Soret-effect induced phase-change in chromium nitride semiconductor film

doi:10.21203/rs.3.rs-2486043/v1

Download PDF

Article

Soret-effect induced phase-change in chromium nitride semiconductor film

https://doi.org/10.21203/rs.3.rs-2486043/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Phase-change materials such as Ge-Sb-Te (GST) exhibiting amorphous and crystalline phases can be used for phase-change random access memories (PCRAM). GST-based PCRAM has been applied as a storage-class memory; however, its relatively low ON/OFF ratio and large Joule heat energy for the RESET process (amorphization) significantly limit the storage density. This study proposes a novel phase-change nitride, CrN, with a much wider programming window (ON/OFF ratio more than 10⁵) and lower RESET energy (a 90 % reduction from GST). High-resolution transmission electron microscopy revealed a unique phase-change from low resistive cubic CrN into high resistive hexagonal CrN₂ induced by the Soret-effect. The proposed phase-change nitride could greatly expand the scope of conventional phase-change chalcogenides and offer a new strategy for next-generation PCRAM, enabling a large ON/OFF ratio, low energy, and fast operation.

Physical sciences/Materials science/Condensed-matter physics/Phase transitions and critical phenomena

Physical sciences/Materials science/Condensed-matter physics/Electronic properties and materials

Physical sciences/Materials science/Materials for devices/Electronic devices

Physical sciences/Materials science/Materials for devices/Information storage

Phase change materials (PCMs) represented by Ge-Sb-Te (GST) are mainly chalcogenides and widely studied on nonvolatile resistive switching memory (NVM) due to their fast and reversible phase change between amorphous and crystalline phases. However, it faces a fundamental challenge of requiring large Joule heat energy to melt in the RESET process (amorphization), which inevitably results in higher operating current and power¹. Besides, the crystallization from the amorphous phase limits the switching speed of the device². Therefore, new-concept melting-free crystalline-to-crystalline PCMs are attracting much attention due to their low writing power and fast switching speed. They include artificial GeTe/Sb₂Te₃ superlattices for iPCM (interfacial phase-change memory) devices, where the resistance contrast between two crystalline phases comes from different bonding states, namely, a covalently bonded state of high resistance and a resonantly bonded state of low resistance³. Another category is polymorphic-change memory devices based on polymorphic compounds such as MnTe, where a phase change via atomic-displacement between NiAs-type and Wurzite-type hexagonal phases, which induces Te coordinate changes, provides a large resistance contrast⁴. Furthermore, two-dimensional (2D) materials such as In₂Se₃ and MoTe₂ can show structural phase changes driven by an electrical field, exhibiting electrical properties ranging from semiconductor to metallic behaviors^5,6. Nonetheless, fabricating a superlattice or single-/few-layered 2D film device is complicated and requires great effort. Moreover, Se- or Te-based materials are not environmentally friendly, and the Te segregation observed after many switching cycles might affect the fatigue property. In this regard, a nitride memory material would offer better processing and chemical compatibility with nitride electrodes, which are generally used in complementary metal–oxide–semiconductor circuits and should be easily applied to power electronics based on nitrides such as GaN⁷. The switching mechanism of nitrides mainly relies on the formation and motion of N vacancies, serving as a conductive filament in a highly insulating nitride matrix, which is known as the resistive random access memory (ReRAM) ^7–9. However, various challenges still exist in the insulating nitride-based memory such as Si₃N₄ and AlN^7,8, hindering its practical applications. First, a high-energy forming process is necessary to generate conductive filaments arising from vacancies in a virgin device under electric bias, which greatly increases power consumption¹⁰. Second, the number of vacancies or defects in insulators are not easy to control; thus, the switching or failure mechanism is still unclear¹¹.

Unlike the abovementioned insulating nitrides, CrN is a well-known electronic material with a conducting behavior varying from metallic to semiconductor-like by tuning defects, and it has been widely studied in fields such as supercapacitors and thermoelectrics ^12,13. Besides, the CrN crystal phase or structure is easily tuned by external stresses such as pressure ¹⁴, electron beam irradiation ¹⁵, or thermal treatment ¹⁶. Therefore, CrN has a high potential for applications in phase-change NVM due to variation in wide electrical properties upon structural changes. Here, we propose a CrN memory material that undergoes a nonvolatile and reversible crystalline-to-crystalline phase change under electrical pulse, enabling environment-friendly NVM devices with large ON/OFF ratio, low power, and fast operation.

We fabricated a conventional T-shaped memory device (Fig. S1), whose cross-section is schematized in Fig. 1a. It consisted of a square W plug (heater electrode) with a side length varying from 34 to 218 nm. After CrN deposition, the top electrode (TE) was sequentially deposited in situ in the same sputtering chamber to avoid surface oxidation of CrN. The CrN layer was characterized by X-ray diffraction and transmission electron microscopy (TEM), revealing a homogeneous growth with a NaCl-like cubic phase (Fig. S2). Based on Rutherford backscattering spectrometry (RBS) measurements, the N/Cr composition ratio was almost 1 with a small amount of unintentionally doped oxygen (O, 8.6 at. %) from the sputter chamber. The as-deposited CrN thin film shows a semiconducting behavior as indicated by the temperature dependence of resistivity from Hall measurement. (Fig. S3) The initial state of the device was the as-deposited low-resistance (~ 10^3–4 Ω) cubic CrN state. The resistance (R) was then read while increasing the amplitude of the applied pulse V (30-ns width) by 0.1 V (Fig. S4). Figure 1b displays the resulting RV characteristics. For the device with a plug size of 37 nm × 37 nm, when we applied a positive voltage pulse to the bottom electrode (BE), the device reached a high-resistive state (HRS) of ~ 10⁸ Ω at 1.0 V, and then followed by the recovery to its initial low-resistive state (LRS) reversibly at a higher voltage (1.3 V). The devices with a larger plug exhibited similar RV trends, indicating nonvolatile and reversible resistive switching, i.e., LRS-to-HRS (RESET process) and HRS-to-LRS (SET process). The highest HRS resistance observed was ~ 10⁹ Ω, and the HRS/LRS resistance ratio could be over 10⁵. The resistance value of both the HRS and LRS show a strong dependence on the plug size, and it decreases with increasing the plug size. Moreover, a typical threshold switching phenomenon in the HRS by a current sweep (pulse width: 500 µS) measurement was observed in the same device with the plug size of 37×37 nm², indicating a similar switching behavior with the traditional chalcogenide-based PCMs (Fig. 1c) The Joule heating energy for the switching operation in the CrN based memory devices were then calculated based on the RV measurements and quantitatively compared with that of the traditional PCM GST (Fig. 1d). For all devices, the SET energy was negligible compared to the RESET energy since the resistance in the RESET state was much higher than in the SET state. Figure 1e plots the RESET energy as a function of the contact area of the devices; for GST memory devices, it decreased with reducing the contact area, following the same scaling trend of other cell types ^17,18. The RESET energy of the CrN-based device was slightly smaller than that of the GST-based one at larger contact size, while it decreases more drastically with decreasing contact area with a larger across a wide range of contact areas. The RESET energy finally decreased to 28 pJ at a plug size of 37 nm × 37 nm, with one order of magnitude reduction compared with the same-sized GST-based device (279 pJ). Figure 1f summarized the programing window of various NVMs including oxide ReRAM/Conducting bridge RAM (CBRAM)^17,19, nitride ReRAM²⁰, PCM³, interfacial PCM (IPCM)³, phase-change heterostructure (PCH)²¹ and Confined PCM²². Among them, the novel nitride PCM in this study exhibits superior programing window comparable to the CBRAM, much larger than most of PCM-based memories, suggesting a good read accuracy and high potential for multi-bit storage of CrN-PCM²³.

We conducted a TEM observation of the CrN device with a plug size of 218 nm × 218 nm to investigate the resistive switching mechanism. Before the TEM analysis, the device sample was switched to an HRS of ~ 10⁶ Ω by applying a voltage pulse of 3.8 V for 50 ns. A distinct bright contrast (active region) was observed in the CrN layer between the W plug and TE electrode (Fig. 2a). The energy-dispersive X-ray spectroscopy (EDX) mapping of its cross-section (Fig. 2b) showed the absence of obvious intermixtures between the CrN layer, W electrode layer, TiN sidewall, and insulating layer, while a poor Cr concentration was clearly observed in the active region of the CrN layer. Figure 2c displays the atomic column image at the boundary region of the matrix (upper part) and active region (lower part) detected by high-resolution TEM (HRTEM) in the area indicated by a yellow dotted box in Fig. 2a. Figure 2d shows the Fast Fourier transform (FFT) image of the matrix area indicated by a red square in Fig. 2c, indicating a cubic phase. The corresponding inverse fast Fourier transform (IFFT) image of the HRTEM micrograph (square area enclosed by a red line in Fig. 2c) showed a clear cubic atomic column image with a distance of the (002) plane of 2.05 Å, which is consistent with the lattice parameter a = 4.17 Å of NaCl-type cubic CrN. For the active region, instead, the FFT pattern (Fig. 2e) revealed a hexagonal structure. The distance between two Cr layers, derived from the IFFT image (bottom inset in Fig. 2c), was 3.7 Å (bright spots).

Let us discuss the details of the hexagonal structure. In Cr–N binary systems, there are two kinds of hexagonal phases: Cr₂N and CrN₂. Since Cr₂N has a metallic property ²⁴, the hexagonal phase formed in the CrN layer with an HRS is unlikely a Cr₂N hexagonal phase. The crystal structures of cubic CrN and hexagonal CrN₂ are displayed in Fig. S5. The IFFT image in the bottom inset in Fig. 2c strongly suggests that the observed hexagonal phase is hexagonal CrN₂ with N–N dimers.

Both density functional theory calculations and experiments have recently revealed the formation of a CrN₂ hexagonal phase under high pressure ^25,26. It is also suggested theoretically that a metal-insulator transition is induced by replacing N with N-N dimer in a WC-type metallic CrN, resulting in a semiconductor characteristics of a WC-type CrN₂.^25,26 Zhao et al. also pointed out that a cooperation of N 2p–Cr 3d hybridization and electron-electron Coulomb repulsion leads to such an insulativity of CrN₂.²⁶ The reported theoretical result strongly suggests that the induced high resistive hexagonal phase in the CrN device is CrN₂. Moreover, the reported simulation has shown that WC-type CrN₂ is stable with the lattice parameters a = b = 2.725 and c = 3.712 Å using Perdew–Burke–Ernzerhof (PBE) exchange-correlation potentials.²⁶ Thus, the observed lattice parameters a = b = 2.719 and c = 3.712 Å of the hexagonal phase formed in the CrN layer in an HRS well agree with those of a WC-type CrN₂ phase; the atomic column image (bottom inset in Fig. 2c) can be well matched with the atomic positions on the (110) plane of WC-type CrN₂. The N atom position could be clearly observed in the IFFT image derived from the FFT pattern in the [121] zone axis of CrN₂ and matched well with the WC-like hexagonal CrN₂ structure (Fig. S6).

To induce the hexagonal CrN₂ phase formation in the CrN matrix, the composition must change simultaneously with the crystal structure, which quite differs from conventional PCMs showing amorphization/crystallization without overall composition change. Thus, the Cr or N atoms should diffuse during the formation of N–N dimers upon phase change in the CrN layer. As confirmed by EDX in Fig. 2b, a poor Cr concentration was clearly observed in the active region of the CrN layer, indicating the diffusion of Cr atoms during phase transition. However, it is difficult to accurately detect the light element such as N using EDX. Therefore, the electron energy loss spectroscopy (EELS) exhibiting high sensitivity for N atoms was employed. Figure 3a, b shows the TEM image and EELS mapping of the CrN layer near the active region, obtained in the energy range of 408–422 eV. The N K-edge mapping clearly indicates a higher N concentration in the active region than in the CrN matrix. Hence, the combination of this EELS analysis with the EDX results (Fig. 2b) confirms a Cr-poor and N-rich composition of the active region. Note that, in our device substrate, there was an ultrathin TiN sidewall adhesion layer (3–5 nm) between the W plug heater and surrounded SiO₂ insulator, as observed from the EDX mapping of Ti (Fig. 2b) and EELS mapping of N (Fig. 3b). As shown in Fig. 3a, the phase-change active region in the CrN layer was just above the TiN thin layer rather than on the whole surface of the W plug, which is quite different from the dome-like active region of conventional PCRAM³. Such a local composition variation under electrical pulses is similar to the switching phenomenon of ReRAM, where the formation/rupture of the conductive filament is realized by ion diffusion due to the application of electrical pulses. Unlike traditional materials for ReRAM, the CrN matrix is noninsulating and, therefore, does not need the formation of a conductive path. Hence, the phase change between LRS and HRS is more like the switching mechanism of traditional materials for PCRAM, where a high-resistance amorphous volume is formed in a conductive crystalline matrix by Joule heating. To rule out the role of unintentionally doped O in the CrN layer on switching properties, the EELS of O K-edge was taken in matrix (Point 1), N-rich (Point 2) and near electrode surface (Point 3) parts, as indicated in Fig. 3a. The peak intensity of O K-edge is equally weak compared to N K-edge, suggesting a negligible influence of O for phase change of the CrN layer. (Fig. 3c) To make it more convincing, we also tried to fabricate a pure CrN film (designated as CrN’ hereafter). The vacuum condition of a sputtering chamber and sputtering condition to obtain a pure CrN film without O contamination are very strict, which has become a remained open question in many fields such as coating and thermoelectric applications.^27–29 Here, we found that the O content decreases by lowering the working pressure of the sputtering chamber. (Supporting information 7). Based on the results, we successfully obtained an O-free CrN’ film exhibiting the same crystal structure (NaCl-cubic) with the CrN film (Fig. S8) The CrN’ memory device exhibits similar RV characteristics and the TEM observation after the RESET operation reveals that the CrN’ device undergoes the phase change from NaCl-type cubic CrN to WC-type hexagonal CrN₂, implying that this phase change is an intrinsic behavior in a CrN film. (Supporting information 9) In addition, interestingly enough, different carrier type was detected in CrN (p-type) and CrN’ (n-type) thin films by both Seebeck and Hall measurements. (Fig. S10) The same switching behavior between the CrN- and CrN’-based devices strongly supports that the defects or the carrier type have a minor effect on the switching properties of a CrN film.

Thus, to understand the driving force of nitrogen atom diffusion and phase change in the CrN devices, we simulated the thermal distribution under the application of a voltage³⁰. In this simulation, thermoelectric effects such as the Thomson effect, Peltier effect, and thermal boundary resistance were ignored, as well as the contact resistance between the low resistive CrN (semi metallic-like) and metal electrodes because of its minor effect on the device operation. The specific structure and size of the simulated CrN-based memory device are presented in Fig. 3d. To understand the effect of the TiN adhesion layer between the W plug electrode and SiO₂ insulator on the thermal distribution when applying a voltage, a single TiN plug-type memory device was also simulated. A positive voltage input of 0.2 V to the BE was simulated, along with a Joule heating generated within the memory structure (Supporting information 11). For the CrN-based memory device, the simulated temperature distribution (Fig. 3e) suggested that when a TiN adhesion layer exists between the W plug and SiO₂ insulator, the hottest region inside the CrN layer is circularly distributed in an arc from near the top of the TiN sidewall layer to the middle of the CrN layer. This hottest region agrees well with the active region showing phase change in the CrN layer, which strongly demonstrates that Joule heating is key in the phase-change phenomenon. For the single TiN plug heater, instead, the hottest region covers the entire interface between the TiN and CrN layers. In this case, the phase change cannot occur just above the W plug heater because the Joule heat dissipates easily toward it due to the higher thermal conductivity of W compared with TiN, resulting in less residual heat above the heater ³¹. We also simulated the electrical field distribution under the application of a voltage in a CrN device with a W plug/TiN sidewall heater or a TiN plug heater (Fig. S11). The electric field was concentrated near the surface of the W plug heater rather than the inside the CrN layer due to the highly conductive W, similar to most electrical distributions in ReRAM devices ³². These results further indicate that the atomic diffusion in the middle part of the CrN layer and the phase change are driven by the thermal effect rather than the electrical field effect. The established thermophoresis/diffusion called Soret–Fick diffusion of atoms or vacancies can well explain the non-electrical field dependency of N atoms in this case ³³. A large radial temperature gradient can be created and the Soret effect can become the dominant factor influencing atomic migration; the sign of the applied electrical field and the carrier types in CrN are not important for the Soret effect ³⁴. It is noteworthy that the phase change after the above thermal diffusion process can be stable and nonvolatile at room temperature as indicated in current-time measurements which shows minor drift. (Fig. 3f) The cyclic resistive switching of the CrN device was also tested to be over 10⁴ cycles, indicating good feasibility of the nitride PCM. (Supporting information 12)

Here, the switching mechanism of phase-change CrN can be shortly summarized in Fig. 4. The SET state originally shows the low resistance of a cubic CrN phase. When applying a voltage pulse, the lighter N atoms can diffuse to the hot active region (RESET-1 state) driven by Soret-effect and, synchronously, the phase change from cubic CrN to hexagonal CrN₂ occurs (RESET-2 state). The hexagonal CrN₂ phase can go back to the cubic CrN phase under mild Joule heating, where the SET operation can be performed with much less energy than the RESET operation. The above process is generally different with the conventional amorphous-crystalline chalcogenides-based PCMs.

In summary, we report a novel memory device based on the fast and reversible phase change of simple toxic-free and easily fabricated CrN. Phase-change CrN exhibits unipolar resistive switching and lower operation energy than traditional GST-based PCMs. Our experiments revealed that the non-volatile resistive switching mechanism is originated from the phase change between low resistive NaCl-type cubic CrN and high resistive WC-type hexagonal CrN₂, which is induced by Soret-effect. The CrN device exhibits a large ON/OFF ratio (more than 10⁵) and a fast switching speed (~ 30 ns) while a low operating energy, superior to the typical GST based memory. This study provides not only new possibilities for the application of crystalline-to-crystalline PCMs as next-generation NVM devices but also a further investigation of low-cost, clean, and chalcogenide-alternative PCMs. Furthermore, the same crystal structure of the n-type and p-type CrN films obtained by tuning O content indicate a great potential in homogeneous pn junction to serve as a diode-type selector element in a memory device ^35–37.

Data and materials availability

The data are available from the corresponding authors upon request.

Acknowledgements

This work was supported by JSPS KAKENHI (Grant Nos. 21H05009, 22K20474) and the Murata Science Foundation. The authors acknowledge the financial support from the Commissioned Research (No. 03701) by National Institute of Information and Communications Technology (NICT), JAPAN. The authors thank Dr. T.Miyazaki, Mr. M.Tanno and Dr. K.Kobayashi (Tohoku University, Japan) for their technical support for the TEM measurements.

Author contributions

Y.Shuang and Y.Sutou conceived the concepts, designed the experiments and co-wrote the manuscript with input from all other authors. Y.Sutou led the project. Y.Shuang carried out film deposition, device fabrication, XRD and TEM observation along with S.M. and D.A. Y.Shuang and Y.Sutou analyzed the results of XRD and TEM observation. Y,Shunag, YH.S., J.H., and Y.Sutou discussed and fabricated the cell structure of memory device. Y.Shuang performed electrical measurements along with S.H. T.Y. simulated the thermal distribution of the device. Y.Shuang, S.M., T.Y., S.H., YH.S., D.A, and Y.Sutou discussed all the data and their interpretations.

Competing interests

Authors declare no competing interests. The present authors, Y.Shuang, S.H and Y.Sutou are inventors on Japanese patent application number 2019-133861, applied for by Tohoku University.

Ahn, C. et al. Energy-Efficient Phase-Change Memory with Graphene as a Thermal Barrier. Nano Lett. 15, 6809–6814 (2015).
Rao, F. et al. Reducing the stochasticity of crystal nucleation to enable subnanosecond memory writing. Science. 358, 1423–1427 (2017).
Simpson, R. E. et al. Interfacial phase-change memory. Nat. Nanotechnol. 6, 501–505 (2011).
Mori, S., Hatayama, S., Shuang, Y., Ando, D. & Sutou, Y. Reversible displacive transformation in MnTe polymorphic semiconductor. Nat. Commun. 11, (2020).
Choi, M. S. et al. Electrically Driven Reversible Phase Changes in Layered In₂Se₃ Crystalline Film. Adv. Mater. 1703568, 1703568 (2017).
Zhang, F. et al. Electric-field induced structural transition in vertical MoTe₂- and Mo_1–xW_xTe₂ -based resistive memories. Nat. Mater. 18, 55–61 (2019).
Choi, B. J. et al. High-Speed and Low-Energy Nitride Memristors. Adv. Funct. Mater. 26, 5290–5296 (2016).
Jiang, P. F. et al. Enhanced Switching Stability in Forming-Free SiN_x Resistive Random Access Memory Devices with Low Power Consumptions Based on Local Pt Doping in a Stacked Structure. Adv. Electron. Mater. 5, 1800739 (2019).
Kim, S. & Park, B. G. Power- and Low-Resistance-State-Dependent, Bipolar Reset-Switching Transitions in SiN-Based Resistive Random-Access Memory. Nanoscale Res. Lett. 11, (2016).
Kim, S., Chang, Y. F., Kim, M. H. & Park, B. G. Improved resistive switching characteristics in Ni/SiN_x/p⁺⁺-Si devices by tuning x. Appl. Phys. Lett. 111, 33509 (2017).
Kim, S., Cho, S., Ryoo, K.-C. & Park, B.-G. Effects of conducting defects on resistive switching characteristics of SiN_x -based resistive random-access memory with MIS structure. J. Vac. Sci. Technol. B 33, 062201 (2015).
Arif, M., Sanger, A. & Singh, A. Sputter deposited chromium nitride thin electrodes for supercapacitor applications. Mater. Lett. 220, 213–217 (2018).
Quintela, C. X. et al. Epitaxial CrN Thin films with high thermoelectric figure of merit. Adv. Mater. 27, 3032–3037 (2015).
Rivadulla, F. et al. Reduction of the bulk modulus at high pressure in CrN. Nat. Mater. 8, 947–951 (2009).
Wan, P. et al. Nitrogen atom shift and the structural change in chromium nitride. Acta Mater. 98, 119–127 (2015).
Lu, F. H. & Chen, H. Y. Phase changes of CrN films annealed at high temperature under controlled atmosphere. in Thin Solid Films vols 398–399 368–373 (2001).
Wong, H. S. P. et al. Stanford memory trends. tech. Rep. (2016).
Fong, S. W., Neumann, C. M. & Wong, H. S. P. Phase-Change Memory - Towards a Storage-Class Memory. IEEE Trans. Electron Devices 64, 4374–4385 (2017).
Zambelli, C., Navarro, G., Sousa, V., Prejbeanu, I. L. & Perniola, L. Phase change and magnetic memories for solid-state drive applications. Proc. IEEE 105, 1790–1811 (2017).
Min, K. P., Li, C. Y., Chang, T. J. & Chu, S. Y. The Effects of Si Doping on the Endurance and Stability Improvement of AlN-Based Resistive Random Access Memory. ACS Appl. Electron. Mater. 3, 5327–5334 (2021).
Ding, K. et al. Phase-change heterostructure enables ultralow noise and drift for memory operation. Science. 366, 210–215 (2019).
Xie, Y. et al. Self-Healing of a Confined Phase Change Memory Device with a Metallic Surfactant Layer. Adv. Mater. 30, 1705587 (2018).
Yin, Y., Noguchi, T., Ohno, H. & Hosaka, S. Programming margin enlargement by material engineering for multilevel storage in phase-change memory. Appl. Phys. Lett. 95, 133503 (2009).
Gharavi, M. A. et al. Microstructure and thermoelectric properties of CrN and CrN/Cr₂N thin films. J. Phys. D. Appl. Phys. 51, (2018).
Niwa, K., Yamamoto, T., Sasaki, T. & Hasegawa, M. High-pressure synthesis, crystal growth, and compression behavior of hexagonal CrN₂ having one-dimensionally aligned nitrogen dimer. Phys. Rev. Mater. 3, 53601 (2019).
Zhao, Z. et al. Potentially superhard hcp CrN₂ compound studied at high pressure. Phys. Rev. B 93, 214104 (2016).
Elangovan, T. et al. Nanostructured CrN thin films prepared by reactive pulsed DC magnetron sputtering. Mater. Sci. Eng. B Solid-State Mater. Adv. Technol. 167, 17–25 (2010).
Nagasawa, S. et al. Changes in the electric resistivity of CrN subsequent to oxygen dissolution. Jpn. J. Appl. Phys. 55, (2016).
Tripathi, Y. et al. Study of phase formulation in CrN thin films and its response to a minuscule oxygen flow in reactive sputtering process. Thin Solid Films 670, 113–121 (2019).
Yamamoto, T., Hatayama, S. & Sutou, Y. Design strategy of phase change material properties for low-energy memory application. Mater. Des. 216, (2022).
Oh, S. I. et al. Effect of Electrode Material on the Crystallization of GeTe Grown by Atomic Layer Deposition for Phase Change Random Access Memory. Micromachines 10, 281 (2019).
Koh, S.-G., Kishida, S. & Kinoshita, K. Extremely small test cell structure for resistive random access memory element with removable bottom electrode. Appl. Phys. Lett. 104, 083518 (2014).
Janek, J. & Timm, H. Thermal diffusion and Soret effect in (U,Me)O_2+δ: the heat of transport of oxygen. J. Nucl. Mater. 255, 116–127 (1998).
Strukov, D. B., Alibart, F. & Stanley Williams, R. Thermophoresis/diffusion as a plausible mechanism for unipolar resistive switching in metal-oxide-metal memristors. Appl. Phys. A Mater. Sci. Process. 107, 509–518 (2012).
Shen, J. et al. Elemental electrical switch enabling phase segregation–free operation. Science. 374, 1390–1394 (2021).
Shuang, Y. et al. Bidirectional Selector Utilizing Hybrid Diodes for PCRAM Applications. Sci. Rep. 9, 20209 (2019).
Sun, L. et al. Self-selective van der Waals heterostructures for large scale memory array. Nat. Commun.10, (2019).

Preparation of CrN thin film. CrN films were deposited on SiO₂ (100 nm)/Si (725 um) or glass (Corning EAGLE XG) substrates by the radiofrequency (RF) magnetron reactive sputtering of Cr (99.99%) pure targets at room temperature in an Ar/N₂(5:3) gas atmosphere. where the substrate holder was rotated during deposition; the RF power was fixed to 50 W, the base pressure of the sputtering chamber was below 5.0 × 10⁻⁵ Pa, and the working pressure was ~4.4 × 10⁻¹ Pa. The film thickness was confirmed with an atomic force microscope (Keyence, VN-8000).

Characterization of CrN thin film. The composition of 30 nm-thick CrN thin films deposited on a SiO₂ (100 nm)/Si (725 um) substrate was assessed by RBS (National Electrostatics Corp., Pelletron 3SDH). The crystal structures were investigated by XRD (Rigaku, Ultima IV); the diffraction patterns were taken in the 2θ range from 35° to 50° with Cu Kα radiation. The cross-sectional microstructures were observed using a TEM system (JEOL, JEM-2100F) at an accelerating voltage of 200 kV. For this analysis, the samples were thinned via an ion milling instrument (Gatan, PIPS). The samples for the XRD and TEM measurements were deposited on a SiO₂ (100 nm)/Si (725 um) substrate with a thickness of ~100 nm.

The electrical properties were evaluated through a Hall effect measurement apparatus (Toyo Corp., ResiTest 8400). The Seebeck coefficient was measured with a ResiTest 8300 (Toyo Corporation) in the temperature range of 260–400 K. The samples for the Hall effect and Seebeck coefficient measurements were grown on glass substrates with a thickness of 100 nm.

Memory device fabrication. T-shaped substrates with W electrode plugs were used, where the W plug had a square shape with the side length varying from 34 to 218 nm (Fig. S1). This substrate was first etched for 115 min by an Ar plasma to remove the surface oxidation of the W plugs. A 100-nm CrN or CrN’ layer was in situ deposited on top of the W plug via conventional lithography. Then, a 250-nm W TE was deposited onto it.

Memory device characterization. The read resistance for all devices were measured using a semiconductor parameter analyzer (Keysight, B1500A). To evaluate the resistive switching properties, a pulse generator (Keysight, B1525A) was used to apply short voltage pulses to the memory cells; the pulse amplitude and width were confirmed by an oscilloscope (Tektronix, TBS 1202B).

The cross-sectional microstructure of the devices after the RESET operation (i.e., from LRS to HRS) was observed with a TEM instrument (JEOL, JEM-2100F) at an accelerating voltage of 200 kV. An EDX system (JEOL, JEM-2100F) was utilized to map the elemental distributions in the T-shaped device. To obtain atomic images, an HRTEM apparatus equipped with an ADF-STEM detector (JEOL, ARM200F) was used. IFFT images were obtained using the Gatan DigitalMicrograph software. An EELS spectroscope (JEOL, ARM200F) was utilized to identify the concentration of N atoms in the cross-section of the devices. The TEM samples of the device cross-sections were thinned by a focused ion beam system (JEOL, JIB-4600F) with a Ga ion beam at 30 keV and polished at 10 keV.

Operation energy calculation. The energy for the RESET process (Q_RESET) was estimated as follows:

𝑄_RESET= (𝑉_RESET²/𝑅_SET)× 𝑡, (1)

where V_RESET is the setting pulse voltage for the RESET process, 𝑅_SET is the resistance at the SET state, and t is the pulse width. In practice, the setting pulse voltage is slightly larger than the applied pulse voltage due to the interior impendence (50 Ω) in the used semiconductor parameter analyzer. Thus, all the voltage values used here for the energy calculation were setting voltage values calibrated by an oscilloscope (Tektronix, TBS 1202B) from the applied ones.

There is NO Competing Interest.

Supplementaryinformation.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Soret-effect induced phase-change in chromium nitride semiconductor film

Status:

Version 1

Abstract

Figures

Introduction

Resistive Switching Behavior Of Crn Device

Mechanism Of Resistive Switching In Crn Device

Nitrogen Atom Diffusion By Soret-effect

Conclusions And Outlook

Declarations

References

Methods

Additional Declarations

Supplementary Files

Status:

Version 1