Mesostructured HfxAlyO2 Thin Films as Reliable and Robust Gate Dielectrics with Tunable Dielectric Constants for High-Performance Graphene-Based Transistors Yunseong Lee,†,‡,§ Woojin Jeon,†,§ Yeonchoo Cho,† Min-Hyun Lee,† Seong-Jun Jeong,*,† Jongsun Park,‡ and Seongjun Park*,† †
Device Lab., Device & System Research Center, Samsung Advanced Institute of Technology, Suwon 16678, Republic of Korea School of Electrical Engineering, Korea University, Seoul 136-713, Republic of Korea
‡
S Supporting Information *
ABSTRACT: We introduce a reliable and robust gate dielectric material with tunable dielectric constants based on a mesostructured HfxAlyO2 film. The ultrathin mesostructured HfxAlyO2 film is deposited on graphene via a physisorbed-precursor-assisted atomic layer deposition process and consists of an intermediate state with small crystallized parts in an amorphous matrix. Crystal phase engineering using Al dopant is employed to achieve HfO2 phase transitions, which produce the crystallized part of the mesostructured HfxAlyO2 film. The effects of various Al doping concentrations are examined, and an enhanced dielectric constant of ∼25 is obtained. Further, the leakage current is suppressed (∼10−8 A/cm2) and the dielectric breakdown properties are enhanced (breakdown field: ∼7 MV/cm) by the partially remaining amorphous matrix. We believe that this contribution is theoretically and practically relevant because excellent gate dielectric performance is obtained. In addition, an array of top-gated metal−insulator−graphene field-effect transistors is fabricated on a 6 in. wafer, yielding a capacitance equivalent oxide thickness of less than 1 nm (0.78 nm). This low capacitance equivalent oxide thickness has important implications for the incorporation of graphene into high-performance silicon-based nanoelectronics. KEYWORDS: graphene, mesostructure, capacitance equivalent oxide thickness, phase transition engineering, atomic layer deposition
T
overcome in order to achieve successful incorporation of graphene into high-performance silicon-based nanoelectronics.11−15 With a view to overcoming this challenge, we recently reported several surface-engineering techniques, such as ebeam-evaporated Hf seed layers,16 and physisorbed-precursorassisted atomic layer deposition (pALD). These techniques have facilitated conformal and uniform deposition of various high-k dielectric films, including Al2O3 and HfO2, on graphene surfaces in a wafer-scale system. However, even the highestquality dielectric films obtained using these previously reported methods exhibited CET values of approximately 1 nm (Hf seed-HfO2 and pALD HfO2 CETs of 1.5 and 1.03 nm,
wo-dimensional (2D) materials are attracting considerable interest across the academic and industrial communities, because of their unusual and fascinating electrical, optical, thermal, and mechanical properties.1−3 In particular, graphene, which is a representative 2D material, has stimulated a considerable amount of study. This material has significant potential as a high-speed channel material in various silicon-based nanoelectronics, because of the associated physical phenomena and its high charge-carrier mobility at room temperature.4−8 Despite these advantages, however, a graphene surface is chemically inert to atomic layer deposition (ALD) precursor molecules.9,10 This characteristic obstructs the integration of high-dielectric-constant (high-k) materials as conformal and pinhole-free ultrathin films, which would ensure excellent electrical performance on the surface. Thus, the downscaling of the capacitance equivalent oxide thickness (CET) of a gate dielectric film to below 1 nm on a graphene surface is one of the most fundamental challenges that must be © 2016 American Chemical Society
Received: March 11, 2016 Accepted: June 29, 2016 Published: June 29, 2016 6659
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
www.acsnano.org
Article
ACS Nano
to control the k values of these films via phase transformation. To predict the relative phase stability for a HfxAlyO2 film with phase transition engineering, first-principles simulations based on density functional theory (DFT) were performed. Note that we considered (i) all of the possible transformed phases, including the monoclinic, tetragonal, and cubic HfO2 phases, and (ii) the formation of oxygen vacancies into HfO2. The energy differences between the phases and the correlated unit cells for the minimum energy configurations of these phases are presented in Figure 1. The energy differences between the
respectively). Therefore, to achieve sub-1 nm CET, ultimately, a gate dielectric film must be downscaled using a material with higher k than Al2O3 and HfO2, while maintaining an extremely low leakage current. Phase transition engineering of a HfO2 film, which can exhibit different k values via allotropic phase transformation from the monoclinic to tetragonal or cubic phases, is considered to be the most promising technique for increasing the k values of dielectric materials (monoclinic, tetragonal, and cubic k = 17, 70, and 29, respectively).17−23 However, the crucial inherent limitation of ALD HfO2 film is that the tetragonal and cubic phases are thermodynamically stable at temperatures significantly higher than the ALD process temperature only (1720 and 2600 °C for the monoclinic-totetragonal and tetragonal-to-cubic transitions, respectively). Meanwhile, the monoclinic phase with lower k value appears at room temperature. In this regard, several phase-transitionengineering techniques, such as cation doping,18−20,23 carbon impurity incorporation,21 and the production of O-deficient phases,22 have been suggested as a means of inducing the higher-k-value phases at the ALD process temperature. Among them, cation doping, which allows phase transformation to occur by decreasing the difference in the volume-free energies, is of considerable interest, because it may constitute a reliable approach to the production of dielectric capacitor materials. Nonetheless, in gate dielectric materials with commercially viable thicknesses of less than 5 nm, the grain boundaries must be eliminated in order to maintain an extremely low leakage current, which remains a fundamental challenge. In addition, for large-scale device integration, CET downscaling must be realized on 2D materials grown via chemical vapor deposition (CVD), which is primarily relevant to large-scale device fabrication. In this work, we introduce an attractive gate dielectric material based on a mesostructured HfxAlyO2 thin film, which has a grain-boundary-free structure and superior k values on CVD-grown monolayer graphene and which has been tailored to overcome the aforementioned problems. The synergistic ALD process is employed to ensure the production of an extremely reliable and robust ultrathin mesostructured HfxAlyO2 thin film on the CVD-grown graphene surface, and the k value of this thin film can be enhanced significantly (to 25), through the intermediates of the tetragonal or cubic phases in an amorphous matrix (cf., HfO2 and Al2O3 k = 17 and 8, respectively, in this work). More significantly, the partially remaining amorphous matrix suppresses the leakage current and enhances the dielectric breakdown properties by eliminating the grain boundaries and their network, which are regarded as the most probable leakage current paths. Thus, excellent gate dielectric performance is obtained, with a high breakdown field of ∼7 MV/cm and a reasonably low leakage current (∼10−8 A/cm2). Further, in this work, we demonstrate a CET of 0.78 nm for an array of top-gated metal−insulator− graphene field-effect transistors (MIG FETs) on a 6 in. wafer.
Figure 1. Relative stabilities between phases for different Al doping concentrations. (a) Difference in energy relative to that of the monoclinic phase as Al doping concentration is varied. O vacancy formation is assumed. Unit cells for minimum energy configurations of (b) single-Al-doped HfO2 with monoclinic phase, (c) triple-Al-doped HfO2 with tetragonal phase, and (d) quadruple-Aldoped HfO2 with cubic phase. (a, inset) Hf Voroni coordination number as a function of Al concentration.
monoclinic and tetragonal or cubic phases in pure HfO2 film (Al doping concentration: 0%) were 0.17 and 0.27 eV, respectively, indicating that the monoclinic phase is the most stable in a pure HfO2 film. However, the energy differences decreased monotonically with increased Al concentration; thus, we confirmed that the stability of the tetragonal and cubic phases increased once a certain Al doping concentration was exceeded (∼7.9% and ∼9.3%, respectively), which corresponded to significant enhancement of the dielectric properties of the HfxAlyO2 film. Interestingly, this stabilization of the tetragonal and cubic phases through phase transition engineering was also affected by the coordination number, which is the number of nearest neighbors in all directions of an atom in a crystal lattice. The inset of Figure 1a presents the Hf Voronoi coordination number24 versus the Al doping concentration. The average values for all phases became similar at Al concentrations of ∼9.3% and higher. This implies that, although there are diverse phases in the HfO2 film, the local chemical boding environments of these phases are similar above a certain Al concentration. In this regard, in order to induce the tetragonal or cubic phase (with a higher k value), a deposition process with controllable Al doping concentration should be designed,
RESULTS AND DISCUSSION The gate dielectric materials based on a mesostructured HfxAlyO2 thin film were produced using (i) the method introduced in our previous work, i.e., the pALD process, which is associated with the conformal and uniform deposition of ultrathin metal oxide films on an inert graphene surface without causing damage to or contamination of the underlying graphene layer, and (ii) phase transition engineering designed 6660
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano
Figure 2. Schematic diagrams and cross-sectional TEM images of (a) HfO2, (b) Al2O3, and (c) HfxAlyO2 films deposited on graphene/SiO2 substrates. Variations in (d) dielectric constant and (e) leakage current density at 1 MV/cm (blue) and breakdown electric field (red), as functions of ALD cycle ratio between HfO2 and Al2O3. All electrical data were obtained from the MIG capacitor devices in the as-fabricated state.
and the Al content of the HfxAlyO2 film should be more than ∼7.9%. In this work, to prepare mesostructured HfxAlyO2 thin films on monolayered graphene in accordance with the results predicted via DFT, we aimed to obtain an intermediate of the tetragonal phase transformation for pALD HfO2 amorphous films through the addition of pALD Al2O3 at an extremely low process temperature of 50 °C, with no postannealing process. In particular, to obtain the maximum k value, the Al concentration in the HfxAlyO2 films was controlled by tuning the cycle ratio of the pALD Al2O3 and HfO2 processes on the CVD-grown graphene layer (which was transferred onto a 6 in. Si wafer covered with a 300-nm-thick thermally grown SiO2 film). The prepared pALD HfO2, Al2O3, and HfxAlyO2 thin films were conformally distributed over the entire graphene surface with no pinholes and were approximately 5 nm thick. These ultrathin films were robust and exhibited high density (ρo) values, as confirmed by cross-sectional transmission electron microscopy (TEM) images (Figures 2a−c) and medium-energy ion scattering (MEIS) analysis (see Table S1; HfO2, Al2O3, and HfxAlyO2 ρo = 9.30, 3.06, and 8.73 g/cm3, respectively). Further, the Al concentration in the HfxAlyO2 films ([Al]/([Al] + [Hf])) was well and uniformly controllable with an Al cycle ratio up to ∼22.7%, according to the X-ray photoelectron spectroscopy (XPS) and energy-dispersive X-ray spectroscopy (EDX) with TEM analysis results shown in Figure S1. These results suggest that, even for a low temperature of 50 °C, the pALD HfxAlyO2 deposition process yields stable film deposition on various heterogeneous
materials, including graphene, Al2O3, and HfO2, during each pALD HfO2 and Al2O3 process sequence. Details of the experimental procedure are given in the Experimental Procedure section. To evaluate the fundamental electrical properties of the HfxAlyO2 films on monolayered graphene as a gate dielectric material, metal−insulator−graphene (MIG) capacitor structures incorporating the various dielectric films were fabricated. First, as shown in Figure 2d, the capacitance-derived k-value evolution in accordance with various Al doping concentrations in the HfO2 was confirmed. Essentially, for pure HfO2/ graphene/SiO2 and Al2O3/graphene/SiO2 samples, the k values were ∼17 and ∼8, respectively. These values were in reasonable agreement with the k values of HfO2 with monoclinic phase (∼18) and Al2O3 with amorphous phase (∼9) on a Si substrate, which are the most likely phases for the ALD deposition temperature.20 Meanwhile, in the case of the HfxAlyO2/ graphene/SiO2 samples, when the Al doping concentration was increased to ∼9.5%, there was a significant increase in the k values. The maximum k value was 25, which was 1.5 times higher than that of the pure HfO2/graphene/SiO2 sample. This increment is well consistent with previously reported results for the typical phase transition engineering technique, in which Al is doped into HfO2 and an additional annealing process is employed to obtain crystallization.20,23 The highest k value observed in this study, however, was slightly lower than the k value (k ≈ 50) previously reported for the above approach. The difference in the k value for the HfxAlyO2/graphene/SiO2 sample will be discussed below. When the Al concentration 6661
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano
Figure 3. GAXRD patterns of HfO2, Al2O3, and HfxAlyO2(9.5% Al) films deposited on a graphene/SiO2 substrate (a) before and (b) after postdeposited annealing. Postdeposited annealing was performed using rapid thermal annealing at 700 °C for 3 min in ultrahigh vacuum. XPS core-level spectra of Hf 4f for as-deposited (c) HfO2 and (d) HfxAlyO2, respectively. Peak deconvolution was conducted with 4f7/2 spectra (right-hand-side peaks).
was further increased to ∼22.7%, the k value rapidly dropped to 15; however, it was confirmed above that a cubic phase with a higher k value has greater stability at Al concentrations of more than ∼10% (in the calculation results shown in Figure 1). This discrepancy between the predicted and measured behavior is most likely due to the degradation of the crystalline structure by the quite large amount of Al dopant.23 Second, the leakage current properties for the various Al doping concentrations of the HfxAlyO2/graphene/SiO2 samples were examined (Figure 2e). As the pALD Al2O3 cycle ratio increased, the leakage current density (at 1 MV/cm) gradually decreased; further, the dielectric breakdown electric field (E field) was enhanced. The results coincide well with the fact that an increase in the concentration of a high-band-gap (Eg) material (such as Al2O3) in a film induces suppression of the leakage current (Al2O3 and HfO2 Eg = 9 and 6 eV, respectively).25 Interestingly, the leakage current level was lower than those previously reported for HfO2 or HfxAlyO2 thin films on a Si substrate, considering the low film thickness of only 5 nm. To characterize the possible origins of the aforementioned superior electrical properties of the HfxAlyO2 films, grazing angle X-ray diffraction (GAXRD) was conducted on three samples (HfO2/graphene, Al2O3/graphene, and HfxAlyO2(9.5% Al)/graphene on SiO2) before and after postannealing (Figure 3a and b, respectively). GAXRD is a useful technique for the estimation of thin-film crystallinity with minimum contribution from the substrate. As shown in Figure 3a, no notable diffraction peaks were apparent for the Al2O3/graphene/SiO2 sample, which implies that the Al2O3 film was completely composed of amorphous phase material. Further, for the as-
deposited HfO2/graphene and HfxAlyO2(9.5% Al)/graphene samples on SiO2, quite broad peaks of approximately 25−40° were observed in the 2θ region, suggesting the presence of a poor mesostructured thin film, which is a partially crystallized structure. The possible origins of the broad peaks of these films could not be determined, because the primary diffraction peaks of the possible phases (including the monoclinic, tetragonal, and cubic phases) are located within the same 2θ region [monoclinic (111) phase peaks at 28.3° and 31.7° and tetragonal (011) and cubic (111) phase peaks at 30.3°]. Therefore, in order to clearly reveal the differences between the origins of these peaks, postannealing was conducted at 700 °C for 3 min and crystalline structures of two of the samples were then confirmed, as shown in Figure 3b. As expected, the crystalline HfO2/graphene/SiO2 sample exhibited diffraction peaks corresponding to the monoclinic phase only. In contrast, the crystalline HfxAlyO2(9.5% Al)/graphene/SiO2 sample exhibited a weak diffraction peak at 36° and two dominant diffraction peaks at 30.3° and 51.0°, which were assigned to the monoclinic (200), tetragonal (011) or cubic (111), and tetragonal (020) or cubic (002) phases, respectively. Thus, the crystalline HfxAlyO2(9.5% Al) film was primarily composed of the tetragonal or cubic phase. The results imply the following: (i) The mesostructured HfO2 film on graphene/SiO2 comprised a mixture of the amorphous and monoclinic phases, whereas the mesostructured HfxAlyO2(9.5% Al) film on graphene/SiO2 comprised the tetragonal or cubic phases in an amorphous matrix; (ii) the enhancement of the k value was due to the tetragonal or cubic phase formation in the mesostructured HfxAlyO2(9.5% Al) films. 6662
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano Further evidence of the mesocrystalline structure of the HfxAlyO3 film was revealed via chemical state analysis using XPS measurements. Figure 3c and d show the collected XPS spectra of the Hf 4f core levels in pure HfO2/graphene/SiO2 and HfxAlyO2(9.5% Al)/graphene/SiO2 samples without postannealing, respectively. Note that the peak position was recalibrated based on the C−C bond peak position, considering a binding energy of 284.5 eV. We also assumed that the deconvoluted Hf 4f peaks were composed of amorphous (17.4 eV), monoclinic (17.9 eV), and tetragonal or cubic (18.1 eV) phase Hf 4f7/2 peaks. As shown in Figure 3c, for the pure HfO2 film on graphene/SiO2, main and minor Hf 4f7/2 peaks appear at 17.4 and 17.9 eV, respectively, suggesting that the film had a mixed structure featuring amorphous and monoclinic phases. In contrast, for the HfxAlyO2(9.5% Al) film on graphene/SiO2 (Figure 3d), an additional peak located at approximately 18.1 eV was identified, which can be assigned to the tetragonal or cubic phase. Hence, we can postulate that the phase transition engineering technique involving addition of pALD Al2O3 allows a “mesocrystalline-structured HfxAlyO2 film” with a partially incorporated tetragonal or cubic phase in an amorphous matrix to be obtained on graphene. Thus, the formation of the mesostructured HfxAlyO2 film is the main origin of the abovementioned enhancement of the electrical properties, including the higher k value, lower leakage current, and superior breakdown E field. We note that after postannealing the ratios of the peaks corresponding to the phases increased, as shown in Figure S4. Thus, the HfO2 film on graphene/SiO2 was fully crystallized to monoclinic phase, whereas the amorphous phase remained in the HfxAlyO2(9.5% Al) film on graphene/SiO2. The effect of the remaining amorphous phase in the crystalline HfxAlyO2(9.5% Al) film on graphene/SiO2 on the material properties will be discussed below. The crystal structure of the mesocrystalline structured HfxAlyO2 thin film was also examined using high-resolution transmission electron microscopy (HRTEM) (see Figure S2). The crystal structure agrees well with the results obtained from the XRD and XPS analyses. To reveal the significant benefits of the mesocrystallinestructured HfxAlyO2 film as a gate dielectric material in graphene-based devices, Raman spectroscopy was performed (Figure 4a). Raman spectroscopy provides information on molecular vibrations that can be used to inspect the graphene integrity and its physicochemical properties in accordance with the adjacent layers. In the obtained Raman spectra, pristine graphene (black line in Figure 4a) exhibits three distinctive G (1586 cm−1), D (1350 cm−1), and 2D (2684 cm−1) peaks and exhibits a typical spectrum of high-quality monolayered graphene, i.e., a 2D peak (∼35 cm−1) with a sharp full width at half-maximum (fwhm) and higher intensity (I2D) than that of the G peak (IG), with negligible D peak intensity (ID/IG < 0.1). For the mesostructured HfO 2 /graphene and the HfxAlyO2(9.5% Al)/graphene (blue and red lines in Figure 4a, respectively) samples, no significant differences from the Raman spectrum of the pristine graphene were observed, indicating that the graphene underwent negligible degrees of chemical and electrical deformation during the pALD processes for those mesostructured films. In contrast, in the spectra of the crystalline HfO2/graphene and HfxAlyO2(9.5% Al)/graphene (pink and green lines in Figure 4a, respectively) samples, ID, ID/ IG, and the fwhm of the 2D peak increased, confirming the occurrence of chemical/mechanical deformation of the graphene [pristine graphene, HfO 2 /graphene, and HfxAlyO2(9.5% Al)/graphene ID/IG = 0.09, 0.66, and 0.22,
Figure 4. (a) Characteristic Raman spectra of CVD-grown graphene covered with variously structured HfO2 and HfxAlyO2 films and (b) corresponding J−E curves obtained for MIG capacitor devices. For comparison, the Raman spectrum of pristine CVD-grown graphene is also shown [black line, a].
respectively; pristine graphene, HfO2/graphene, HfxAlyO2(9.5% Al)/graphene fwhm 2D = 19.3, 36.1, and 28.8 cm −1 ; respectively]. Note that the increases in ID and ID/IG are attributed to the Raman scattering process through intervalley double resonance of graphene with a considerable amount of structural defects.26 Further the widening of fwhm2D is attributable to interruption of the graphene Raman scattering due to chemical deformation.27,28 Therefore, the crystallinestructured dielectric film formation process results in severe degradation of the graphene, probably because of mechanical strain and chemical reaction caused by the adjacent oxide layers during the postannealing process. The superiority of the mesocrystalline-structured HfxAlyO2 as an ultrathin gate dielectric film for the fabrication of graphenebased devices was further confirmed by characterizing the leakage current in the MIG capacitor. Figure 4b shows the leakage current density versus the applied electric field (J−E) curves of MIG capacitors with 5-nm-thick HfO2/graphene and HfxAl yO 2(9.5% Al)/graphene samples before and after postannealing. As mentioned above (Figure 2e), the mesostructured HfxAlyO2(9.5% Al)/graphene sample (red line in Figure 4b) was vastly superior in terms of the breakdown E field and leakage current [breakdown voltage = 6.5 MV/cm]. 6663
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano
Figure 5. (a) MIG FET arrays fabricated on a 6 in. Si wafer. (b) Optical microscope image of fabricated MIG-FET unit device and (c) schematic diagram of a MIG FET device. Representative electrical characteristics measured for fabricated MIG FET devices with 5-nm-thick mesostructured HfxAlyO2: (d) dielectric constant (red) and gate capacitance density (blue) as functions of frequency and (e) transfer curve (drain current ID vs gate voltage VG).
phase remained in the crystalline HfxAlyO2(9.5% Al)/graphene sample. This could explain the greater leakage-current degradation of the crystalline HfO2/graphene sample compared to that of the crystalline HfxAlyO2(9.5% Al)/graphene sample. The amorphous phase remaining in the crystalline HfxAlyO2(9.5% Al)/graphene sample was most likely due to the Al dopant suppression of crystallization. These results indicate that the very low leakage current and the sufficiently high breakdown E field of the mesostructured HfxAlyO2(9.5% Al)/graphene sample originated from the sample structure itself, which was composed of a mixture of amorphous phase and tiny crystals. These crystals may have been of intermediate size, i.e., with sizes between those of crystal seeds and fully crystallized grains. Finally, we fabricated an array of MIG FETs using the mesostructured HfxAlyO2 film as a gate dielectric on a wafer scale. Details of the fabrication process are given in the Experimental Procedure section. Note that the thickness of the mesostructured HfxAlyO2(9.5% Al) film was less than 5 nm. As shown in the optical microscopy images and the schematic diagram (Figure 5a−c), MIG FET units in a top-gated geometry were built on a 6 in. SiO2/Si wafer. The gate capacitance of the MIG FETs was measured at alternating current (ac) frequencies ranging from 10 kHz to 1 MHz (blue line in Figure 5d). As mentioned above, the capacitance-derived k value was ∼25 (red line in Figure 5d), which confirms that the CET of the mesostructured HfxAlyO2(9.5% Al) gate dielectric was approximately 0.78 nm. Note that the gate capacitance of MIG FETs was included in the quantum capacitance of graphene.34,35 Thus, the real k value of the HfxAlyO2 dielectric film connected in series with the quantum
Meanwhile, for the crystalline HfxAlyO2(9.5% Al)/graphene sample (green line in Figure 4b), the leakage current was severely degraded from that of the mesostructured sample; that is, the leakage current density level was increased by at least 2 orders of magnitude in the initial stage of the voltage sweep, and the breakdown E field was also decreased to ∼4.0 MV/cm. This behavior is most likely related to two mechanisms: (i) crystallization of the dielectric film and (ii) graphene degradation. First, considering the increased level, the crystallization of the HfxAlyO2 film was assumed to be the dominant factor of the leakage current degradation, because the grain boundary in the ultrathin film can function as the main leakage current path.29−31 Second, the MIG capacitor has a back-to-back diode structure, in which the cathode interface plays a dominant role in controlling the electron carrier injection, provided the majority carriers are electrons. Thus, in that system, the interface between the bottom electrode (graphene) and the dielectric film governs the electron injection when a positive bias is applied to the top electrode. This implies that the degraded leakage properties of the crystalline HfxAlyO2/graphene sample can be attributed to the graphene degradation caused by the postannealing, which agrees well with the results of the Raman analysis (Figure 4a). For the crystalline HfO2/graphene sample (pink line in Figure 4b), the leakage current was more sensitively affected by crystallization. This is because the dominant leakage conduction mechanism in HfO2 is Poole−Frenkel conduction, which is influenced by the defects in the dielectric film.32,33 Moreover, as mentioned above with regard to the XPS analysis (Figure S4), the crystalline HfO2/graphene sample was fully crystallized to the monoclinic phase, whereas an amorphous 6664
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano capacitance of graphene was ∼34, which agrees well with the result obtained from the metal−insulator−metal structure (see Figure S3). To evaluate the FET performance, the representative drain current (ID) versus the gate bias (VG) curve of a fabricated MIG FET was obtained, as shown in Figure 5e. The Dirac voltage was observed at ∼2.2 V, and the on/off ratio value for the hole was approximately 2.75 when the on-current level at the gate voltage shifted from the Dirac voltage by −1 V. The within-wafer and wafer-to-wafer uniformity were also evaluated to demonstrate the reproducibility of the HfxAlyO2(9.5% Al) film and MIG FETs in wafer scale (see Figure S5). The deviation of the dielectric constant of the HfxAlyO2 film was only 1.22 in the 6 in. wafer, which is in reasonable agreement with the results obtained from HfxAlyO2(9.5% Al) films in Figure 2d. The transfer curves (ID−VG) of MIG FETs in different test patterns and different wafers exhibited an almost identical property, indicating that the HfxAlyO2 films in a 6 in. wafer and batch-to-batch were reproducible. These results suggest that the mesostructured HfxAlyO2 film is the most appropriate material for an MIG FET gate dielectric, because of the superior electrical properties of the film, along with its process compatibility with graphene on a wafer scale.
the Si wafer using the adhesive tape. The Cu layer was etched using a Cu etchant; then, the Cu etchant residue was rinsed using DI water. Finally, the separated graphene layer was transferred onto a 6 in. Si wafer covered with a thermally grown SiO2 film (300 nm). pALD of HfO2 and Al2O3 Films on Graphene. The HfO2 and Al2O3 films were deposited using a traveling-wave-type ALD reactor (NCD Co., Ltd.) at a wafer temperature of 50 °C by separately injecting tetrakis(ethylmethylamino)hafnium (TEMAHf, UP Chemical), trimethylaluminum (TMA, UP Chemical), and H2O vapor, with N2 purging steps being performed in between. TEMAHf and TMA were used as the Hf and Al precursors, respectively, and H2O was used as the O source. The Al doping concentration in the HfxAlyO2 film was controlled by adjusting the cycle ratio of the Al2O3 and HfO2 pALD processes. The purge time of each precursor and reactant was sufficiently long to prevent reactant mixing and subsequent chemical reactions. Raman Characterization. The Raman spectra were obtained using a micro-Raman (Renishaw InVia) spectrometer in a dark room. An Ar-ion laser of 514 nm wavelength was used as an excitation source, and the incident power was controlled to under 2 mW to avoid heating effects. The typical spatial resolution was lower than 1 cm−1, and peak analysis was conducted safely using WiRE 3.3 software. MIG Capacitor and Graphene FET Fabrication. After the graphene was transferred onto the SiO2 wafer, the graphene channel was defined using photolithography and O2-reactive ion etching processes with a Au hard mask. The 10 nm Cr/100 nm Au source/ drain electrodes were patterned using photolithography and a lift-off method after deposition using an e-beam evaporation method. To form the gate dielectric, HfO2, Al2O3, and HfxAlyO2 films were deposited on the CVD-grown graphene using the pALD method. Finally, a 10 nm Cr/100 nm Au top gate electrode was deposited and defined in a similar manner to the source/drain formation process. Then 10 nm Ti/100 nm Au metal pads were subsequently formed on each metal electrode for stable electrical measurement. For the MIG capacitor fabrication, the top electrode was defined by the same process and material as for the FET gate electrode, using Cr/Au with a lift-off process. After the top electrode formation, postmetalization annealing was performed at 700 °C for 3 min under an Ar atmosphere, via rapid thermal annealing. Film and FET Characterization. The chemical compositions of the films were characterized using XPS (Quantum 2000). The crosssectional structures were analyzed using a TEM (Tecnai Osiris, FEI). GAXRD (D8 Advance, Bruker; wavelength λ = 1.788 97 nm, at 40 kV and 100 mA) was used to examine the crystal structures of the thin films. The film thickness of all samples was set to 20 nm, in order to obtain sufficient diffraction peak intensity. The crystal structures of the films were also analyzed using XPS (Quantum 2000). The electrical characteristics of the fabricated MIG capacitor and FET devices were measured using a Keithley 4200-SCS semiconductor parameter analyzer.
CONCLUSION A mesostructured HfxAlyO2 film deposited on graphene via a pALD process was evaluated as a reliable and robust gate dielectric material for graphene-based device applications. The distinguishable crystal structure of the mesocrystalline-structured HfxAlyO2 thin film consisted of an intermediated state with small immaturely crystallized parts in an amorphous phase matrix. The crystallized part of the mesostructured HfxAlyO2 film was induced via a phase transition from the monoclinic to tetragonal or cubic phases of HfO2, which was achieved through crystal phase engineering using Al dopant. This yielded an enhanced dielectric constant of ∼25 and improvement of the leakage current property, with no chemical or mechanical deformation in the graphene layer. Note that the effects of the phase transition engineering process on the HfxAlyO2 film were predicted using DFT calculations. Finally, the developed gate dielectric material was applied to wafer-scale metal−insulator− graphene field-effect transistors. Hence, it was found that incorporation of a 5-nm-thick mesostructured HfxAlyO2 gate dielectric film yields a CET of 0.78 nm and superior FET performance.
ASSOCIATED CONTENT
EXPERIMENTAL PROCEDURE
S Supporting Information *
Computational Details. The DFT calculations were performed using the Vienna ab initio simulation package (VASP).36 The generalized gradient approximation was employed, along with the projector-augmented-wave method. The electronic wave functions were expanded using the plane wave basis set with a cutoff energy of 550 eV. K-points were sampled on a 2 × 2 × 2 uniform grid. Note that this setting has been proven successful in previous calculations for doped Hf. We used a 2 × 2 × 2 supercell with 32 Hf and 64 O atoms for all calculations. The atomic positions and lattice parameters were fully relaxed for each doping concentration, and doped configurations were generated by considering every inequivalent substitution of Al atoms for Hf atoms. Graphene Synthesis and Transfer. Monolayered graphene was synthesized on a Cu-evaporated 6 in. Si wafer using a CVD process, under a flowing reaction gas mixture of H and CH4. Adhesive tape was attached to the sample tightly and was peeled off under a deionized (DI) water flow. Then, the graphene/Cu layer could be detached from
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.6b01734. Figures S1, S2, S3, S4, and S5 and Tables S1 and S2 (PDF)
AUTHOR INFORMATION Corresponding Authors
*E-mail (S.-J. Jeong):
[email protected]. *E-mail (S. Park):
[email protected]. Author Contributions §
Y. Lee and W. Jeon contributed equally to this work.
Notes
The authors declare no competing financial interest. 6665
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666
Article
ACS Nano
(19) Tomida, K.; Kita, K.; Toriumi, A. Dielectric Constant Enhancement due to Si Incorporation into HfO[sub 2]. Appl. Phys. Lett. 2006, 89 (14), 142902. (20) Park, P. K.; Kang, S.-W. Enhancement of Dielectric Constant in HfO[sub 2] Thin Films by the Addition of Al[sub 2]O[sub 3]. Appl. Phys. Lett. 2006, 89 (19), 192905. (21) Cho, D.-Y.; Jung, H. S.; Yu, I.-H.; Yoon, J. H.; Kim, H. K.; Lee, S. Y.; Jeon, S. H.; Han, S.; Kim, J. H.; Park, T. J.; Park, B.-G.; Hwang, C. S. Stabilization of Tetragonal HfO 2 under Low Active Oxygen Source Environment in Atomic Layer Deposition. Chem. Mater. 2012, 24 (18), 3534−3543. (22) Jang, J. H.; Jung, H.-S.; Kim, J. H.; Lee, S. Y.; Hwang, C. S.; Kim, M. Investigation of Oxygen-Related Defects and the Electrical Properties of Atomic Layer Deposited HfO[sub 2] Films Using Electron Energy-Loss Spectroscopy. J. Appl. Phys. 2011, 109 (2), 023718. (23) Yoo, Y. W.; Jeon, W.; Lee, W.; An, C. H.; Kim, S. K.; Hwang, C. S. Structure and Electrical Properties of Al-Doped HfO2 and ZrO2 Films Grown via Atomic Layer Deposition on Mo Electrodes. ACS Appl. Mater. Interfaces 2014, 6 (24), 22474−22482. (24) O’Keeffe, M. A Proposed Rigorous Definition of Coordination Number. Acta Crystallogr., Sect. A: Cryst. Phys., Diffr., Theor. Gen. Crystallogr. 1979, 35 (5), 772−775. (25) Jeon, W.; Chung, H.-S.; Joo, D.; Kang, S.-W. TiO[sub 2]/Al[sub 2]O[sub 3]/TiO[sub 2] Nanolaminated Thin Films for DRAM Capacitor Deposited by Plasma-Enhanced Atomic Layer Deposition. Electrochem. Solid-State Lett. 2008, 11 (2), H19. (26) Saito, R.; Hofmann, M.; Dresselhaus, G.; Jorio, A.; Dresselhaus, M. S. Raman Spectroscopy of Graphene and Carbon Nanotubes. Adv. Phys. 2011, 60 (3), 413−550. (27) Bruna, M.; Ott, A. K.; Ijäs, M.; Yoon, D.; Sassi, U.; Ferrari, A. C. Doping Dependence of the Raman Spectrum of Defected Graphene. ACS Nano 2014, 8 (7), 7432−7441. (28) Casiraghi, C. Probing Disorder and Charged Impurities in Graphene by Raman Spectroscopy. Phys. Status Solidi RRL 2009, 3 (6), 175−177. (29) Buchanan, D. A. Scaling the Gate Dielectric: Materials, Integration, and Reliability. IBM J. Res. Dev. 1999, 43 (3), 245−264. (30) Chan, A. C. K.; Cheng, C. F.; Chan, M. Effects of Dopants on the Electrical Behavior of Grain Boundary in Metal-Induced Crystallized Polysilicon Film. IEEE Trans. Electron Devices 2005, 52 (8), 1917−1919. (31) Shubhakar, K.; Pey, K. L.; Raghavan, N.; Kushvaha, S. S.; Bosman, M.; Wang, Z.; O’Shea, S. J. Study of Preferential Localized Degradation and Breakdown of HfO2/SiOx Dielectric Stacks at Grain Boundary Sites of Polycrystalline HfO2 Dielectrics. Microelectron. Eng. 2013, 109, 364−369. (32) Zhu, W. J.; Tamagawa, T.; Kim, J.; Di, Y. Current Transport in Metal/hafnium Oxide/silicon Structure. IEEE Electron Device Lett. 2002, 23 (2), 97−99. (33) Ma, T. P.; Bu, H. M.; Wang, X. W.; Song, L. Y.; He, W.; Tseng, H.-H.; Tobin, P. J. Special Reliability Features for Hf-Based High-/spl Kappa/ Gate Dielectrics. IEEE Trans. Device Mater. Reliab. 2005, 5 (1), 36−44. (34) Zhang, Z.; Xu, H.; Zhong, H.; Peng, L.-M. Direct Extraction of Carrier Mobility in Graphene Field-Effect Transistor Using CurrentVoltage and Capacitance-Voltage Measurements. Appl. Phys. Lett. 2012, 101 (21), 213103. (35) Xu, H.; Zhang, Z.; Peng, L.-M. Measurements and Microscopic Model of Quantum Capacitance in Graphene. Appl. Phys. Lett. 2011, 98 (13), 133122. (36) Kresse, G. Efficient Iterative Schemes for Ab Initio Total-Energy Calculations Using a Plane-Wave Basis Set. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54 (16), 11169−11186.
ACKNOWLEDGMENTS The authors are grateful for the assistance provided by their colleagues at the Samsung Advanced Institute of Technology. The authors specifically thank D.-S. Ko and X. Li for help with the TEM analyses, D. Yu for assistance with the XRD analyses, and the Nano Fabrication Group for assistance with device fabrication. REFERENCES (1) Wang, L.; Meric, I.; Huang, P. Y.; Gao, Q.; Gao, Y.; Tran, H.; Taniguchi, T.; Watanabe, K.; Campos, L. M.; Muller, D. A.; Guo, J.; Kim, P.; Hone, J.; Shepard, K. L.; Dean, C. R. One-Dimensional Electrical Contact to a Two-Dimensional Material. Science 2013, 342 (6158), 614−617. (2) Geim, A. K.; Grigorieva, I. V. Van Der Waals Heterostructures. Nature 2013, 499 (7459), 419−425. (3) Xu, M.; Liang, T.; Shi, M.; Chen, H. Graphene-like TwoDimensional Materials. Chem. Rev. 2013, 113 (5), 3766−3798. (4) Kim, K.; Choi, J.-Y.; Kim, T.; Cho, S.-H.; Chung, H.-J. A Role for Graphene in Silicon-Based Semiconductor Devices. Nature 2011, 479 (7373), 338−344. (5) Schwierz, F. Graphene Transistors. Nat. Nanotechnol. 2010, 5 (7), 487−496. (6) Geim, A. K. Graphene: Status and Prospects. Science 2009, 324 (5934), 1530−1534. (7) Yang, H.; Heo, J.; Park, S.; Song, H. J.; Seo, D. H.; Byun, K.-E.; Kim, P.; Yoo, I.; Chung, H.-J.; Kim, K. Graphene Barristor, a Triode Device with a Gate-Controlled Schottky Barrier. Science 2012, 336 (6085), 1140−1143. (8) Heo, J.; Byun, K.-E.; Lee, J.; Chung, H.-J.; Jeon, S.; Park, S.; Hwang, S. Graphene and Thin-Film Semiconductor Heterojunction Transistors Integrated on Wafer Scale for Low-Power Electronics. Nano Lett. 2013, 13 (12), 5967−5971. (9) Xuan, Y.; Wu, Y. Q.; Shen, T.; Qi, M.; Capano, M. A.; Cooper, J. A.; Ye, P. D. Atomic-Layer-Deposited Nanostructures for GrapheneBased Nanoelectronics. Appl. Phys. Lett. 2008, 92 (1), 013101. (10) Wang, X.; Tabakman, S. M.; Dai, H. Atomic Layer Deposition of Metal Oxides on Pristine and Functionalized Graphene. J. Am. Chem. Soc. 2008, 130 (26), 8152−8153. (11) Bai, J.; Liao, L.; Zhou, H.; Cheng, R.; Liu, L.; Huang, Y.; Duan, X. Top-Gated Chemical Vapor Deposition Grown Graphene Transistors with Current Saturation. Nano Lett. 2011, 11 (6), 2555− 2559. (12) Lin, Y.-M.; Valdes-Garcia, A.; Han, S.-J.; Farmer, D. B.; Meric, I.; Sun, Y.; Wu, Y.; Dimitrakopoulos, C.; Grill, A.; Avouris, P.; Jenkins, K. A. Wafer-Scale Graphene Integrated Circuit. Science 2011, 332 (6035), 1294−1297. (13) Xu, H.; Zhang, Z.; Xu, H.; Wang, Z.; Wang, S.; Peng, L.-M. Top-Gated Graphene Field-Effect Transistors with High Normalized Transconductance and Designable Dirac Point Voltage. ACS Nano 2011, 5 (6), 5031−5037. (14) Xu, H.; Zhang, Z.; Wang, Z.; Wang, S.; Liang, X.; Peng, L.-M. Quantum Capacitance Limited Vertical Scaling of Graphene FieldEffect Transistor. ACS Nano 2011, 5 (3), 2340−2347. (15) Wang, Z.; Xu, H.; Zhang, Z.; Wang, S.; Ding, L.; Zeng, Q.; Yang, L.; Pei, T.; Liang, X.; Gao, M.; Peng, L.-M. Growth and Performance of Yttrium Oxide as an Ideal High-Kappa Gate Dielectric for CarbonBased Electronics. Nano Lett. 2010, 10 (6), 2024−2030. (16) Jeong, S.-J.; Gu, Y.; Heo, J.; Yang, J.; Lee, C.-S.; Lee, M.-H.; Lee, Y.; Kim, H.; Park, S.; Hwang, S. Thickness Scaling of Atomic-LayerDeposited HfO2 Films and Their Application to Wafer-Scale Graphene Tunnelling Transistors. Sci. Rep. 2016, 6, 20907. (17) Lee, J. H.; Yu, I.-H.; Lee, S. Y.; Hwang, C. S. Phase Control of HfO2-Based Dielectric Films for Higher-K Materials. J. Vac. Sci. Technol. B 2014, 32 (3), 03D109. (18) Lee, C.-K.; Cho, E.; Lee, H.-S.; Hwang, C. S.; Han, S. FirstPrinciples Study on Doping and Phase Stability of HfO 2. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78 (1), 012102. 6666
DOI: 10.1021/acsnano.6b01734 ACS Nano 2016, 10, 6659−6666