Neutron-Induced Soft Error Analysis in MOSFETs from a 65nm to a 25 nm Design Rule using Multi-Scale Monte Carlo Simulation Method

Transcription

1 Neutron-Induced Soft Error Analysis in MOSFETs from a 65nm to a 25 nm Design Rule using Multi-Scale Monte Carlo Simulation Method Shin-ichiro Abe, Yukinobu Watanabe, Nozomi Shibano 2, Nobuyuki Sano 2, Hiroshi Furuta 3, Masafumi Tsutsui 3, Taiki Uemura 3, and Takahiko Arakawa 3 ( abe@aees.kyushu-u.ac.jp). Departments of Advanced Energy Engineering Science, Kyushu University, Kasuga, Fukuoka, , Japan 2. Institute of Applied Physics, University of Tsukuba, Ibaraki , Japan 3. Semiconductor Technology Academic Research Center, STARC, Kanagawa , Japan

2 Purpose To perform SER analyses and compare with the scaling trend To investigate the impact of secondary ions on soft errors To clarify the incident energy range in which terrestrial neutrons have significant influence on soft errors Abstract We have analyzed terrestrial neutron-induced soft errors in MOSFETs from a 65 nm to a 25 nm design rule by means of multi-scale Monte Carlo simulation using PHITS-HyENEXSS code system. The resulting scaling trend of SERs per bit is still decreasing similar to other predictions. From this analysis, it is clarified that secondary He and H ions provide a major impact on soft errors with decreasing critical charge. It is also found that terrestrial neutrons with energies up to several hundreds of MeV have a significant contribution to soft errors regardless of design rule and critical charge. 2/2

3 Background Soft Error: Temporary malfunction of microelectronic devices. They are mainly due to terrestrial cosmic-ray neutrons on the ground. Neutron MOSFET Gate Drain Source Scale [m] -6-9 (2) Charge Deposition Soft Errors -2-5 () Nuclear Reaction (3) Charge Collection Time [sec] To simulate the soft error rate with high reliability, it is required to employ highly reliable physical models. to synthesize each physical model at the difference stages. 3/2

4 Our previous work Ref.) S.Abe et.al., IEEE Trans. on Nucl. Sci., vol.59, no.4, pp , 22. We have developed a multi-scale Monte Carlo simulation code system by linking PHITS and HyENEXSS. This code system enables integrated simulation from nuclear reaction to charge collection PHITS Interface HyENEXSS () Nuclear Reaction (2) Charge Deposition (3) Charge Collection 4/2

5 PHITS (Particle and Heavy Ion Transport code System) Ref.) K. Niita, et.al., PHITS: Particle and Heavy Ion Transport code System, Version 2.23, JAEA- Data/code D Mote Carlo simulation code to calculate Nuclear Reaction and Transport of particles and heavy ions Nuclear reaction model contained in the PHITS code Dynamic Process Evaporation Process Intra Nuclear Cascade (INC) model Quantum Molecular Dynamics (QMD) mode Statistical Decay Model (SDM) Generalized Evaporation Model (GEM) etc etc Neutron-induced reaction below 2 MeV event generator mode (e-mode) with evaluate nuclear data library 5/2

6 Information: Accuracy of nuclear reaction model Ref. of Comparisons) S. Abe et.al., Proc. of Int. Nucl. Phys. Conf. 2. etc Ref. of MQMD) Y. Watanabe and D. N. Kadrev, EDP Science, pp. 2-24, 28. Cross Section [mb] d /de [mb/mev] CS of Si(n,x )X Exp. INC+GEM QMD+GEM MQMD+GEM e-mode (JENDL-3.3) Incident Energy [MeV] DX of Si(n,x )X, E in = 96MeV Exp. INC+GEM QMD+GEM MQMD+GEM Alpha Particle Energy [MeV] d 2 /de/d [mb/mev/sr] DDX of Si(n,x )X E in = 96MeV Exp. INC+GEM QMD+GEM MQMD+GEM 2deg 4deg (x -2 ) 6deg (x -4 ) 8deg (x -6 ) deg (x -8 ) Alpha Particle Energy [MeV] 6/2

7 Information: Accuracy of stopping power calc. Ref.) S. Abe, Thesis for master s degree, Department of Advanced Energy Engineering Sciences, Kyushu University, Japan, 2. (in Japanese) Stopping Power [kev/( g/cm 2 )].5.5 H into Si SH.Z.Izmailov+ (98) E.I.Sirotinin+ (984) P.Mertens+ (988) A.Ikeda+ (996) SRIM-28 PHITS (ATIMA) PHITS (SPAR) 4 Incident Ion Energy [kev] Stopping Power [kev/( g/cm 2 )] He into Si W.K.Lin (973) D.C.Santry (98) J.D.Pearce (98) D.C.Santry (984) SRIM-28 PHITS (ATIMA) PHITS (SPAR) 4 Incident Energy [kev] 7/2

8 Interface tool takomesh Ref.) N. Shibano, Thesis for master s degree, Institute of Applied Physics, University of Tsukuba, Japan, 2. (in Japanese) Secondary ions are emitted at arbitrary directions. takomesh generate the mesh structure optimized for soft error simulation. Octree mesh method Subdivide rectangular parallelepiped shaped mesh into 8 small meshes recursively. e.g.) Spallation reaction of 28 6 n Si O 2 2p 3n The initial charge distribution is generated normally even though many secondary ions were passed. 8/2

9 HyENEXSS (Hyper Environment etc ) Ref.) 3-D TCAD simulator HyENEXSS, developed by Selete: Semiconductor Leading Edge Technologies Inc.; 3-D TCAD simulator that consisting of Process Simulator and Device Simulator Transient analysis is performed by Drift-Diffusion method. 2 32nm Tech. NMOS I drain 25. drift diffusion Q drain 2. I drain [ma] Qdrain [fc] time [sec]. 9/2

10 Simplified flowchart of SER analysis Incident particles Device structure Distribution of impurity PHITS LET distributions (each event) dump file PHITS dump file (after filtering) dump files (each event) Mesh structure Input file for takomesh PHITS takomesh Interface Interface HyENEXSS Input file for HyENEXSS HyENEXSS Results of analysis /2

11 How to calculate SER F A SER ( Q c ) N( q) dq N N Q c in bit -9 F : total neutron flux [cm -2 s - ] A : surface area of device [cm 2 ] N in : number of incident neutrons in PHITS calculation N bit : number of bit cells 25nm tech. NMOSFET 25nm tech. NMOSFET SER [FIT/Mbit] 4 3 Count [/neutron/cell/fc] Critical Charge [fc] Collected Charge [fc] /2

12 Structure and parameter of NMOSFETs z [μm].3.2 Si SiO 2 poly Si x [μm] TABLE : PARAMETER OF NMOSFET FOR EACH TECHNOLOGY Technology [nm] Critical Charge [fc] Bias voltage of drain [V] Gate oxide thickness [nm] Active area length [μm] Active area width [μm] STI depth [μm] /2

13 Test device structure to make dump file Cosmic-ray Neutrons μm Substrate (Si) Analysis Volume Detection Volume Metal Layer (Cu & SiO 2 ) Package Layer (SiO 2 ) Insulation Layer (SiO 2 ) μm 5 μm 3. μm.35 μm 5 μm for 25nm Tech. NMOS 2 μm STI:.25μm. μm μm 3 μm.5 μm The event was detected when secondary ions pass through Analysis Volume, or generated in the Analysis Volume. The number of incident neutrons: 4,,, The number of detected events: 3,8, It is necessary to select the analyzed events. 3/2

14 How to choose event = Condition of Event Selection = If the carge deposited in the analysis volume is less than critical charge, the event is rejected. 2 If any secondary ions do not pass through the detection volume, the event is also rejected. Analysis Volume 2 3. μm STI 2 μm μm.25 μm.5 μm Detection Volume 4/2

15 Results of SER analyses SER (FIT/Mbit) nm Tech. NMOSFET 45nm Tech. NMOSFET 32nm Tech. NMOSFET 25nm Tech. NMOSFET Critical Charge (fc) The calculated SERs per bit have a decreasing trend as the device technology scales down. 5/2

16 2 Scaling trend of SER Normalized SER.5 Normalized PHITS-HyENEXSS.5 P2 E.Ibe+ 2 (Sim.) P3 H.Kobayashi+ 29 (Sim.) P4 P.Shivakumar+ 22 (Sim.) P5 D.Lambert+ 24 (Sim.) P6 N.Seifert+ 28 (Exp.) Technology node (nm) The present result shows a decreasing trend similar to other SER predictions. 6/2

17 Fraction Fraction Impact of secondary ion species on SER Fraction Fraction H ion He ion other ions Critical Charge [fc] Critical Charge [fc] (a) 65nm Tech. NMOSFET (Q c = 2.fC) (b) 45nm Tech. NMOSFET (Q c =.5fC) Critical Charge [fc] Critical Charge [fc] (c) 32nm Tech. NMOSFET (Q c =.95fC) (d) 25nm Tech. NMOSFET (Q c =.6fC) He ions have a major cause of soft errors. H ions impact SER significantly when Q c is reduced. 7/2

18 Information: Cosmic-ray neutron spectrum Ref.) JEDEC Standard JESD89A, Oct. 26.; Neutrons [/MeV/cm 2 /hour] Secondary Cosmic-ray Neutron in NewYork Energy Range fraction ~MeV.355 ~5MeV.4 5~4MeV.49 4~8MeV.47 8MeV~GeV.7-4 Neutron Energy [MeV] 8/2

19 Fraction Fraction Dependence of SER on incident neutron energy Fraction Fraction ~MeV ~5MeV 5~4MeV 4~8MeV Critical Charge [fc] Critical Charge [fc] (a) 65nm Tech. NMOSFET (Q c = 2.fC) (b) 45nm Tech. NMOSFET (Q c =.5fC) 8MeV~GeV Critical Charge [fc] Critical Charge [fc] (c) 32nm Tech. NMOSFET (Q c =.95fC) (d) 25nm Tech. NMOSFET (Q c =.6fC) The contribution of neutrons below MeV is a few percent. Incident neutrons up to 4MeV account for about 8%. 9/2

20 Conclusion The results show that the scaling trend of SERs per bit is still decreasing similar to the other predictions. He ions have a major cause of soft errors, and H ions impact soft errors considerably if the critical charge is further reduced. About 8% of terrestrial neutron-induced soft errors can be caused by neutrons up to 4 MeV. Accurate estimation of production of H and He ions in the neutron energy range up to several hundreds of MeV plays a key role in terrestrial neutron-induced soft error simulation. Acknowledgement This works was financially supported by STARC Joint Research Program No. 89 and partially by the Sasakawa Scientific Research Grant from the Japan Science Society. 2/2