Atomic layer deposition of Sn-doped germanium diselenide for an As-free Ovonic threshold switch with low off-current

Abstract

This work introduces the atomic layer deposition (ALD) of Sn-doped GeSe₂ (SnGS2) films for developing arsenic-free Ovonic threshold switch (OTS) devices with low off-current (I_off) and desirable threshold voltage (V_t). The undoped GeSe₂ film exhibits high V_t and low endurance characteristics despite its low I_off originating from a large mobility gap. Such problems could be overcome by appropriate doping, and thus, Sn was doped into the GeSe₂ film using a supercycle consisting of SnN_x and GeSe₂ subcycles. The co-injection of NH₃ with Se-precursor ([(CH₃)₃Si]₂Se) during the GeSe₂ subcycle generated a reactive H₂Se intermediate, which enhanced the reactivity and transformed the pre-deposited SnN_x to SnSe₂. The overall cation-to-anion ratio of the deposited films remained at approximately 1 : 2, with amorphous structure and smooth surface morphology. The planar OTS device using SnGS2 films with a Sn concentration of 9.6%, for which the crystallization temperature was >400 °C, showed the lowest V_t of ∼3.5 V and I_off of ∼12.5 nA at half V_t and demonstrated switching endurance over 10⁶ cycles. Furthermore, a vertical-type OTS device was fabricated using the same SnGS2 film using the excellent step coverage of ALD, exhibiting similar switching characteristics to its planar counterpart.

---

I. Introduction

The Ovonic threshold switch (OTS), discovered by Ovshinsky in the 1960s, is a two-terminal volatile switch device based on amorphous chalcogenide film.¹ It exhibits threshold switching behavior with a high on-current (I_on), large selectivity, moderate threshold voltage (V_t), and rapid switching speed. These properties have garnered significant interest in the OTS for its potential application as a selector device in non-volatile memory-based crossbar arrays, where it can effectively suppress sneak currents, and as a neuron device in spiking neural networks.^2–6 Additionally, a selector-only memory that utilizes the non-volatile polarity-dependent V_t shift of the OTS itself as memory is also being investigated.^7,8 However, the increasing demand to further enhance integration density drives the evolution of applications towards more complex three-dimensional structures, such as vertical crossbar array (V-CBA) architectures.^9,10 Therefore, developing a method that deposits OTS materials conformally on complex structures is necessary to maintain their high performance, which the currently prevalent physical deposition method, such as sputtering, cannot achieve.^11–14

Among various potential OTS materials, selenide-based OTS materials have attained significant attention due to their low off-current (I_off) and moderate switching voltages. Ge_xSe_1−x has been extensively studied for its high amorphous stability and stable OTS characteristics, observed in various compositions.^15,16 The Ge-rich compositions result in a decreased mobility gap (E_g), leading to higher I_off. In contrast, Se-rich compositions have larger E_g and lower I_off but higher forming voltage (V_f) and V_t. Various strategies for doping into the Ge_xSe_1−x matrix have been explored to enhance the OTS properties further. Arsenic (As) doping has been actively pursued for its superior OTS properties.^17,18 However, the high toxicity of As poses significant environmental concerns, requiring the exploration of As-free OTS materials. Sb and N doping into GeSe₂ was explored to lower V_t while maintaining the amorphous stability of the Se-rich composition.¹⁹ Despite these advances, research on the deposition method of OTS materials has been predominantly focused on sputtering.^20–22 While sputtering is suitable for planar structures, it has limitations in depositing films on the V-CBA structure, where the OTS film must be conformally deposited on the sidewalls of the etched holes. Therefore, the atomic layer deposition (ALD) method, providing excellent step coverage in high-aspect-ratio structures through self-limited surface reaction, is required to meet the demands of these advanced structures. However, compared with the ALD of oxide materials, the ALD process of chalcogenides requires a more careful selection of precursors and process design to facilitate efficient ALD reaction.²³ Although various materials have been explored for As-free OTS, few have demonstrated promising electrical performance when deposited via ALD, highlighting the need for further research.

This work demonstrated the ALD of Sn-doped GeSe₂ (SnGS2) for an As-free OTS with low I_off properties. GeSe₂ was the base material due to its large E_g and high crystallization temperature.^15,24 Previous research showed that doping with Sn, an Earth-abundant and eco-friendly element, increases the density of trap states, thereby lowering V_f.²⁵ In this study, Sn doping was utilized to enhance the GeSe₂-based OTS performance further while maintaining high amorphous stability. Additional N doping was also attempted to decrease I_off further. The deposition of SnGS2 was attempted by utilizing an ALD supercycle composed of GeSe₂ and SnN_x subcycles, using tetrakis(dimethylamino) germanium(IV) (Ge[N(CH₃)₂]₄, TDMA-Ge), tetrakis(dimethylamino) tin(IV) (Sn[N(CH₃)₂]₄, TDMA-Sn), and bis(trimethylsilyl) selenide ([(CH₃)₃Si]₂Se, BTMS-Se) as the precursors for Ge, Sn, and Se, respectively. NH₃ injection was essential for GeSe₂ deposition since it formed a reactive intermediate, H₂Se, by co-injection with BTMS-Se. By adjusting the subcycle ratio of SnN_x and GeSe₂, SnGS2 films of various compositions were feasibly deposited. The SnGS2 films exhibited excellent morphology and were conformally deposited on high-aspect-ratio structures, indicating their applicability as a device for V-CBA architecture. Notably, during the subcycles of GeSe₂, the thin SnN_x layer was converted to SnSe₂ due to highly reactive H₂Se, diminishing the effect of N doping. Despite this compositional change, the resulting SnGS2 films maintained their amorphous structure even at 400 °C without crystallization at Sn concentrations below ∼10 at%. The chemical properties of three films with compositions within this Sn concentration range were examined by X-ray photoelectron spectroscopy (XPS), Raman spectroscopy, and spectroscopic ellipsometry (SE). The electrical characterizations were performed on the planar and vertical OTS device structures. The V_f, V_t, and current for on, off, and initial states were measured and analyzed for different Sn doping concentrations. A pulsed cyclic switching endurance test was conducted to evaluate the performance of the OTS devices, demonstrating endurance up to 10⁶ cycles with an I_on/I_off ratio exceeding 10⁴. The feasible fabrication and analysis of the vertical-type device confirmed the suitability of ALD SnGS2 films for applications in complex three-dimensional architectures.

II. Experimental

Atomic layer deposition of Sn-doped GeSe₂ films

Film deposition was performed on SiO₂ and TiN film substrates at a substrate temperature of 110 °C. 100 nm-thick SiO₂ film substrates were prepared by thermal oxidation of a p-type Si (100) wafer, and TiN film substrates were prepared by sputtering 5 nm-thick Ti and 50 nm-thick TiN layers on the SiO₂ substrates (Applied Materials, Endura 5500). Before the ALD process, the substrates were sonicated in acetone and isopropyl alcohol to remove organic contaminants. The films were deposited using showerhead-type ALD process equipment with a 12-inch diameter showerhead and an 8-inch-wafer-scale-compatible substrate heater (CN1, Plus-200). All ALD lines were heated to 100 °C to prevent precursor gas condensation. The Ar gas carried precursors into the ALD chamber with a flow rate of 50 standard cubic centimeters per minute (sccm). The TDMA-Ge and TDMA-Sn precursors, whose canisters were maintained at 40 °C, were carried into the chamber using the bubbling method. The BTMS-Se precursor, kept in a canister at room temperature, was delivered via the vapor draw method. NH₃ gas with a flow rate of 200 sccm was directly injected into the chamber. After each precursor injection sequence, 200 sccm of Ar gas was injected to purge the remaining unreacted species and byproducts. The working pressure of the chamber was maintained within the range of 1.2–1.5 torr and 4.0–5.0 torr during the SnN_x ALD and GeSe₂ ALD processes, respectively.

Film characterization

The composition and layer density of the deposited films were determined using energy-dispersive X-ray fluorescence (EDXRF, Thermo Scientific, Quant'X EDXRF). The bulk density of the film was measured through X-ray reflectivity (XRR, PANalytical, X'Pert Pro MPD), and the thickness of the film was obtained by dividing the layer density by the bulk density. Due to the detection limits of EDXRF for nitrogen, the ALD behavior of SnN_x films was evaluated by measuring the layer density of Sn. The composition of the SnN_x film was confirmed to be SnN_0.29 by X-ray photoelectron spectroscopy (XPS, Versaprobe III, Ulvac-PHI) through a comparison of the peak areas of the Sn and N spectra. Atomic force microscopy (AFM, Park Systems, NX10) was used to observe the morphology of the film. The thickness and composition of the film deposited on the hole pattern of 20 : 1 aspect ratio were confirmed by transmission electron microscopy (TEM, JEOL, JEM-ARM200F(NEOARM)) at 200 kV accelerating voltage and energy-dispersive X-ray spectroscopy (EDS, Oxford Instruments, Ultim Extreme), respectively. Before the TEM sampling, an aluminum oxide layer was deposited on the chalcogenide film using ALD to inhibit its oxidation by air. The K-line signal for Ge and Se and the L-line signal for Sn were used for the elemental mapping. The impurity levels and the chemical state of the elements in the films were analyzed by XPS with an Al Kα source. Raman spectra of the films were obtained to analyze the bonding structures of the films. A laser of wavelength 633 nm (HORIBA, LabRAM HR Evolution) with a focused spot size of 200 μm was used. The spectra were calibrated using the characteristic peak of Si at 520.7 cm⁻¹ from an undoped Si wafer. Spectroscopic ellipsometry (SE, J.A. Woolam, ESM-300) was used to measure the optical E_g of the film. The thermal stability of the films was tested through thermal annealing under an air atmosphere and temperature range of 320 °C to 400 °C. The crystallization temperature of the films was examined by an X-ray diffractometer (XRD, PANalytical, X'Pert PRO MPD) with a Cu Kα radiation source. The grazing angle incidence XRD (GIXRD) scans covered 2θ values ranging from 10° to 60°, with an incidence angle of 2°.

Device fabrication and electrical measurement

The planar OTS device for fundamental electrical characterization was fabricated from a sputtered 100 nm-thick W bottom electrode covered by a 50 nm-thick plasma-enhanced chemical vapor-deposited (PECVD, Oxford Instruments, PlasmaPro System100) SiO₂ layer. To make the electrical contact between the bottom electrode and the OTS film, a 1 μm diameter hole was patterned using photolithography (NanoSystem Solutions. Inc., DL-1000 HP) followed by dry etching of the SiO₂ layer (Oxford Instruments, RIE 80 plus). In the case of a vertical device, a stack comprising a 100 nm-thick W wordplane sandwiched between two dielectric layers (SiO₂) was etched to make a 1 μm diameter hole. The stack of the alternating metal wordplane and SiO₂ layers was precisely etched using different etch recipes and facilities (for SiO₂, Gigalane, NeoS-MAXIS 200L; for the W layer, Oxford Instruments, PlasmaPro System100 Cobra). Subsequently, the ALD SnGS2 film was deposited on both substrates with a thickness of ∼15 nm by adjusting the number of supercycles. The top electrode was formed by depositing a 50 nm-thick amorphous carbon layer and a 50 nm-thick Mo film through sputtering. A TiN thin film load resistor was also integrated into the bottom electrode to suppress the overshoot current during threshold switching. The resistor with a value of 19.7 kΩ was employed for all the measurements. The electrical contact pads were fabricated via the lift-off process using an evaporated Ti (5 nm)/Al (50 nm) film deposited by an electron-beam evaporator (Sorona, SRN200i). Fig. S1 of the ESI† shows the schematic diagrams of the planar and vertical OTS devices. The bird's-eye view images of the OTS device were obtained using field emission scanning electron microscopy (FE-SEM, Hitachi, S-4800) with a 25° chuck tilt.

The electrical properties were measured using a Keysight B1500A facilitated by a B1530A waveform generator/fast measurement unit with a remote-sense and switch unit. Due to the sensitivity of the OTS device toward DC electrical stress, all AC current–voltage (I–V) measurements were performed using triangular pulses with a rise/fall time of 2.5 μs. The amplitude of the voltage applied to the OTS device was 6 V (10 V for initial electroforming only), and the V_f and V_t were extracted when the current was increased above 50 μA. The I_on and I_off were also measured by the B1530A unit using the appropriate current measurement range for the pulsed cyclic endurance test. The low current of the subthreshold conduction was measured in DC voltage sweep mode using a B1517A high-resolution source/measurement unit. Unless otherwise indicated, all measurements were performed at room temperature.

III. Results and discussion

Fig. 1(a) schematically shows the chemical structures of the precursors for the ALD of SnN_x and GeSe₂ films. SnN_x was deposited at 110 °C using TDMA-Sn and NH₃.²⁶ Fig. S2 in the ESI† shows the ALD-specific saturation behaviors for the injection and purge times of TDMA-Sn and NH₃. The appropriate injection and purge times of each element were determined as being 3/20 s (TDMA-Sn injection/purge) and 4/30 s (NH₃ injection/purge). Fig. 1(b) shows the linear growth behavior according to the number of cycles under saturation conditions. These data confirm that the adopted ALD conditions can deposit SnN_x film stably.

Fig. 1 (a) Chemical structures of TDMA-Sn, TDMA-Ge, and BTMS-Se with NH₃ for SnGS2 ALD. Linear growth behavior of (b) SnN_x and (c) GeSe₂ films as a function of the number of ALD cycles under saturation conditions. The left-hand y-axis represents layer density, and the right-hand y-axis represents atomic composition. (d) Gas injection sequence for the ALD of SnGS2 films.

Next, the ALD behavior of the GeSe₂ films was examined. The co-injection of NH₃ with BTMS-Se during the deposition of GeSe₂ was essential, as the conventional ALD sequence without NH₃ gas injection could not grow films due to the limited reactivity between TDMA-Ge and BTMS-Se. These findings are consistent with the previous results that deposited SnSe_x film using TDMA-Sn and BTMS-Se.²⁷ The following reactions (1) and (2) represent the proposed mechanism for the deposition of GeSe₂ film:

H₂Se + Ge[N(CH₃)₂]₄ → GeSe₂ + 4(NH)(CH₃)₂↑ (2)

Reaction (1) represents the gas-phase reaction, where NH₃ reacts with BTMS-Se to form H₂Se, a highly reactive intermediate that is subsequently adsorbed onto the surface. This highly reactive species easily interacts with the subsequently injected TDMA-Ge via the ALD-specific ligand exchange reaction, resulting in the GeSe₂ film growth following the reaction (2). Fig. S3 in the ESI† shows the self-limited growth of the ALD for GeSe₂ film under sufficient injection- and purge-time conditions. The optimal injection and purge times of each element were determined to be 3/30 s (TDMA-Ge injection/purge) and 3/70 s (BTMS-Se with NH₃ injection/purge). The constant composition ratio of Ge : Se = 1 : 2 under saturation conditions indicates that the valences of Ge and Se ions were well maintained at +4 and −2, respectively, following the reaction eqn (2). Fig. 1(c) illustrates the variations of the layer density and composition of the GeSe₂ film as a function of cycle numbers. On both SiO₂ and TiN substrates, the films grew linearly with a fitted slope of 28.2 ng cm⁻² per cycle (0.71 Å per cycle) and 25.8 ng cm⁻² per cycle (0.65 Å per cycle), respectively, while maintaining the 1 : 2 composition. The incubation cycles for each substrate were 29 cycles and 40 cycles, respectively, indicating the presence of the nucleation barrier. The optimized deposition conditions for SnN_x and GeSe₂ films were used to deposit SnGS2 films using each cycle as a subcycle for a single supercycle, as illustrated in Fig. 1(d).

Fig. 2(a) and (b) show that the composition of the films could be precisely controlled by altering the ratio of SnN_x to GeSe₂ subcycles, particularly regarding the Ge and Sn concentrations. In contrast, the Se concentration remained relatively constant at ∼67% regardless of the subcycle ratio, indicating that the ratio of (Sn + Ge) : Se is maintained at 1 : 2. Fig. 2(c) shows the composition of SnGS2 films in an Sn–Ge–Se ternary diagram when changing the SnN_x : GeSe₂ subcycle ratio, as shown in Fig. 2(a) and (b). Most compositions lie on the GeSe₂–SnSe₂ pseudobinary tie line, not the GeSe₂–Sn (or SnN_x) tie line. If the film growth in each subcycle were unaffected by the other, an increase in the SnN_x subcycle ratio would have increased the overall cation (Sn + Ge) concentration. However, the constant (Sn + Ge) concentration suggests that the highly reactive precursors of the GeSe₂ subcycle affect the chemical state of the underlying thin SnN_x layer, leading to a deviation from the expected behavior. A detailed analysis of this phenomenon is presented in the subsequent section.

Fig. 2 Variation in film composition as a function of (a) the number of SnN_x subcycles (m) from 1 to 3 with the number of GeSe₂ subcycles fixed to 5 and (b) the number of GeSe₂ subcycles (n) from 5 to 15 with the number of SnN_x subcycles fixed to 1. (c) Ternary phase diagram of the ALD SnGS2 films as a function of subcycle ratio. The red and blue dashed lines indicate the pseudobinary tie lines of the GeSe₂–SnSe₂ and GeSe₂–Sn(N_x) systems, respectively.

Fig. 3(a) shows the variation in layer density for each film as a function of the number of supercycles. The SnN_x and GeSe₂ subcycle ratios were 1 : 10, 1 : 5, and 2 : 5, referred to as [1−10], [1−5], and [2−5]. The corresponding Sn atomic compositions were approximately 3.5, 5.5, and 9.6%, respectively. The three films exhibit linear growth on both SiO₂ and TiN substrates, with a growth rate of 11, 6.2, and 8.4 Å per supercycle, respectively, based on the bulk density of each film, as shown in Fig. S4 of the ESI.†Fig. 3(b) shows the root-mean-squared (RMS) roughness values of the 15 nm-thick GeSe₂ and SnGS2 films of three different compositions measured by AFM. Fig. S5 in the ESI† shows the AFM images for all films. Although GeSe₂ films exhibited a smooth morphology with RMS values of 0.51 and 0.63 nm on SiO₂ and TiN substrates, respectively, SnGS2 films showed improved roughness with RMS values below 0.21 and 0.36 nm on the respective substrates.²⁸ The improved roughness of the SnGS2 films is attributed to the shorter incubation cycles required for the nucleation, which leads to relatively lower RMS values than those for GeSe₂ films.

Fig. 3 (a) Variation in layer density of the ALD SnGS2 films as a function of the number of supercycles. (b) Comparison of film roughness of the GeSe₂ and SnGS2 films measured by AFM. The pattern of the bar graph distinguishes the substrates for SiO₂ (none) and tin (hatch).

The step coverage of the SnGS2 films produced by the ALD process over the deep hole structure was confirmed by the cross-sectional TEM and EDS analyses. Fig. 4(a) shows that the SnGS2 film was conformally deposited on the ∼1 : 20 aspect ratio hole pattern (120 nm hole diameter and 2500 nm depth). The film deposited was an SnGS2 film with a [2−5] subcycle ratio, and an additional layer of aluminum oxide was deposited via ALD to prevent oxidation during atmospheric exposure prior to analysis. Fig. 4(b) presents the magnified images of the top, middle, and bottom parts of the hole structure, demonstrating the conformal growth of the film. Fig. 4(c) illustrates the EDS line profile analysis results along the horizontal line in Fig. 4(b). Although the data were noisy, the values approximately coincided with the overall composition expected from the subcycle ratio of [2−5]. Fig. 4(d) shows the fast Fourier transform result obtained from the selected area of the SnGS2 film, confirming its amorphous nature.

Fig. 4 (a) Cross-sectional TEM image of SnGS2 film deposited on a contact hole structure with an aspect ratio of 20 : 1. Scale bar, 300 nm. (b) Magnified images of the top, middle, and bottom parts of the hole structure to confirm the conformal growth of the SnGS2 film. Scale bar, 50 nm. (c) EDS line profile obtained from the yellow line in (b) indicated in the magnified TEM image. (d) Fast Fourier transform result of the ALD SnGS2 film obtained from the selected area indicated in the middle image of (b).

Fig. 5 shows the Ge 3d, Sn 3d_5/2, Se 3d, and N 1s XPS spectra identifying the chemical state of the ALD SnGS2 films. The analysis was performed after in situ Ar⁺ ion sputtering in the XPS chamber to remove the oxidized layer from the surface of the film. Fig. S6 of the ESI† shows that impurity-induced signals were negligible, except for the Se Auger signal detected within the C 1s spectral range. This low impurity level indicates that the ligands of the precursors were sufficiently removed during film growth. The Ge 3d peaks of all SnGS2 films were located at 31.9 (3d_3/2) and 31.3 eV (3d_5/2) regardless of the Sn concentration, corresponding to the binding energy of GeSe₂.²⁹ Notably, the Sn 3d peaks for all compositions were located at 486.5 eV, consistent with the binding energy of SnSe₂, while no N 1s peaks were observed in any of the films.²⁷ Also, the Se 3d peaks of all SnGS2 films were deconvoluted into two chemical states corresponding to the binding energies of GeSe₂ (3d_3/2: 55.7 eV; 3d_5/2: 54.9 eV) and SnSe₂ (3d_3/2: 55.2 eV; 3d_5/2: 54.4 eV).^27,29,30 The area ratio of these two peaks was consistent with the atomic concentration ratio of Ge and Sn. These findings suggest that the pre-deposited SnN_x layer undergoes a chemical reaction with the Se precursor during the subsequent GeSe₂ subcycle, resulting in SnSe₂. This transition is also supported by comparing the XPS spectra of SnN_x and SnGS2 films shown in Fig. S7 of the ESI.† The Sn 3d peak is located at 484.6 eV, and an evident N 1s peak was observed from the SnN_x film. However, the Sn 3d peak was shifted toward higher binding energy, and the N 1s signal disappeared in the SnGS2 film, indicating that SnN_x was converted to SnSe₂. As described in reaction (1), co-injection of BTMS-Se with NH₃ leads to the formation of H₂Se, which has been reported to be an effective selenizing agent to promote the deposition of various selenide films in previous studies.^31–33 Considering the same Se concentration and the XPS results of the ALD SnGS2 films, the highly reactive H₂Se selenizes the SnN_x layer during the subsequent GeSe₂ subcycle. This selenization process was further analyzed by varying the number of co-injection and purge cycles of BTMS-Se and NH₃ after the initial deposition of the SnN_x layer of 50 cycles, as shown in Fig. S7(c) of the ESI.† The Se injection and purge times were consistent with those used during the GeSe₂ subcycle. As the number of co-injection cycles increased, the increase in Se layer density was initially rapid, converting 10% of the Sn in SnN_x to SnSe₂ within 5 cycles. Subsequently, the incorporation rate of Se gradually slowed down after 20% of the Sn in the SnN_x layer was converted to SnSe₂. In the SnGS2 films studied in this work, only 1 or 2 SnN_x subcycles were employed, which render the thin SnN_x layers readily selenized during the GeSe₂ subcycle.

Fig. 5 XPS spectra of the ALD SnGS2 films. (a) Ge 3d, (b) Sn 3d_5/2, (c) Se 3d, and (d) N 1s peaks, respectively.

The formation of SnSe₂ also affects the amorphous network structure of SnGS2 films. Fig. 6(a) and (b) illustrate the vibration modes and the Raman spectra of 15 nm-thick GeSe₂ and SnGS2 films. The main feature, located near 200 cm⁻¹ and designated as I, II, and III in Fig. 6(b), corresponds to the vibration modes of XSe₂ (X = Ge, Sn). The peak located at 185 cm⁻¹ corresponds to the out-of-plane A_1g vibrational mode of SnSe₂ (I).³⁴ The peaks at 198 cm⁻¹ and 216 cm⁻¹ are attributed to GeSe₂, specifically to the A₁ vibrational mode of the corner-sharing GeSe_4/2 tetrahedral unit (II) and the A_c1 vibrational mode of the edge-sharing Ge₂Se_8/2 bi-tetrahedral unit (III), respectively.^35–39 As the Sn doping concentration increased, the intensity of peaks II and III decreased while the intensity of peak I increased. Consequently, the central position of the main peak shifted to a lower wavenumber. This trend coincides with the Se 3d peak observed in the XPS spectra, where the SnSe₂ peak became prominent with increasing Sn concentration. The broad peak between 250 and 280 cm⁻¹ (IV) and the peak at 302 cm⁻¹ (V) correlate with the homopolar bonds of Ge and Se, which can be identified as the Se chain and Ge LO mode.^35,38,40,41 The intensity of these peaks, especially that for V, reveals a decreasing trend with increasing Sn concentration. This trend suggests that Sn doping reduced the homopolar bonds and generated other stable heteropolar bonds, Sn–Se, compared with undoped film.

Fig. 6 (a) Schematic illustration of the vibration modes corresponding to the Raman peaks of GeSe₂ and SnSe₂. (b) Raman spectra of GeSe₂ and SnGS2 films. The characteristic peaks are denoted by roman numerals. (c) Se measurement results of GeSe₂ and SnGS2 films, with the mobility gap estimated from the Tauc plot. (d) GIXRD measurement results of the 15 nm-thick SnGS2 films annealed at 400 °C.

Structural changes in the film caused by varying the subcycle ratio also affected the optical E_g. Fig. 6(c) shows the SE measurement results for the 15 nm-thick GeSe₂ and SnGS2 films following the Tauc method. The optical E_g values were estimated by extrapolating the Tauc plot to the x-axis. The GeSe₂ film exhibited a large E_g of 2.10 eV. As the Sn concentration increased, the E_g values decreased to 1.98, 1.90, and 1.83 eV for the [1−10], [1−5], and [2−5] films, respectively. The decrease in optical E_g with increased Sn concentration can be attributed to the higher concentration of SnSe₂ species, which possess a narrower E_g compared to that of GeSe₂. Although the E_g value decreased, it remained higher than that of Ge-rich Ge_xSe_1−x and GeS, suggesting the potential for low I_off characteristics.^13,15,22

Fig. 6(d) shows the GIXRD spectra of the 15 nm-thick SnGS2 films with the three Sn concentrations after the annealing at 400 °C for 30 minutes under an ambient atmosphere. A 5 nm-thick TiN layer and 20 nm-thick Al layer were sequentially sputtered as capping layers to prevent selenium evaporation during annealing. Consequently, all XRD spectra showed an Al peak at 38.5°, marked by a gray dotted line. Diffraction peaks from the chalcogenide material were not observed in all the films. However, when the Sn concentration was further increased to 15.3 at% by adopting a subcycle ratio of [3−5], the film exhibited a diffraction peak at 30.7°, corresponding to SnSe₂, after annealing at 360 °C for 30 minutes, indicating the onset of crystallization (Fig. S8 in the ESI†).⁴² Therefore, the films grown by the [1−10], [1−5], and [2−5] cycle ratios exhibit high amorphous stability, which is sufficient to withstand the back-end-of-the-line (BEOL) thermal budget.

The electrical OTS performances of the [1−10], [1−5], and [2−5] SnGS2 films were examined using the test setup depicted in Fig. 7(a). All devices require the electroforming step due to the initial state of high insulation. Fig. 7(b) and (c) show the impact of varying Sn concentrations on the electrical characteristics. Both V_f and V_t decrease as the SnN_x subcycle ratio increases, suggesting that the increase of Sn concentration in the SnGS2 film decreases the V_f and V_t values (Fig. 7(b)). The mean values of V_t for each device are 3.95, 3.73, and 3.45 V, with increasing Sn concentration. The current level is also correlated with the Sn concentration (Fig. 7(c)). The current value in the pristine state, observed before electroforming, increases with increasing Sn concentration due to the decrease in the E_g, as shown in Fig. 6(c). The I_off after electroforming also increased with the increasing Sn concentration. These trends are presented in Fig. S9 of the ESI,† which shows representative I–V data for three different SnGS2 OTS devices, highlighting V_t and the I_off at half V_t.

Fig. 7 (a) Schematic diagram of a measurement setup for electrical characterization of OTS devices. Comparison of electrical characteristics as a function of subcycle ratio. (b) V_f and V_t for 10 different devices, with V_t extracted as the mean value from 20 cycles for each device. (c) Off-current measured for the initial state (I_initial, black) at 2.0 V and after forming (I_off, red) at half V_t. On-current (I_on, blue) was measured at 6.0 V. (d) Energy barrier fitted from the subthreshold current and the mobility gap. (e) Pulsed cyclic endurance results for different subcycle ratios.

The carrier transport in the OTS device was modeled using the trap-limited conduction model with the current given by

where I₀ is a pre-exponential factor, E_a,0 is the energy (potential) barrier at zero bias, u_a is the thickness of the OTS film, and β_PF is the barrier-lowering constant according to the Poole–Frenkel model.⁴³Eqn (3) predicts an exponential current decrease with 1/kT. Therefore, the derivative of ln(I) with respect to 1/kT shows a linear dependency on the square root of voltage :

From the best-linear-fitting of the derivative values concerning , E_a,0 can be calculated from eqn (4) at zero bias.⁴⁴ The calculated E_a,0 is plotted as a function of the subcycle ratio in Fig. 7(d). For easier comprehension, E_g values are also plotted on the right-hand y-axis. It can be seen that the E_a,0 after electroforming is smaller than half of the E_g. This observation suggests that the conduction mechanism undergoes a significant change during the electroforming process. Despite the decrease in the energy barrier height, the activation energy decreases with increasing Sn concentration, consistent with the observed trend in V_t and I_off levels shown in Fig. 7(b) and (c). The strong correlation between the temperature-dependent I–V curve and the values derived from eqn (4) is further supported by the fitting results presented in Fig. S10 of the ESI.†Fig. 7(e) shows the impact of doping concentration on pulsed cyclic endurance results. For undoped GeSe₂, the stable threshold switching operation could not be maintained beyond 10³ cycles due to the I_off increase, as shown in Fig. S11 of the ESI,† consistent with the previous literature on OTS using Se-rich films.^19,45,46 In contrast, all SnGS2 compositions surpassed 10³ cycles with an I_on/I_off ratio exceeding 10⁴. In particular, the composition with Sn concentration exceeding 5% enhanced the cyclic endurance to >10⁶ cycles. This enhancement is due to decreased Ge homopolar bonds and the predominance of heteropolar Ge–Se and Sn–Se bonds, which suppress segregation under repeated electrical stress.¹⁹ Furthermore, the endurance results for three devices of each SnGS2 composition are presented in Fig. S12 of the ESI,† providing statistical evidence to support the reliability of the results.

The vertically stackable OTS device using SnGS2 film was demonstrated to examine the viability of a V-CBA architecture. Fig. 8(a) shows the SEM image of the vertical-type OTS device (V-OTS) using 15 nm-thick [2−5] SnGS2 film, a 100 nm-thick W wordplane (or bottom electrode), and a stacked 50 nm-thick amorphous carbon/50 nm-thick Mo vertical bitline (or top electrode). Fig. 8(b) depicts a representative electrical characteristic during a triangular pulse application. The abrupt transition between the on and off state was clearly observed, with a time constant below 40 ns. Fig. 8(c) presents the repeated AC and DC I–V data for the V-OTS device. The device showed a V_t with a mean value of 3.54 V and an I_off of 4.83 nA at half V_t. The off-current density (J_off) for the V-OTS is 1.54 A cm⁻², which is consistent with the value observed in the planar device (1.59 A cm⁻²), although the effective area for the vertical structure is 5 times smaller than that of the planar device. Table S1 in the ESI† summarizes the electrical characteristics of As-free OTS devices fabricated using Se-rich chalcogenide film. The V-OTS device demonstrated in this study shows comparable electrical properties to other As-free OTS materials previously limited to planar devices, including excellent selectivity, low J_off, and robust endurance. These results confirm the suitability of the ALD SnGS2 films for advanced three-dimensional architecture applications.

Fig. 8 (a) The FE-SEM image of the SnGS2 film-based V-OTS device with an inset (yellow dashed line) of a magnified image of the hole structure. The white dashed line indicates the bottom electrode (wordplane) of 100 nm-thick W, while the red dashed line indicates the SnGS2 film covered with the top electrode (vertical bitline). The scale bars of the main and inset images are 5 μm and 500 nm, respectively. (b) Representative threshold switching on/off behavior under the application of triangular voltage pulses with rise and fall times of 2.5 μs. (c) Repeated AC I–V characteristics during the 100 triangular pulses and DC I–V results (inset) after 100 cycles of AC measurement. The AC I–V data were collected using a triangular 6 V input.

IV. Conclusions

This study demonstrated the ALD of Sn-doped GeSe₂ (SnGS2) using the supercycle composed of SnN_x and GeSe₂ subcycles. The co-injection of NH₃ with BTMS-Se was critical for enhancing the reactivity of BTMS-Se with TDMA-Ge and adjusting the composition of SnN_x to SnSe₂ through formation of the reactive intermediate, H₂Se. As a result, although the concentration of Sn could be adjusted by the subcycle ratio of SnN_x and GeSe₂, the overall cation (Ge and Sn) and anion (Se) compositional ratio remained at approximately 1 : 2. Despite the removal of N during the GeSe₂ subcycles, the crystallization temperature of SnGS2 with Sn concentrations below 10% remained above 400 °C, due to the high amorphous stability of the GeSe₂. Three different subcycle ratios of [1−10], [1−5], and [2−5] produced the films with the Sn concentrations of 3.5, 5.5, and 9.6%, respectively. All three SnGS2 films exhibited linear growth behavior for the number of supercycles, with growth rates of 11, 6.2, and 8.4 Å per supercycle, respectively. All three SnGS2 films, including GeSe₂, showed an RMS roughness of less than 1 nm at a film thickness of 15 nm. The 15 nm-thick SnGS2 film with a [2−5] subcycle ratio demonstrated excellent step coverage when deposited on a hole structure with an aspect ratio of 1 : 20.

The concentration of Sn had an impact on the electrical characteristics of the OTS devices using SnGS2 films. As the Sn concentration increased, the current for the pristine and electroformed states increased while the V_f and V_t values decreased. These trends closely correlated with the modulation of E_g and the energy barrier height, E_a,0, derived from the trap-limited conduction model. The film with 9.6% Sn concentration showed cyclic endurance over 10⁶ cycles while maintaining selectivity of over 10⁴, which can be attributed to the decrease of Ge–Ge homopolar bonds. This study demonstrates the potential to overcome the endurance limitations of GeSe₂ films through ALD-based Sn doping, significantly improving performance and enabling the deposition of high-quality, As-free OTS materials. Moreover, the feasible operation of the V-OTS device based on ALD SnGS2 film highlights its potential for application in complex three-dimensional architectures.

Author contributions

B.P. and J.W.J. contributed equally. J.W.J conceived the concept and designed the experiment with B.P. B.P. performed the ALD of SnGS2 films by optimizing the ALD conditions with the assistance of W.K. B.P. and J.W.J. performed the characteristics of SnGS2 films with the assistance of C.Y. J.W.J. fabricated the OTS device using SnGS2 films and performed the electrical measurements. W.C., G.J., S.J., and S.K. participated in the hardware maintenance for the ALD equipment. J.L., Y.S., and D.A. contributed to data analysis and provided insightful advice throughout the project. B.P., J.W.J., and C.S.H. wrote the manuscript. C.S.H. supervised the overall research.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts of interest to declare.

Acknowledgements

The financial support for material analysis of Sn-doped GeSe₂ films and fabrication of the Ovonic threshold switching device was provided by the National Research Foundation of Korea (No. 2020R1A3B2079882). The process for realizing Sn-doped GeSe₂ films through ALD and the methodology for evaluating the electrical properties in vertical structures were developed with support from SK hynix Inc. The authors would like to thank Soulbrain, Inc. for supplying tetrakis(dimethylamino) tin(IV) precursor. Raman, XPS, and TEM analysis results were obtained using the instruments installed at the Research Institute of Advanced Materials (RIAM) at Seoul National University.

References

1. S. R. Ovshinsky, Phys. Rev. Lett., 1968, 21, 1450–1453.
2. D. C. Kau, S. Tang, I. V. Karpov, R. Dodge, B. Klehn, J. A. Kalb, J. Strand, A. Diaz, N. Leung, J. Wu, S. Lee, T. Langtry, K.-W. Chang, C. Papagianni, J. Lee, J. Hirst, S. Erra, E. Flores, N. Righos, H. Castro and G. Spadini, in 2009 IEEE International Electron Devices Meeting (IEDM), IEEE, 2009, pp. 1–4.
3. G. Navarro, A. Verdy, N. Castellani, G. Bourgeois, V. Sousa, G. Molas, M. Bernard, C. Sabbione, P. Noe, J. Garrione, L. Fellouh and L. Perniola, in 2017 Symposium on VLSI Technology, IEEE, 2017, pp. T94–T95.
4. M. Alayan, E. Vianello, G. Navarro, C. Carabasse, S. La Barbera, A. Verdy, N. Castellani, A. Levisse, G. Molas, L. Grenouillet, T. Magis, F. Aussenac, M. Bernard, B. DeSalvo, J. M. Portal and E. Nowak, in 2017 IEEE International Electron Devices Meeting (IEDM), IEEE, 2017, pp. 2.3.1–2.3.4.
5. K. Li, J. Yao, P. Zhao, Y. Luo, X. Ge, R. Yang, X. Cheng and X. Miao, Mater. Horiz., 2024, 11, 2106–2114.
6. J. Lee, S. Kim, S. Park, J. Lee, W. Hwang, S. W. Cho, K. Lee, S. M. Kim, T. Seong, C. Park, S. Lee and H. Yi, Adv. Mater., 2022, 34, 2201608.
7. S. Hong, H. Choi, J. Park, Y. Bae, K. Kim, W. Lee, S. Lee, H. Lee, S. Cho, J. Ahn, S. Kim, T. Kim, M. Na and S. Cha, in 2022 International Electron Devices Meeting (IEDM), IEEE, 2022, pp. 18.6.1–18.6.4.
8. T. Ravsher, D. Garbin, A. Fantini, R. Degraeve, S. Clima, G. L. Donadio, S. Kundu, H. Hody, W. Devulder, J. Van Houdt, V. Afanas'ev, R. Delhougne and G. S. Kar, IEEE Trans. Electron Devices, 2023, 70, 2276–2281.
9. K. J. Yoon, Y. Kim and C. S. Hwang, Adv. Electron. Mater., 2019, 5, 1800914.
10. J. W. Jeon, B. Park, Y. H. Jang, S. H. Lee, S. Jeon, J. Han, S. K. Ryoo, K. Do Kim, S. K. Shim, S. Cheong, W. Choi, G. Jeon, S. Kim, C. Yoo, J.-K. Han and C. S. Hwang, ACS Appl. Mater. Interfaces, 2024, 16, 15032–15042.
11. S. Yoo, C. Yoo, E.-S. Park, W. Kim, Y. K. Lee and C. S. Hwang, J. Mater. Chem. C, 2018, 6, 5025–5032.
12. W. Kim, S. Yoo, C. Yoo, E.-S. Park, J. Jeon, Y. J. Kwon, K. S. Woo, H. J. Kim, Y. K. Lee and C. S. Hwang, Nanotechnology, 2018, 29, 365202.
13. C. Yoo, W. Kim, J. W. Jeon, E.-S. Park, M. Ha, Y. K. Lee and C. S. Hwang, ACS Appl. Mater. Interfaces, 2020, 12, 23110–23118.
14. V. Adinolfi, M. Laudato, R. Clarke, V. K. Narasimhan, L. Cheng and K. Littau, J. Vac. Sci. Technol., A, 2020, 38, 052404.
15. S.-D. Kim, H.-W. Ahn, S. Y. Shin, D. S. Jeong, S. H. Son, H. Lee, B.-k. Cheong, D. W. Shin and S. Lee, ECS Solid State Lett., 2013, 2, Q75–Q77.
16. H. Li and J. Robertson, Sci. Rep., 2019, 9, 1867.
17. H. Y. Cheng, W. C. Chien, I. T. Kuo, C. W. Yeh, L. Gignac, W. Kim, E. K. Lai, Y. F. Lin, R. L. Bruce, C. Lavoie, C. W. Cheng, A. Ray, F. M. Lee, F. Carta, C. H. Yang, M. H. Lee, H. Y. Ho, M. BrightSky and H. L. Lung, in 2018 IEEE International Electron Devices Meeting (IEDM), IEEE, 2018, pp. 37.3.1–37.3.4.
18. V. Adinolfi, M. Laudato, R. Clarke, S. Jewhurst, M. E. McBriarty, S. Hoang, V. K. Narasimhan, L. Cheng and K. A. Littau, ACS Appl. Electron. Mater., 2020, 2, 3818–3824.
19. A. Verdy, G. Navarro, V. Sousa, P. Noe, M. Bernard, F. Fillot, G. Bourgeois, J. Garrione and L. Perniola, in 2017 IEEE International Memory Workshop (IMW), IEEE, 2017, pp. 1–4.
20. M.-J. Lee, D. Lee, S.-H. Cho, J.-H. Hur, S.-M. Lee, D. H. Seo, D.-S. Kim, M.-S. Yang, S. Lee, E. Hwang, M. R. Uddin, H. Kim, U.-I. Chung, Y. Park and I.-K. Yoo, Nat. Commun., 2013, 4, 2629.
21. J. Yoo, Y. Koo, S. A. Chekol, J. Park, J. Song and H. Hwang, in 2018 IEEE Symposium on VLSI Technology, IEEE, 2018, pp. 207–208.
22. S. Jia, H. Li, T. Gotoh, C. Longeaud, B. Zhang, J. Lyu, S. Lv, M. Zhu, Z. Song, Q. Liu, J. Robertson and M. Liu, Nat. Commun., 2020, 11, 4636.
23. Y. K. Lee, C. Yoo, W. Kim, J. W. Jeon and C. S. Hwang, J. Mater. Chem. C, 2021, 9, 3708–3725.
24. T. T. Nang, M. Okuda, T. Matsushita, S. Yokota and A. Suzuki, Jpn. J. Appl. Phys., 1976, 15, 849–853.
25. J. Lee, S. W. Cho, Y. W. Lee, J. Y. Kwak, J. Kim, Y. Jeong, G. W. Hwang, S. Park, S. Kim and S. Lee, J. Mater. Chem. C, 2022, 10, 18033–18039.
26. M. Z. Ansari, D. K. Nandi, P. Janicek, S. A. Ansari, R. Ramesh, T. Cheon, B. Shong and S.-H. Kim, ACS Appl. Mater. Interfaces, 2019, 11, 43608–43621.
27. J. W. Jeon, C. Yoo, W. Kim, W. Choi, B. Park, Y. K. Lee and C. S. Hwang, Dalton Trans., 2022, 51, 594–601.
28. B. J. Choi, S. Choi, T. Eom, S. W. Ryu, D.-Y. Cho, J. Heo, H. J. Kim, C. S. Hwang, Y. J. Kim and S. K. Hong, Chem. Mater., 2009, 21, 2386–2396.
29. T. Ueno, Jpn. J. Appl. Phys., 1983, 22, 1349.
30. T. Takahashi, T. Sagawa and H. Hamanaka, J. Non-Cryst. Solids, 1984, 65, 261–267.
31. A. Rockett and R. W. Birkmire, J. Appl. Phys., 1991, 70, R81–R97.
32. V. Alberts, J. H. Schön, M. J. Witcomb, E. Bucher, U. Rühle and H. W. Schock, J. Phys. D: Appl. Phys., 1998, 31, 2869–2876.
33. H. Park, J. H. Hwang, S. H. Oh, J. J. Ryu, K. Jeon, M. Kang, H.-J. Chai, A. Ham, G. H. Kim, K. Kang and T. Eom, ACS Nano, 2024, 18, 22071–22079.
34. D. G. Mead and J. C. Irwin, Solid State Commun., 1976, 20, 885–887.
35. P. Boolchand and W. J. Bresser, Philos. Mag. B, 2000, 80, 1757–1772.
36. R. K. Pan, H. Z. Tao, H. C. Zang, X. J. Zhao and T. J. Zhang, Appl. Phys. A, 2010, 99, 889–894.
37. E. Petracovschi, B. Bureau, A. Moreac, C. Roiland, J.-L. Adam, X.-H. Zhang and L. Calvez, J. Non-Cryst. Solids, 2016, 431, 16–20.
38. J. Keukelier, K. Opsomer, T. Nuytten, S. Sergeant, W. Devulder, S. Clima, L. Goux, G. S. Kar and C. Detavernier, J. Mater. Chem. C, 2021, 9, 117–126.
39. B. Yuan, H. Chen and S. Sen, J. Phys. Chem. B, 2022, 126, 946–952.
40. J.-H. Fournier-Lupien, S. Mukherjee, S. Wirths, E. Pippel, N. Hayazawa, G. Mussler, J. M. Hartmann, P. Desjardins, D. Buca and O. Moutanabbir, Appl. Phys. Lett., 2013, 103, 10–15.
41. M. Qi, C. A. Stephenson, V. Protasenko, W. A. O'Brien, A. Mintairov, H. Xing and M. A. Wistey, Appl. Phys. Lett., 2014, 104, 1–6.
42. N. D. Boscher, C. J. Carmalt, R. G. Palgrave and I. P. Parkin, Thin Solid Films, 2008, 516, 4750–4757.
43. P. Fantini, N. Polino, A. Ghetti and D. Ielmini, Adv. Electron. Mater., 2023, 2300037, 1–8.
44. D. Ielmini and Y. Zhang, J. Appl. Phys., 2007, 102, 054517.
45. B. Song, H. Xu, S. Liu, H. Liu, Q. Liu and Q. Li, Appl. Phys. A:Mater. Sci. Process., 2019, 125, 1–6.
46. S. Jun, S. Seo, S. Park, T. H. Kim, M. Lee, S. M. Hong, T. Kim, S. Chung, T. Lee, M. Kim and H. Kim, J. Alloys Compd., 2023, 947, 169514.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4dt02946a

‡ These authors contributed equally.

---