Estimates of SEU For Semiconductors Using MC50 Cyclotron and GEANT4 Simulation
Estimates of SEU For Semiconductors Using MC50 Cyclotron and GEANT4 Simulation
Estimates of SEU For Semiconductors Using MC50 Cyclotron and GEANT4 Simulation
20222025
J. -S. Chai
Department of Energy Science and School of Information and Communication Engineering,
Sungkyunkwan University, Suwon 440-746, Korea
(Received 26 April 2010)
The Single Event Upset (SEU) on semiconductor RAM memories due to terrestrial neutrons is
measured by using the MC50 Cyclotron at KIRAMS. Proton beams impinging on a thick Be target
produce neutrons, which then bombard the semiconductors to cause SEU. Due to the thickness of
the Be target used in this experiment, the neutron beam has a broad energy spectrum. Thus, the
SEU could not be obtained directly as a function of neutron energies. To solve this problem, we
propose an approximate method of estimating the SEU as a function of the neutron energy when
the neutron beams have a broad range of energy. In this work, three different energies of proton
beams of 30, 35, and 40 MeV were used. By extracting the difference of the neutron yields at these
energies, we can make a rough estimate of SEU as a function of neutron energies. Monte Carlo
simulations are performed to obtain the neutron flux from the MC50 cyclotron. Our results for the
SEU turn out to be comparable to the previous results.
PACS numbers: 61.80.Hg, 29.20.Hm
Keywords: SEU, MC-50, Neutron, Monte Carlo simulation, GEANT4
DOI: 10.3938/jkps.59.2022
I. INTRODUCTION
Random access memories (RAMs) are sensitive to
high-energy particles such as energetic neutrons in the
atmosphere. The excessive charges generated by high
energy particles incident on reverse-biased PN junctions
can induce an instant current flow which may disrupt
normal system function [1]. This single-event effect
(SEE) can change the content of the memory elements,
which is called Single Event Upset (SEU). In our previous work, we proposed an approximate subtraction
method to take into account the non-mono energetic
nature of the neutron beam from MC-50 cyclotron at
KIRAMS (Korea Institute of Radiological Medical Sciences) [2], where we calculated neutron spectra with
GEANT4 (v.9.1) with G4binary cascade model and ob E-mail:
swhong@skku.ac.kr
1
dNSEU (E)
Nbits dF Ln (E)
(1)
-2022-
Estimates of SEU for Semiconductors Using MC50 Cyclotron and GEANT4 Simulation J. W. Shin et al.
-2023-
in C++ language, which enables the simulation of propagation of particles that interact with materials and/or
other particles. It is widely used in many different fields
such as high energy and nuclear physics [4-5], medical
physics [6-7] and space science [7-8].
A simulation of the electromagnetic processes is well
standardized, since the underlying theories (i.e., Classical and Quantum Electrodynamics) are well established.
On the contrary, there is a huge diversity for the hadronic
processes, depending on the particles under consideration and on their energies. To accommodate this diversity, many packages (or physics models) are builtin in GEANT4. Users should then invoke the relevant
physics models suitable for the process [9-10], which we
will briefly discuss in the next subsection. In this work,
GEANT4 9.3 version is used.
4. Hadronic Models in GEANT4
-2024-
Journal of the Korean Physical Society, Vol. 59, No. 2, August 2011
Table 1. SEU cross sections ( 1014 ) [cm2 /bit].
6
10
= 30 MeV
G4LE
( n / cm / MeV )
G4Binary
10
En
G4Berti
6
10
= 35 MeV
Previous
results [2]
= 40 MeV
32.5
1.83 0.72
1.99 0.81
1.51 0.79
1.79 0.88
37.5
2.77 1.12
2.84 1.17
2.62 1.15
2.48 1.20
32.5
3.97 0.59
4.24 0.79
3.50 0.69
4.00 0.76
37.5
4.88 0.91
5.01 0.95
4.50 0.82
4.35 1.03
(MeV)
10
10
10
dFL /dE
G4LE
G4Binary
G4Bertini
10
SRAM0
10
10
10
15
20
25
30
35
40
45
(MeV)
x 10
difference spectrum
(a)
SRAM1
G4LE
G4Binary
G4Berti
0
9
(b)
10
15
20
25
30
35
40
45
(MeV)
Fig. 4.
(Color online) Difference spectra between two neutron fluencies are plotted.
(a) represents the difference spectra of (dF Ln /dEn )Ep = 35 MeV (dF Ln /dEn )Ep = 30 MeV and (b) represents the difference
spectra of (dF Ln /dEn )Ep = 40 MeV - (dF Ln /dEn )Ep = 35 MeV .
ence spectrum between two neutron fluences from two incident proton energies. In our analysis, neutrons having
the energy lower than 1 MeV are excluded, because low
energy neutrons cause little contribution to SEU cross
sections [12]. By subtracting (dF Ln /dEn )Ep = 30 MeV
from (dF Ln /dEn )Ep = 35 MeV we obtain the spectrum in
Fig. 4(a) which looks similar to the quasi-mono energetic
spectrum at En = 32.5 MeV with a width of 5 MeV. The
same can be done at 35 < En < 40 MeV to get the spectrum in Fig. 4(b). We may take Fig. 4 (a) as a virtual
quasi-mono energetic neutron beam at En = 32.5 MeV.
Similarly Fig. 4 (b) can be taken as a virtual quasi-mono
energetic neutron beam at En = 37.5 MeV. Figure 4 also
shows that the difference spectra from different hadronic
models are more or less the same.
3. SEU cross section
Estimates of SEU for Semiconductors Using MC50 Cyclotron and GEANT4 Simulation J. W. Shin et al.
-13
(cm /bit)
10
-14
SEU
10
TSL
TSL (Raw)
KIRAMS with G4LE
KIRAMS with G4Binary
KIRAMS with G4Berti
10
-15
20
40
60
80
100
120
140
160
180
200
(MeV)
Fig. 5. (Color online) SEU cross sections for SRAM0 device. The filled circles and triangles represent TSL (with the
full neutron spectrum) and TSL (Raw) (with only the peak
of neutron spectrum) results, respectively [12]. The open
squares, circles and triangles represent KIRAMS results with
LE, Binary and Bertini models, respectively. The open diamonds denote our previous results [2].
-13
(cm /bit)
10
10
-14
SEU
TSL
TSL (Raw)
KIRAMS with G4LE
KIRAMS with G4Binary
KIRAMS with G4Berti
10
-2025-
-15
20
40
60
80
100
120
140
160
180
200
(MeV)
Fig. 6. (Color online) SEU cross sections for SRAM1 device. The filled circles and triangles represent TSL (with the
full neutron spectrum) and TSL (Raw) (with only the peak
of neutron spectrum) results, respectively [12]. The open
squares, circles and triangles represent KIRAMS results with
LE, Binary and Bertini models, respectively. The open diamonds denote our previous results [2].
[1] R. C. Baumann, Int. J. High Speed Electronics and Systems 14, 299 (2004).
[2] J. K. Park, S. G. Kwon, S. W. Lee, J. T. Kim, J. S.
Chai, J. W. Shin and S. W. Hong, J. Korean Phys. Soc.
58, 1511 (2011).
[3] S. Agostinelli et al., Nucl. Instrum. Methods Phys. Res.
Sect. A 506, 250 (2003).
[4] V. N. Ivanchenko, Nucl. Instrum. Methods Phys. Res.
Sect. A 494, 514 (2002).
[5] J. T. Rhee, M. Jamil, B. Hong, K. S. Lee, S. Park, K. S.
Sim, Y. J. Kim, Y. U. Kim, S. K. Nam, D. G. Koo, S.
J. Lee, S. Y. Bahk and S. J. Hong, J. Korean Phys. Soc.
48, 33 (2006).
[6] D. Visvikis, M. Bardies, S. Chiavassa, C. Danford,
A. Kirov, F. Lamare, L. Maigne, S. Staelens and R.
Taschereau, Nucl. Instrum. Methods Phys. Res. Sect. A
569, 335 (2006).
[7] V. N. Ivanchenko, Nucl. Instrum. Methods Phys. Res.
Sect. A 525, 402 (2004).
[8] V. S. Makhmutov, L. Desorgher, G. A. Bazilevskaya, E.
Fl
uckiger and J.-P. Raulin, Adv. Space Res. 39, 1458
(2007).
[9] J. P. Wellisch, Nucl. Instrum. Methods Phys. Res. Sect.
A 502, 669 (2003).
[10] A. Khan, K. Schofield and D. Wright, IEEE Nucl. Sci.
Sym. Conf. Rec. (Wyndham, Puerto Rico, 2005).
[11] H. C. Fesefeldt, Technical Report PITHA 85-02, 1985.
[12] T. Granlund and N. Olsson, IEEE T. Nucl. Sci. 53, 1871
(2006).