figure 2).For this purpose, we grew undoped and 1.2 at % Mg doped, In 2 O 3 films, each 1 μm thick, on 200 nm thick, Sn-doped (1 wt % SnO 2 ) In 2 O 3 back contact layers to create vertical Schottky barrier diodes (cf. Figure 1a).We investigated the influence of the main plasma treatment parameters, namely background pressure p and a plasma power P, by varying them both to 3 different levels, that is, 0.027, 0.053, and 0.133 mbar and 20, 50, and 100 W. In order to keep the same plasma conditions, the power ratio between the IC and the RF plasma source was kept constant at 3:1.
In the top row of Figure 4 I-V, characteristics of plasma treated samples are shown for different background pressures and a plasma power of P HF = 100W.It can be seen from the untreated samples that the incorporation of Mg into In 2 O 3 films has a positive effect on the rectification of the fabricated diodes.This is in agreement with existing literature. [11,18]It is further evident that any oxygen plasma treatment prior to the deposition of the Schot-  [20] tky contact improves the diode characteristics.This ports original hypothesis that the plasma treatment saturate vacancies near the surface. [14]rom the characteristics, the diode parameters were Rectification ratio S U = ±1V = |j(1V)/j(− 1V)|,  and barrier  B, eff .The average values and standard deviation are shown in Figure 4 mid row.By fitting with the thermionic emission model as used in Ref. [11], the effective barrier height  B, eff of the SBDs and their ideality factor  were determined.Scatter plots of single devices values are presented in Figure 4 bottom row.Due to variation in device quality within one sample, the spread in the  B - plot can be used to determine the homogeneous barrier height via extrapolation to  if ≈ 1.01. [20,21]This turns out to be  B, hom = 1.05 ± 0.10eV for both, undoped and doped In 2 O 3 SBDs, which agrees with the reports on MBE-grown In 2 O 3 Pt-SBDs in Ref. [11].
It is evident that the best diodes can be fabricated with low pressure and high power oxygen plasma treatments.Both tendencies are related to high intensity processes.Higher gas pressure leads to more collisions of the accelerated particles on their way to the sample, reducing both their kinetic energy and particle flux.Lower plasma source power leads to both lower plasma density and lower particle energies.These parameters were used to fabricate FET devices as a final step.
After the optimization of oxygen plasma treatment parameters (most intense process: lowest pressure, highest power), we applied those to fabricate MESFET devices.Thin (25-27 nm), PLD grown In 2 O 3 films were used as channel material.We screened for an optimal Mg doping concentration by using the same three targets (with 0 wt%, 0.5 wt%, and 1 wt% MgO) as before.Devices were fabricated as illustrated in Figure 2 and as described in Section 2. Thus, two plasma steps were used during device fabrication: once before sputtering of the rectifying gate contacts and once after completion of the device.For the entire set of devices per sample, first the source-gate and then the drain-gate diode characteristics were measured.The devices were measured with a voltage sweep that always began with a negative voltage bias.The characteristics of each ensemble were then logarithmically averaged and their standard deviation calculated.The results are presented in Figure 5.This way, any deterioration processes that occur, for example, a breakdown of the diode, can clearly be identified.
As opposed to the experience from the diode experiments, oxygen plasma treatment of the area under the gate contact before its deposition does not lead to any positive change in the characteristics.On the contrary, these samples consistently show smaller rectification than devices without this treatment.The exponentially increasing current under reverse bias indicates charge transport via tunneling through the Schottky barrier. [22]or a nearly ideal MESFET gate contact, saturation of the reverse current is expected for gate voltages below a certain threshold value.At this point the depletion region extends across the entire channel cross-section.In our samples, this is especially pronounced for doped samples without pre-gate plasma process and indicates very small space charge layer width for all other cases.
A post-gate plasma treatment, on the other hand, is revealed to be crucial for device functionality.As shown in Figure 2, this plasma process affects the exposed parts of the channel between the metalized contacts.All devices not subjected to this plasma treatment show low performance caused by a drastic current increase under reverse bias.An explanation for this behavior is that the SEAL, present in the non-metallized and untreated regions of the channel, leads to the formation of parallel surface conduction channels.Thus, part of the current can bypass the depletion zone, resulting in regions with a low parallel resistance.This explanation is supported by the fact that this change is permanent, that is, once it occurs it does not regenerate over time.
The transfer characteristics are shown in Figure 6.We are proud to report high performance MESFET devices for the first time with up to 6.5 orders of magnitude current swing.Since the off-current of MESFETs is directly dependent on the leakage current of the gate diode, a lower bound of the leakage current in off-state is given by reverse saturation current of the gate diode.Thus, only samples where this current is small and stable show a distinct on-off switching.The best performance is observed for devices with at least 1.2 at % Mg and which were plasma treated exclusively after the gate deposition.Low sub-threshold swing values of 120-150 mV dec -1 could be achieved.Compared to devices with the same Mg content, this fabrication strategy improves on-off current ratio by 3 orders of magnitude and channel conductivity by a factor of 3-100.

Conclusion
We presented strategies to realize and optimize the fabrication of vertical Schottky barrier diodes (SBDs) and lateral metalsemiconductor field-effect transistors (MESFETs) based on In 2 O 3 The influence of plasma treatment on In 2 O 3 thin films has been investigated by Hall effect measurements in a first step.Our samples exhibited typical electrical properties in an as-grown state: Mg acts as compensating dopant and reduces the charge carrier density significantly. [17]The results suggest that both plasma treatments, with Ar much more than O 2 , increase the formation of an already existing surface electron accumulation layer (SEAL).For O 2 plasma, this contradicts previous publications on the influence of oxygen plasma on In 2 O 3 . [4,14]We could find evidence that the plasma exposure was too intense and introduced additional surface-near defects instead of saturating oxygen deficiencies.
With this in mind, we optimized plasma process parameters on the basis of rectifying SBDs.Films with a Sn-doped backcontact layer were employed to ensure vertical devices (cf. Figure 1a).We could show a positive correlation of the plasma power to the rectification of our diodes.In general oxygen plasma of the interface between semiconductor and metal Schottky contact (Figure 2, step 3) was found to improve the device quality, as also reported in Ref. [12].The effect intensifies with increasing power and decreasing gas pressure.
Finally we show only by combining both approaches, compensating acceptor doping and oxygen plasma treatment, highperforming MESFETs could be fabricated.Thin films of 25-27 nm thickness were used as channel material and metal contacts were sputtered in front-front design (cf. Figure 1c).This time we distinguished between the plasma treatment of the area under the gate contact and the exposed areas between source/gate or gate/drain (Figure 2, steps 3 and 5).We could not observe any positive effect of the former.On the contrary, a post-gate plasma treatment reduces the rectification of the gate diodes by introducing a significantly higher reverse current and thus degrades the transmission characteristics of the FETs.However, on samples on which there was no plasma treatment of the area between the contacts, we observed an irreversible deterioration of the diode characteristics.These results indicate that a parallel conduction channel is being introduced, although this mechanism is not yet fully understood.It can be presumed that the origin of these dissimilarities between vertical SBDs and lateral MESFETs may lie in the different thicknesses of the functional layers.
Further device improvements could be achieved by using channels with a vertical doping gradient, which can be achieved by the segmented PLD target method. [23]By doping only the near-surface portion of the channel material, one could compensate for the lower overall conductivity caused by Mg doping and optimize the on-current while still maintain an improved semiconductor-metal interface.

Figure 1 .
Figure 1.Visualization of different sample structures.Thicker In 2 O 3 films were utilized for Hall effect measurements a) and thinner ones for ringshaped field-effect transistors (ringFETs, c).b) In 2 O 3 films with a Sndoped back contact layer for vertical Schottky barrier diodes.d) Microscopic top view of a group of ringFETs with their respective gate-ring diameter.White arrows represent the current path through the devices.

Figure 2 .
Figure 2. Flow diagram of the fabrication process of an In 2 O 3 -based fieldeffect transistor (FET) with two different plasma treatment steps.

Figure 3 .
Figure 3.Comparison of resistivity , sheet resistance R s , charge carrier mobility  and density n of In 2 O 3 films in an as deposited and Ar/O 2 plasma treated state as determined by Hall effect measurements.Films were grown on r-Al 2 O 3 substrates with various film thickness d.Lines represent power-law fits of the respective sample series.

Figure 4 .
Figure 4. Influence plasma process parameters, oxygen pressure p and plasma power P, on diode characteristics of Pt-In 2 O 3 SBDs.a) Average I-V characteristic (line) and their standard deviation (shaded area) of 10-20 devices b) Average values of rectification ratio S V = ±1V = |j(+ 1V)/j(− 1V)|.c) Scatter plot of ideality factor  and effective barrier height  B, eff .Extrapolation to  ≈ 1.01 can used to estimate homogeneous barrier height  B, hom .[20]

Figure 5 .
Figure 5. Influence of O 2 plasma treatment on the Source-Gate (S, red, first measurement) and Drain-Gate (D, blue, second measurement) characteristic of In 2 O 3 -Pt ringFET gate diodes for varying Mg content and different plasma treatment strategies.The mean values of the measured ensambles (solid line) and their standard deviation (colored areas) are displayed.The measurement sweep direction is marked by light (increasing) and dark (descreasing sweep) colors.All measurements started with a rising voltage sweep.

Figure 6 .
Figure 6.Typical transfer characteristics for In 2 O 3 -Pt ringFETs depending on different Mg-contents and oxygen plasma treatment steps under 2 V source-drain bias.The measurement sweep direction is marked by light (increasing) and dark (descreasing sweep) colors.All measurements started with a rising voltage sweep.