Project Abstract:

AMOS aims to build a map for the discovery of novel chalcogenide materials with ovonic threshold switching (OTS), enabling the creation of high density crossbar memory arrays. OTS is a volatile electrical characteristic of a chalcogenide material, which can rapidly swap from a low to a high conductivity state, by applying a voltage that exceeds the threshold voltage. First, a database of materials properties, using a combination of computed and measured data will be constructed. Atomic properties such as electronegativity, orbital radii and bond enthalpies will be used to compute hybridization, ionicity and glass transition temperature, whereas measured data will be generated by combinatorial deposition and thorough physical characterization of thin films to evaluate the band gap and crystallization temperature. This database will be visualized as a map. Statistical methods will be employed for the systematic identification of new OTS materials with specific properties. Selected materials from the predicted OTS class, belonging to binary and ternary chalcogenide systems, will be used to build test devices by photolithography. From the evaluation of their current-voltage behavior, the OTS will be experimentally verified. Well established sub-threshold conduction models will be applied in order to extract material properties, such as trap height and density of defects in the band gap. The interplay between the trap density and trap depth in chalcogenides could be the key for designing OTS materials with application specific electrical characteristics. The rational, data-driven search for materials has the potential to mitigate the costs, risks, and time involved in the current trial-and-error approaches. Systematic exploration of the materials space can significantly accelerate the discovery of new chalcogenide materials and contribute to solving current data storage needs.

Project Objectives:

AMOS goal is the development of a map for discovering novel chalcogenide materials exhibiting OTS. Given the staggering compositional and configurational degrees of freedom possible in materials, a fundamental impediment to an efficient materials discovery process is the lack of suitable methods to rapidly and accurately predict the properties of a vast number of new materials. To date, trial and error strategies guided by intuition have dominated the identification of materials suitable for a specific application.

The key innovative factor of AMOS is the use of a data-driven approach for materials discovery and design, applied to the class of Te-based chalcogenide materials. An innovative combination of material properties for identifying materials for non-volatile memories is used. The database of chalcogenides properties constructed from computed and measured data will constitute an original and highly valuable dataset.

Visualizing and exploring the chalcogenides materials space is extremely useful for further prediction and identification of trends that might otherwise not be obvious, even to experts who have worked in the field for many years. An original and unique map of chalcogenides will contribute to the advance of the domain. It allows researchers to make informed decisions regarding screening and optimization.

Team:

The project team consists of:

1. Dr. Alin Velea - project leader
2. Dr. Claudia Mihai - postdoctoral researcher
3. Dr. Aurelian-Cătălin Gâlcă - young researcher
4. Dr. Cristina Besleaga Stan - postdoctoral researcher
5. Cristina Mozaceanu - master student

Research and Results

Stage 1: Development and validation of the map

Deadline: 31.12.2018

This first stage of the project aimed to develop and validate a map of chalcogenide materials with ovonic threshold switching. Several activities have been accomplished to achieve this goal.

A database was built using a combination of experimental and DFT data. The calculated data set comes from the Open Quantum Materials Database and contains the DFT predictions of the formation energy, bandgap, and volume for about 3,326 compositions. The experimental data set was developed from aggregated data from the literature. It's a compilation of about 478 records from multiple experimental references. A vector of characteristics was constructed based on the composition of the material containing 147 attributes. This vector uniquely identifies a material in the database and is associated with the physics and chemistry that influence the properties of interest. These attributes are designed to allow automated learning algorithms to build general rules to predict new materials. Map1.0 of the entire dataset (3,849 materials) was implemented in the dual descriptor space, of ionicity and covalence.

To predict new OTS materials, using Map1.0 previously developed, we used clustering algorithms. These algorithms refer to a set of unsupervised learning techniques, and are used to detect subgroups or groups in a data set. The K-means clustering algorithm and hierarchical clustering were applied to the data set. The results of the consecutive application of the K-means algorithm showed the partitioning of data into four groups with a subgroup of possible new Te-based OTS materials, containing 154 compositions.

It was started the construction of a dataset for chalcogenide glasses domain formation.  We expect OTS materials to be located in the central area because they are known as good glass formers. This attribute of glass forming ability will be added to the data set used for new OTS materials prediction.

The clustering algorithms predicted that the binary systems Ge-Te and Si-Te have a high probability to contain compositions of materials with OTS. From these two systems we have experimentally prepared the following compositions: SiTe, SiTe2, SiTe3, SiTe4, SiTe6, GeTe4 and GeTe6, which are found in the glass forming domains of these systems.

To study the electrical properties of the materials, we manufactured "cross-point" devices with two terminals. Experimental measurements revealed the presence of ovonic threshold switching effect in all tested compositions, thus confirming the predictive power of the map. Of these, the best material for use in selector devices is GeTe6 due to its high non-linearity of three orders of magnitude, low threshold voltage and low threshold current.

In conclusion, the objectives of this research phase have been fully achieved.

Stage 2: Major map improvement

Deadline: 31.12.2019

This second stage of the project was aimed to the major improvement of the map for chalcogenide materials exhibiting ovonic threshold switching. In order to reach this objective, several activities have been carried out.

The glass transition temperature is the temperature at which the glass undergoes a transition from a rigid to a 'polymeric type' soft form. At normal heating rates, amorphous chalcogenide materials generally crystallize close but above the glass transition temperature. Therefore, this temperature can be regarded as the lower limit of the crystallization temperature. A code has been developed in the Python programming language that implements a glass transition temperature computation model. The developed program can be used to compute the glass transition temperature in binary, ternary and quaternary chalcogenide compositions.

To validate the implemented model of the glass transition temperature, we compared the crystallization temperature (Tc) measured using in-situ X-ray diffraction (IS-XRD) for ten compositions experimentally prepared by magnetron sputtering, with the calculated glass transition temperature (Tg). The results confirm that Tg is always lower than Tc. In addition to our measurements, we have collected a number of literature compositions for the Ge-As-Te system and compared the experimental and calculated temperatures. We obtained an almost linear correlation between these values. We can conclude that the calculated glass transition temperature can be considered as the lower limit for the crystallization temperature.

Since the glass transition temperature model has been validated, it can be used to find new materials with high thermal stability. Using the Lankhorst model, we computed the thermal stability for all the possible AxB1-xTey materials combinations. A and B are any of the following elements: Cu, Ag, Au, Zn, B, Al, In, C, Si, Ge, Sn, N, P, As and Sb. We found a number of 60 chalcogenide compositions with a crystallization temperature above 400 °C. To test the predictive power of the model, we deposited by magnetron sputtering and evaluated by IS-XRD the thermal stability of several predicted  materials in the C-Si-Te and Sn-Si-Te systems. All the tested materials have crystallization temperatures above the calculated glass transition temperatures.

For a more detailed study from a compositional, structural, optical and electrical point of view, we have chosen three ternary systems that we have prepared by combinatorial deposition: Si-Ge-Te, Ag-Si-Te and Sn-Si-Te. The composition of the obtained films was determined by Rutherford Backscattering Spectrometry (RBS) using the Tandem accelerator (3 MV Tandetron) from the IFIN-HH and by X-ray fluorescence (XRF) at the Swiss Light Source synchrotron from the Paul Scherrer Institute in Switzerland. The structure of the deposited materials was tested by X-ray diffraction (XRD) using both laboratory and synchrotron radiation sources. The bandgap of the materials was measured by spectroscopic ellipsometry.

The thermal stability of the three investigated systems was evaluated using two different approaches. Within the first system, Si-Ge-Te, all the samples resulting from the three depositions were structurally measured using XRD. Then the samples were annealed at 400 oC in Ar atmosphere for 15 minutes. Most of the analyzed samples were completely amorphous in the initial state, and after annealing all the samples crystallized. The crystalline phases that are formed are mainly GeTe and crystalline Te, but in some samples we found crystallites of Ge and even observed the formation of a ternary SiGeTe. For the other two systems (Ag-Si-Te and Sn-Si-Te), thermal stability was tested by microXRD measurements performed at the synchrotron. microXRD was performed on the initial samples, but also on samples annealed at 200 oC and 350 oC. It was observed in both systems that the samples in the initial state are amorphous with crystalline inclusions, and the crystalline phases increase with the increase of the annealing temperature.

Lateral devices were built on Si\SiO2 substrates using photolithography and lift-off process. The layers were deposited by magnetron sputtering between TiN electrodes. Several compositions, namely Si0.03Ge0.14Te0.83, Si0.05Ge0.14Te0.81, Si0.07Ge0.13Te0.80, Si0.10Ge0.13Te0.77 and Si0.15Ge0.12Te0.73 were tested. A volatile switching was obtained for Si0.03Ge0.14Te0.83 but from Si0.05Ge0.14Te0.81 a combination of volatile switching and non-volatile switching was observed, and the non-volatile part increased with the increase of  the Si amount.

For a better identification of the possible threshold switching materials, a new version of the map, namely Map2.0, has been developed. Since threshold switching has been observed only in chalcogenide glasses, it is critical to identify in the compositions space those materials that have a stable glassy state. Thus, we have built a valuable data set with the glass formation domain based on experimental data from specialized literature. The constructed data set contains 3348 compositions containing one of the chalcogen elements S, Se and Te, which are in the amorphous (AM), crystalline (CR) phase or in an amorphous matrix with crystalline inclusions (AC). The compositions in the dataset are grouped into 91 ternary systems having one or two chalcogens. The glass transition temperature model was used to estimate Tg for a total of 2734 compositions in the glass formation domain dataset.

Moreover, for these 2734 compositions the bond orbital coordinates, namely the degree of ionicity and the degree of covalency or hybridization were calculated. The identification of potential threshold switching materials was done by focusing the map in the area of ​​Tellurium based materials. Candidates are identified by applying the conditions of formation of stable compounds in the amorphous phase and having a glass transition temperature greater than 400 oC. Following this filtering, 235 compositions were selected, of which 42 are based on Tellurium. Thus, Map2.0 was obtained, which indicates Tellurium based materials with a high probability of showing threshold switching. 

In conclusion, the objectives of this second research phase have been successfully achieved.

The results obtained until now in this project, have enabled the participation at three international conferences with one invited lecture, three oral presentations and four posters. One poster was awarded the the “Ioan Ursu” prize for the original contribution of a young researcher.

1. Invited lecture entitled: “Chalcogenide materials for emerging memories”, authors: A. Velea,  at the “9th International Conference on Amorphous and Nanostructured Chalcogenides (ANC-9)”, which was held on 30 June – 4 July, 2019 in Chisinau, Republic of Moldova.
2. Oral presentation the title: “Optical properties of binary and ternary chalcogenides”, authors: A. C. Galca, F. Sava, I. D. Simandan, G. Socol, A. Velea,  at the “9th International Conference on Amorphous and Nanostructured Chalcogenides (ANC-9)”, which was held on 30 June – 4 July, 2019 in Chisinau, Republic of Moldova.
3. Oral presentation entitled: “Study of phase change in stacked chalcogenide films”, authors: A. Velea, F. Sava, C. Borca, G. Socol, A. C. Galca, C. Mihai, D. Grolimund, at the international conference “12th International Conference on Physics of Advanced Materials (ICPAM-12)”, which was held on September 22 – 28, 2018, in Heraklion, Crete, Greece.
4. Oral presentation entitled: “Physical properties of optimized amorphous Ge-Te alloy thin films for memory applications“, authors: A.C. Galca, C. Besleaga, V. Dumitru, C. Bucur, F. Sava și Alin Velea, at the “18th International Balkan Workshop on Applied Physics (IBWAP 18)”, which took place on July 9 – 14, 2018, in Constanta, Romania.
5. Poster with the title: “Optimized amorphous GeTe films for memory applications”, authors: I. D. Simandan, A. C. Galca, F. Sava, C. Bucur, V. Dumitru, C. Porosnicu, C. Mihai, A. Velea,  at the “9th International Conference on Amorphous and Nanostructured Chalcogenides (ANC-9)”, which was held on 30 June – 4 July, 2019 in Chisinau, Republic of Moldova.
6. Poster with the title: “Chalcogenide materials screening for Ovonic Threshold Switching”, authors: C. Mihai, A. Velea,  at the “9th International Conference on Amorphous and Nanostructured Chalcogenides (ANC-9)”, which was held on 30 June – 4 July, 2019 in Chisinau, Republic of Moldova.
7. Poster with the title: “Characterization of CZTS thin films obtained by magnetron co-deposition from binary sputtering targets”, authors: O. Diagne, A. C. Galca, F. Sava, I. D. Simandan, A. Velea,  at the “9th International Conference on Amorphous and Nanostructured Chalcogenides (ANC-9)”, which was held on 30 June – 4 July, 2019 in Chisinau, Republic of Moldova.
8. Poster with the title: “Cellular automata model of phase change in stacked chalcogenide films”, authors: C. Mihai, A. Velea,  at the “12th International Conference on Physics of Advanced Materials (ICPAM-12)”, which was held on September 22 – 28, 2018, in Heraklion, Crete, Greece. Awarded with “Ioan Ursu” prize for the original contribution of a young researcher.

Four articles have been written and sent for publication in ISI journals.

1. Low power non-volatile memory switching in monolayer-rich 2D WS2 and MoS2 devices
   Authors: C. Mihai, F. Sava, A.-C. Galca, A. Velea
   Submitted for publication to Applied Physics Letters.
2. The Influence of Stoichiometry and Annealing on the Properties of Combinatorial Cu2S - ZnS - SnS2 Thin Films Obtained by Magnetron Sputtering
   Authors: O. Diagne, F. Sava, I.-D. Simandan, A.-C. Galca, M. Burdusel, C. Mihai, A. Velea
   Submitted for publication to Physica Status Solidi B.
3. Structural and electrical characterization of metal diffusion in chalcogenide thin films
   Authors: F. Sava, I. D. Simandan, V. Dumitru, A. C. Galca, I. Stavarache, C. Mihai, A. Velea
   Submitted for publication to Chalcogenide Letters.
4. Multilevel memristive GeTe devices
   Authors: V. Dumitru, C. Besleaga, A. C. Galca, A. Velea
   Submitted for publication to Journal of Ovonic Research.

Project Contact Person:

Dr. Alin Velea

alin.velea@infim.ro