Memory Chip Upsets
Development of Methods for Reliable Prediction of the Memory Chip Upsets in Radiation Fields
Tech Area / Field
3 Approved without Funding
Khlopin Radium Institute, Russia, St Petersburg
- Brookhaven National Laboratory / Associated Universities, Inc., USA, NY, Upton\nPower Reactor and Nuclear Fuel Development Corporation (PNC), Japan, Tokyo\nLawrence Livermore National Laboratory, USA, CA, Livermore\nCollege of Engineering & Science, USA, SC, Clemson\nUniversity of Jyväskylä, Finland, Jyväskylä
Краткое описание проектаExploitation of information systems with memory chips built of semiconductor elements has demonstrated that distortion of the information at the level of one bit is highly probable in fields of radiation. Such a logical inversion process, after which the irradiated memory chip still operates, is commonly cited as Single Event Upset (SEU). The history of space exploration confirms that memory upsets of such kind are leading to unpredictable and sometimes disastrous consequences.
There exists also a certain probability of the Multi-Bit Upsets of the memory chips - situation when upsets occur in several memory cells simultaneously. In this case it is impossible to correct information with the help of special codes. The overall tendency for the size decreasing of the memory cells enhance the probability of the upsets of this kind. These factors make the development of the methods for the predictions of SEU/MBU in the memory chips affected by the radiation essential.
A number of models simulating the phenomenon have been proposed, among these is the model of V.V.Miroshkin and M.G.Tverskoy (IEEE Trans. Nucl. Sci., 41, 1994, pp. 2085-2092). This model of Project participants includes two parameters: the effective volume of the memory cell V and a threshold energy Ethr needed to be deposited in the volume V in order to produce upset. By performing two or more experiments for the SEU rates at different initial conditions (type or energy of the projectile) V and Ethr can be determined. Once these parameters are defined, prediction of the probability of the memory cells upset may be evaluated for radiation fields of different particle types and energies for which direct experimental study is not possible. The results of calculations show that suggested approach has quite a good predictive power and accuracy.
The essential part of the model consists of predictions for the energy deposit in the memory cell volume by the reaction products. This is calculated within the frame of a Monte-Carlo method. Such calculations are accurate enough only if the details of secondary interactions of the nuclear reaction products with memory cell media atoms and nuclei are known. Experimental measurements of the energy loss spectra in Si and GaAS layers and energy loss spectra due to particles leakage into detectors from different targets (C, Al, Ti, Cu, Pb) are therefore needed for comparing with the developed theoretical approach. In such experiments semiconductor detector, acts as the "enlarged" prototype memory cell of the chip.
The goal of the proposed Project is to develop reliable methods for predicting upset rates of the memory chips in different radiation fields and to perform experimental tests of basic ingredients of the approach in order to increase the reliability of the calculations.
The main tasks of the Project are as follows:
- Irradiation of the Si and GaAs detectors with different thickness in the proton beam of the PNPI synchrotron (incident proton energy 1 GeV), measurements of the energy losses by the reaction products in thin Si and GaAs layers and energy transfer spectra from C, Al, Ti, Cu and Pb targets.
The proposed systematic measurements of energy losses in thin GaAs and Si layers and linear energy transfer spectra from different targets with detectors of different thickness will be carried out for the first time. This research is important, both for the fundamental knowledge and possible applications. It is especially evident for the energy losses in thin GaAs layers for which there are no experimental data available.
- Development of the calculation algorithm, which will account for the variation of the effective charge of the ion, fluctuations of the energy losses during the ion stopping process in the memory cell media, and subsequent charge recombination. The nuclear fragmentation and slow negative pion reabsorbtion in addition to other inelastic interaction channels will be included within the developed approach.
- The development of methods for reliable prediction of the probability for the Multi-Bit Upsets (MBU).
- Testing of the developed theoretical approach and the estimation of the predictive power of the proposed algorithm by comparing theoretical calculations and experimentally measured data on SEU; complete investigation of the impact of the values of the constants and parameters used within the theoretical approach for the description of the nuclear and atomic processes on the finally obtained results.
This Project will result in the development of a reliable method for prediction of SEU and MBU of the semiconductor memory chips in different radiation fields realized as a set of computer codes. Results obtained will be important for the nuclear and material science and will have a certain commercial interest. This interest will be based on the necessity in such predictions which might lead to a significant decrease of the costs needed otherwise for the special experimental memory chip testing. Besides, this Project will open alternative job opportunity for the Russian scientists previously involved in the military related projects.
The Project is open for foreign partners to participate. Foreign collaborators would visit our laboratory, discuss results, receive first-hand information of latest developments and, if necessary, jointly develop suitable applications. We also hope to use their experience in suggesting ways for our investigations, analyzing obtained experimental data, and also in mutual scientific and commercial using of the results of Project realization.