Radiation-induced errors in memory chips

Radiation-induced errors in memory chips

R. J. Peterson

University of Colorado, Boulder, CO 80309-0446 USA

ABSTRACT

We have measured probabilities for proton, neutron and pion beams from accelerators to induce temporary or soft errors in a wide range of modern 16 Mb and 64 Mb DRAM memory chips, typical of those used in aircraft electronics. Relations among the cross sections for these particles are deduced. Measurement of alpha particle yields from pions on aluminum, as a surrogate for silicon, indicate that these reaction products are the proximate cause of the charge deposition resulting in errors.

I Introduction

In 1978 a serious industrial problem was causing errors in the memory contents of Intel 2107 16Kb DRAM (Dynamic Random Access Memory) chips.[1] These temporary or soft errors did not damage the chip, but an erroneous 'zero' had replaced a 'one', or vice versa, spoiling the contents of the memory. This problem was traced to small amounts of alpha-emitting impurities in the materials. These 5 MeV alpha particles caused ionization in the Si of the chip, and the collected charge was enough to redefine the memory state.

Later, in repair records from 1984, IBM found a correlation between these errors, here called 'soft error upsets' or SEU, and altitude. These errors were then shown to be closely correlated with cosmic ray intensities.[2,3] In Denver, at 1700m, the SEU rate was ten times the national average.

It is easy to understand how heavily-ionizing low energy alpha particles, with a range of 25 microns in Si, could deposit disruptive charge, but how would high energy neutrons, protons and pions from cosmic ray reactions induce errors? At sea level, most cosmic rays are muons, with only small cross sections for any reactions, but evidently energetic hadrons can cause reactions within microcircuits to produce short range heavily ionizing reaction products.

A series of experiments with intense accelerator beams has been used to measure the probabilities or cross sections for hadrons to induce SEU in a wide range of modern memory chips. These measurements are particularly important for commercial and military aircraft, with their high altitude operations.

Here we define

II Accelerator testing

Commercially available memory chips were placed directly in our beams. Proton beams were mostly from the Harvard cyclotron, at energies from 50 to 150 MeV.[4] Pion beams from LAMPF [5] and TRIUMF [6] and a 14 MeV neutron generator [7] have been used. Appropriate beam monitoring gave beam fluxes, for instance 0.1 - 1 × 10⁶ pions/cm² sec.

A special driver box alternately wrote and read patterns of zeros and ones into the memory chips. Discrepancies were counted as errors. The 16 Mb samples used three technologies - Trench External Charge (TEC), Stacked Capacitor (SC) and Trench Internal Charge (TIC). Samples from several manufacturers were chosen for TEC and SC technologies.

For 16 Mb TEC samples the pion cross sections showed clear evidence, in the macroscopic chip, of the familiar 3-3 pion-nucleon resonance.[5] See Fig. 1. We also note the expected charge symmetry between p⁺ and p^– beams for reactions upon symmetric ²⁸Si. The curve shows the computed p⁺/p^– averaged ²⁸Si reaction cross section.[8] Proton SEU cross sections are lower, and show no resonant effect; proton reaction cross sections show no resonances in this high range of energies.[9] A single datum shows the SEU cross section for 14 MeV neutrons. A smaller range of pion energies was used for the 64 Mb DRAM samples, and the 3-3 resonance was not obvious, as seen in Fig. 1.[6]

III Comparisons

SEU cross sections for 16 Mb samples showed a very wide range, strongly dependent upon the chip construction. Cross sections are shown in Fig. 2. For some manufacturers, several different chips of the same model were studied, with little scatter. The large differences can be explained by the smaller volume of Si available to deposit stray charge in the newer generations.

Only SC and TIC technologies were used for the 64 Mb samples we tested, and again a strong difference is found for cross sections for these technologies. Cross sections for several samples from several manufacturers are shown in Fig. 3.

A similar pion experiment measured SEU in a commercial 64 kb SRAM, with results shown in Fig. 4. Largest pion SEU cross sections for 16 Mb and 64 Mb DRAM samples were 6 × 10^-13 and 4 × 10^-14 cm², comparable to these for the SRAM sample.

To understand these effects, we need to compare the charge left in Si by ionizing particles to the charge that determines a memory state. The critical charge that distinguishes a '0' from a '1' in a modern device is about 50 fC or about 0.3 million ion pairs. One MeV deposited in Si creates 0.28 million pairs, or 45 fC, very near a critical charge. A minimum ionizing particle deposits only about 0.47 keV per micron, or 130 pairs per micron of its passage. Since memory cells are only a few microns on a side, this dE/dx is not the cause of errors. In contrast the 5 MeV alpha particles from natural radioactivity have a range of 25 microns, and deposit hundreds of keV per micron as they slow. If charge is collected from different volumes of Si by the several chip types, greater or lesser amounts of charge may be collected.

The set of identical 16 Mb samples used for Fig. 2 was studied by us with the same equipment for 148 MeV protons, 150 MeV pions and 14 MeV neutrons.[4] Pions of this energy sit atop the fundamental 3-3 resonance where reaction cross sections are largest. The scatter plot to compare these is shown in Fig. 5. Note that the resonant energy pions have cross sections larger than for protons or neutrons.

With a similar set of identical 64 Mb chip samples, we made a similar close comparison of SEU for 148 MeV protons, 14 MeV neutrons and 154 MeV pions. The scatter plot for 14 different 64 Mb chips is shown in Fig. 6. These pion data were also shown in Fig. 3.

In both scatter plots the solid line indicates proton cross sections equal to those for the neutrons or the pions. At this resonance energy pion s_SEU are about three times larger than proton data. We note the wide range of cross sections, and thus susceptibility to errors in the circuits, with TIC devices less sensitive than SC devices. The TEC technology is not used for 64 Mb chips.

One might expect s_SEU to be related to s_R, the reaction cross section for all that happens other than elastic scattering. These s_R are known or calculated reliably. Pion s_SEU for 16 Mb TEC devices show a close proportionality to s_R from 50 to 400 MeV, across the important resonance region, as shown in Fig. 1. This ratio is very different for other samples, but the resonance is visible in all samples studied for a range of pion beam energies. The ratio of proton s_SEU to s_R increases only slightly from 60 to 300 MeV.[6]

These data show smaller SEU cross sections for 64 Mb samples than for 16 Mb samples. If the chip area is the same, there would be a factor of four difference between the cross-sectional areas per bit for these two densities. Older chips with lower densities have also had soft error cross sections measured for cosmic rays.[10] Fig. 7 shows those SEU cross sections and those currently measured for pions with the largest cross sections. These cross sections drop more quickly than just the cross sectional area, and more like the linear dimension to the third power, as if the sensitive volume is decreasing, not just the cross sectional area.

For these typical modern devices, we have thus found empirical connections among data for neutrons, protons and pions. These suffice for modeling responses to cosmic ray fluxes for many current applications. Further, our results indicate that we need measure s_SEU for only one set of these beams to infer responses to the others. This can lead to a dramatic decrease in the cost of testing.

Experts can model the fluxes of protons, neutrons and pions from cosmic rays at a range of altitudes and latitudes. Together with our data, we can now calculate rates of soft errors in complex avionics packages for commercial and military aircraft, depending on the type and manufacturer of the memory chips used.[11]

IV Reaction mechanisms

This empirical knowledge of chip SEU cross sections should be based upon an understanding of the means by which lighly ionizing protons, neutrons and pions interact to deposit charge within devices. Extensive modeling studies have been carried out for many electonic components by those in the industry.[12] Because of their high rate of energy loss, low energy alpha particle production is a likely direct means to induce errors, since it has long been known that 5 MeV alpha particles from natural radioactivity are dangerous sources of SEU.[1] Reaction model calculations for protons and pions on Si provide spectra for emerging alpha particles that peak near this energy of 5 MeV [13], but measurements have not gone to such low energies.

We have carried out an early measurement of alpha particles in the 5 MeV energy range emerging from Al, as a simple substitute for Si, using pion beams from TRIUMF and commercial plastic track detector technology. Error bars are large, and experimental control was difficult. Our method was also not able to distinguish between alpha particles and low energy protons which also leave heavy ionization in the material. Our method has no real energy determination, but is expected to be sensitive for alpha particle energies from 0.5 to 6 MeV.

Nonetheless, we present in Fig. 8 the cross sections we have measured.[14] Comparison is made to an integral over the calculated spectra [13] across the alpha particle energy range we expect to be able to sense, and to the computed reaction cross sections. Yields are large and seem to reflect the 3-3 resonance as expected. The measured cross sections are near the reaction cross sections, indicating a multiplicity greater than one, possible for the very high excitation energies that can be available from pion absorption. We are thus confident, even from this rough measurement, that pion (and presumably proton and high energy neutron) reactions produce alpha particles within the body of the microcircuits, and these short range and heavily ionizing products deposit the energy and charge needed to induce a transition among ones and zeroes to cause the SEU.

Acknowledgements

This work was supported in part by the USDOE, with the strong collaboration of IBM and the US Naval Academy.

References

[1] T. May and N. Woods, IEEE Trans. Electron Devices 26, 2 (1979).

[2] J. F. Ziegler and W. A. Lanford, Science 200, 776 (1979).

[3] J. F. Ziegler et al., IBM J. Research and Devel. 40, 3 (1996).

[4] J. F. Ziegler et al., IEEE Trans. Solid State Circuits 33, 246 (1997).

[5] C. J. Gelderloos et al., IEEE Trans. Nucl . Sci. 44, 2237 (1997).

[6] G. J. Hofman et al., IEEE Trans. Nucl. Sci. 47, 403 (2000).

[7] J. D. Shell et al., 1997 HEART/GOMAC Conference, Las Vegas.

[8] A. A. Ebrahim and R. J. Peterson, Phys. Rev. C54, 2499 (1996); R. J. Peterson, Few Body Sys. Suppl. 9, 17 (1995).

[9] R. F. Carlson, At. Data Nucl. Data Tables 63, 93 (1996).

[10] cited in Ref. 3.

[11] M. E. Nelson et al., Jour. Radiation Effects, Research and Engin.19, 100 (2001).

[12] G. R. Srinivasan, IEEE Trans. Nucl. Sci.41, 2063 (1994).

[13] G. Srinavasan, H. Tang and P. Murley, IEEE Trans. Nucl. Sci. 41, 2083 (1994); H. Tang, IBM Jour. Res. Dev.40, 91 (1996).

[14] R. J. Peterson et al., Radiation Measurements 35, 565 (2002).

Received on 30 October, 2002

Publication Dates

• Publication in this collection
  25 Aug 2003
• Date of issue
  June 2003

History

• Received
  30 Oct 2002
• Accepted
  30 Oct 2002