The CHEMAPH project

Chemical Vapor Deposition of Chalcogenide Materials for Phase-change Memories- EU IST Project # 027561

Project in progress supported by the European Commission by the FP6 IST Programme Project Coordinator: Dr. Claudia Wiemer, CNR-INFM National Laboratory MDM, Milan, Italy

Chalcogenide-based phase change materials (which contain at least one element from Group VI in the periodic table: Te, Se,or S) have been studied since the 1960s and they were one of the subjects of the research effort on amorphous materials during the 1970s. However, it was only recently that their special properties generated stronger attention for electronic memory applications, due to their success as optical storage media. Phase-change memories (PCM) are one of the most promising candidates for next-generation non-volatile memories, having the potential to improve the performance compared to Flash memories as well as to be scalable beyond Flash technology.

One technological issue is the phase-change layer deposition process. Phase-change films are currently grown by sputtering, a physical vapor deposition technique, which has yielded demonstrator chips for high-density memories. However, for continued down-scaling for nanoelectronic device architectures, greater control of film deposition over non-planar structures than is possible with sputtering is necessary. This would allow lower programming currents, which leads to improved performance and lower costs. Despite this need, no other deposition routes are available nor widely studied.

The project therefore aims at the development of a film manufacturing process based on a chemical-based technique, metal-organic chemical vapor deposition (MOCVD). MOCVD enables the production of thin films with superior quality compared to those obtained by sputtering, especially in terms of conformality, coverage, and stoichiometry control, and allows implementation of phase-change films in nanoelectronic devices. The main phase-change chalcogenide material system that will be investigated is Ge₂Sb₂Te₅ (GST), as it is already the basis of optical storage media and prototype PCM devices. This objective will require extensive studies of the thermochemical properties of a variety of possible precursor materials, and their interactions, and the investigation of the optimal process conditions to achieve desired film properties, such as resistivity, phase-transition temperature, roughness, density. The project also aims at an in-depth characterization of the electrical, optical, structural, chemical, and functional properties of MOCVD-prepared Ge₂Sb₂Te₅ and similar materials, in order to obtain a link between the phase-change switching behaviour and the chemical-structural composition. The scientific knowledge obtained will enable design of higher-performance materials.

Once a suitable deposition process is developed (i.e. one which yields high-quality films with proper stoichiometry, low impurity level, good adhesion, and high conformal coverage), it will be applied for fabricating state-of-the-art electrical memory cells at the 90/65nm node. The device performance will be evaluated using standard sputter-deposited devices as a benchmark, in terms of parameters such as the programming current, cyclability, and retention.

The development of the deposition technology will proceed with the investigation of the precursor materials, and the deposition conditions in a research-size deposition tool. The target is to demonstrate the feasibility of an MOCVD technique for Ge₂Sb₂Te₅ by the end of the first year, as this compound is currently being used in device development. In the second year, the deposition technique will be extended to cover modified compounds. At the end, devices based on optimized materials will be demonstrated.

The use of chalcogenide materials for electronic memory applications has gained much attention in recent years, due to their successful application as materials for optical data storage. Non volatile memories based on phase change materials are actually the most promising candidate to replace the actual Flash memory technology based on poly-silicon floating gate and on SiO2 as inter-poly-silicon dielectric. Phase change memories (PCM) have been demonstrated to be scalable beyond the limit of actual Flash memory technology. One of the main issues related to the introduction of chalcogenide materials is the choice of the deposition method. In particular, chalcogenides are now deposited by physical vapour deposition methods, but, with the down-scaling of the electronic device architectures, a more conformal deposition method is required. This project aims at the feasibility study of growing the main chalcogenide material, Ge2Sb2Te5 (GST), by metal organic chemical vapour deposition (MOCVD), with equal or superior thickness, stoichiometry and coverage control than those achieved by sputtering techniques. This requires the test and development of suitable precursors for MOCVD, together with the optimisation of process conditions, either by the development of a specific-designed MOCVD system, or by the optimisation of the deposition parameters. The in-depth investigation of structural, chemical and functional properties of MOCVD chalcogenide materials is also an aim of the project. Once the deposition process has been optimised, state of the art memory cells at the 90/65 nm node will be fabricated. The device perfomances will be compared with those of cells based on sputtering-deposited GST as benchmark. In parallel with the device optimisation, modified compounds will be investigated.

The development of a research-size MOCVD reactor, and the demonstration of the feasibility to grow GST by MOCVD is the goal of the first year of the project. The process of MOCVD-GST based devices, together with the study of new compounds is the target of the second year. At the end, devices based on MOCVD-grown, optimised materials will be demonstrated.

This project is coordinated by MDM laboratory, INFM-CNR (Agrate, Brianza, I), a National Laboratory situated inside the site of STMicroelectornics, one of the five main semiconductor manufactures world-wide and having one of the main Research and Development department devoted to the study of innovative non-volatile memories situated at the Agrate Brianza’s site. The consortium gathers Europe-wide experts in the specific fields required for the success of the Chemaph project. The precursor development is assigned to Epichem (UK), the development of the new MOCVD reactor to Aixtron (G), the test of new precursor and preliminary MOCVD growth to Vilnius University (LT), the test of functionality of new materials to CSIC (Sp), and the device development, together with the exploitation and dissemination of results to STMicrolectronics (I). Despite the co-ordination activity, CNR takes also care of the MOCVD growth with the new reactor, and of the structural and chemical characterisation, as well as help in device manufacturing, based on compounds issued from the new reactor.

During this first year of the project, the main effort was concentrated on the design and development of a new MOCVD system for the deposition of Ge2Sb2Te5, on the development of new precursors for the MOCVD growth and on the test with a research system of the deposition conditions of single elements, binary and ternary compounds of the GexSbyTez family.

The design and fabrication of the new system, due to different restrictions and requirements presented in the Wp2 section, has absorbed more resources and time than expected in the DoW. The delivery date of the MOCVD system to CNR-MDM, scheduled for month 6, was shifted to month 11. The development, test and delivery of the new precursors were timely accomplished. The modification of the VU research system in order to fulfil the requirements necessary for the growth of GexSbyTez has been timely accomplished and a lot of work has been devoted to the test of different precursor combinations. Both structural (VU) and functional (CSIC) characterisation have been performed on well selected samples and switching properties has been demonstrated on the GexSby system. Up to now, although the growth of Ge2Sb2Te5 has not been demonstrated by liquid injection MOCVD, a large window of deposition parameters and precursor combinations has been tested. This preliminary work constitutes a very solid starting point for the growth of GexSbyTez . The new MOCVD reactor has been accepted on the Aixtron site at month 11 and promptly moved to the MDM Laboratory. The set up of the apparatus at the MDM laboratory took place at the middle of month 12. Due to some delays of the installation team at ST site, the hook up and installation of the MOCVD reactor, started the end of month 12, is not accomplished at the time of this report.

During month 7, the CHEMAPH coordinator, Dr. Andrew Teren, left CNR-MDM. Dr Claudia Wiemer, CNR-MDM, is the present coordinator of the CHEMAPH project and had the chance to meet all the partners and to become promptly involved and updated on the project achievements during the first Partner meeting held in Vilnius on the July 3th-14th 2006.

After the first year, the exploitable results are mostly related to innovation in the field of precursor selection for chalcogenide material production. Due to the close relation between the new designed MOCVD system and a real industrial system, once the process will be accessed, STMicroelectronics will be able to easily industrialized it. However, due to the early stage of the project and to delay in the availability of the prototype equipment for the MOCVD deposition, it has not been yet possible to define a detailed exploitation plan and to assess the environmental, health and safety issues related to this equipment.

The second reporting Period of the Chemaph project was devoted to the development of phase change materials by MOCVD, together with their structural, chemical and functional characterisation.

Following the indications of the Reviewers, the timing was strictly monitored in order to manage to install the Aixtron MOCVD reactor in the MDM laboratory in the scheduled time frame. Several deposition runs were performed and different chemical compositions including Ge2Sb2Te5, GeTe and Sb2Te3 were tried. Unfortunately, although the right chemical composition and crystallographic structures are demonstrated, the deposition of continuous, smooth chalcogenide thin films was not achieved till now. The achieved results have been discussed and a workplan has been defined to improve the deposition process.

At VU, the modification of the existing substrate holder allowed us to avoid the problem of material accumulation at the reactor walls. With this cold walls reactor, the results were very similar to those obtained by MDM. A second major modification of the VU reactor included a sample holder perpendicular to the vapour flow, including both an internal heating, and a hollow pre-heating furnace able to enhance precursor decomposition. In this way, the first, reproducible, thin and smooth layers of chalcogenide material were obtained at VU. The optimisation of the chemical composition is now running, and the spread of sample for deeper structural, functional, and electrical characterization is scheduled for the first half of December.

Meantime, successful functional characterisation of sputtered deposited Ge2Sb2Te5 was performed. The results where presented at conferences and gathered in one submitted publication.

Following the time schedule fixed by the new Description of Work, alternative chalcogenide materials and dopants for Ge2Sb2Te5 were evaluated by the consortium. Both usefulness in terms of reported electrical performance, the availability of precursors, their compatibility with the actual selected process, together with the safety issues set by the selected scrubbing system and sensors apparatus have been taken into account. Three doping elements have been selected, with the following order of priority, to be integrated within the Aixtron reactor:

1. N
2. Si
3. Se

The deposition by different MOCVD techniques, together with the exploration of new chemistries for phase change materials to be implemented in next generation non volatile memory devices are the subject of current dissemination of this project. The expected final results, i.e 90 or 45 nm demonstrator of phase change memory device including MOCVD chalcogenide material will have a strong impact on the microelectronic community, since devices including MOCVD grown chalcogenide materials were solely recently reported by Samsung Electronics Co. (“Highly Scalable Phase Change Memory with CVD GeSbTe for Sub 50nm Generation”, J.I. Lee, H. Park, S.L. Cho, Y.L. Park, B.J. Bae, J.H. Park, J.S. Park, H.G. An, J.S. Bae, D.H. Ahn, Y.T. Kim, H. Horii, S. A. Song , J.C. Shin, S.O. Park, H.S. Kim, U-In. Chung, J.T. Moon, and B.I. Ryu, VLSI Tech dig. P1203 2007).

The last period of the Chemaph project,-extended by one additional month in order to include the time for an interesting synchrotron radiation experiment on selected, optimised, GeTe samples to take place- was characterized by an intense experimental activity on improvement of growth and characterization of chalcogenide materials. Following the Reviewers’ suggestions, a project flow chart for the remaining time of the project was set in January, including recovering plans in case of failure of the main plan. The effort on deposition by MOCVD was carried out on three fronts: N2-based, standard MOCVD (horizontal tube reactor at MDM), H2 activated liquid injection MOCVD (close coupled showerhead reactor at AIXTRON) and hot wire – liquid injection MOCVD using N2 (vertical tube reactor at VU). A strict interaction between AIXTRON and MDM was set, in order to both evaluate the effectiveness of hydrogen in the process activation and to exchange data/information/characterization results, to reciprocally sustain advances in the respective deposition processes. The final aim was obviously to overcome the difficulties found in the achievement of bidimensional growth of uniform GST layers, so that all the subsequent milestones of the project could be achieved.

Although the effectiveness of hydrogen in improving the morphology cannot be assessed optimized layers were obtained at AIXTRON by conventional thermal MOCVD using low pressure, low temperature processes. Best layers showed reversible phase change behaviour after nanopuls laser irradiation. At MDM, after a wide investigation of the deposition parameters within the limits of the system, an upgrade was decided in order to reach low deposition pressures (p 10 mbar). The upgrade consisted in the heating/isolation of the gas lines carrying the metalorganic precursors and letting them into the reaction chamber. Such installation was performed by Aixtron/MDM personnel in July 2008 and was necessary to allow the working temperature of 45°C for the thermostat baths containing the precursor bubblers. In such a way it could be possible to prevent the condensation of the precursor vapors along the gas lines, thus a sensitive decrease of the source molar flow. Improved layers were then deposited with morphology similar to the one reached at AIXTRON.

At VU, after several reactor reconstructions, the hot wire -liquid injection MOCVD method was used for the deposition of Ge2Sb2Te5 and GeTe, finalized to the fabrication of prototype memory cells. Unfortunately, although an effort was made in order to optimize the handling and timing of the different steps involved in the fabrication of the device, no better results were obtained. By the optimized MOCVD process for Ge2Sb2Te5, depositions were also tried in VU with the aim of demonstrating the MOCVD capability of filling very aggressive, high aspect ratio (AR) structures supplied by Numonyx. Preliminary results show that the MOCVD layer has a very good conformality on structure with an AR around 1:1, while for a very aggressive AR (5:1) the film thickness at the bottom of the trench is reduced. In any case some chalcogenide has been deposited also at the bottom of the high AR trenches, proving the expected well superior performances of the MOCVD approach in term of film conformality with respect to the PVD technique.

During this reporting period, the precursor supply by SAFC totally satisfied the needs of the growth teams, with an impressive effort due to the increasing needs related to the large out put of the deposition tool (working on 12 inches wafers) at AIXTRON. The precursor for doped-GST, namely tBuHNNH2, was also supplied in time to MDM and VU. Growths at VU of N-doped chalcogenide materials just started in October 2008.

The material characterization, assessed during the first two reporting periods, supported the growth teams with on time responses. An increased involvement was set at MDM and AIXTRON, in order to face increasing needs for chemical, structural and morphological characterization, also including sample exchanges and reciprocal visits.

A strong effort was also made at CSIC for providing functional characterization of new materials produced at AIXTRON, MDM and VU. The results are encouraging in terms of demonstrating phase change in samples produced by each of the three MOCVD systems.

Moreover, two additional experimental studies were promoted by CSIC in order to contribute to a better understanding of the structural dynamics underpinning the physics of phase-change memory devices:

• One ultrafast optical experiment, performed in collaboration with Davide Boschetto and Barbara Mansart from LOA-ENSTA in Palaiseau, France, where the coherent phonon signal was measured via optical reflectivity in single and double pump-probe experiments on GeTe films.
• A combined ultrafast optical-X-ray experiment, to be performed in November at PSI, Villigen, Switzerland, aiming at the detection of the coherent phonon and related structural dynamics of the ferroelectric phase transition in crystalline GeTe.

Due to the high surface roughness over the wide area pieces necessary for the 3 ω method, the thermal conductivity of chalcogenide materials produced at VU was not evaluated at MDM. Since the consortium decided that it could be interesting to have an estimate of the thermal conductivity of some materials deposited by HW-CVD, it was agreed to perform some measurements in collaboration with the University of Bordeaux (TREFLE laboratory) that has a world-famous experience in thermal characterization of materials. The measurements were performed by photothermal radiometry and the results turned out to be in agreement with those found in the literature for the same class of materials.

Due to the novelty of the subject, the results obtained during this last reporting period, although not at the expected level, where published in international journals and were presented at different conferences. In particular, the Chemaph Consortium was deeply involved in the organization of Symposium H Materials and Emerging Technologies for Non-Volatile-Memory Devices of EMRS 2008, Spring Meeting, Strasbourg, France and was also represented by two invited talks, one oral and one poster contribution.

The success of SAFC in developing advanced precursor technologies has resulted in a new product range specifically targetted at the PCM market.  Individual Ge, Sb and Te sources have been identified and robust synthesis and purification processes suited to a production environment have been established.  A press release detailing the new products was well received and customer interest has been high.  SAFC will continue to work to bring the optimum chemicals to the semiconductor industry to meet current and future requirements.

Ahead of schedule AIXTRON has realized a first production prototype for the deposition of GST on 300 mm wafers. Analyzing the status of the process development after the second year AIXTRON decided to use the existing Tricent® technology as basis for the GST production tool. The systems provide for production necessary high deposition rate, high throughput and lower cost of ownership by employing unique processing technique enabled by a proprietary pulsed injection vaporizer called TriJet. In a first step the precursor delivery system was adapted to the GST precursors and up coming additional safety issues have been addressed.

The demonstrated good conformality of the MOCVD film on Numonyx high aspect ratio structures represents an important achievement of the project and it proves that this technique is mandatory for the fabrication of alternative PCM cell architecture (full-confined PCM cells). On the other side, the electrical properties of the MOCVD deposited film are still in the validation phase and only a preliminary assessment has been achieved. It follows that the PVD deposited GST still remains the reference material for the PCM cell fabrication, even if there is interest to continue the evaluation of the MOCVD technique through demo activities from equipment suppliers.