Search

High quality, large-grained AlN films are of interest for use as either heat spreading layers or as a buffer/templating layer for further film growth. As the typical deposition temperature for high-quality crystalline AlN is usually in excess of 800 oC, there is a large amount of thermal strain in the films and it is advantageous to develop processes to lower the deposition temperature so that it is compatible with back end of line (BEOL) processing. In this work, the low-temperature (oC) deposition of polycrystalline AlN films is demonstrated by atomic layer annealing (ALA) which is a variant of ALD that utilizes a third pulse of low-energy inert gas ions in addition to the usual metal and co-reactant pulses [1]. Using trimethyl aluminum (TMA) or tris(dimethylamido) aluminum (TDMAA) with the highly reactive nitrogen-containing precursor hydrazine (N2H4), AlN can be deposited at ~200 oC [2]; however, these films are amorphous and would have low thermal conductivity due to a high degree of phonon scattering. Using TDMAA with N2H4 or NH3 at temperatures >350 oC, polycrystalline films can be deposited in a purely thermal process; however, the reported grain sizes are small ( In the present study of AlN ALA on Si (111) (a non-lattice matched substrate), two metal precursors (TMA and TDMAA) were compared using ultra-high purity anhydrous N2H4 as a co-reactant and argon ions with tuned energy for the third pulse. It was found that deposition using TDMAA as the Al precursor resulted in high-quality AlN films with large grain size (>20 nm) and low C/O contamination (5%), owing to its thermal instability at 400 oC. Transmission electron micrographs for a 40 nm film grown using TDMAA and N2H4 show vertical grain structure with grains spanning the entire thickness of the film. As heat spreading layers often need to be in excess of 1 micron thick in order to be relevant for use in high volume manufacturing, it is also demonstrated that these 40 nm films successfully transfer the large grain size of the ALA film to a thick sputtered AlN film (which can be deposited relatively quickly) by acting as a template layer. [1] H-Y. Shih et al, Scientific Reports 7:39717. [2] M. Mizuta et al, Japanese J. Appl. Phys. 25(12), L945-L948 (1986). [3] R. G. Gordon, U. Riaz, and D. M. Hoffman, J. Mater. Res., 7(7) (1992). [4] A. I. Abdulagatov et al Russian Microelec. 47(2), 118-130 (2018). [5] W-C. Kao et al, RSC Adv. 9, 12226-12231 (2019). [6] W-H. Lee et al ACS Sustainable Chem. Eng. 7,1, 487-495 (2019). This work was supported in part by the Semiconductor Research Corporation. Figure 1

Three-dimensional integration of microelectronic devices and high-power radio frequency devices requires the integration of electrically insulating heat spreaders. Aluminum nitride is one potential material, as it can be deposited within backend thermal budgets and conducts heat isotopically, as opposed to diamond which requires deposition temperatures in excess of 900 °C and hexagonal boron nitride which only conducts heat laterally. Growth of crystalline aluminum nitride on non-lattice matched substrates is possible using bias-enhanced atomic layer annealing, in which a plasma treatment step with controlled ion energy follows each cycle of traditional atomic layer deposition precursor dosing. Tris(dimehtylamido) aluminum was used as the aluminum precursor to minimize carbon contamination, and anhydrous hydrazine was used to prevent oxygen contamination. Plasma treatment with neon, krypton, and argon at substrate biases of -10V, -25V, and -40V were investigated. Varying the gas identity and substrate bias provides control over the momentum and kinetic energy of ions bombarding the surface, resulting in selective control over the crystalline orientation. Across all gas and bias combinations ion flux is held constant by adjusting the power applied to the plasma. In-situ x-ray photoelectron spectroscopy (XPS) is used to study the elemental composition of the samples. Films deposited with this technique routinely show carbon and oxygen content less than 3% by XPS. Ex-situ grazing-incidence X-ray diffraction (GI-XRD) and x-ray reflectivity (XRR) are used to analyze the crystallinity, thickness, and density of the deposited material. 40 nm thick films were deposited onto HF-etched Si(111) and SiC substrates. GI-XRD scans of each condition are shown in Figure 1. Atomic layer annealing on Si(111) substrates with neon plasma produced films with poor crystallinity and preferential (002) orientation. For these films, the AlN (002) peak position is shifted relative to the peak position in films treated with argon and krypton, indicating the deposition of strained films. This may be due to embedding of neon ions, causing relatively small crystallite sizes of 5 nm to 9 nm. Films deposited with krypton plasma treatment also showed preferential AlN (002) deposition, though crystallite sizes are large and range from 9 nm to 11 nm. Optimal crystallization occurred when using argon plasma treatment with a -25V substrate bias, producing a film with selective (200) orientation and an average crystallite size of 27 nm. This crystallinity resulting from this condition may be the result of enhancing the preferential (200) crystallinity afforded by the purely thermal atomic layer deposition of the same precursors. Accordingly, the impinging argon ions would have significant momentum to increase surface adatom mobility, but kinetic energy sufficiently low as to not change the preferential orientation. Figure 1

Uniform and defect-free Al2O3 films were grown on highly oriented pyrolytic graphite (HOPG) terraces by thermal atomic layer deposition (ALD) at low temperature (50 oC) without any functionalization of the surface. Controlling the pulse times of trimethylaluminum (TMA) and H2O while using short Ar purge times, 1-2 nanometer size Al2O3 particles were formed on the HOPG terraces. The particles provided a layer of nanoscale nucleation centers on the HOPG terraces. Capacitance-voltage measurements of Al2O3 films grown at 50 oC using 50 ALD cycles showed an areal capacitance 1.17 mF/cm2 with very small frequency dispersion, consistent with the absence of induction cycles and formation of a high quality interface. The leakage current of the Al2O3 was ~10-5 A/cm2 in large area devices (1900 mm2) which is comparable with results from devices prepared using an identical ALD process for Al2O3 on Si0.7Ge0.3(001) substrates, and is consistent with the absence of pinholes.

Channel passivation of SiGe and InGaAs requires removal of dangling bonds and formation of a thermally stable interface. On InGaAs, one solution is the deposition by self-limiting CVD or ALD of a few monolayer of silicon, silicon oxide, or silicon nitride. To form an Si-OH terminated In0.53Ga0.47As(001) surface, two monolayers of Si were deposited at 350°C with self-limiting CVD using Si2Cl6 dosing following by dosing anhydrous HOOH(g). Scanning tunneling microcopy (STM) shows and ordered surface is form while scanning tunneling spectroscopy (STS) shows the surface Fermi level position moves towards midgap due to a surface dipole formation from –OH groups and oxygen bonding to the surface. TMA was dosed at 250°C onto the Si-OH terminated InGaAs(001), and XPS shows the emergence of the Al 2p and C 1s peaks indicative of TMA surface nucleation. The TMA shifts the InGaAs Fermi level back towards the conduction band, consistent with unpinning. MOSCAPs were fabricated to further demonstrate Fermi level unpinning using the Si-OH termination of InGaAs. While this same Si deposition process could be employed on SiGe, a simple annealing process can be employed to form a silicon terminated SiGe surface. A saturation dose of H2O2(g) at 25oC chemisorbs to on Si0.5Ge0.5(110) and Si0.47Ge0.53(001) surfaces leaving the Fermi level on the surface unpinned, and the surface is functionalized with mainly Si-OH, Ge-OH and Si-O-Ge bond. After a subsequent TMA dose at 25oC and annealing at 300oC, XPS and STM verify that a thermally stable and well-ordered monolayer of Al2O3 is formed on SiGe(110) and (001) surfaces with only Al-O-Si bonds and no detectable Ge-O-Si bonds. The H2O2(g) functionalization provides 3 times higher O sites on the surface and 3 times higher TMA nucleation density than H2O(g) at 25°C and 120°C. A variety of strongly bound passivating atoms can induce formation of a nearly pure Si interfacial layer between high-k dielectrics and SiGe. This annealing procedure was employed to form MOSCAP on SiGe(001) using Al2O3 and HfO2 gate oxides after either (NH4)2S or NH3 plasma passivation. For NH3 plasma passivation, angle-resolved XPS showed that the annealing allows formation of nearly Ge-free SiON layer. For higher Ge concentrations, the annealing process is unlikely to form a nearly Ge-free SiON layer so a novel plasma-free chemistry has been develop to deposit SiON using ALD with N2H4(g) as a precursor. Figure 1

High resolution STM images of In0.53Ga0.47As(001)-(2x4) were obtained and surface defects were quantified as a function of sample preparation technique. Published STM images of InGaAs(001)-(2x4) samples of varying In composition were examined and missing dimer unit cells, adatom trough defects, and incomplete atomic terraces were quantified for comparison with the In0.53Ga0.47As(001)-(2x4) surface. Density Functional Theory (DFT) modeling of α2(2x4) and β2(2x4) unit cell constructions and electronic structures show that the missing dimer defect creates conduction band edge states not readily passivated by trimethy aluminum; therefore, the density of dimer defects may cause trap state formation at oxide/InGaAs(001) interfaces.

Air exposed III-V surfaces nearly always have electronic defects which prevent full modulation of the Fermi level thereby impeding their use in practical semiconductor devices such as metal oxide field effect transistors (MOSFETs). For a high speed device, the air induced defects and contaminants need to be removed to reduce trap states while maintaining an atomically flat surface to minimize interface scattering thereby maintaining a high carrier mobility. Using in-situ atomic scaling imaging with scanning tunneling microscopy, a combination of atomic hydrogen dosing, annealing and trimethyl aluminum dosing is observed to produce an ordered passivation layer on air exposed InGaAs(001)-(4×2) surface with only monatomic steps.

Kelvin probe force microscopy (KPFM) is a unique technique that can provide two dimensional potential profiles inside a working device. A procedure is described to obtain high-resolution KPFM results on ultra-high vacuum (UHV) cleaved III-V MOSCAPs. Two tip preparation methods: field emission and Cr coating show reproducible high spatial and energy resolution KPFM images. A unique sample design has been developed which is compatible with UHV cross-sectional KPFM (x-KPFM). Key design features are high density of devices on the cleave face, a buried device interface, and a cleavable gate contact. Using x-KPFM, the first UHV cleaved MOSCAP surface potential mapping is demonstrated.

The monolayer passivation of Ge(100) surface via formation of Ge-N and Ge-O surface species was studied using scanning tunneling microscopy (STM) and density functional theory (DFT). Direct niridation using an electron cyclotron resonance (ECR) plasma source formed an ordered Ge-N structure on a Ge(100) surface at 500oC. DFT calculation found the hydrogen passivation on this Ge-N ordered structure could reduce the bandgap states by decreasing the dangling bonds and the bond strain. Oxidation of Ge(100) using H2O produced an -OH and -H terminated surface with very few Ge ad-atoms, while e-beam evaporation of GeO2 formed semi-ordered Ge-O structures and Ge ad-species at room temperature. Annealing above 300oC formed suboxide rows on both H2O and GeO2 dosed surfaces, and the scanning tunneling spectroscopy (STS) showed that the Fermi level was pinned near the valence band edge on the n-type Ge surfaces covered by suboxides.

Titanium nitride (TiN) has been extensively studied in semiconductor devices because of its ideal thermal, mechanical, and electrical properties, along with its ability to act as a diffusion barrier to WF6 during W metal fill. Similarly, tantalum nitride (TaN) is being utilized as a diffusion barrier on SiOCH to Cu, as Cu can readily diffuse, causing device reliability issues. Metal halide precursors are typically preferred over organometallic grown films when there is no concern about substrate etching; however, organometallic-grown films usually contain higher levels of carbon and oxygen contamination, which has been correlated with an increase in film resistivity. Plasma enhanced-ALD TiN has been shown to achieve optimal growth rates with lower contamination at temperatures below 350°C, but the film and underlying substrate can suffer from plasma-induced damage. In this study, low temperature thermal ALD TiNx from anhydrous N2H4 and TiCl4 was performed on SiO2, and the deposited films were studied using x-ray photoelectron spectroscopy (XPS) and atomic force microscopy (AFM). TaNx films were grown utilizing N2H4 and tris(diethylamido)(tertbutylimido) tantalum(V) (TBTDET) and characterized similarly. In situ XPS was employed to demonstrate pulse saturation during ALD for TiCl4 + N2H4 à TiNx. AFM showed that 40 cycles of TiCl4 + N2H4 at 300°C produces pinhole-free, conformal films. Additionally, the effect of air exposure on TiNx films was studied; O attacked the TiNx film, evidenced by observing a 1.5 eV BE shift of the Ti 2p 3/2 peak. A comparison was made of the TiNx films grown at 300ºC and 400ºC with NH3 vs. N2H4; even at 400ºC there was approximately 2x more O and C and 50% more Cl in NH3 grown films. Furthermore, N2H4 films showed lower resistivities, attributed to lower contamination and likely better nucleation density, especially at 300oC. In situ XPS was also used to study TaNx films grown using TBTDET and N2H4. To study the nucleation on SiO2, an initial TBTDET exposure was performed; the Ta 4d peak position confirmed the nucleation and formation of Si-O-Ta bonds. After subsequent cycles, the Ta 4d peak shifts ~2eV toward lower binding energy upon forming Ta-N bonds. While the films were of extremely high purity with stoichiometry resembling Ta3N5, the N2H4 was unable to reduce the Ta oxidation state from films grown between 100oC and 300oC showing the need for some metals for the use of low oxidation state precursors or a stronger, highly energetic reducing agent. These studies with anhydrous N2H4 show the potential to form high purity barrier layers using low temperature thermal ALD as long as highly energetic reactants are used.

Silicon germanium (SiGe) channels for CMOS are favorable due to high intrinsic carrier mobility and band gap tunability[1] To utilize the superior properties of SiGe, low interface defect density (Dit) must be obtained at the SiGe-high k interface. Germanium oxide (GeOx) is the primary source of interface defects[2]. In comparison with GeOx, SiOx has a higher heat of formation. Using this difference, it is possible to reduce interface defects with selective reduction or diffusion of GeOx by employing an oxygen-scavenging metal as a gate material[4,5] such as aluminum. ALD was performed with TDMAH and H2O to grow 50 cycles (5nm) of HfO2 at 250C. 50 nm thick, 150 um diameter Ni or Al gate metal was deposited by thermal evaporation. C-V measurement were employed to characterized the electronic defects. The interface trap density (Dit) integrated across the band gap was more than an order of magnitude lower for Al vs Ni gated MOSCAPs. From TEM-EELS-EDS measurement, Al gated MOSCAPs were observed to have a very thin interfacial oxide layer (2/SiOx/SiGe interfaces can be defect-free. DFT models also show that two 3- and 5-fold coordinated interfacial Si atoms do not create any mid-gap or band-edge states. These results indicate the need to form a SiOx interface between HfO2 and SiGe, but the same interface may be created with other techniques. The alternative techniques include a selective oxidation of SiGe form SIOx using remote O3 oxidation and ALD of an Si rich layer on SiGe. The key issue is which of the three techniques is manufacturable and can produce both a scale oxide and scaled interlayer. [1] Liu, C. et. al, MRS Bull. 39, 658–662 (2014). [2] Zhang, L. et al, ACS Appl. Mater. Interfaces 8, 19110–19118 (2016). [3] Zhang, L. et al, ACS Appl. Mater. Interfaces 7, 20499–20506 (2015). [4] Kim, H. et al, J. Appl. Phys. 96, 3467–3472 (2004). [5] Frank, M. et al, ECS Solid State Lett. 2, N8–N10 (2012). [6] Sardashti, K. et al. Appl. Surf. Sci. 366, 455–463 (2016)

The mechanism of stability of the phases of HfO2, ZrO2, and HZO (HfxZr1-xO2) were systematically investigated with density functional theory molecular dynamics (DFT-MD) to determine the mechanism for HZO having a much larger process window for formation the ferroelectric phase as compared to doped HfO2 or ZrO2. For the bulk states, the monoclinic phase (“m”) is about 80 mV per formula unit more stable than either the orthorhombic ferroelectric (“f”) phase or tetragonal (t-phase) for all three oxides. The surface free energies of the (001), (110), and (111) surfaces of all three oxides were calculated using an identical DFT technique. For all three oxides, the (111) face has the lowest surface free energies consistent with experimental data on columnar HZO grains showing [111] is the preferred growth direction. However, the surface free energy for all direction are nearly degenerate between HfO2, ZrO2, and HZO; therefore, even for nanocrystal formation the surface free energy does not favor f-phase formation. The effect of stress/strain was calculated by determining the free energy of formation as a function of the volume of the unit cell. When the oxides are grown in the low density amorphous phase but a post deposition anneal is perform for crystallization. The crystalline forms are more dense than the amorphous forms and the DFT calculation show that a higher surface area per unit cells will greatly favor f-phase formation. However, the effect is nearly identical for HfO2, ZrO2, and HZO; this is consistent with experiments showing the molar volumes of HfO2 and ZrO2 being within 2%. Instead, formation of nanocrystalites is hypothesized to be the source of the enhanced processing window for HZO. Experimental data is consistent with partial phase separation in HZO. Atom probe tomography imaging of the chemical composition of TiN/5 nm HZO/Si(001) ferroelectric films show an asymmetric distribution of the Hf and Zr within the HZO layer with the Zr being concentrated near the TiN/HZO interface; this is consistent with ZrO2 having a 100C lower crystallization temperature than HfO2 and therefore initiate the crystallization starting on the TiN(111) surface. It is hypothesized that the nanocrystals which template on TiN(111) can produce the interfacial stress/strain needed to stabilize f-phase formation; high resolution TEM shows regions of epitaxial alignment between HZO and TiN consistent with this mechanism. In addition atom probe tomography (APT) was performed on TiN/HZO/Si structures to determine the film composition of the interfaces for indication of possible phase separation of HZO since phase separation could promote nanocrystal formation. Funding by LAM Research is gratefully acknowledged.

The surface reconstructions of decapped In0.53Ga0.47As(001) have been studied using scanning tunneling microscopy (STM). It is shown that the As-rich α2(2×4) and β2(2×4) reconstructions, predicted by density function theory (DFT) (1-3) for GaAs(001)-(2×4), InAs(001)-(2×4) and InGaAs(001)-(2×4) surfaces, were observed to coexist on In0.53Ga0.47As (001). In contrast to molecular beam epitaxy (MBE) grown In0.53Ga0.47As(001), the STM results on decapped In0.53Ga0.47As(001) do not show the existence of the heterodimer In0.53Ga0.47As(001)-(4×3) structure (4-5). At the intermediate annealing temperature ranges of 400 - 440oC, a (2×4)-(4×2) mixed surface reconstruction was observed. When In0.53Ga0.47As(001)/InP sample was annealed between 440oC and 470{degree sign}C, a pure In/Ga-rich (4×2) surface reconstruction was observed.

The rapid scaling of complementary metal oxide semiconductor (CMOS) technology requires substituting the traditional gate oxide, SiO2, with highdielectrics, which can maintain the same capacitance with much lower leakage current. Amorphous aluminum and zirconium oxides (a-Al2O3 and a-ZrO2) are leading candidates for such highgate oxide materials on Ge. Ge is one of a few semiconductors that offer significantly higher hole mobility than silicon and is being extensively investigated for p-channel high-k MOSFETs. (1-3).

Molecular and electronic structures of In2O and Ga2O bonding to the As-rich In0.37Ga0.63As(001)-(2×4) surface were investigated using density functional theory (DFT) modeling. Calculated enthalpies of adsorption revealed that the bonding geometries of In2O and Ga2O on In0.37Ga0.63As(001)-(2×4) were essentially identical to those on GaAs(001)-(2×4). However, the calculated electronic structures resulting from the adsorbates/semiconductors systems exhibited subtle differences. Both oxides induced state density in the band gap consistent with Fermi level pinning on In0.37Ga0.63As(001)-(2×4) and GaAs(001)-(2×4). In addition, the state density worsened with increasing adsorbate coverage. However, a greater adsorbate density was required for the onset of Fermi level pinning for InGaAs than GaAs. This effect was attributed to reduction in band gap for In0.37Ga0.63As(001)-(2×4) vs. GaAs(001)-(2×4). Variations in the alloy structure of InGaAs (i.e. subsurface locations of In and Ga atoms) were found to play only a minor role on the resulting oxide adsorbate bonding geometries and electronic structures.

not Available.

not Available.

not Available.

Oxide monolayers and submonolayers formed by vapor deposition of In2O and SiO oxides on InAs001-4 2 were studied by scanning tunneling microscopy. At low coverage, In2O molecules bond to the edges of the rows and most likely form new In–As bonds to the surface without any disruption of the clean surface structure. Annealing the In2O/InAs001-4 2 surface to 380°C results in the formation of flat ordered monolayer rectangular islands. The annealed In2O no longer bonds with just the As atoms at the edge of row but also forms new O–In bonds in the trough. SiO chemisorption on InAs001-4 2 is completely different than In2O chemisorption. At room temperature, even at low coverage SiO adsorbates bond to themselves and form nanoclusters. For SiO/InA001-4 2 postdeposition annealing does not disperse the nanoclusters into flat islands. Both In2O and SiO depositions on the InAs001-4 2 surface do not displace surface atoms during both room temperature deposition and postdeposition annealing.

not Available.

not Available.

not Available.