1. Introduction

As a major reliability factor in space-borne microelectronics ^[1], single event effects (SEEs) are caused by heavy ions or protons in space radiation environment. With a higher flux than heavy ions, protons are capable of inducing SEEs through both direct and indirect ionization. SEEs from proton direct ionization have been observed in advanced technologies (below 100 nm) primarily at proton energies below 5 MeV, and have received widespread attention in recent years ^[2-8].

On the other hand, many studies have been published in recent years concerning SEEs from proton indirect ionization, focusing mainly on investigating proton testing methods ^[9,10], proton energy effects ^[11], Monte Carlo simulations ^[12-15], and relations between proton and heavy ion SEE data ^[12,15-18]. In addition, several standards have been developed in recent decades to guide ground-based testing. Some of these standards specify requirements on the proton energy range for qualified SEE evaluation. For example, in 2002 the European Space Components Coordination (ESCC) basic specification No. 25100 stated that the accelerator should be capable of delivering protons in the energy range of 20 to 200 MeV ^[19]. The Sandia National Laboratories document SAND 2008-6983P claimed that the radiation source must be capable of providing protons with energies over a range from at least 20 to 180 MeV. Ideally, the radiation source should be capable of producing protons with energies as high as the maximum energy of protons in the system environment ^[20]. For trapped protons in space, this is 400 MeV ^[21]. More recently, JESD234, issued in 2013, suggested that for a non-destructive single event upset (SEU) test, testing above 200 MeV is not considered necessary. However, for single event latchup (SEL), where some systems forbid parts with any realizable SEL cross-section, testing to energies of 40–500 MeV may be necessary ^[22]. It can be observed that existing standards exhibit a clear discrepancy in the specification of the maximum proton energy for qualified ground-based SEE evaluation. In addition, with the constant downscaling of technology, new materials, processes, and structures are being employed in modern microelectronics, which may lead to changes in the traditional maximum proton energy. Determining the maximum proton energy for qualified SEE testing may benefit the community by 1) avoiding underestimations of SEE sensitivity through using protons with unqualified energies; 2) enabling low-energy proton accelerators qualified for complete ground-based SEE tests in certain circumstances, which will reduce costs and increase the number of suitable facilities; and 3) guiding proton energy selection for in-construction proton facilities and terminals. However, limited data have been published relating to the physical nature of the maximum proton energy and the underlying mechanisms.

Thus, this study focuses on an in-depth investigation of the maximum proton energy for qualified accelerator-based SEE evaluation. Typical SRAM devices were selected, with heavy metal materials (e.g., tungsten (W) and Titanium (Ti)) residing in close proximity to the sensitive volume (SV). By comparing the energy-deposition spectrums in SV and SEE cross-sections for protons with different energies (ranging from tens of MeV to 500 MeV), the maximum proton energy can be determined for different kinds of SRAM devices with various threshold LETs (LET_th). The inner mechanisms were further revealed by analyzing the impact of the top metallization layers, especially high-Z materials (W, Ti, and so on). It was found that the threshold LET for a device is a critical parameter for the determination of the maximum proton energy.

2. Analysis of space proton spectrum

As an example, Fig. 1 depicts the particle flux-energy spectrum on International Space Station (ISS, 500 km, 51.6°) orbit. The CREME96 model was employed to calculate the space particle spectrum ^[23,24]. The AP8MIN model was utilized for trapped protons. The magnetic weather condition was set as "magnetic quiet." The space weather is solar minimum, with 3 mm Al shielding. Particles including protons, alpha particles, and heavy ions are present. It can be seen that (a) compared to other particles, the proton flux is higher; (b) trapped protons contribute to an obvious increase of the proton flux in the energy range of below several hundred MeV, and the flux of trapped protons reaches its maximum at an energy of 40 MeV; and (c) the proton flux decreases rapidly with increasing energy.

Fig. 1.Full-screen View

(Color online) Particle flux-energy spectrum on ISS orbit (calculation conditions: solar minimum, 3 mm Al shielding).

Proton energy is the primary variable in ground-based evaluation of single event effects. However, the energies utilized in a test do not necessarily reflect the proton spectrum in space. The limits of the test energy range versus the actual environment must be taken into consideration during data analysis.

3. Monte Carlo (MC) simulations

The technology evaluated in this study consists of a typical 4 Mbit, 3.3 V CMOS SRAM ^[25,26]. The 3D device model is presented in Fig. 2. A rectangular parallelepiped (RPP)-shaped SV of 2.00 × 2.00 × 2.25 μm³, suitable for the MC simulations, is constructed. The 4 μm² surface of the SV is centrally located beneath the top metallization layers with respect to the beam direction. The surface area of the device model is 10 × 10 μm², which is considered sufficiently large to include the impact of all surrounding protons. Note that although only one memory cell is constructed, the charge-sharing effects between cells were indirectly included in the MC calculations. The reasons for this are described as follows. During the simulations, all the protons were normally and uniformly incident at the surface of the device model, and all the energy depositions in the central SV induced by incident protons were recorded. For protons striking the surrounding area of the central single SV, their energy depositions in the central SV induced by either direct or indirect ionization were also taken into account. However, protons striking the surrounding area of the central single SV will bombard other SVs close to the central one, which means that charge sharing between nearby SVs was indirectly considered.

Fig. 2.Full-screen View

(Color online) The device model (not to scale).

Furthermore, to investigate the proton energy effect on destructive SEEs, the depth of the constructed SV was expanded to 30 μm, because some CMOS devices may have sensitive volume depths of 30 μm or more. Accordingly, the device model in Fig. 2 was modified to fit the change of SV depth. Specifically, the depth of the substrate silicon layer was changed to 50 μm. After the simulations, the SEE responses of the two SVs were compared.

The Geant4 ^[27] and CRÈME-MC ^{[23,24,28,29]} toolkits were utilized. For most of the simulation runs, the ion fluence was between 10¹² p/cm² and 10¹⁴ p/cm², allowing for sufficient statistics. Direct ionization and detailed nuclear reaction processes were both computed, excluding the details of δ-rays. After each run, the deposited energy spectrums in the SV and cross-sections were extracted and analyzed.

4. Results and Analysis

4.1 Energy dependence

Figs. 3 and 4 present the spectrums of the deposited energy in the device SVs with depths of 2.25 μm and 30 μm, respectively. Protons with various energies, ranging from 20 MeV to 500 MeV, were utilized. The energy points were chosen based on existing test standards including ESCC 25100, SAND 2008-6983P, and JESD234. It can be observed that the spectrums exhibit a wide distribution of energy deposition, resulting from both direct and indirect ionization. The left peaks in Figs. 3 and 4 are caused by proton direct ionization. As the proton energy increases, the left peak moves toward the low-deposited-energy side, resulting from a decrease in the electric stopping power of protons.

Fig. 3.Full-screen View

(Color online) Spectrums of deposited energy in the device SV with a depth of 2.25 μm, for protons with various energies.

Fig. 4.Full-screen View

(Color online) Spectrums of deposited energy in the device SV with depth of 30 μm, for protons with various energies.

Conversely, the opposite trend is observed on the high-deposited-energy side, marked by red dashed circles in Figs. 3 and 4.

(1) In Fig. 3, for the 20 MeV protons the maximum deposited energy in the SV is only around 4 MeV. This value increases for protons with higher energies. For the 500 MeV protons, the maximum event can reach as high as 20 MeV. The equivalent LET for proton-induced secondary recoils can be calculated as follows:

${\text{LET}}_{\text{EQ}}=\frac{E_{\text{d}}}{R_{\text{SV}}\times {\rho }_{\text{Si}}}.$

Here, LET_EQ represents the equivalent LET of proton recoils in the SV, as defined by R. Ladbury in 2015 ^[16], E_d denotes the deposited energy of secondary recoils in the SV, and R_SV denotes the travel distance of secondary recoils in the SV. After a nuclear reaction process with a DUT nucleus, proton-induced secondary recoils can strike the SV from all directions. To determine the minimum LET_EQ value of this maximum event, the largest R_SV, i.e., the diagonal incidence should be utilized. Consequently, it can be concluded that the maximum LET of secondary recoils induced by 500 MeV protons in the device should be no less than 24 MeV·cm²/mg.

(2) In Fig. 4, a similar trend is observed as in Fig. 3, which verifies the proton energy effect in devices with different sensitive volumes.

By reverse integrating the counts of energy-deposition events (see Figs. 3 and 4) divided by the ion fluence, the cross-section can be obtained as a function of the critical energy (see Figs. 5 and 6) as

Fig. 5.Full-screen View

(Color online) SEU cross-section as a function of the critical energy for the device SV with a depth of 2.25 μm for protons with various energies.

Fig. 6.Full-screen View

(Color online) SEU cross-section as a function of the critical energy for the device SV with a depth of 30 μm for protons with various energies.

$\sigma =\frac{{\displaystyle \underset{E_{\text{c}}}{\int }N(E_{\text{d}})}}{F}$

where σ represents the cross-section, E_c denotes the critical energy of the device SV, N denotes the count of energy-deposition events in Figs. 3 and 4, and F is the ion fluence per cm².

By comparing Figs. 5 and 6, it can be observed that (1) the trends of the proton energy effect on the SEE response are similar, and (2) the proton-induced cross-sections of the device with an SV depth of 30 μm are higher by approximately one order of magnitude in comparison with those of the device with an SV depth of 2.25 μm.

By determining the critical energy in Fig. 5, the plot of the SEU cross-section against the proton energy, which is usually the end product of accelerator-based SEE testing, can be obtained (as shown in Fig. 7). The threshold LET in the plot is calculated as

Fig. 7.Full-screen View

(Color online) Plot of proton-induced SEU cross-section against proton energy under various device threshold LETs. The depth of the device SV is 2.25 μm.

${\text{LET}}_{\text{th}}=\frac{E_{\text{c}}}{D_{\text{SV}}\times {\rho }_{\text{Si}}}.$

Here, D_SV represents the depth of the device SV.

In Fig. 7, for a small LET_th of 1 MeV·cm²/mg the SEU cross-section appears to be constant, even for a proton energy of 20 MeV. With an increase of LET_th to 4 MeV·cm²/mg, the SEU cross-section is saturated at a proton energy of 50 MeV. As the LET_th increases further, the SEU cross-section appears to increase constantly by several orders of magnitude as the proton energy increases to 500 MeV. Another notable phenomenon is that as the LET_th increases above 8 MeV·cm²/mg, the SEU cross-section–E_p curve becomes "shorter." Protons with an energy below 200 MeV cannot induce SEU when LET_th is below 20 MeV·cm²/mg. In Fig. 8, a similar trend is observed as in Fig. 7, which verifies the threshold LET effect in devices with different sensitive volumes. Note that in Fig. 8, the critical energy rather than the threshold LET is used to define the device SEE sensitivity. The reason for this is that for the device with an SV depth of 30 μm, the relation between the critical energy and threshold LET can change, given that proton-induced secondary particles can penetrate through the device SV from all directions, and the resulting path lengths in the device SV can be very different.

Fig. 8.Full-screen View

(Color online) Plot of proton-induced SEU cross-section against proton energy under various device critical energies. The depth of the device SV is 30 μm.

Thus, it can be concluded that the threshold LET is an important parameter in determining the curve shape and maximum proton energy. For electronic devices with a low threshold LET, which are usually unhardened, testing under 200 MeV is sufficient. However, for high-threshold LET devices, which are usually hardened, tests using insufficient energy may 1) underestimate the saturated cross-section, and thus the device sensitivity; and 2) miss the resulting effects, which may lead to catastrophic consequences owing to incorrect immunity results.

4.2 Metallization dependence

To further investigate the mechanisms of proton-induced SEE, especially the impact of high-Z materials in the metallization layers, the device model used in the MC simulations was modified by replacing the W via layer by silica or replacing all the metallization layers by silica (see Fig. 9). By comparing the simulation results, the metallization dependence can be distinguished and quantitatively analyzed.

Fig. 9.Full-screen View

(Color online) Illustration of the change in the device model: (a) original, (b) "no W" represents the result of a W via layer replaced by a silica layer, (c) "simplified" represents the result of all metallization layers replaced by silica layers. (Not to scale.)

Fig. 10 depicts the spectrums of deposited energy for 500 MeV protons in the device SV. Three cases are compared: original, no W, and simplified. By comparing (a) original and (b) no W, we observe that high-energy deposition events (>10 MeV) result from interactions between the 500 MeV protons and W. It appears that changing the metallization layers has no impact on the deposited energy spectrum in the low-energy region (<10 MeV), which is thus attributed to the interaction between protons and Si and O elements, including direct and indirect ionization processes.

Fig. 10.Full-screen View

(Color online) Spectrums of deposited energy for 500 MeV protons in the device SV with a depth of 2.25 μm. Three cases are compared: original, no W, and simplified.

5. Conclusion

In this work, the maximum proton energy for qualified earth-based SEE testing is investigated using Monte Carlo simulations for SRAM devices with various threshold LETs. The deposited energy spectrums for protons in the device SV and the cross-sections are obtained and analyzed. The maximum deposited energy in the SV increases for protons with higher energies, along with the generation probability. For 500 MeV protons, the maximum event can reach as high as 20 MeV. It is found that the threshold LET of DUT is an important parameter for determining the σ~E_p curve shape and maximum proton energy. For high-threshold LET devices, which are usually hardened, tests using insufficient energy may 1) underestimate the saturated cross-section, and thus the device sensitivity; and 2) miss the resulting effects, which may lead to catastrophic consequences owing to incorrect immunity results. Finally, the mechanisms of proton-induced SEE are further revealed, showing that high-energy deposition events (>10 MeV) in the SV result from interactions between 500 MeV protons and W. It appears that changing the metallization layers has no impact on the deposited energy spectrum in the low-energy region (<10 MeV), which is thus attributed to the interaction between protons and Si and O elements, including direct and indirect ionization processes.