Doping Density Extraction of Plasma Treated Metal Oxide Thin Film Diodes by Capacitance–Voltage Analysis

High quality thin film p‐n junction diodes with high rectification ratios and low ideality factors have been fabricated from metal oxides, such as amorphous oxide semiconductors (AOSs), and characterized. Plasma treatment of interfaces has been demonstrated to improve devices made from AOSs, using current–voltage (I–V) measurements. However, capacitance–voltage (C–V) measurements of the devices have been scarcely reported in the literature. Therefore, the focus of this work is characterization of cuprous oxide (Cu2O)/amorphous zinc‐tin oxide (a‐ZTO) thin film heterojunction diodes using C–V analysis. Performance differences of plasma‐treated and untreated diodes that are difficult to observe in I–V analysis are more prominent in C–V analysis. Moreover, C–V analysis allows extraction of charge density profiles, which is a measure of the defect state density that led to intrinsic doping. The variation of doping densities of the untreated diode across the full range of applied reverse bias is shown to be up to 2 orders of magnitude, while those of the treated diodes are within a factor of 10 only. Junction charge profiles, interfacial charge depletion, and accumulation that are key features of rectifying diodes are shown to be clearly distinct between untreated, nitrogen‐treated, and oxygen‐treated diodes, thus explaining why oxygen‐treated diodes are superior.


Doping Density Extraction of Plasma Treated Metal Oxide Thin Film Diodes by Capacitance-Voltage Analysis
Yin Jou Khong hydrogenated amorphous silicon (a-Si:H) for the active channel layer used in thin film transistors (TFTs) for display back planes. [1] This is due to the higher mobility of AOSs compared to a-Si:H, and AOSs can be deposited at room temperature with excellent uniformity, thus opening potential applications such as flexible electronics, sensors, displays, and radio frequency identification tags. In addition to TFTs, AOSs have been incorporated in fundamental devices such as p-n junction diodes, with a variety of combinations of n-type and p-type oxide films. Example of oxide-based diodes include NiO x /ZnO fabricated using pulsed laser deposition (PLD) at up to 650 °C [2] showing rectification of 2.6 × 10 10 , Cu 2 O/Ga 2 O 3 made by thermal oxidation of copper at 1010 °C and PLD at room temperature (RT) exhibiting rectification of 4 × 10 4 , [3] Cu 2 O/indiumgallium-zinc oxide (IGZO) by magnetron sputtering at RT rectifying with ratio 3.4 × 10 4 , [4] zinc-cobalt oxide (ZCO)/ZTO by PLD at RT with rectifying ratio of 5 × 10 4 , [5] and platinum/ZTO/ titanium/gold Schottky diode deposited using electron beam evaporation (for the metals) and magnetron sputtering (for the oxide) at RT showing rectification of ≈30. [6] One commonly used and very effective method to improve performance of metal oxide thin film p-n junction diodes is to plasma treat interfaces. [7][8][9][10] Surface treatment of metal oxides or the interface between metal contacts and metal oxides alters the electronic characteristics of the junctions, as reported by several authors. [10][11][12][13][14][15] Carlos et al has also reported that their indium-tin oxide (ITO)/ZTO/poly (3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) Schottky diodes fabricated in the pilot-scale roll-to-roll (R2R) printing line exhibited rectification ratio of 20 without plasma treatment, but showed rectification of up to 1000 after plasma treatment of all 3 ZTO layers. [16] Different metal contact deposition methods and the order of deposition of semiconductors and metals have a strong impact on the performance of the device due to oxidation or reduction processes at the interface. Oxygen treatment of the surface of the metal oxide or certain metal contacts before the deposition of the next layer of oxide can benefit the formation of an interface by chemically stabilizing the contact interface, removing any highly conductive surface accumulation layer or reducing any interfacial trap states. [11][12][13][14][15] As the surface of metal oxides is abruptly terminated, it is very likely to be reactive to contaminants in the atmosphere as well as having a larger High quality thin film p-n junction diodes with high rectification ratios and low ideality factors have been fabricated from metal oxides, such as amorphous oxide semiconductors (AOSs), and characterized. Plasma treatment of interfaces has been demonstrated to improve devices made from AOSs, using current-voltage (I-V) measurements. However, capacitance-voltage (C-V) measurements of the devices have been scarcely reported in the literature. Therefore, the focus of this work is characterization of cuprous oxide (Cu 2 O)/ amorphous zinc-tin oxide (a-ZTO) thin film heterojunction diodes using C-V analysis. Performance differences of plasma-treated and untreated diodes that are difficult to observe in I-V analysis are more prominent in C-V analysis. Moreover, C-V analysis allows extraction of charge density profiles, which is a measure of the defect state density that led to intrinsic doping. The variation of doping densities of the untreated diode across the full range of applied reverse bias is shown to be up to 2 orders of magnitude, while those of the treated diodes are within a factor of 10 only. Junction charge profiles, interfacial charge depletion, and accumulation that are key features of rectifying diodes are shown to be clearly distinct between untreated, nitrogen-treated, and oxygentreated diodes, thus explaining why oxygen-treated diodes are superior.

Introduction
Devices based on thin film metal oxide materials have become an increasingly important part of the electronics industry over recent years. For example, amorphous oxide semiconductors (AOSs) have been widely investigated as an alternative to www.advmatinterfaces.de number of defects such as oxygen vacancies which contributes to charge carrier generation in n-type metal oxides.
In a previous paper, the effect of the plasma treatment on p-n heterojunction thin film diodes fabricated at room temperature using p-type cuprous oxide (Cu 2 O) and n-type amorphous zinc-tin oxide (a-ZTO) has been discussed by current-voltage (I-V) analyses. [17] We have shown that without any plasma treatment, the device performance was poor with a rectification ratio ≈ 20 only, but plasma treatment on the a-ZTO surface improved the device performance significantly. Amongst various plasma treatments investigated (such as nitrogen (N 2 ), argon (Ar), and oxygen (O 2 ) plasma), the best devices were obtained with O 2 plasma treatment with rectification ratio of 10 4 at ±1 V. The performance of these devices are in the same level as the room temperature sputtered ZTO Schottky diodes recently reported by Branca et al. [6] The effectiveness of the surface treatment on metal oxides p-n junctions have been widely reported by the I-V measurements. However, there are very few reports on the C-V measurement of the p-n junction. In this paper, we have performed such C-V measurements and consequently the donor density of the ZTO at the junction of the diodes were extracted analytically. The doping profiles between untreated and plasma treated were compared and the processes that is happening at the p-n junctions were hypothesized. Given the relevance of ZTO diodes to the state of the art and even flexible electronic devices such as those demonstrated by Carlos et al., [16] we believe that the C-V analysis can highlight the differences in performance of plasma-treated and untreated diodes which are difficult to observe in I-V analysis. In addition, C-V analysis allows extraction of charge density profiles, which is a measure of the defect state density that led to intrinsic doping, which in turn plays a significant role in electronic device characteristics.

Results and Discussion
The p-n heterojunction diodes were fabricated using p-type Cu 2 O and n-type a-ZTO, in a stacked-sandwiched layout, as shown by the schematic (Figure 1a) and cross sectional electron micrograph ( Figure 1b). More information is provided in the experimental section. A quick summary of the key metrics for the room-temperature-fabricated devices reported in the literature in comparison with our work is shown in Table 1 for easy reference. The I-V analyses including bias stress measurements, and X-ray photoelectron spectroscopy (XPS) characterization of these devices have been detailed in a previous publication. [17] Figure 2 shows the C-V measurements of the diodes performed at various frequencies a) without plasma, b) with nitrogen plasma and c) oxygen plasma treatments. When measured at 100 kHz, the untreated diode exhibits a capacitance of ≈0.4 nF during the forward sweep from −1 to 0.5 V, and it decreases to ≈0 F from 0.5 V to 1 V. On the other hand, the capacitance increases from ≈0.1 to ≈0.9 nF during the reverse sweep from 1 to 0 V and decreases again to 0.4 nF from 0 to −1 V, indicating the modulation of charge by applied bias. The difference in the C-V curves between the forward sweep and reverse sweep also suggests the presence of some trap states which are being filled as the positive voltage is applied, resulting in a very large hysteresis. Moreover, in the C-V response the capacitance decreases with increasing measurement frequency, showing a sign of broad distribution in energy of trap states. [21] For the nitrogen-plasma treated diode plotted in Figure 2b, the capacitance measured at 100 kHz increases in the forward sweep from ≈3 nF at −1.5 V to ≈9 nF at 0.75 V, then decreases to ≈0.6 nF at 1.5 V The C-V profile of the reverse sweep is very similar to that of forward sweep with a very small hysteresis. This is a significant difference from the untreated diode indicating the effect of the nitrogen plasma treatment. It should also be noted that the capacitance of the nitrogen treated diode is higher than that of the untreated diode by an order of magnitude. However, the nitrogen-plasma treated diode required a larger bias sweep (± 1.5 V) to see the charge depletion and accumulation in the C-V profiles. Finally, plotted in Figure 2c is the C-V response of the oxygen-plasma treated diode which shows further improved characteristics. Again, note the extended bias sweep range in the x-axis of the plot. The oxygen-plasma treated diode exhibits steady capacitance of ≈1 nF from −1 to −0.5 V, then a sharp rise to ≈8 nF from −0.5 to 0.25 V, followed by a sharp fall to ≈0.1 nF from 0.25 to 1 V. This seemed to closely follow the theoretical model of depletion capacitance of a p-n junction's space charge region with consistent forward and reverse sweep curves in C-V and C −2 -V plots. Moreover, the decrease of capacitance with the measured frequencies is limited to the range between 0 and 0.75 V, and no hysteresis is observed.  The average built-in voltages between 100 and 300 kHz, the range of frequencies used in these measurements, are 0.1 ± 0.05 V, 0.5 ± 0.20 V, and 0.0 ± 0.03 V for the untreated, nitrogen-plasma treated, and oxygen-plasma treated diodes respectively. The oxygen-treated diode exhibits consistent builtin voltage (0 V) for different frequencies. This indicates that the oxygen-treated diode has an abrupt junction, similar to that exhibited by a homojunction diode. On the other hand, it is noted that the nitrogen-plasma treated diode exhibits higher built-in voltage and larger standard deviation, indicating that it is inferior even to the untreated diode. Overall, the nitrogentreated diode exhibits many unwanted characteristics.

C-V Analysis
To further elucidate the effect of plasma treatment, Figure 3 shows the overlay of the profiles of the a) C-V, and b) the corresponding I-V measurements. In Figure 3a, the untreated diode shows the lowest capacitance and no effective modulation by applied bias, meaning no distinct junction. That corresponds well to the almost similar current density between forward and reverse applied bias in Figure 3b. On the other hand, the diode with nitrogen treatment shows higher capacitance and a gradual modulation effect since the capacitance decreases overall as the voltage changes from 1 to −1 V. This shows that the nitrogen-treated diode still has a significant quantity of traps present at the interface but a lower concentration than that of the untreated diode. This may also explain why the forward current of the nitrogen-treated diode did not reach similar levels to that of untreated and oxygen-treated diode at ≈1 V, despite having lower reverse current and currents at small applied bias (Figure 3b). Finally, the diode with oxygen treatment shows the most well-defined and abrupt junction, as clearly illustrated by the distinct C-V peak with the voltage being swept through forward and reverse applied bias. In addition, the hysteresis of the C-V characteristics observed in the oxygen-treated diode is also very small, further demonstrating the improved performance of the oxygen-treated diode over the untreated one. Overall, the oxygen-treated diode shows significant improvement over both the untreated and nitrogen-plasma treated diodes.
Even though the C-V profiles between the nitrogen treated and oxygen treated diodes are very different, that is not the case with their corresponding I-V profiles as shown in Figure 3b. The I-V characteristics in the off-state (reverse bias) between the nitrogen and oxygen treated diodes are very similar, and those in the on-state (forward bias) are also somewhat similar except for the current density level at applied bias between 0.3 and 1 V. The capacitance (charge) profiles elucidate why the nitrogen plasma treatment is not as good as the oxygen plasma treatment. It also clearly shows how C-V measurements are very important to analyse the junction characteristics of the diodes, in addition to the standard I-V that is typically reported in the literature.

Extraction of Donor Density
Following the analytical analysis of the C-V measurement, as detailed in the supporting information, the charge density of the p-n junction is extracted using Equation 1 (also Equation S13, Supporting Information), where N A and N D are the intrinsic acceptor density and donor density respectively. ε p and ε n are the electrical permittivities of p-type Cu 2 O and n-type a-ZTO respectively, A is the area of the diode, q is the elemental charge, V is the applied voltage and C is the capacitance. Hall effect measurements on the same Cu 2 O gives a carrier concentration of 3 × 10 17 cm −3 in the untreated diode and 3 × 10 19 cm −3 in the plasma treated diodes, which are good approximations of the intrinsic acceptor density N A in the respective p-type Cu 2 O layers in the respective diodes, assumed to be uniform throughout the devices. The permittivities used

www.advmatinterfaces.de
in the calculations are ε p = 7 and ε n = 19. [22,23] Rearranging Equation 1 to make donor density the subject, Thus Equation 2 allows the a-ZTO donor density variation with applied bias to be plotted in Figure 4. The donor density profile is clearly depth dependent as the space charge width of a p-n junction is voltage dependent.
As shown in Figure 4a, the a-ZTO donor density of the untreated diodes under small reverse bias was measured to be ≈10 16 cm −3 which corresponds well to the carrier density measured using the Hall effect in the same a-ZTO film of ≈2 × 10 16 cm −3 . For a negative-to-positive voltage (forward) sweep the extracted doping density is always higher than that of a positive-to-negative voltage (reverse) sweep, despite the diode always being subjected to a reverse bias. The charge and release rate of the associated trap states are likely to be slower than the rate of change of applied voltage, so the delayed action manifested as hysteresis. It is interesting to note that the higher the measurement frequency, the higher the doping density that

www.advmatinterfaces.de
is extracted from the untreated diode, which will be discussed later. Figure 4b shows the doping density profile on the a-ZTO films in the heterojunction of plasma treated diodes. It can be observed that nitrogen-plasma treated a-ZTO showed relatively uniform doping densities ≈1 × 10 18 cm −3 for a large range of applied bias between −0.4 and −1.5 V, and it increases slightly to ≈3 × 10 18 cm −3 closer to the junction. For the oxygen-plasma treated a-ZTO, relatively uniform doping densities ranging between ≈1 × 10 17 and ≈3 × 10 16 cm −3 for applied bias between −0.4 and −1.5 V, and then increases sharply to ≈5 × 10 17 cm −3 closer to the junction. The donor density of the nitrogen-plasma treated diodes is an order of magnitude higher than the oxygen plasma treatment ones. The difference in I-V and C-V profiles between nitrogen and oxygen treatment can be quantified in terms of this donor density.
For the untreated diode, it is noted that the higher the measurement frequency, the higher the doping density that is extracted from the untreated diode. This observed outcomes may be explained by the shift in equilibrium in the conversion process between ionised oxygen vacancies and electron-filled states. [17] The majority of charge carrier contributors in the a-ZTO layer of the untreated diode are states in the metal oxide constantly shifting between oxygen vacancy states and electronfilled states depending on Fermi level position favouring the generation of either one. When the alternating current signal from C-V measurement increases, these states are no longer able to follow, and the apparent doping density increases. The donor state is favoured instead of electron-filled likely because the associated time constant to shift from electron-filled state into donor state is shorter than the opposite process. The convergence of extracted doping densities performed in the reverse sweep confirms this, noting that as applied bias increases in magnitude (in the negative direction) more donor states are emptied.
In contrast to the untreated diode, lower doping densities were extracted with increasing measurement frequencies in plasma treated diodes. This may be attributed to majority charge carrier contributors stemming from states that are more stable, and do not get converted between vacancy states and filled states due to the aforementioned ion dynamics of metal oxides. This is a much more desired characteristic of an electronic device as its performance do not depend on volatile behaviour of the materials and is more predictable. Notably, the variation of doping densities of the untreated diode across the full range of applied reverse bias is between 1 and 2 orders of magnitude, while those of the treated diodes are typically much lower, within a factor of 10 only.
From Figure 4b it was clear that only the oxygen-treated diode can be assumed one-sided as Cu 2 O has a charge density over 100 times that of oxygen-treated a-ZTO. By neglecting the change of depletion width at the Cu 2 O side as the junction was assumed one-sided, the charge density profile for oxygen-treated a-ZTO was also plotted against the depletion region depth in Figure 5. No similar plot can be produced for nitrogen-treated a-ZTO or even the untreated a-ZTO because the assumption of a one-sided junction is not met, unlike that with the oxygentreated a-ZTO. Due to the high resistivity of the treated films, no meaningful Hall effect measurements can be done on them for comparison. However, the plot of doping density against depth for a-ZTO layer in Figure 5 illustrates that a-ZTO has a more uniform doping in depth after the oxygen plasma treatment,  and that the influence of oxygen-vacancy-associated trap states are greatly reduced, improving the performance of the diode.

Conclusion
The improvements to Cu 2 O/a-ZTO thin film diodes due to plasma surface treatment on the a-ZTO layer before p-n heterojunction formation has been characterized in terms of C-V measurements and doping density profile extraction. The a-ZTO donor density of the untreated diodes was measured to be ≈10 16 cm −3 which is very close to that measured using the Hall effect, the nitrogen-plasma treated a-ZTO showed doping densities ≈1 × 10 18 cm −3 , while the oxygen-plasma treated a-ZTO exhibited doping densities ranging between ≈1 × 10 17 and ≈3 × 10 16 cm −3 . These values lie close to the range of doping densities reported by Dhara et al. [24] for a-ZTO films, which confirms the viability of the C-V analysis used to extract the doping densities. The results show that the surface accumulation layer full of charge carriers, which negatively impacts rectifying junction formation, have been effectively suppressed by plasma treatments, thus improving the performance of the rectifying junction and the diode. It was also shown that oxygen plasma improved the diode the most with the most minimal amount of hysteresis between sweeps and most consistent built-in potential.

Experimental Section
The heterojunction diodes were fabricated on Corning 7059 glass substrates measuring 40 mm × 40 mm. The bottom contact, a 100 nm thick molybdenum layer was sputtered first, then the active layers, a-ZTO and Cu 2 O of 300 nm thick each were sputtered. Finally, the top contact, a 60-100 nm gold layer was thermally evaporated. The patterning process involved a 3-mask lift-off photolithography, with the devices subjected to the highest temperature of 120 °C for 2 min during post baking. The diodes characterized and discussed in this work were 1 mm × 1 mm in size. The compositions of both oxides along with their X-ray diffraction patterns have been discussed in separate publications. [25,26] The precise fabrication and plasma treatment procedures of the diodes, along with the process and equipment involved were detailed in the previously published work. [17] Many samples were previously fabricated, tested, and optimized prior to the devices involved in this publication. However, in this study only one device per category was fabricated and characterized. To the best of the knowledge, the repeated measurements on the same devices under similar conditions were consistent.
I-V and C-V measurements were performed using an Agilent B1500 semiconductor parameter analyzer connected to a Cascade probe station. I-V measurements were carried out with step size of 20 mV, interval between steps of 1 s, swept from −1 to 1 V (forward/ up sweep) and then 1 V back to −1 V (reverse/down sweep). C-V measurements were carried out with AC signal of 30 mV at frequencies of 100-300 kHz or as specified, DC step size of 1 mV, step interval of 1 ms, swept up from −1 to 1 V then down to −1 V, in similar way to I-V measurements. All the electrical characterizations were done under normal atmospheric conditions, in the dark and under room temperature unless specified.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.