Monolithic integration of perovskites on Ge(001) by atomic layer deposition: a case study with SrHfxTi1−xO3

This work reports the growth of crystalline SrHfxTi1−xO3 (SHTO) films on Ge (001) substrates by atomic layer deposition. Samples were prepared with different Hf content x to explore if strain, from tensile (x = 0) to compressive (x = 1), affected film crystallization temperature and how composition affected properties. Amorphous films grew at 225 °C and crystallized into epitaxial layers at annealing temperatures that varied monotonically with composition from ∼530 °C (x = 0) to ∼660 °C (x = 1). Transmission electron microscopy revealed abrupt interfaces. Electrical measurements revealed 0.1 A/cm leakage current at 1 MV/cm for x = 0.55. Introduction As the transistor feature sizes continue to scale ever smaller there has been a transition away from the use of conventional materials for the channel, Si, and the gate oxide, SiO2. [1] Germanium has electron and hole mobility of 3900 and 1900 cm/Vs, respectively, compared with 1400 and 470 cm/Vs at 300 K, respectively, in Si. For this reason, Ge is being considered for p-type metal-oxide semiconductor field effect transistors (MOSFETs). There are multiple considerations in selecting a gate oxide material including the dielectric constant, band offset, leakage current, interface trap density (Dit), and ease of manufacturing. [3,4] Various groups have reported gate oxides on Ge in MOSFETs, including TiO2/ Al2O3, ZrO2, LaAlO3 on an interfacial layer of SrGex, HfO2 on an interfacial layer of Y2O3-doped GeO2, Y2O3 on a GeOx interfacial layer, and HfO2 with Al2O3 to suppress HfO2–GeOx intermixing. [5–10] Amorphous oxides generally have lower dielectric constants than the crystalline form. However, the absence of grain boundaries in amorphous films is a potential advantage as grain boundaries can serve as defect trap sites. Crystalline oxides have been reported on silicon and germanium. These crystalline oxides on semiconductors (COS) can offer high dielectric constants, perfection of the crystal structure at the oxide/semiconductor interface, and the possibility to coherently bond across the interface and minimize dangling bonds. Many COS are grown using molecular beam epitaxy (MBE). We recently reported an all-chemical growth process for SrTiO3 (STO) and SrHfO3 (SHO) on Ge (001) using atomic layer deposition (ALD) that illustrates a potentially scalable integration route to crystalline oxides on germanium. SHO has a large band gap of 6.1 eV with favorable conduction band offset (∼2.2 eV) and valence band offset (∼3.2 eV) with Ge. This is in contrast with Ti-based perovskites, where the Ti 3d states yield negligible conduction band offsets with Si and Ge (∼0.1–0.5 eV). Incorporation of Hf into the SrHfxTi1−xO3 alloy provides an upward shift of the d-states, which improves the conduction band offset, and increases the lattice constant, which may affect epitaxy. The dielectric constants were k∼ 90 and k∼ 20 for thin films of STO and SHO, respectively. Capacitor structures showed the leakage current for STO was around 10 A/cm at 0.7 MV/cm with equivalent oxide thickness (EOT) of 0.7 nm and the leakage current for SHO was less than 10 A/cm at 1.0 MV/cm with EOT of 1.0 nm. In the previous studies amorphous films were deposited during ALD and the crystalline films formed after annealing at temperatures from 530 to 660 °C. The present study was †Current Address: Towerjazz Texas, San Antonio, TX 78251, USA. ‡Current Address: Intel Corporation, Chandler, AZ 85226, USA. MRS Communications (2016), 1 of 8 © Materials Research Society, 2016 doi:10.1557/mrc.2016.36 MRS COMMUNICATIONS • www.mrs.org/mrc ▪ 1 http://dx.doi.org/10.1557/mrc.2016.36 Downloaded from http:/www.cambridge.org/core. University of Texas Libraries, on 15 Sep 2016 at 15:05:35, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms. undertaken to explore the role of strain on the crystallization temperature and composition on the film properties by growing alloys of SrHfxTi1−xO3 (SHTO) by ALD on Ge (001). The lattice constants of 3.905 Å and 4.069 Å for bulk STO and SHO, respectively, lead to 2.2% and −1.9% strain with the Ge (001) substrate for fully-strained, commensurate films affording a composition for which the lattice constant will match Ge (001) surface spacing along the [110] direction. Experiment The Ge substrates (18 × 20 mm) are diced from a 4-in Gewafer (n-type, Sb-doped, 0.029–0.054 Ω·cm resistivity from MTI Corp.) The sample preparation procedure and experimental system are described in previous work. The wafer pieces are cleaned with acetone, isopropyl alcohol, and deionized water in an ultrasonic bath for 10 min each. After dryingwith nitrogen the sample is exposed to UV/ozone for 30 min to remove residual carbon contamination. The sample is mounted on a molybdenum puck and loaded into the vacuum system immediately and transferred to aMBE chamber. The sample is annealed and deoxidized in vacuum (<2 × 10−9 Torr) by heating from 200 to 500 °C at 20 °C/min and then from 550 to 650 °C at 10 °C/ min, annealed at 650 °C for 1 h, and finally cooled to 200 °C at 30 °C/min. This procedure produces the 2 × 1-reconstructed clean Ge (001) surface, which is essential as the starting surface for perovskite ALD. In situ reflection high-energy electron diffraction (RHEED) is used to verify the surface order; Supplementary Material Fig. S1 shows a representative surface after this procedure. The Ge substrate with the 2 × 1-reconstructed surface is transferred in situ to the ALD chamber where it is allowed to equilibrate for ∼15 min at the growth temperature of 225 °C. Film growth is performed at 1Torr using strontiumbis(triisopropylcyclopentadienyl) [Sr(Pr3Cp)2] (HyperSr), hafnium formamidinate [Hf(fmd)4] (Hf-FAMD), titanium tetraisopropoxide [Ti(O-Pr)4] (TTIP), and water as the co-reactant (oxygen source). The Sr, Hf, and Ti precursors were heated to 130, 115, and 40 °C, respectively. The quaternary compound was grown using the dosing and purging times indicated in Supplementary Material Fig. S2. For each metalorganic precursor, a 2-s dose time saturates the surface; the co-reactant water is dosed for 1 s. Following each precursor or water dose, a 15-s Ar purge is required. Our previous work showed that excess Sr was required to initiate the growth of STO on Ge. For this reason the Sr:Ti cycle ratio is 2:1 and we adopted that protocol herein.Similarly, SHOgrowthonGeusedaSr:Hf cycle ratio of 1:1 and that was used herein. The cycle ratios in Fig. S2 produce SHTO films that are slightly Sr-rich (Sr/(Sr + Hf + Ti)) and a 1:1 (Sr:Ti + Hf) film requires some of the Sr:Ti cycle ratios to be 1:1 rather than 2:1. The quaternary SrHfxTi1−xO3 oxide was grown by adjusting the number of m SHO subcycles and n STO subcycles to vary the Hf content (x). Different film thickness is controlled by the total number of supercycles l. Following ALD growth, the sample is transferred back to the MBE chamber and annealed while monitoring the surface in real time with RHEED to follow the transformation from amorphous to crystalline. The substrate is heated from 200 to 500 °C with a 20 °C/min ramp rate, followed by a ramp rate of 10 °C/min as the temperature is increased further. Fig. S3 in Supplementary Material shows a representative transformation for an 11.4-nm SrHf0.34Ti0.66O3 film. The temperature at which spots in the RHEED pattern emerge is monitored and this is referred to as the crystallization temperature. The temperature is then increased by 20 °C and held at this as annealing temperature for 5 min to fully crystallize the sample. The sample is then cooled to 200 °C at 30 °C/min and transferred from the MBE chamber. The films were characterized by in situ x-ray photoelectron spectroscopy (XPS) to analyze the composition and uniformity using monochromatic Al Kα source at 1486.6 eV and a VG Scienta R3000 analyzer, which is calibrated by a silver foil. High-resolution spectra are measured five times and summed up for the Sr (3d, 3p) Ti 2p, O 1s, C 1s, Hf 4f, and Ge 3d features. The measurement settings were 50 meV steps with 157 ms/step dwell time and 100 eV pass energy with a 0.4 mm analyzer slit width, which resulted in 350 meV effective resolution. The stoichiometry and Hf content x, which is defined as the ratio of Hf to (Hf + Ti), for the SrHfxTi1−xO3 films are calculated by the integrated area of the Sr 3d, Ti 2p, and Hf 4f peaks. The atomic sensitivity factors for Sr 3d, Ti 2p, and Hf 4f are set as 1.843, 2.001, and 2.639, respectively. The thickness and crystallinity of the SHTO films were measured by x-ray reflectivity (XRR) and x-ray diffraction (XRD) on a Rigaku Ultima IV system with a Cu Kα source. The interface of selected samples was examined by cross-sectional transmission electron microscopy (TEM). The samples were prepared via standard cross-section method with Ar ion milling. Aberration-corrected scanning transmission electron microscopy (STEM) was used for further interface study. The TEM images were taken with a JEOL 2010F and STEM images were taken with a JEOL ARM 200F. Electron-energy-loss spectroscopy (EELS) composition mapping was also applied to investigate the elemental distributions in the growth direction. Electrical properties (dielectric constant k and leakage current I ) were established for some samples by fabricating standard metal oxide semiconductor capacitor (MOS capacitor) structures. The films had a top electrode of TaN applied by sputtering and 15 μm photolithographic features were defined with a SF6-based plasma etch. After building up the MOS capacitor structure the back side of the wafer was scratched and silver paste was applied to form the bottom electrode. The capacitance–voltage (C–V ) and current–voltage (I–V ) measurements were performed on an Agilent B1500A semiconductor device parameter analyzer with a Cascade Microtech probe station. Results and Discussion Deposition and crystallization of SHfxTi1−xO3 films Films with Hf content x distributed from 0 to 1 and with thickness between 8.0 and 14.0 nm were grown. SHTO films were deposited with subcycle ratios (Fig. S2) m:n from 1:3 to 7:1. 2▪ MRS COMMUNICATIONS • www.mrs.org/mrc http://dx.doi.org/10.1557/mrc.2016.36 Downloaded from http:/www.cambridge.org/core. University of Texas Libraries, on 15 Sep 2016 at 15:05:35, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms. STO (x = 0) and SHO (x = 1) films were deposited with around 100 cycles producing films that were 9.7 and 12.6 nm thick, respectively. SHTO films were grown with a total of l × (2m + 3n) cycles as the composition was varied. The XRD and rocking curve measurements confirmed the crystallization that was indicated by RHEED. Figure 1 shows the θ–2θ XRD and rocking curve around the film (002) reflection at 2θ = 45.45 ± 0.5° for a film with a composition of SrHf0.47Ti0.53O3. This sample was grown with a m:n = 2:1 subcycle ratio and l = 14 supercycles. The crystallization temperature was found to be 612 °C. Ex situ XRR measurement indicates the thickness is 13.9 nm. The (002) reflection at 2θ = 45.45 ± 0.5° leads to an out-of-plane lattice constant of c = 3.989 ± 0.005 Å. The rocking curve scan around the SHTO (002) reflection reveals a full-width half-maximum (FWHM) of 1.3°. The best quality STO and SHO films grown by ALD had FWHM of 0.8° and 1.2°, respectively. Figure 2(a) illustrates that the temperature for crystallization onset increases monotonically with increasing Hf content from 510 °C for STO to 640 °C for SHO. The bulk lattice constants for SHO and STO are 4.069 Å and 3.905 Å, respectively. By assuming SHTO forms a substitutional alloy, and by applying Vegard’s law the bulk SHTO lattice constants (a) can be estimated as a function of x and these are represented by the red line in Fig. 2(b). From Ge surface spacing of 3.992 Å along the <110> direction we compute the in-plane strain that should result from a fully-strained, commensurate film at room temperature and present this as the green line in Fig. 2(b). The fullystrained films vary from tensile (2.2% for x = 0) to compressive (−1.9% for x = 1). At room temperature a value of x∼ 0.53 should give zero strain. The coefficient of thermal expansion for Ge is approximately one order of magnitude less than a composition-averaged value for SHTO alloys; over the 500–650 °C annealing window the x-value that matches the Ge separation distance decreases from x∼ 0.4 to x∼ 0.3. We sought to determine how interface strain influenced crystallization. The monotonically increasing crystallization temperature with x [Fig. 2(a)] suggests that the annealing temperature does not depend on the strain between the substrate and the SHTO alloy but is rather dependent on the atomic mass of the elements that comprise the alloy. Since the atomic mass of Hf is much heavier than Ti, it will require more thermal energy to move Hf to the correct perovskite crystal structure location compared with Ti, similar to what has been found for A-site cations in ATiO3 perovskites. Both Sr-rich or Sr-lean stoichiometry in STO films grown on STO produced an out-of-plane lattice constant that was greater than expected for a fully strained film. Similar results are reported herein for SHTO alloys. The SHTO out-of-plane lattice constants c are determined from the (002) XRD reflections and are also presented in Fig. 2(b). In general the experimental c-values are greater than a [the red line in Fig. 2(b)] for compressive films and less than a for tensile films, consistent with expectations for commensurate films. The squares Figure 1. X-ray diffraction pattern (a), and rocking curve (b), for a 13.9-nm SrHfxTi1−xO3 (x = 0.47) film grown on Ge (001) by ALD, and annealed at 632 °C for 5 min in vacuum. The peak of the SHTO (002) reflection is at 2θ = 45.45 ± 0.5° and the rocking curve for the (002) reflection has a full width at half maximum (FWHM) of 1.3°. Figure 2. (a) Crystallization temperature T versus Hf content x for 8–14-nm thick SrHfxTi1−xO3 films. The dashed line is drawn to guide the eye. (b) Predicted dependence of the bulk lattice constant and film with Hf content x at room temperature. The red line indicates the cubic lattice constant a of bulk SrHfxTi1−xO3 based on Vegard’s law. The green line presents the strain for commensurate and fully-strained SHTO films on Ge. The squares are the experimental out-of-plane lattice constants c for different x; the squares with letters indicate that samples with Sr-rich compositions of around 55% and other squares indicate samples for which the Sr composition varied from 49% to 51%. Functional Oxides Research Letter MRS COMMUNICATIONS • www.mrs.org/mrc ▪ 3 http://dx.doi.org/10.1557/mrc.2016.36 Downloaded from http:/www.cambridge.org/core. University of Texas Libraries, on 15 Sep 2016 at 15:05:35, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms. with letters in Fig. 2(b) correspond to samples that have Sr-enrichment around 55% and could be expected to have out-of-plane lattice constants that are greater than fully-strained stoichiometric films. A fully-strained stoichiometric STO film on Ge should have an out-of-plane lattice constant of 3.852 Å at room temperature based on a Possion ratio for STO of 0.232. The experimental STO value [Square (a) in Fig. 2(b)] is 3.883 Å due to Sr-enrichment. Similarly Squares (b), (c), and (d) correspond to samples with out-of-plane lattice constants that are greater than the bulk, cubic lattice constant. Whereas, films with similar compositions as (b), (c), and (d), which should be under tension if fully strained, display out-of-plane lattice constants less than a. In situ XPS study and composition uniformity In situ XPS was performed on the SHTO films before and after annealing. Figure 3 presents results for a SrHf0.56Ti0.44O3 film grown with an m:n = 3:1 subcycle ratio and l = 11 supercycles. This film started to crystallize at 633 °C had an out-of-plane lattice constant of 4.04 Å. Figures 3(a)–3(e) present the Sr 3d, Ti 2p, O 1s, C 1s + Sr 3p, and Hf 4f core levels, respectively. The Sr 3d3/2 and 3d5/2 peaks are located at binding energies of 135.5 eV and 133.8 eV, respectively, which indicates that the Sr is fully oxidized (Sr) in the SHTO film. Similarly, the Ti 2p and Hf 4f features in Figs. 3(b) and 3(e), respectively, correspond to fully oxidized Ti (Ti) and Hf (Hf). Figure 3(d) shows that there is no carbon peak at the C 1s position of 285 eV. The Hf content x = 0.56 and stoichiometry are determined by integrating the areas of the Sr 3d, Ti 2p, and Hf 4f features. The Sr:(Hf + Ti) ratios for films in this study reveal a stoichiometry that is consistent with an ABO3 perovskite. Previous work in our group has shown that slightly Sr-rich films (i.e., A-rich) crystallize more readily on Ge (001) than B-rich films. The Sr composition (viz., Sr/(Sr + Hf + Ti)) of all crystallized films falls in the range from 49% to 56%. Some films outside this range, such as 47% Sr, are still observed to crystallize. However, the RHEED images for such films (not shown) suggest rough surfaces and imply lower crystalline quality. The sample in Fig. 3 has the ratio of A:B of 55:45. The value of x = 0.56 for a m:n = 3:1 subcycle ratio suggests that Ti is more readily incorporated into the SHTO alloy during ALD. Compositional uniformity across the Ge substrates and throughout the films was probed with XPS. Figure S4 in Supplementary Material presents results for the film discussed in Fig. 3 after annealing at 653 °C for 5 min. The composition is uniform across the film and constant with gas flow direction, as would be expected for an ALD process. Angle-resolved XPS (AR-XPS) was performed at wafer position Number 5 to examine uniformity of the composition through the thickness of the film. Based on the universal escape depth curve we estimate nearly 97% of the signal comes from a depth of no more than ∼6.2 nm from the SHTO surface when the sample is fixed on the horizontal plane (i.e., 0°). The analysis angles [Fig. S4 (b)] of 0°, 15°, 30°, 45°, and 60° correspond to sampling depths of 6.2, 6,0, 5.4, 4.4, and 3.1 nm, respectively. For the Figure 3. X-ray photoelectron spectra for Sr 3d (a), Ti 2p (b), O 1s (c), C 1s and Sr 3p (d), and Hf 4f (e) in a SrHfxTi1−xO3 (x = 0.56) film grown by ALD on Ge (001). The blue line in each figure corresponds to the spectrum post-deposition and the red line corresponds to the spectrum after annealing at 654 °C for 5 min. 4▪ MRS COMMUNICATIONS • www.mrs.org/mrc http://dx.doi.org/10.1557/mrc.2016.36 Downloaded from http:/www.cambridge.org/core. University of Texas Libraries, on 15 Sep 2016 at 15:05:35, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms. AR-XPS scans that sampled depths of 6.2–3.1 nm from the free surface, the Sr, Hf, and Ti compositions were 54.8 ± 0.9%, 25.0 ± 0.6%, and 20.3 ± 0.6%, respectively, confirming the composition uniformity with depth of the film.