Charge Transport in Low-Temperature Processed Thin-Film

5 hours ago - The influence of the composition within multilayered heterostructure oxide semiconductors has a critical impact on the performance of th...
0 downloads 0 Views 4MB Size
Research Article www.acsami.org

Cite This: ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Charge Transport in Low-Temperature Processed Thin-Film Transistors Based on Indium Oxide/Zinc Oxide Heterostructures Jan Krausmann,† Shawn Sanctis,† Jörg Engstler,† Martina Luysberg,§ Michael Bruns,‡ and Jörg J. Schneider*,† †

Fachbereich Chemie, Eduard-Zintl-Institut, Fachgebiet Anorganische Chemie, Technische Universität Darmstadt, Alarich-Weiss-Straße 12, 64287 Darmstadt, Germany § Forschungszentrum Jülich GmbH, Ernst Ruska-Centre (ERC) and Peter Grünberg Institute (PGI), Wilhelm-Johnen-Straße, 52428 Jülich, Germany ‡ Institute for Applied Materials (IAM-ESS), Karlsruhe Institute of Technology, Hermann-von-Helmholtz-Platz 1, B 321, D-76344 Eggenstein-Leopoldshafen, Germany S Supporting Information *

ABSTRACT: The influence of the composition within multilayered heterostructure oxide semiconductors has a critical impact on the performance of thin-film transistor (TFT) devices. The heterostructures, comprising alternating polycrystalline indium oxide and zinc oxide layers, are fabricated by a facile atomic layer deposition (ALD) process, enabling the tuning of its electrical properties by precisely controlling the thickness of the individual layers. This subsequently results in enhanced TFT performance for the optimized stacked architecture after mild thermal annealing at temperatures as low as 200 °C. Superior transistor characteristics, resulting in an average field-effect mobility (μsat.) of 9.3 cm2 V−1 s−1 (W/L = 500), an on/off ratio (Ion/Ioff) of 5.3 × 109, and a subthreshold swing of 162 mV dec−1, combined with excellent long-term and bias stress stability are thus demonstrated. Moreover, the inherent semiconducting mechanism in such multilayered heterostructures can be conveniently tuned by controlling the thickness of the individual layers. Herein, devices comprising a higher In2O3/ZnO ratio, based on individual layer thicknesses, are predominantly governed by percolation conduction with temperature-independent charge carrier mobility. Careful adjustment of the individual oxide layer thicknesses in devices composed of stacked layers plays a vital role in the reduction of trap states, both interfacial and bulk, which consequently deteriorates the overall device performance. The findings enable an improved understanding of the correlation between TFT performance and the respective thin-film composition in ALD-based heterostructure oxides. KEYWORDS: indium oxide, zinc oxide, thin-film transistor, atomic layer deposition, heterostructure

1. INTRODUCTION Main group metal oxides have received intense attention based on their broad field of applications in areas like, for example, transparent electronics, thermoelectrics, or photonics.1 As such, large efforts have been dedicated toward the exploitation of ZnO-based transparent conductive oxides and their implementation as electrode materials with favorable properties, which allow the tunability of their electronic properties, such as high charge carrier mobility and optical transparency.2,3 Besides their application as passive materials, metal oxides have been largely explored as active semiconductor layers due to their feasibility for next-generation thin-film transistors (TFTs) in large-area electronics. This relies on the fact that they show enormous potential to surpass performances typically achieved by conventional low-temperature polysilicon-based technology even after fabrication at room temperature or postdeposition processing at low temperature.4,5 To this end, research in the © XXXX American Chemical Society

past decade has largely been focused on amorphous ternary and quaternary metal oxide systems like indium zinc oxide (IZO), zinc tin oxide (ZTO), indium gallium oxide (IGO), gallium tin oxide (GTO), or indium gallium zinc oxide (IGZO), demonstrating remarkable TFT performances.6−9 However, in recent years, directed efforts have been dedicated toward the generation of materials based on heterostructured stacks comprising different individual metal oxide layers.10−15 Although this concept is well established for nitride-based semiconductors and dielectrics, investigations pertaining to the use of such multilayered heterostructured oxides are still in their nascent stages and remain largely unexplored.16,17 Recent investigations on heterostructure oxide TFTs by Anthopoulos Received: February 26, 2018 Accepted: May 7, 2018

A

DOI: 10.1021/acsami.8b03322 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

conduction mechanism is investigated, based on compositional changes in the stacked heterostructures. The reported findings provide a rational tool to optimize the performance of ALDbased multilayered heterostructure semiconductor stacks with an attractively low thermal budget of 200 °C.

and co-workers, via solution-processed deposition of layered crystalline In2O 3/ZnO films, displayed superior device mobilities when compared to that of their single-oxide counterparts. Their significantly improved performance was attributed to the formation of a two-dimensional electron gas (2 DEG) region at the heterointerface between the two individual oxides, indicated by the temperature-independent behavior of the device mobility.10,18,19 Under an applied field, such 2 DEG transport plays a crucial role, wherein the electron density is largely confined at this critical heterointerface.15,18,19 This phenomenon is well reflected in the inherent charge carrier conduction mechanism, which is mainly governed by the percolation conduction (PC) mechanism and thus shows a fundamentally different behavior compared to that of singlelayer oxide transistors, which are based on trap-limited conduction (TLC). Owing to these significant changes in the conduction mechanism, heterostructured semiconductors enable a bandlike electron transport, resulting in superior TFT performance parameters.19,20 The crucial step in controlling the heterointerface relies on well-defined oxide layers and precisely controlled layer deposition. This control is largely dependent on the fabrication process, which is able to tailor the desired compositional conditions and morphology of the deposited films close to the heterointerface. Such precise depositions can generally be achieved by sophisticated techniques like molecular-beam epitaxy or pulsed laser deposition to ensure high-quality heterointerfaces.11−13,21 It is noteworthy that such heterostructured high-performance TFTs can even be achieved by controlled solution-based processes as well.18,19,22 However, atomic layer deposition (ALD) offers a unique alternative to obtain a controlled, tailored composition of active materials over complex substrate geometries and structured scaffolds.23,24 Surprisingly, reports on ALD-based heterostructured oxides are rather scarce to date, considering the fact that it offers a feasible strategy and high potential toward the fabrication of such lowdimensional multilayered structures.25 It facilitates conformal material deposition with a high degree of reproducibility of thin layered films with a superior control over film thickness due to its intrinsic reaction-controlled film deposition. This advantage allows the possibility of precise deposition over large-area substrates, either planar or three-dimensional structured with precise thickness control at angstrom levels, which is enabled by surface-limited reactions of the molecular precursors employed.25−27 In this context, heterostructure devices based on aluminum oxide/zinc oxide and aluminum oxide/indium oxide compositions have been demonstrated, showing considerable evidence of tunability of their properties in favor of highperformance semiconductor/conductor behavior.14,28 Thus, the ALD technique offers its largely underutilized potential to take advantage of such phenomena arising at the heterointerface and enabling their fine-tuning on a more or less atomic scale.14,28,29 In this realm, we have recently reported on multilayered heterostructures obtained by ALD, consisting of alternating layers of In2O3 and ZnO.30 Herein, we now report a strategy toward significantly enhancing the overall TFT device performance of a multilayered In2O3/ZnO structure by employing an ALD-based process. This approach is primarily aimed toward understanding the influence of the design of the stack composition on the resulting TFT performance by tuning individual layer thicknesses of the deposited oxides within the heterostructure stack. In addition, the temperature-dependent behavior of the field-effect mobility as well as the nature of the charge carrier

2. EXPERIMENTAL SECTION 2.1. ALD Process. In2O3/ZnO heterostructures with different compositions were generated in a Savannah S 100 system (Cambridge Ultratech). All depositions were carried out at 200 °C. Prior to deposition, the substrates were kept at 200 °C in the ALD chamber for 20 min under an argon flow of 20 sccm (background pressure 0.8 Torr). The metal sources for the thin-film depositions were trimethyl indium 98+% (99.9% In, Strem Chemicals) and diethyl zinc (min. 95%, Strem Chemicals) for In2O3 and ZnO, respectively. Water (highperformance liquid chromatography (HPLC) grade, Sigma-Aldrich) was used as an oxidant. All precursors as well as the oxidant were kept at room temperature. Argon (99.9999%, α Gaz) was used as a carrier gas and set to a flow rate of 20 sccm during the depositions. To vary the composition of the heterostructures, an ALD supercycle procedure was employed, where the cycle ratio of In in fraction of the total ALD cycles was modified between 0.8 and 0.5. The supercycles were repeated three times to achieve a desirable stack architecture. The detailed deposition parameters for the ALD process have been reported previously.30 2.2. Material Characterization. UV−vis measurements were performed on quartz substrates (Thermo Scientific-Evolution 600). Ellipsometry (Accurion EP3 System) at a wavelength of 632.8 nm was used to determine the respective thicknesses of the thin films. Transmission electron microscopy (TEM) and energy-dispersive spectroscopy (EDS) analyses were performed on an FEI Titan G2 80−200 Crewley instrument.31 Samples for TEM investigation were directly prepared by the thin-film deposition on TFT substrates (15 × 15 mm2) and annealed at 200 °C for 2 h. Further processing included focused ion beam (FIB) preparation with a gallium-focused ion beam (FEI Helios-400) and subsequent coating with a platinum layer.32 To investigate the roughness and texture of the heterostructures, atomic force microscopy (AFM) was conducted on an MFP-3D (Asylum Research) system equipped with silicon cantilevers. X-ray photoelectron spectroscopy (XPS) measurements were carried out on a KAlpha XPS system (Thermo Fisher Scientific, East Grinstead, U.K.) with a microfocused, monochromated Al Kα X-ray source, with all spectra referenced to the C 1s peak of hydrocarbons at 285.0 eV. The measurement inaccuracy was ±0.1 eV. 2.3. Thin-Film Transistor Characterization. For TFT characterization, commercial substrates (Fraunhofer IMPS, Dresden) were used. The substrates consist of a highly n-doped silicon that represents the gate electrode with silicon oxide (90 nm) as the dielectric. Source− drain electrodes are employed as an interdigital structure in a bottomgate-bottom-contact architecture and gold electrodes (40 nm) with an indium tin oxide adhesion layer (10 nm). To avoid an overestimation of mobility, a channel length of L = 20 μm and a channel width of W = 10 mm constituting a high width-to-length ratio (W/L) of 500 was used.33,34 For each stack variation, 12 devices were measured. The general cleaning procedure for the substrates comprised a sequence of ultrasonication for 10 min in acetone, deionized water, and isopropanol (all HPLC grade). Prior to TFT characterization, all thin films were annealed at 200 °C, after deposition for 2 h. TFT characterization was carried out in an inert and dark environment in a glovebox using an HP 4155A Semiconductor Parameter Analyzer (Agilent). Low-temperature measurements of the TFT characteristics were obtained via cooling with liquid nitrogen in the temperature range between room temperature and 78 K, where the samples were maintained for 30 min at the respective temperatures. Long-term stability tests were performed on nonencapsulated samples, stored for 6 months in ambient atmosphere under dark conditions. Gate bias stress measurements were performed on aged samples under constant stress voltages of VGS = 20 V (positive bias stress) and VGS = −20 V (negative bias stress) for 120 min. The stress voltage was interrupted B

DOI: 10.1021/acsami.8b03322 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Scheme 1. Schematic Representation of the Number of Deposited In2O3/ZnO Heterostructure Stacks Deposited by ALD for Application as Active Semiconductor Layers in Thin-Film Transistors

in intervals of 20 min to obtain the TFT characteristics. To evaluate the device stability under light, aged TFTs were illuminated with lightemitting diodes (LEDs; Roithner Laser Technik GmbH) of different wavelengths from the visible to UV region. The LEDs were calibrated to achieve an intensity of 1 mW cm−2. The charge carrier mobilities, μ, were determined from the transfer characteristics in the saturation regime of the transistor device. The threshold voltage, Vth, was extracted from a linear fit of the square root of the source−drain current, IDS, versus the gate−source voltage, VGS.

Table 1. Summary of the Deposition Rates Expected from the Rule of Mixture for Different In Cycle Ratiosa stack In/Zn In/Zn In/Zn In/Zn a

3. RESULTS AND DISCUSSION To fabricate the desired In2O3/ZnO heterostructure stacks investigated herein, a supercycle process, utilizing an alternating deposition of In2O3 and ZnO layers, was employed.30 To change the overall composition of the heterostructures, the In cycle ratio was varied between 0.5 and 0.8 in fraction of the total number of ALD cycles. A typical supercycle utilizing an In precursor cycle ratio of 0.5 constitutes five cycles of In2O3, followed by a subsequent deposition of five cycles of Zn precursor or multiples of that. The number of supercycle iterations was kept constant at three. A schematic of the generated heterostructures is shown below (Scheme 1) with the actual ALD cycle ratios employed. An increase in the number of In cycles is accompanied by an increased thickness of the resulting In2O3 layers and a simultaneous decrease of the ZnO-layer thickness incorporated in the stack. The heterostructure with a cycle ratio of In/Zn 28:20 was intentionally generated to study the role of ZnO-layer thickness, resulting in an In cycle ratio of 0.58. It must be noted that a variation of the In cycle ratio consequently changes the overall film thickness due to the different deposition rates of the individual oxides (0.4 and 1.45 A per cycle for In2O3 and ZnO, respectively), as indicated in the schematic representation. The actual deposition rates of the heterostructure stacks are lower than the expected growth rates from the rule of mixture, where the corresponding values determined are summarized below (Table 1). This nature of metal oxide growth in the ALD process has been observed previously and was attributed to a retarded adsorption of precursors on metal oxide surfaces as well as changes in surface chemistry during deposition, leading to a reduction of reaction sites on the surface.23,35 To evaluate the levels of homogeneity and uniformity and to assess the crystalline nature of the deposited thin films, a crosssectional TEM investigation was performed. The bright field STEM image obtained for In/Zn 28:12 reveal excellent film homogeneity with a uniform thickness (Figure 1a), owing to the conformal deposition process. Additionally, the highresolution image (Figure 1b) reveals the crystalline nature of

20:20 28:20 28:12 32:8

rule of mixture 0.92 0.83 0.71 0.61

deposition rate (A per cycle)

In cycle ratio

± ± ± ±

0.5 0.58 0.7 0.8

0.65 0.61 0.52 0.44

0.01 0.01 0.01 0.01

thickness (nm) 7.8 8.8 6.2 5.3

± ± ± ±

0.08 0.07 0.05 0.07

Actual deposition rates were determined by ellipsometry.

the deposited layer. Clearly, lattice planes with a distance of 0.28 nm can be seen extending parallel to the interface/surface. Furthermore, the elemental composition of the thin film was investigated via EDS. The line scan (Figure 1c) exhibits three alternating distinct maxima and minima for the elemental distributions of In and Zn, thus verifying the heterostructurebased design and the number of individual oxide layers to obtain the final multistack heterostructure. To assess the morphological evolution and the resulting film quality based on the compositional variations of the heterostructures, atomic force microscopy (AFM) studies were carried out on the generated In 2 O 3 /ZnO stack architectures after postdeposition annealing at 200 °C. The AFM micrographs exhibit a systematic decrease of the surface roughness with higher In cycle ratios, which is well displayed by the obtained RRMS values, where the stack comprising the lowest In ratio (In/Zn 20:20) exhibits a highly textured surface (RRMS ∼ 0.32 nm). A further increase of the In cycle ratio results in a gradual reduction of the surface roughness for In/ Zn 28:12 (RRMS ∼ 0.26 nm) and 32:8 (RRMS ∼ 0.22 nm) (Figure 2). Interestingly, the heterostructure incorporating 28 cycles of In2O3 with an increased number of cycles of its ZnO counterpart (In/Zn 28:20) displays a nearly similar morphology, accompanied only by a slightly increased RRMS value of 0.28 nm. To obtain insight into the optical properties of the thin films, UV−vis measurements were performed on the heterostructures with varied metal oxide compositions. All heterostructures investigated herein exhibit good optical transparency (>80%) in the visible region (Figure S1a). Further Tauc analysis of the investigated films reveals a widening of the optical band gap for layers with higher In/Zn (28:12 and 32:8) ratios (Figure S1b). The observation in the latter is attributed to the Burstein−Moss effect that occurs from the transition of electrons from the valence band states to energy levels localized beyond the conduction band minimum (CBM), indicating that the energy C

DOI: 10.1021/acsami.8b03322 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 1. (a) STEM bright field image displaying the alternating layers of the deposited film, (b) magnified STEM image of (a) showing the full crystallinity, (c) EDS-line scan obtained for the alternating heterostructure composition In/Zn 28:12, annealed at 200°C for 2 hrs. (Dashed lines serve as guide for the eyes).

this regard, X-ray photoelectron spectroscopy (XPS) was performed for heterostructure stacks with varying indium and zinc layer thicknesses, corresponding to the number of deposition cycles (Figure 3). The films investigated herein consist of four stack variations as stated earlier, that is, In/Zn of 20:20, 28:12, 32:8, and 28:20, cf. Scheme 1. All of the samples were subjected to thermal annealing for 2 h at 200 °C in ambient atmosphere. The obtained O 1s core spectra for all of the given stacks primarily comprise two peaks at ∼530.0 and ∼532.0 eV, where the former is assigned to oxygen originating from well-coordinated metal−oxygen (M−O) bonds and the latter is assigned to hydroxyl species present on the surface as well as within the volume of the thin film.42,43 Because the TFTs generally fabricated from stacks with higher indium content exhibited an improved TFT performance, it is interesting to observe the changes in the O 1s spectra related to the fabricated stacks where the zinc concentration is significantly varied. In all four stack variations, the relative M−O content is much higher for stacks with comparatively lower zinc contents (Table 2). This is clearly observed for stacks with In/Zn of 28:12, 32:8, and 28:20 (Figure 3b−d). On the contrary, when the zinc layer thickness is increased, the contributions from M−OH are rather higher than those from M−O, as is evident for In/Zn of 20:20 (Figure 3a). These observations serve as a primary indication that the source of defect-related species within the different stack variations is significantly sensitive to the increase in the zinc layer thickness, based on the precursor and deposition conditions employed. Systematic quantifications of the oxygen-related species from the O 1s core spectra show that the amount of species related

Figure 2. AFM micrographs and the corresponding root-mean-square roughness (RRMS) values for different heterostructure stack compositions comprising In/Zn (a) 20:20, (b) 28:20, (c) 28:12, and (d) 32:8.

levels in the conduction band minimum of In2O3 are prefilled by free electrons.36−38 Changes in the nature of the compositional species within metal oxides are critical as far as active TFT performance is concerned.34,39 More importantly, the change in the nature of the oxygen-related species has been known to have significant influence on the critical TFT performance parameters.40,41 In D

DOI: 10.1021/acsami.8b03322 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 3. Deconvoluted oxygen O 1s core spectra for the different heterostructure stack variations of In/Zn of (a) 20:20, (b) 28:20, (c) 28:12, and (d) 32:8. All of the samples were annealed to 200 °C for 2 h.

ZnO is incorporated in the stack architecture, with the most distinct shifts being observed for those stacks comprising lower ZnO contents (Figure S3b). However, the observed spectral shifts are of a minor extent, which makes it difficult to base a discussion regarding changes in the chemical environment solely on the present XPS investigations. Nevertheless, a similar trend was observed for solution-processed In2O3/ZnO heterostructures, where the spectral shifts were attributed to an improved oxidation, correlated to the passivation of oxygen vacancy defects and the potential interaction of oxygen-related species at the heterointerface between the metal oxides.19,22 All of the aforementioned In2O3/ZnO heterostructures were incorporated in a TFT device to evaluate its active semiconductor performance. All depositions were carried out with an initial deposition of an In2O3 layer at the dielectric− semiconductor interface, followed by the subsequent deposition of a ZnO layer and two additional iterations to achieve the desired heterostructure. This approach of employing a highly conductive oxide forming the semiconductor−dielectric interface is well established to realize an improved device performance, whereby the enhanced TFT performance is attributed to a better passivation of interfacial trap states as well as increased conductivity at the dielectric−semiconductor interface.30,44−46 In our previous work, we have demonstrated that TFTs fabricated solely from indium oxide exhibit a highly conductive behavior with a current on/off ratio close to 1.30 Hence, the use of indium oxide as the initial interfacial layer

Table 2. Atomic Concentrations of Oxygen (M−O) and Hydroxyl Species (M−OH) Derived from the Deconvoluted XPS O 1s Spectra for Different In2O3/ZnO Heterostructures stack In/Zn In/Zn In/Zn In/Zn

M−O (atom %)

M−OH (atom %)

47.7 51.1 55.2 56.4

52.3 48.9 44.8 43.6

20:20 28:20 28:12 32:8

to the M−OH signals is significantly high (52.3 atom %) for the films with the relatively highest ZnO content (In/Zn of 20:20) and was systematically lowered (43.6 atom %) for the stack with the least ZnO content. This was further confirmed by investigating the O 1s spectra of the individual In2O3 and ZnO thin films with thicknesses similar to those of the heterostructure stacks (Figure S2a,b). In this case, for the individual ZnO film, the M−OH contributions (49.1 atom %) are considerably higher than those of the individual In2O3 films (36.8 atom %), clarifying the intrinsic chemical nature of the individual oxides; this is in good agreement with the changes occurring in the stacks with increasing ZnO content. Moreover, the core In 3d5/2 and Zn 2p3/2 spectra of the individual oxides and the different heterostructure stacks were investigated. Interestingly, whereas the In 3d5/2 peak position remains mostly unchanged (Figure S3a), the Zn 2p3/2 position exhibits a spectral shift toward higher binding energy when

Table 3. TFT Performance Parameters and Their Corresponding Standard Deviations Derived for the Heterostructures with Compositional In/Zn Ratio Variations stack

mobility, μsat. (cm2 V−1 s−1)

In/Zn In/Zn In/Zn In/Zn

± ± ± ±

20:20 28:20 28:12 32:8

3.1 6.5 9.3 12.2

0.25 0.12 0.13 0.19

on-voltage, Von (V) −2.2 −2.3 0.0 −1.7

± ± ± ±

0.25 0.11 0.12 0.22

threshold voltage, Vth (V) 8.0 5.1 2.1 −1.4

± ± ± ±

E

0.21 0.14 0.22 0.31

current on/off ratio, Ion/Ioff 2.5 4.7 5.3 6.1

(±2.1) (±1.9) (±2.5) (±1.8)

× × × ×

7

10 106 109 107

subthreshold slope, SS (mV dec−1) 621 565 162 442

± ± ± ±

31.4 18.9 9.1 21.5

DOI: 10.1021/acsami.8b03322 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX

Research Article

ACS Applied Materials & Interfaces

Figure 4. (a) Collective transfer characteristics of the heterostructure stacks with different In/Zn variations and output characteristics for the corresponding stacks with In/Zn of (b) 20:20, (c) 28:12 (d) 28:12, and (e) 32:8.

adjusting the thickness of the individual layers, wherein the thickness of the second layer can be adjusted to counter the detrimental effects of the former, thereby achieving an overall improved transistor performance, without significantly compromising other performance metrics. At this point, it is worth mentioning that the overall thickness of the active semiconductor is crucial to obtain an optimized TFT performance. As such, for sputter-based amorphous IZO, Barquinha et al. have shown that for ultrathin films (