DOI: 10.1007/s10967-008-0914-1 Journal of Radioanalytical and Nuclear Chemistry, Vol. 278, No.2 (2008) 509–512
0236–5731/USD 20.00 Akadémiai Kiadó, Budapest
© 2008 Akadémiai Kiadó, Budapest Springer, Dordrecht
Thermal neutron induced soft error rate measurement
in semiconductor memories and circuits
C. Çelik,
1
K. Ünlü,
1
* K. Ramakrishnan,
2
R. Rajaraman,
2
V. Narayanan,
2
M. J. Irwin,
2
Y. Xie
2
1
Radiation Science and Engineering Center, Penn State University, University Park, PA 16802, USA
2
Department of Computer Science and Engineering, Penn State University, University Park, PA 16802, USA
(Received July 10, 2008)
Soft error rate (SER) testing and measurements of semiconductor circuits with different operating voltages and operating conditions have been
performed using the thermal neutron beam at the Radiation Science and Engineering Center (RSEC) at Penn State University. The high neutron
flux allows for accelerated testing for SER by increasing reaction rate densities inside the tested device that gives more precision in the
experimental data with lower experimental run time. The effect of different operating voltages and operating conditions on INTEL PXA270
processor has been experimentally determined. Experimental results showed that the main failure mechanism was the segmentation faults in the
system. Failure response of the system to the operating conditions was in agreement with the general behavior of SERs.
Introduction
A single-event upset or soft error is the sensitivity of
semiconductor memories and circuits to environmental
radiation. The environmental radiation creates excess
charge carriers that directly or indirectly induces
localized ionization, which can change the state of the
internal values of the memory cells and cause wrong
data to get latched onto the output of the combinational
circuits.
Initial experimental results of SER measurements for
thermal neutron-induced SER on semiconductor
memories have been published previously.
1
The
majority of detected errors in memory chips originate
from transient failures and results can be found in the
literature.
2,3
As the technology improves, SER will
significantly affect reliability and dependability of the
semiconductor memories. SER measurements and
testing of devices have been reported in the literature
extensively.
4–6
In the past, experiments were carried out
with alpha-particles to study their effects on memories
and showed that the effects of borophosphosilicate glass
(BPSG) layers on SER could be avoided by reducing the
SER due to
10
B containment.
7–11
In this paper, SER measurements for PXA270
microprocessor at the Penn State Breazeale Reactor
(PSBR) using thermal neutrons are discussed. Currently,
fast neutron effects are being explored to observe effects
of higher-energy neutrons on circuit operation and SER.
In addition, simulations of SER in semiconductor
circuits for verifying the experimental results are also
being studied.
* E-mail: k-unlu@psu.du
Experimental setup
The Radiation Science and Engineering Center
(RSEC) at Pennsylvania State University served as the
test facility for the SER measurement. The Penn State
Breazeale Reactor (PSBR) [TRIGA] was used as the
neutron source in the experiments. The maximum rated
power of the reactor is 1 MW in the continuous mode,
and 2000 MW in the pulse mode. The reactor power is
adjusted from 10 W to 1 MW to observe the soft error
rate dependence on neutron flux. Figure 1 shows the
experimental setup. In order to get a well-thermalized
neutron beam, a D
2
O tank was used between the core
and beam port.
The average thermal flux at the exit of the beam port
is about 3
.
10
7
n
.
cm
–2.
s
–1
, and the fast flux is about three
magnitudes smaller. The neutron spectrum, which was
measured using a slow neutron chopper, was compared
with the corresponding Maxwell Boltzmann distribution
and was found to be fairly matching. The different
power levels at PSBR allow us to characterize the soft
error rate and categorize the different types of errors
effectively. PSBR has the ability to produce fast
neutrons (energy >1 MeV) and epithermal or thermal
neutrons (energy <0.1 MeV) depending on the two types
of tests namely the in-core and out-of-core tests. The
neutron flux near the core is dominated by fast neutrons
while the flux at the end of the beam port is dominated
by the thermal neutrons due to the loss of energy during
the passage of neutrons through a heavy water tank. The
upper energy limit of the fast neutron flux is less than
the maximum atmospheric neutron spectrum, but
corresponds to dominant portion of the atmosphere
neutron flux.