Stacked indium oxide/zinc oxide heterostructures as semiconductors in thin film transistor devices: a case study using atomic layer deposition

Shawn Sanctis a, Jan Krausmann a, Conrad Guhl b and Jörg J. Schneider *a
aFachbereich Chemie, Eduard-Zintl-Institut, Fachgebiet Anorganische Chemie, Technische Universität Darmstadt, Alarich-Weiss-Straße 12, Darmstadt, 64287, Germany. E-mail:
bFachgebiet Oberflächenforschung, Technische Universität Darmstadt, Jovanka-Bontschits-Straße 2, Darmstadt, 64287, Germany

Received 16th August 2017 , Accepted 4th November 2017

First published on 6th November 2017

Multi-layer heterostructure oxide semiconductors employing a layer-by-layer deposition of alternating indium oxide and zinc oxide thin films generated via atomic layer deposition (ALD) are investigated for their feasibility into high performance thin film transistors (TFT). The successful deposition of uniform film thickness across the alternating indium oxide and zinc oxide deposition at 200 °C is achieved using trimethyl indium (TMI), diethyl zinc (DEZ) and water as oxidizing agent. The as-prepared polycrystalline material shows a conductive behaviour which upon additional mild annealing between 250–300 °C demonstrates a high TFT device performance. In addition, insights into the dependency of the defect passivation gradient within the multilayer upon thermal annealing of the oxide stack are presented. Studies towards an optimised film thickness result in a high device performance in enhancement mode with a saturation field-effect mobility (μsat.) of 6.5 cm2 V−1 s−1 and an on/off ratio (Ion/off) of 4.6 × 107 using a deliberately large width to length channel ratio (W/L = 500). The TFT performance turned out to be dependent on the position of the individual oxide layers within the stack and the number of heterostructure stacks. These findings on the influence of semiconductor stack formation allow for a better understanding on the formation of the active semiconductor channel and serve towards the applicability of ALD based heterostructure metal oxide semiconductors in next generation electronics.

1. Introduction

Transparent thin film transistors based on semiconducting metal oxides have become the focal point of next generation display technology since their emergence as superior counterparts to the traditional silicon based transistor technology. Multinary amorphous oxides have been intensively investigated for their superior TFT performance since its inception.1,2 A variety of ternary and quaternary metal oxides, namely indium zinc oxide (IZO), zinc tin oxide (ZTO) and indium gallium zinc oxide (IGZO) among others, have evolved as promising candidates for integration into commercially relevant next-generation display technology.3,4 Deposition of such multinary metal oxide combinations have conventionally been realised via radio-frequency (RF)/direct-current (DC) sputtering as well as solution-processing especially over the last years, representing an economically feasible approach towards large area fabrication techniques.5–9 However, ALD based deposition of oxide semiconductors have gained increasing momentum in the past decade. This is largely due to the fact that the gas-phase technique ALD relies on a sequential, self-limiting surface reaction mechanism that enables it to offer precise thickness, reproducibility and compositional control of the fabricated layers.10–12 Furthermore, ALD enables the growth of high quality films with high uniformity and conformity which makes it a promising candidate for the fabrication of heterostructure based TFTs. However, ALD based deposition of metal oxides has largely been employed for the fabrication of high-k dielectrics (e.g. Al2O3, HfO2), since it ensures a facile conformal coating over desired substrates.13,14 While employing ALD based oxide dielectrics in combination with solution processed organic dielectrics it was demonstrated that device performance could be significantly enhanced in organic TFTs as well.15–17 Research in metal oxide semiconductors, research has largely focused on ZnO based devices due to their high mobility, scalability and low temperature processing. Especially, the investigation of polycrystalline ZnO based TFTs prepared via atomic layer deposition (ALD) has become a fast growing area of research which is related to the feasible tuning of electronic properties.18,19 Efforts towards the fabrication of ALD based ZnO TFTs combined with other metal oxides were mainly dedicated to enhance the stability of ZnO TFTs under bias stress. To this end, cations like Al or Hf have been utilised, aiming to control the density of defects like oxygen vacancies, resulting in high performance AlZnO (AZO) or ZnHfO TFTs.20–22 Moreover, high performance ALD based multinary amorphous InGaZnO (IGZO), employing sophisticated precursor co-injection systems, as well as amorphous IZO, obtained by using a more chemically sophisticated indium precursor in combination with diethyl zinc (DEZ), have shown promising potential in terms of amorphous semiconductor oxide deposition.23,24 However, in recent times, multilayered heterostructure semiconductors, such as ZnO in combination with various dielectric oxides have allowed access to high performance transistors, wherein the engineered insulating oxide layer facilitated the controlled passivation of defects as well as an oriented growth direction of the ZnO component.20,25,26 Interestingly, ALD deposited multilayer heterostructures based on individual binary semiconductor/semiconductor structures towards TFT applications are rather scarce, where pulsed laser deposition (PLD) based ZnO/SnO2 stacks consisting of amorphous layers has been reported.27

Herein, we show a straightforward approach towards fabrication of a multilayered heterostructure comprising polycrystalline In2O3/ZnO, using conventional ALD precursors and H2O as an oxidant, for the generation of high performance heterostructure semiconductors. Within the intended scope of the current work, we have investigated the deposition of such heterostructures, the change in device performance based on a variation of the position and number of In2O3/ZnO heterostructure interfaces as well as the influence of a reasonable post-deposition annealing process. Improved performance in such ALD based polycrystalline heterostructure semiconductor layered stacks would certainly promote interest towards their potential integration into advanced display technology as an alternative to the multinary amorphous oxide semiconductors.

2. Results and discussion

2.1. ALD based thin film deposition

To obtain the desired In2O3/ZnO stacks supercycles comprising of individual In2O3 and ZnO layers with a fixed In2O3 ratio of 0.6 in fraction to the total ALD cycles were deposited.31 In general, a supercycle or complete stack constitutes the sequential deposition of these different metal oxide layers where the cycle ratio is derived from the number of ALD cycles for the individual oxides. Consequently, a supercycle with a In2O3 ratio of 0.6 in fraction can be composed of six cycles of In2O3 and four cycles of ZnO or multiples of those. The general procedure of the aforementioned supercycle deposition and its utilisation to generate stacked heterostructures of In2O3/ZnO interfaces is shown below [Scheme 1]. Since 200 °C reflects the lower limit of the temperature window for deposition of In2O3 from TMI and favourably coincides with the temperature window for ZnO deposition from DEZ, this temperature was chosen for the deposition of the In2O3/ZnO heterostructure stacks.
image file: c7tc03724d-s1.tif
Scheme 1 Schematic representation of a single supercycle performed at 200 °C for the deposition of a In2O3/ZnO heterostructure stack from the precursors TMI and DEZ. Steps (1) and (2) depict the indium precursor (TMI) deposition and partial reaction, followed by removal of volatile by-products with an Ar purge and final conversion to the indium oxide with a H2O pulse. Steps (4) and (5) depict the deposition of the zinc precursor (DEZ) on the indium oxide layer, with analogous reaction steps as for the indium precursor TMI. Steps (3) and (6) indicate multiple iterations of the individual precursor cycles to achieve the desired deposition thicknesses.

The first half of the supercycle (Step 1 and 2) comprises the exposure of the substrate to a TMI pulse (1 s) accompanied with an extended precursor exposure (1.5 s) and the subsequent oxidization with water. In the first step, TMI molecules react with surface OH-groups according to the following equation:32

(OH)x* + In(CH3)3 (g) → (–O–)xIn(CH3)3−x + xCH4 (g)
where the surface species are denoted by an asterisk (*). After purging the reactor with carrier gas (Ar), water is introduced into the chamber as an oxidizing agent. As examined in earlier reports, an extremely large Langmuir exposure (∼2 Torr s−1) is required for the surface reaction to convert In–CH3* molecules into InOH* groups.33 To provide sufficient conditions for this conversion, an extended water pulse time of 0.1 s was used, followed by Ar purging (20 s). After repeated utilization of the described sequence of TMI and water pulses (Step 3), surface In–OH groups are further subjected to ZnO deposition (Step 4 and 5) to finally obtain the desired In2O3/ZnO interfaces. The sequence of ZnO deposition includes the pulsing of DEZ (0.015 s) in the first step, adopting an additional extended exposure of the precursor (1.5 s) with a subsequent Ar purge (15 s), assuming the following reaction of surfacial In–OH*-groups with DEZ:
In–OH* + Zn(C2H5)2 → In–O–Zn–C2H5* + C2H6
The latter exposure was used to provide sufficient adsorption time and to overcome adsorption delay of the precursor which is known to cause reduced deposition rates when mixed metal oxides are deposited via an ALD route.34 In order to oxidize the resulting Zn–C2H5* bound surface species, converting them into Zn–OH* species and providing new adsorption sites for the successive deposition step, a water pulse (0.015 s) was applied (Step 5), accompanied by Ar purge (15 s). A full supercycle to obtain the desired heterostructures is completed via reiteration of the latter sequence (Step 6). The described supercycle was repeated to vary the number of In2O3/ZnO interfaces or in more general terms to adjust the thickness of the thin films. Regarding ALD of mixed metal oxides in a supercycle fashion as described, the growth rate of the thin films per ALD cycle can be estimated from the deposition rates of the individual binary oxides according to the rule of mixture:33
G(In2O3/ZnO stack) = R(In2O3) × G(In2O3) + [1 − R(In2O3)] × G(ZnO)
where G denotes the growth rate per cycle for the corresponding metal oxide and R describes the fraction of In2O3 cycles in the supercycle.

In order to estimate the deposition rate for the current In2O3/ZnO heterostructures, the growth rates of the individual oxides were determined. These could be assigned to 0.4 Å cycle−1 and ∼1.45 Å cycle−1 for In2O3 and for ZnO, respectively. Based on these results, the expected growth rate derived from the rule of mixture for the heterostructures was ∼0.82 Å cycle−1 when an In2O3 cycle ratio of 0.6 was used. Compared to this estimation, the actual deposition rate of the stacked heterostructures is significantly lower, exhibiting a value of ∼0.54 Å cycle−1. The observed trend of a deteriorated growth rate was also reported for amorphous IZO thin films that were generated via utilization of TMI, DEZ and water as the metal precursors and oxidizing agent, respectively wherein the lowered growth rate was ascribed to a retarded adsorption of DEZ during ALD growth of the amorphous ternary metal oxide.33 Similar observations have been made for ALD based zinc tin oxide (ZTO) thin films where it was reported that a change in surface chemistry due to a ligand conversion led to a reduction of –OH groups, consequently reducing the reaction sites for subsequent ZnO deposition.34

Cross-sectional transmission electron microscopy (TEM) micrographs obtained by focused ion beam (FIB) preparation of the ultra-thin stacks display a defined deposition of four In2O3/ZnO stacks, with In2O3 as the first deposited layer on the Si/SiO2 substrate followed by ZnO deposition. The TEM investigations indeed reflect the growth of an ultrathin layer (∼9 nm), comprising a layer-by-layer architecture (Fig. 1a) with distinct individual layers of In2O3 (darker contrast) and ZnO (lighter contrast). These layers exhibit a microcrystalline nature of the deposited oxides (Fig. 1b). In order to verify the gradient of the Zn2+ and In3+ cationic species, Auger electron spectroscopy (AES) was performed on the as-prepared sample. The Auger depth profile exhibits multiple peak profiles indicating the expected distribution of the respective oxides along the vertical gradient of the film showing alternating peaks and troughs related to the atomic concentration assignable to the varying film deposition according to the deposition protocol of the supercycles (Fig. 1c). The variation of chemical composition for both the In3+ and Zn2+ across the stacks corroborate well with the findings from the TEM investigation of the In2O3/ZnO heterostructure. Additionally, no significant carbon contamination from undecomposed precursors was observed within the film, showing only carbon related signals arising from surface species, but not from intrinsic contamination of the deposited layers. This clearly indicates a successful conversion of the precursor by solely using H2O as an oxidant. Presence of minor amounts of adsorbed surface moieties on the thin film can be expected, since the devices were not encapsulated.7 In order to gain further insight into the evolution of the film quality of the resultant stacked layer, atomic force microscopy (AFM) was performed for the individually deposited In2O3 and ZnO films and for the In2O3/ZnO stacks with all three films possessing the same thickness (∼9 nm) and post calcination at 300 °C (Fig. S1, ESI). The individually deposited In2O3 films (RRMS ∼ 0.34 nm) showed a higher degree of roughness in comparison with that of the individually deposited ZnO films (RRMS ∼ 0.22 nm). The In2O3/ZnO stack layers (RRMS ∼ 0.28 nm) exhibit an intermediate degree of roughness compared to their individual oxide counterparts. These observations display that In2O3 films obtained from TMI contribute significantly to the overall roughness of the stack and are slightly rougher than the ZnO films obtained using DEZ. Although film roughness might be an important issue to improve the TFT device performance, based on the current investigations it appears that in such ultrathin heterostructure stack films (<10 nm), film roughness does not seem to affect the overall TFT performance. Instead, an even larger effect might arise from the fact that increasing the local charge carrier concentration closer to the dielectric interface, by deposition of a relatively rough In2O3 film first, at the dielectric SiOx interface, instead of a smoother ZnO film. The deposition of a smoother ZnO film as the first layer followed by an In2O3 film just by switching the sequence of the deposited layers within the stack indeed decreased the overall TFT performance drastically (see Section 2.2 – thin film transistor performance). In order to assess the optical transparency, individual In2O3, ZnO films as well as the (In2O3/ZnO) stacked layer films were investigated (Fig. S2, ESI). All three films display desirable optical properties, with good optical transparency (>80%) in the visible region.

image file: c7tc03724d-f1.tif
Fig. 1 Cross-sectional TEM micrographs obtained via FIB sample preparation of (a) as-deposited, (b) high-resolution image and (c) characteristic auger electron spectroscopy (AES) based depth profile of as-deposited four In2O3/ZnO heterostructure stacks.

To investigate possible changes in the chemical bonding of the of the In2O3 and ZnO layers within a stack in its as-prepared state and after post-deposition annealing, X-ray photoelectron spectroscopy (XPS) was performed for the as-prepared stack as well as for those which were post annealed at 250 °C and at 300 °C respectively (Fig. S3 and Table S1, ESI). Based on the peak position obtained from the core spectra, the In3d5/2 and the In3d3/2 for all three samples are observed at 444.4 eV and 452 eV respectively, with no significant peak shifts. The observations are in good agreement with typical values reported for In2O3, indicating that the indium oxide did not undergo any significant chemical changes due to this post annealing within the investigated temperature regime. However, interestingly significant shifts towards higher binding energies for the Zn peaks are observed when comparing the as-prepared sample with the post annealed sample. Herein, for the as-prepared and annealed sample, the Zn2p3/2 and Zn2p1/2 values shifted from 1021.3 eV to 1021.8 eV and 1044.4 eV to 1044.9 eV respectively. Similar shifts towards higher binding energies were observed in recent literature for ultrathin bilayer In2O3/ZnO films, where the peak shifts with increased processing temperature were attributed to an improved passivation of oxygen vacancy related defects as well as the potential interplay between the oxygen related species at the bilayer In2O3/ZnO interface, arising from the formation of individual oxides during the post-deposition annealing.35 The correlation of the chemical composition of the thin films, with respect to the nature and concentration of oxygen related species across the depth of the stacks were thus further studied via angle resolved XPS (ARXPS). The core oxygen O1s spectra are deconvoluted into two primary peaks at ∼530 eV and ∼532 eV. The former is attributed to the co-ordinated metal–oxygen bonds, while the latter is observed due to the residual hydroxide species present within the film, due the conditions employed for film growth.36 Deconvoluted O1s spectra of the as-prepared film as well as the 300 °C annealed films across different sample tilt angles (STA) however reveal distinct changes in the ratio of oxygen related species of the observed O1s spectrum (Fig. 2 and Table 1).

image file: c7tc03724d-f2.tif
Fig. 2 Angle resolved O1s spectra of (a–d) as-deposited In2O3/ZnO heterostructures (four stacks) and (e–h) In2O3/ZnO heterostructures annealed at 300 °C for 2 hours for different sample tilt angles (STA). An increasing STA (30° → 75°) reflects oxygen contributions arising from the region close to the surface (30°), graduating towards the interface at 75°.
Table 1 Atomic % ratio of co-ordinated metal to oxygen bonds (M–O) and hydroxide species (M–OH) of the O1s spectra based on ARXPS of heterostructure [In2O3] films together with the corresponding processing conditions. An increasing STA from 30° to 75° reflects oxygen contributions mainly arising from the surface and graduating towards the interface
Sample In2O3/ZnO STA 30° STA 45° STA 60° STA 75°
M–O (at%) M–OH (at%) M–O (at%) M–OH (at%) M–O (at%) M–OH (at%) M–O (at%) M–OH (at%)
As-prepared 72.6 27.4 67.7 32.3 63.2 36.8 55.7 44.3
Annealed at 300 °C 93.2 6.8 87.1 12.9 84.8 15.2 84.5 15.5

A lower STA reveals contributions mainly from the surface and a gradual increase in STA accounts for contributions from the surface along with the increasing contributions closer to the dielectric interface (semiconductor/SiOx). The ARXPS analysis hints towards the higher concentration of residual hydroxyl groups at the interface, which also correlates with the conductive behaviour observed in thicker films and the need for post-deposition annealing temperatures.36,37

One possible reason for this observation can be allocated to the fact that if the films are stronger oxidized on the surface, the density of the formed oxide at the surface may serve as a strong oxygen diffusion barrier which however, could be overcome with an additional thermal annealing step.36 This idea is well supported by the fact that a high pressure annealing of sputter based IGZO in different environments, while maintaining a low thermal budget, has indeed shown a remarkable improvement in performance as well a substantial reduction in processing temperature.38,39

2.2. Thin film transistor performance

In order to understand the influence of each of the individual oxides on the overall TFT performance, devices with bottom gate/bottom contact (BGBC) set-up were fabricated to evaluate the performance of the active layers. A sufficiently large source–drain electrode geometry with a channel width (W) to length (L) ratio (W/L = 500) was deliberately chosen to avoid any overestimation of field-effect mobility for the measured TFTs. Choice of electrode dimensions with extremely small W/L ratio could easily lead to an overestimation of the effective TFT mobility of up to 400%.40,41 TFTs based on indium oxide as well as zinc oxide (both ∼9 nm) were fabricated, resembling similar thickness conditions as that of the optimised heterostructure layer with four stacks.

After post-annealing (300 °C for 2 hours), the indium oxide based TFTs exhibited a conductive behaviour, with a Ion/off ratio ∼1, while the zinc oxide based TFT exhibited a clear semiconducting behaviour with an average saturation field effect mobility (μsat.) of 0.48 cm2 (V−1 s−1), Vth of 12.7 V, Ion/off ratio of 3 × 106 and a subthreshold swing (SS) of 1.61 V dec−1 (Fig. 3). This serves as a good indication that in comparison with zinc oxide, indium oxide generated, possesses a higher defect concentration, wherein further thermal annealing (300 °C), did not passivate the defects sufficiently to exhibit a semiconductive behaviour when integrated into a TFT device.42 This is also in agreement with the XPS results discussed earlier, where the peak shifts (Zn2p, In3d) were observed only for ZnO and not for In2O3 within the stack.

image file: c7tc03724d-f3.tif
Fig. 3 Individual transfer characteristics of TFTs based on individual In2O3, ZnO layers, [(In2O3/ZnO), four stacks] and on [(ZnO/In2O3), four stacks] each sample annealed at 300 °C for 2 hours.

Additionally, based on the metal oxide employed and its observed semiconducting properties, deposition (either sputtering or solution processing) of semiconducting layers with a higher charge carrier concentration at the dielectric interface, followed by subsequent deposition of layers with reduced charge carrier concentration has shown substantial improvement in the overall transistor performance.43,44 Our preliminary examination of reversing positions of the interlayers, starting with deposition of the zinc oxide layer first, followed by indium oxide layer deposition (→ ZnO/In2O3, four stack arrangement) yielded only poor TFT performance with an average field-effect mobility (μsat.) of 1.07 cm2 V−1 s−1, Vth of 10 V, Ion/off ratio of 5.2 × 106 and a sub-threshold swing (SS) of 1.85 V dec−1. However, when the deposition of the indium oxide layer is first, followed by the zinc oxide layer (→ In2O3/ZnO, four stack arrangement) a comparatively high device performance with an average field-effect mobility (μsat.) of 6.5 cm2 V−1 s−1, Vth of 8.9 V, Ion/off ratio of 5 × 107 and a substantially reduced subthreshold swing (SS) of 0.7 V dec−1 is observed (Fig. 3). This observation is in good agreement with reported observations based on synergistic effects of device performance arising from interplay of intrinsic vacancy defects of the individual layers, which are based on their position with respect to the dielectric interface, within a heterostructure architecture.45–47

Based on the TFT performance of the individual oxides, preliminary experiments were then performed to determine the critical number of deposition cycles for In2O3 and the subsequently needed cycles for ZnO in order to understand the required stack architecture for which active semiconductor behaviour in a relatively low temperature regime at 300 °C could be observed. Once this was achieved, device optimisation was performed in order to assess the number of In2O3/ZnO stacks needed, followed by a subsequent impact of post deposition annealing temperature on TFT performance. Active film thickness was increased by performing multiple deposition iterations of a single supercycle via ALD, followed by post processing of all the TFTs at 300 °C. Film thicknesses for three, four, five and six stacks were 7 nm, 9 nm, 11 nm and 13 nm respectively, as determined via ellipsometry. The measured transfer and output characteristics for an increasing number of stacks and its corresponding evaluation of the critical device performance parameters are shown (Fig. 4 and Table 2).

image file: c7tc03724d-f4.tif
Fig. 4 (a) Schematic representation of the TFTs (not to scale) fabricated employing ALD based In2O3/ZnO heterostructure stacks as an active semiconducting layer. (b) Collective transfer characteristics of three, four, five and six stacks and (c–f) output characteristics of devices with three, four, five and six In2O3/ZnO heterostructure stacks, respectively.
Table 2 Evaluated TFT performance parameters for devices with three, four, five and six In2O3/ZnO heterostructure stacks, respectively
Number of stacks Mobility μsat. [cm2 (V−1 s−1)] Threshold voltage Vth [V] Current on/off ratio Ion/off Subthreshold slope SS [V dec−1]
3 2.3 14.8 8 × 106 1.3
4 6.5 8.9 5 × 107 0.7
5 7.1 5 1 × 106 1.1
6 4.6 3.3 1 × 105 1.9

From the evaluation of the transfer characteristics, a systematic increase in film conductivity is observed, as one increases the number of deposited stacks. A primary effect observed for an increasing thickness of the semiconducting layer is a gradual increase of the off-current and a negative shift of the threshold voltage Vth. The need for a compromise between the device mobility and other performance parameters such as Vth, Ion/off ratio and subthreshold swing is a well reported observation with inorganic oxide based TFTs.

Several recent strategies to control the local carrier concentration gradient across the film-thickness by manipulating the oxygen environment during sputter based thin film deposition or by modulating the cation species within metal oxide systems (ranging from binary to quaternary oxides) have demonstrated their efficacy in substantially improving device performance.43,44,46–48 The latter approach is also in accordance with the results demonstrated within the current work, whereby increasing the overall charge carrier concentration at the interface (by In2O3 deposition as the first layer of the stack) appears to be a straight-forward fabrication route using ALD.

The TFT characteristics of the thinnest sample (three stacks) indicate this thickness to be the threshold limit of critical thickness of the active layer herein, where sufficient charge carriers are not present within the sub-optimal stack thickness with a visibly lower on-current (Ion) in the saturation regime (Fig. 4b and c). For the optimal stack (four stacks), the best trade-off between mobility, threshold voltage, current on/off ratio and sub-threshold swing is achieved, with modest μsat. and Vth values accompanied by improved subthreshold swing and high Ion/off. Hence, we consider the four stacks as an optimum stack within the current work. Moreover, the mobility and Vth does improve with increased film thickness (four and five stacks), but the subthreshold swing and Ion/off are compromised, which is undesired. Additionally, for an increasing film thickness, an increase in off-current and a negative shift in the Vth has been attributed to deep-donor like states arising from oxygen vacancy related defects, thereby increasing the overall trap density.26,49,50 A similar correlation of TFT performance parameters with increasing thickness has been reported for sputter-based IZO films, wherein deterioration of μsat., Ion/off and SS are observed for incremental film thickness beyond the optimal range.51 The observed changes of these performance parameters with increasing film thickness were attributed mainly to charge scattering processes arising from the bulk portion in the thicker films. This begins to add up significantly to the existing interfacial trap states as the film thickness is increased beyond the optimum range. This could account for the notably decreased μsat., Ion/off (higher Ioff) and increased SS (six stacks) in the current investigation, which is indeed reflected in the stack thickness based evolution of the transfer characteristics (Fig. 4b).

Since the active layer with four stacks displayed the best overall TFT characteristics, compromising a mobility of 6.5 cm2 V−1 s−1, a threshold voltage of 8.9 V and a reasonably high Ion/off ratio of 4.6 × 107, we investigated the influence of annealing temperature on the TFT characteristics of this device configuration in order to further study the role of oxygen vacancy passivation by thermal annealing on the TFT characteristics. For that reason the TFT performance of devices annealed for two hours at 250 °C and 300 °C in air is compared (Fig. 5 and Table 3). Comparing the TFT output characteristics, films post annealed at 300 °C exhibit a reduced saturation drain current and an improved gate modulation, indicating a more efficient reduction of free charge carriers via defect passivation at higher annealing temperature compared to the as-prepared ones. In comparison with a TFT device annealed at 250 °C, the ones annealed at 300 °C also show an improvement in mobility, accompanied by an increased Ion/off of about a decade and a substantial reduction in SS from 1.9 V dec−1 to 0.7 V dec−1. Further improvement of device performance by annealing at higher temperatures (350 °C) was not observed.

image file: c7tc03724d-f5.tif
Fig. 5 (a) Transfer characteristics, (b) and (c) output characteristics for the TFT devices annealed to 250 °C and 300 °C, respectively.
Table 3 Summary of the TFT parameters electron mobility, threshold voltage and Ion/off ratio for devices annealed at 250 °C and 300 °C
Annealing temperature [°C] Mobility μsat. [cm2 V−1 s−1] Threshold voltage Vth [V] Current on/off ratio Ion/off Subthreshold slope SS [V dec−1]
250 4.8 7.6 4 × 106 1.9
300 6.5 8.9 5 × 107 0.7

Additionally, average performance parameters of the mentioned devices annealed at 250 °C (four stacks) are similar to those of devices based on six stacks, obtained at an annealing temperature of 300 °C, with respect to the obtained mobility and SS values. These observations clearly exhibit that, apart from achieving an optimum stack In2O3/ZnO thickness, thermal annealing plays an equally important role in improving device performance, whereby the synergistic effect of heterostructure architecture alone may not be sufficient to render a high TFT performance.

3. Conclusion

To conclude, an ALD based In2O3/ZnO heterostructure design yielding high performance TFTs has been successfully demonstrated using trimethyl indium and diethyl zinc as molecular precursors. Employment of a layer by layer deposition of individual semiconducting oxides provides a potential alternative pathway to conventional amorphous semiconductors which are otherwise rather difficult to fabricate via ALD based deposition systems. TFT devices fabricated solely on indium and zinc oxide thin layer films were not found to be feasible towards generation of high performance device characteristics. However, generation of an optimised heterostructure based on sequential deposition of the aforementioned individual oxides, processed at a reasonably low temperature regime (250–300 °C) deliver high performance TFTs. Devices based on such a fabrication process demonstrated an average saturation field-effect mobility (μsat.) of 6.5 cm2 V−1 s−1 and a high current on/off ratio of 4.6 × 107 and a low subthreshold swing (SS) of 0.7 V dec−1 respectively, at a reasonable processing temperature of 300 °C. Such ALD based semiconductor heterostructure architecture presents itself as a facile strategy towards a cost-effective fabrication technique with potential applications in the field of large-area oxide electronics.

4. Experimental section

4.1. ALD process

The heterostructure thin film depositions of (In2O3/ZnO) stacks were performed in a Savannah S 100 system (Cambridge Ultratech) at a base pressure of 0.8 Torr. The precursors used for the depositions were trimethyl indium 98+% (99.9% In, Strem Chemicals), diethyl zinc (min. 95%, Strem Chemicals) and water (HPLC grade, Sigma-Aldrich). All thin films were deposited via ALD supercycles with a cycle ratio of In of 0.6 in fraction. To obtain the desired film-thickness, the numbers of supercycle iterations were varied accordingly. All precursors were maintained at room temperature. Argon (99.9999%, Alpha Gaz™) was used as a carrier gas, for the depositions and maintained at a constant flow rate of 20 sccm. For the deposition of the thin-films a custom-configured precursor exposure recipe was developed. In particular, for the In2O3 deposition, the sequential deposition cycle is as follows: TMI pulse of 1 s, precursor exposure of 1.5 s, argon purge for 10 s, H2O pulse of 0.1 s and a final argon purge for 20 s. For the ZnO deposition, the sequence is as follows: DEZ-pulse of 0.015 s, precursor exposure of 1.5 s, argon purge for 15 s, H2O pulse of 0.015 s and a final argon purge for 15 s. It must be noted that the exposure time is the additional time, provided for sufficient adhesion of the precursor molecules to the substrate to undergo partial reaction after being introduced in the ALD chamber.

4.2 Material characterization

Transmission electron microscopy (TEM) was carried out using a Tecnai G2 F20 (FEI), with an operating voltage of 200 kV.28 Samples for the TEM investigations were prepared by thin-film deposition on Si/SiO2 substrates (10 × 10 mm2) and annealed at the desired temperatures. Samples for focussed ion beam (FIB) were prepared using a gallium focussed ion beam (Helios-400, FEI) and coated with a platinum layer.29 Film thicknesses were obtained via ellipsometry (Accurion EP3 System) at a wavelength of 632.8 nm. Atomic force microscopy (AFM) measurements were carried out with MFP-3D™ (Asylum Research), using silicon cantilevers. UV-vis measurements were carried out on clean quartz substrates (Thermo Scientific-Evolution 600). XPS measurements were carried out in an integrated ultra-high vacuum (UHV) system with a base pressure 1 × 10−9 mbar, equipped with a PHI 5000 VersaProbe photoelectron spectrometer.30 Auger spectroscopy was performed on a PHI 680 (Physical Electronics) scanning Auger nanoprobe operated under ultra-high vacuum (3 × 10−10 mbar) at an acceleration voltage of 10 keV and a current of 20 nA. Sputtering was carried out under ultra high vacuum (5 × 10−9 Torr) with an argon ion gun operated at 250 eV and 500 nA.

4.3 Thin film transistor characterization

TFT substrates with pre-fabricated source–drain electrodes in a bottom-gate-bottom-contact (BGBC) device geometry were obtained from the Fraunhofer IMPS, Dresden. The substrates consist of highly n-doped silicon with a 90 nm silicon-oxide dielectric layer. The source–drain electrodes, with a channel length L = 20 μm and a channel width W = 10 mm (W/L = 500), comprise 40 nm of gold (interdigital structure) with a 10 nm intermediate adhesion layer of indium tin oxide (ITO). Devices were isolated via a controlled mechanical scribing process post fabrication. The FET substrates, Quartz slides (1.5 × 1.5 cm2) and bare Si/SiO2 substrates were all cleaned sequentially via ultrasonication in acetone, DI-water and iso-propanol (all HPLC grade) for 10 minutes each. TFT characteristics were determined with an HP 4155A Semiconductor Parameter Analyzer (Agilent) in a glove box under exclusion of air and moisture in the dark. Charge carrier mobility (μsat.) and threshold voltage Vth were derived from a linear fitting of the square root of the source–drain current (√Ids) as a function of gate-source voltage VGS.

Conflicts of interest

There are no conflicts to declare.


S. S. and J. K. contributed equally towards this work. S. S. and J. J. S. acknowledge financial support through the DFG SPP 1569 program. TEM investigations were performed at ERC Jülich under contract ERC-TUD1. Auger measurements were carried out at Karlsruhe Nano Micro Facility (KNMF proposal number 2016-015-010549) at Karlsruhe Institute of Technology (KIT). We thank Tobias Weingärtner (KIT) for Auger measurements and Dr Jörg Engstler (TUDa) for TEM studies.


  1. K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano and H. Hosono, Nature, 2004, 432, 488–492 CrossRef CAS PubMed.
  2. H. Yabuta, M. Sano, K. Abe, T. Aiba, T. Den, H. Kumomi, K. Nomura, T. Kamiya and H. Hosono, Appl. Phys. Lett., 2006, 89, 112123 CrossRef.
  3. T. Kamiya and H. Hosono, NPG Asia Mater., 2010, 2, 15–22 CrossRef.
  4. E. Fortunato, P. Barquinha and R. Martins, Adv. Mater., 2012, 24, 2945–2986 CrossRef CAS PubMed.
  5. J. K. Jeong, J. H. Jeong, H. W. Yang, J.-S. Park, Y.-G. Mo and H. D. Kim, Appl. Phys. Lett., 2007, 91, 113505 CrossRef.
  6. W. Lim, S. Kim, Y.-L. Wang, J. W. Lee, D. P. Norton, S. J. Pearton, F. Ren and I. I. Kravchenko, J. Electrochem. Soc., 2008, 155, H383–H385 CrossRef CAS.
  7. K. K. Banger, Y. Yamashita, K. Mori, R. L. Peterson, T. Leedham, J. Rickard and H. Sirringhaus, Nat. Mater., 2011, 10, 45–50 CrossRef CAS PubMed.
  8. R. M. Pasquarelli, D. S. Ginley and R. O'Hayre, Chem. Soc. Rev., 2011, 40, 5406–5441 RSC.
  9. J. S. Park, H. Kim and I.-D. Kim, J. Electroceram., 2014, 32, 117–140 CrossRef CAS.
  10. M. Ritala, K. Kukli, A. Rahtu, P. I. Räisänen, M. Leskelä, T. Sajavaara and J. Keinonen, Science, 2000, 288, 319 CrossRef CAS PubMed.
  11. H. Kim and P. C. McIntyre, J. Korean Phys. Soc., 2006, 48, 5–17 CAS.
  12. L. Niinistö, M. Nieminen, J. Päiväsaari, J. Niinistö, M. Putkonen and M. Nieminen, Phys. Status Solidi A, 2004, 201, 1443–1452 CrossRef.
  13. S. Gieraltowska, L. Wachnicki, B. S. Witkowski, M. Godlewski and E. Guziewicz, Thin Solid Films, 2012, 520, 4694–4697 CrossRef CAS.
  14. M. Leskelä and M. Ritala, Thin Solid Films, 2002, 409, 138–146 CrossRef.
  15. X. F. She, J. Wang, Q. Xue and W. Xu, RSC Adv., 2016, 6, 114593–114598 RSC.
  16. W. Xu and S.-W. Rhee, J. Mater. Chem., 2011, 21, 998–1004 RSC.
  17. W. Xu, F. Wang and S.-W. Rhee, J. Mater. Chem., 2012, 22, 1482–1488 RSC.
  18. L. Grzegorz, G. Marek, G. Elzbieta, S. Pavlo, C. Vladyslav and V. Dmytro, Semicond. Sci. Technol., 2012, 27, 074006 CrossRef.
  19. K. Semyung, B. Seokhwan, L. Seungjun, J. Sunyeol, J. Wooho, K. Hyungchul, G. Su Cheol, C. Ho Jung, P. Hyung-ho and J. Hyeongtag, Semicond. Sci. Technol., 2009, 24, 035015 CrossRef.
  20. P. K. Nayak, Z. Wang, D. H. Anjum, M. N. Hedhili and H. N. Alshareef, Appl. Phys. Lett., 2015, 106, 103505 CrossRef.
  21. Y. J. Chung, W. J. Choi, S. G. Kang, C. W. Lee, J.-O. Lee, K.-J. Kong and Y. K. Lee, J. Mater. Chem. C, 2014, 2, 9274–9282 RSC.
  22. C. H. Ahn, M. G. Yun, S. Y. Lee and H. K. Cho, IEEE Trans. Electron Devices, 2014, 61, 73–78 CrossRef CAS.
  23. A. Illiberi, B. Cobb, A. Sharma, T. Grehl, H. Brongersma, F. Roozeboom, G. Gelinck and P. Poodt, ACS Appl. Mater. Interfaces, 2015, 7, 3671–3675 CAS.
  24. J. Sheng, H.-J. Lee, S. Oh and J.-S. Park, ACS Appl. Mater. Interfaces, 2016, 8, 33821–33828 CAS.
  25. C. H. Ahn, K. Senthil, H. K. Cho and S. Y. Lee, Sci. Rep., 2013, 3, 2737 CrossRef PubMed.
  26. P. K. Nayak, Z. Wang and H. N. Alshareef, Adv. Mater., 2016, 28, 7736–7744 CrossRef CAS PubMed.
  27. S.-J. Lee, C.-S. Hwang, J.-E. Pi, J.-H. Yang, C.-W. Byun, H. Chu, K. I. Cho and S. Cho, ETRI J., 2015, 37, 1135–1142 CrossRef.
  28. M. Luysberg, M. Heggen and K. Tillmann, J. Large-Scale Res. Facil., 2016, 2, 77 CrossRef.
  29. D. Meertens, M. Kruth and K. Tillmann, J. Large-Scale Res. Facil., 2016, 2, 60 CrossRef.
  30. S. Sanctis, R. C. Hoffmann, R. Precht, W. Anwand and J. J. Schneider, J. Mater. Chem. C, 2016, 4, 10935–10944 RSC.
  31. D.-J. Lee, J.-Y. Kwon, J. I. Lee and K.-B. Kim, J. Phys. Chem. C, 2011, 115, 15384–15389 CAS.
  32. A. U. Mane, A. J. Allen, R. K. Kanjolia and J. W. Elam, J. Phys. Chem. C, 2016, 120, 9874–9883 CAS.
  33. D. J. Lee, J. Y. Kwon, J. Kim, K.-J. Kim, Y.-H. Cho, S.-Y. Cho, S.-H. Kim, J. Xu and K.-B. Kim, J. Phys. Chem. C, 2014, 118, 408–415 CAS.
  34. J. T. Tanskanen, C. Hägglund and S. F. Bent, Chem. Mater., 2014, 26, 2795–2802 CrossRef CAS.
  35. K. Tetzner, I. Isakov, A. Regoutz, D. J. Payne and T. D. Anthopoulos, J. Mater. Chem. C, 2017, 5, 59–64 RSC.
  36. T. Kamiya and H. Hosono, ECS Trans., 2013, 54, 103–113 CrossRef.
  37. H. J. Kim, S. Y. Park, H. Y. Jung, B. G. Son, C.-K. Lee, C.-K. Lee, J. H. Jeong, Y.-G. Mo, K. S. Son, M. K. Ryu, S. Lee and J. K. Jeong, J. Phys. D: Appl. Phys., 2013, 46, 055104 CrossRef.
  38. W.-G. Kim, Y. J. Tak, B. Du Ahn, T. S. Jung, K.-B. Chung and H. J. Kim, Sci. Rep., 2016, 6, 23039 CrossRef CAS PubMed.
  39. H. S. Shin, Y. S. Rim, Y.-G. Mo, C. G. Choi and H. J. Kim, Phys. Status Solidi A, 2011, 208, 2231–2234 CrossRef CAS.
  40. K. Okamura, D. Nikolova, N. Mechau and H. Hahn, Appl. Phys. Lett., 2009, 94, 183503 CrossRef.
  41. M. G. Kim, H. S. Kim, Y.-G. Ha, J. He, M. G. Kanatzidis, A. Facchetti and T. J. Marks, J. Am. Chem. Soc., 2010, 132, 10352–10364 CrossRef CAS PubMed.
  42. J. Sheng, D.-w. Choi, S.-H. Lee, J. Park and J.-S. Park, J. Mater. Chem. C, 2016, 4, 7571–7576 RSC.
  43. Y. S. Rim, H. Chen, X. Kou, H.-S. Duan, H. Zhou, M. Cai, H. J. Kim and Y. Yang, Adv. Mater., 2014, 26, 4273–4278 CrossRef CAS PubMed.
  44. S. Yoon, S. J. Kim, Y. J. Tak and H. J. Kim, Sci. Rep., 2017, 7, 43216 CrossRef CAS PubMed.
  45. G. X. Liu, A. Liu, F. K. Shan, Y. Meng, B. C. Shin, E. Fortunato and R. Martins, Appl. Phys. Lett., 2014, 105, 113509 CrossRef.
  46. M. Nag, A. Chasin, M. Rockele, S. Steudel, K. Myny, A. Bhoolokam, A. Tripathi, B. van der Putten, A. Kumar, J.-L. van der Steen, J. Genoe, F. Li, J. Maas, E. van Veenendaal, G. Gelinck and P. Heremans, J. Soc. Inf. Disp., 2013, 21, 129–136 CrossRef CAS.
  47. X. Yu, N. Zhou, J. Smith, H. Lin, K. Stallings, J. Yu, T. J. Marks and A. Facchetti, ACS Appl. Mater. Interfaces, 2013, 5, 7983–7988 CAS.
  48. Y. H. Lin, H. Faber, J. G. Labram, E. Stratakis, L. Sygellou, E. Kymakis, N. A. Hastas, R. Li, K. Zhao, A. Amassian, N. D. Treat, M. McLachlan and T. D. Anthopoulos, Adv. Sci., 2015, 2, 1500058 CrossRef PubMed.
  49. G. Wakimura, Y. Yamauchi, T. Matsuoka and Y. Kamakura, J. Adv. Simulat. Sci. Eng., 2014, 2, 201–210 CrossRef.
  50. E. N. Cho, J. H. Kang and I. Yun, Microelectron. Reliab., 2011, 51, 1792–1795 CrossRef CAS.
  51. P. Barquinha, A. Pimentel, A. Marques, L. Pereira, R. Martins and E. Fortunato, J. Non-Cryst. Solids, 2006, 352, 1749–1752 CrossRef CAS.


Electronic supplementary information (ESI) available: Additional AFM, UV-vis and XPS data. See DOI: 10.1039/c7tc03724d
Authors with equal contribution.

This journal is © The Royal Society of Chemistry 2018