⇦Chinese Physics Letters, 2017, Vol. 34, No. 11, Article code 118504 Influence of Tilted Angle on Effective Linear Energy Transfer in Single Event Effect Tests for Integrated Circuits at 130 nm Technology Node * Le-Qing Zhang(张乐情)1,2, Jian Lu(卢健)3, Jia-Ling Xu(胥佳灵)3, Xiao-Nian Liu(刘小年)1,2, Li-Hua Dai(戴丽华)1,2, Yi-Ran Xu(徐依然)1,2, Da-Wei Bi(毕大炜)1**, Zheng-Xuan Zhang(张正选)1 Affiliations 1State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050 2University of Chinese Academy of Sciences, Beijing 100049 3Shanghai Engineering Center for Microsatellites, Chinese Academy of Sciences, Shanghai 201203 Received 14 August 2017 *Supported by the Key Laboratory of Microsatellites, Chinese Academy of Sciences.
**Corresponding author. Email: davidb@mail.sim.ac.cn Citation Text: Zhang L Q, Lu J, Xu J L, Liu X N and Dai L H et al 2017 Chin. Phys. Lett. 34 118504 Abstract A heavy-ion irradiation experiment is studied in digital storage cells with different design approaches in 130 nm CMOS bulk Si and silicon-on-insulator (SOI) technologies. The effectiveness of linear energy transfer (LET) with a tilted ion beam at the 130 nm technology node is obtained. Tests of tilted angles $\theta =0^{\circ}$, 30$^{\circ}$ and 60$^{\circ}$ with respect to the normal direction are performed under heavy-ion Kr with certain power whose LET is about 40 MeVcm$^{2}$/mg at normal incidence. Error numbers in D flip-flop chains are used to determine their upset sensitivity at different incidence angles. It is indicated that the effective LETs for SOI and bulk Si are not exactly in inverse proportion to $\cos \theta$, furthermore the effective LET for SOI is more closely in inverse proportion to $\cos \theta$ compared to bulk Si, which are also the well known behavior. It is interesting that, if we design the sample in the dual interlocked storage cell approach, the effective LET in bulk Si will look like inversely proportional to $\cos \theta$ very well, which is also specifically explained. DOI:10.1088/0256-307X/34/11/118504 PACS:85.40.-e, 29.27.Fh © 2017 Chinese Physics Society Article Text Earth-based testing is of primary importance to guarantee the reliability of integrated circuits for space missions.[1] Thus devices are irradiated with different ions at various values of linear energy transfer (LET). Limited to the cost of beam time and the actual availability of ion cocktails in accelerators, single event effect (SEE) testing procedures often include the possibility to use tilted ion strikes to imitate normally incident ions of higher LET.[2,3] This practice is known as the effective LET method and it relies on the assumption that the total energy deposited in the sensitive volume of the device is proportional to the ion path length, which obeys an inverse-cosine law. This law assumes a slab-shaped rectangular parallelepiped (RPP) sensitive volume. However, this law is known to suffer from inherent failure as usual.[3-5] Reasons for these discrepancies are still discussed. The commonly proposed explanation is that the depth-to-width aspect ratio of the device sensitive volume is not small enough for the inverse cosine law to apply,[6] in other words, the RPP becomes box-shaped rather than slab-shaped.[7] Few data reporting angular effects for silicon-on-insulator (SOI) devices are available.[8,9] It would be interesting to quantify the angular effects in SOI transistors. The SOI technology is an ideal candidate to reduce the sensitive volume from which ion-induced charge can be collected,[10-12] hence it is a very popular and efficient hardening method against SEU. In this work, to evaluate whether effective LET (LET$_{\rm eff}$) still obeys an inverse-cosine law when the beam is tilted in 130 nm CMOS bulk Si and SOI technologies, specially designed experimental samples with the same usage are fabricated on 130 nm SOI and 130 nm bulk Si CMOS technologies separately. Heavy-ion irradiation experiments are carried out with tilted incidence, and the variability in SEU response between samples is studied. These samples used in experiment were composed of four chains of 1000 D-flip-flops (DFF) which are one of the most significant cells for digit ICs. A large number of DFFs in digital ICs can directly affect their SEU robustness. According to the heavy ion SEE test outline, the highest fluence is only $1\times10^{7}$ ions/cm$^{2}$. In other words, there is probably just 1 ion per 10 μm$^{2}$, which means that one DFF can be struck by merely less than one ion averagely. Hence, to guarantee the measurement accuracy, the test samples have to contain a great number of DFFs. As shown in Fig. 1, there are numerous DFF series connected together in a line named DFF chain. The SOI technology is an efficient hardening method against SEU and a dual interlocked storage cell (DICE) is one of the most commonly used redundancy designing methods.[13] In reality, these two hardening methods are frequently used together or separately in radiation hardening ICs. To evaluate the influence of the tilted angle to LET$_{\rm eff}$ in the SEE test for radiation hardening ICs, samples were designed using both the hardening methods together and separately. The sample named as DICE_SOI means using DICE and SOI hardening methods together, DICE_Si means using the DICE hardening method singly, NRH_SOI means using the SOI hardening method singly where NRH represents non-radiation hardening, and NRH_Si represents the commercial device without any hardening method. The I/O voltage of all samples is 3.3 V and their core voltage is 1.2 V. The SOI wafer used in this study is a kind of partially depleted SOI (PDSOI) with a 100-nm-thick top silicon and a 145-nm-thick buried oxide (BOX), which has been applied widely. To minimize their floating body effect (FBE) and to improve their RH performance,[14] body contact was used in all SOI transistors. The samples were irradiated by heavy ions at the HI-13 tandem accelerator in China's Institute of Atomic Energy, Beijing and the Heavy Ion Research Facility in Lanzhou (HIRFL) in the Institute of Modern Physics, Chinese Academy of Sciences. As shown in Fig. 1, samples were measured dynamically during irradiation. The data 0101 was written in and then read out continuously during irradiation at a frequency of 20 MHz. If an upset occurred, the data read would not be 0101. The upset cells were recorded and upsets were counted. The test would be stopped if there were no less than 100 upsets detected or the fluence arrived at $1\times10^{7}$ ions/cm$^{2}$, which means that the fluence is $1\times10^{7}$ ions/cm$^{2}$ when there was less than 100 upsets occurred. The particle type and corresponding tilted angle are summarized in Table 1. The figure of experimental setup to demonstrate the configuration of incident angle and sensitive volume is shown in Fig. 2.
cpl-34-11-118504-fig1.png
Fig. 1. The structure diagram of the DFF chain and its measurement process. The data 0101 was written in and then read out continuously during irradiation. If an upset occurred, the data read would not be 0101.
Table 1. Particle type, energy, penetration range (PR), tilted angle, effective LET (LET$_{\rm eff}$) and samples.
Ions Energy PR Tilted angle LET$_{\rm eff}$ Vehicles
(MeV) (μm) (MeV$\cdot$ cm$^{2}$/mg)
Kr 909 114 60$^{\circ}$ 77.1 NRH_SOI, DICE_SOI
30$^{\circ}$ 42.2 NRH_SOI, DICE_SOI
698 84 0$^{\circ}$ 39.8 NRH_SOI, DICE_SOI, NRH_Si, DICE_Si
1175 154 60$^{\circ}$ 75.8 NRH_Si, DICE_Si
30$^{\circ}$ 37.2 NRH_Si, DICE_Si
Bi 1500 78 0$^{\circ}$ 98.3 NRH_SOI, DICE_SOI
C 90 154 0$^{\circ}$ 1.7 NRH_Si, DICE_Si
F 125 98 0$^{\circ}$ 4.1 NRH_Si, DICE_Si, NRH_SOI
Si 110 39 0$^{\circ}$ 12.0 NRH_Si, DICE_Si, NRH_SOI
cpl-34-11-118504-fig2.png
Fig. 2. The figure of experimental setup to demonstrate the configuration of incident angle and sensitive volume. (a) The relationship between incident angle and volume. (b) The schematic diagram of experimental setup. The aluminum is used to adjust beam energy, and the angle $\theta$ is set by rotating the clamp which fixes the sample.
The LET of every ion in the sensitive area of the test sample is calculated by the stopping and range of ions in matter (SRIM) software[15] according to its types, energy, space distance from the surface of the sample and the thickness of each material constituting the routing layer of the sample. The effective LET for the tilted beam is calculated by $$\begin{align} {\rm LET}_{\rm eff} =\frac{{\rm LET}_0 }{\cos\theta},~~ \tag {1} \end{align} $$ where LET$_{\rm eff}$ is the effective LET for an ion incident at the angle of $\theta$ from the normal incidence, and LET$_{0}$ is the particle's normal incidence LET. The cross section is calculated based on the upsets measured by $$\begin{align} \sigma =\frac{N_{\rm u}\times \cos\theta }{U\times N_{\rm b}},~~ \tag {2} \end{align} $$ where $N_{\rm u}$ is the upset number induced by heavy ions, $U$ is the fluence of heavy ions, $N_{\rm b}$ is the cell number, and $\theta$ is the angle of incidence. Figure 3 shows the SEU cross section versus heavy ions LET for NRH_Si, DICE_Si and NRH_SOI. The dashed lines are the corresponding fitting curves with the Weibull function. The Weibull function has become the de facto standard for describing SEE cross section curves.[16] The four parameters provide a versatile least-square fit to the data (if the fit converges), and the resulting fitting parameters are commonly used for the integral rectangular parallel piped (IRPP) calculation of space upset rates.[17] To acquire fit convergence, there have to be at least five LET$_{\rm eff}$ values tested per one sample and the values should distribute appropriately so that it can converge the entire range from threshold to saturation. The error bars here refer to the 95% confidence level.[18] Because there was no error occurring in the samples of DICE_SOI, its test result is not shown here. To study the influence of tilted angle to the LET$_{\rm eff}$ based on Eq. (1), samples were tested by tilted angles of 30$^{\circ}$ and 60$^{\circ}$ in addition to vertical incidence when Kr was chosen as the test heavy ion. The LET here refers to LET$_{\rm eff}$ for the titled beam.
cpl-34-11-118504-fig3.png
Fig. 3. The SEU cross sections for NRH_Si, NRH_SOI and DICE_Si versus heavy ions LET$_{\rm eff}$. (a) The solid triangles present experimental data for NRH_Si. (b) The solid squares present experimental data for NRH_SOI. (c) The solid circles stand for the experimental data for DICE_Si. Here dashed lines are the corresponding fitting curves with the Weibull function.
In Fig. 3(a), it could be observed that the expected $\sigma$ to the LET$_{\rm eff}$ of 37.2 Mev$\cdot$ cm$^{2}$/mg at 30$^{\circ}$ tilted angle tends to be smaller than its Weibull function fit value. This means that the actual LET$_{\rm eff}$ for Kr at the 30$^{\circ}$ tilted angle tends to be smaller than its theoretical value based on Eq. (1) in 130 nm bulk Si technology. This phenomenon can also be found in Fig. 3(b) but much more slightly. The expected $\sigma$ to the LET$_{\rm eff}$ of 42.2 Mev$\cdot$ cm$^{2}$/mg at the 30$^{\circ}$ tilted angle is only slightly smaller than its Weibull function fit value. It looks like Eq. (1) will fail to work well in bulk Si if taking it seriously. This is because Eq. (1) is based on the slab-shape RPP model, which supposes that the depth-to-width aspect ratio of the device sensitive volume is small enough. However, this hypothesis is not applicable for the 130 nm CMOS bulk Si technology and the box-RPP model should be taken into account. According to the box-RPP model, the depth size of sensitive volume should not be ignored and $\sigma$ of the tilted beam is smaller than that of normal incidence. Hence, the actual LET$_{\rm eff}$ for a tilted beam is smaller than its theoretical value based on Eq. (1) in the 130 nm bulk Si technology. Because the BOX in SOI decreases the depth size of its sensitive volume obviously, Eq. (1) can still perform well in SOI (100/145 nm). However, it seems that LET$_{\rm eff}$ of the tilted beam still obeys an inverse-cosine law at the 60$^{\circ}$ tilted angle here. The possible explanation is that its real LET$_{\rm eff}$ is so large that the corresponding SEU cross section reaches saturation. Thus the phenomenon could not be observed at the 60$^{\circ}$ tilted angle from the test result here. As shown in Fig. 3(b), it could be observed that $\sigma$ to LET$_{\rm eff}$ of 77.1 Mev$\cdot$ cm$^{2}$/mg at the 60$^{\circ}$ tilted angle is almost similar to the LET of 98.3 Mev$\cdot$ cm$^{2}$/mg which was realized by the high energy heavy ion of Bi at normal incidence. From the entire fit line, it was found that $\sigma$ to LET of 98.3 Mev$\cdot$ cm$^{2}$/mg has reached saturation. Thus $\sigma$ to LET$_{\rm eff}$ of 77.1 Mev$\cdot$ cm$^{2}$/mg at the 60$^{\circ}$ tilted angle reaches saturation, too. As shown in Fig. 3(c), it is found that Eq. (1) seems to perform very well in the DICE structure. The DICE structure makes it harder to upset by the redundancy designing methods that two coupling sensitive nodes must hit simultaneously and both the charges will be received up to their threshold value. However, the tilted beam makes it easier to hit multiple sensitive nodes simultaneously and leads to a higher SEU cross section. It might be explained that this characteristic effectively eliminates the above phenomenon. Hence, $\sigma$ to LET$_{\rm eff}$ of 37.2 Mev$\cdot$ cm$^{2}$/mg at the 30$^{\circ}$ tilted angle is near its Weibull function fit value in the test result of DICE_Si. However, it is hard to obtain accurately whether it still obeys an inverse-cosine law well in the DICE structure by the test result here. It needs more data to verify. In summary, taking a DFF cell as the test object, special samples are designed based on the 130 nm technology platform and their heavy ion irradiation experiments with tilted angles are completed in this study. It can be concluded that the actual LET$_{\rm eff}$ for the tilted beam tends to be smaller than its theoretical value based on Eq. (1) in the 130 nm bulk Si technology. The explanation is that the slab-shape RPP model is not applicable well for the 130 nm bulk Si technology because the depth size of sensitive volume cannot be ignored compared with its width size, and the box-RPP model is more likely to be appropriate for it. The effective LET for SOI is more closely in inverse proportion to $\cos \theta$ compared to bulk Si (100/145 nm). Equation (1) still performs well in the 130 nm SOI (100/145 nm) technology due to the fact that the BOX in SOI decreases the depth size of sensitive volume obviously. Hence, the tilted beam should be used with certain attention for samples in 130 nm or smaller bulk Si technology. It looks like that LET$_{\rm eff}$ of the tilted beam still obeys an inverse-cosine law very well in the DICE structure by the test result here, but it needs more data to be verified. References Ion Photon Emission Microscope for Single Event Effect Testing in CIAESingle-event effects ground testing and on-orbit rate prediction methods: the past, present, and futureGeometrical factors in SEE rate calculationsRecent trends in single-event effect ground testingCharge collection and SEU from angled ion strikesIn-flight observations of multiple-bit upset in DRAMsSuggested Single Event Upset Figure of MeritImpact of Heavy Ion Energy and Nuclear Interactions on Single-Event Upset and Latchup in Integrated CircuitsImpact of Ion Energy and Species on Single Event Effects AnalysisExperimental study on heavy ion single event effects in SOI SRAMsRadiation effects in SOI technologiesProduction and propagation of single-event transients in high-speed digital logic ICsUpset hardened memory design for submicron CMOS technologyTotal-Ionizing-Dose-Induced Body Current Lowering in the 130 nm PDSOI I/O NMOSFETsRate prediction for single event effects-a critiqueParametric and Threshold Studies of Single Event Sensitivity
[1] Zhang Y W, Guo G, Liu J C et al 2017 Chin. Phys. Lett. 34 073401
[2] Reed R A, Kinnison J, Pickel J C et al 2003 IEEE Trans. Nucl. Sci. 50 622
[3] Petersen E L, Pickel J C, Smith E C et al 1993 IEEE Trans. Nucl. Sci. 40 1888
[4] Duzellier S and Ecoffet R 1996 IEEE Trans. Nucl. Sci. 43 671
[5] Dodd P E, Shaneyfelt M R and Sexton F W 1997 IEEE Trans. Nucl. Sci. 44 2256
[6] Swift G M and Guertin S M 2000 IEEE Trans. Nucl. Sci. 47 2386
[7] Petersen E L, Langworthy J B and Diehl S E 1983 IEEE Trans. Nucl. Sci. 30 4533
[8] Dodd P E, Schwank J R, Shaneyfelt M R et al 2007 IEEE Trans. Nucl. Sci. 54 2303
[9] Reed R A, Weller R A, Mendenhall M H et al 2007 IEEE Trans. Nucl. Sci. 54 2312
[10] Li Y, He C, Zhao F et al 2009 Nucl. Instrum. Methods Phys. Res. Sect. B 267 83
[11] Schwank J R, Ferlet-Cavrois V, Shaneyfelt M R et al 2003 IEEE Trans. Nucl. Sci. 50 522
[12] Dodd P E, Shaneyfelt M R, Felix J A et al 2004 IEEE Trans. Nucl. Sci. 51 3278
[13] Calin T, Nicolaidis M and Velazco R 1996 IEEE Trans. Nucl. Sci. 43 2874
[14] Liu X N, Dai L H, Ning B X et al 2017 Chin. Phys. Lett. 34 016103
[15][Online]. Available: http://www.srim.org
[16] Petersen E L, Pickel J C, Adams J H et al 1992 IEEE Trans. Nucl. Sci. 39 1577
[17] Petersen E L 2007 IEEE Trans. Nucl. Sci. 54 1392
[18]Hang G M and Zhu G Q 2014 J. Math. Med. 27 48