Noise and Random Telegraph Signals in Nanoelectronic Devices Zeynep Çelik-Butler Electrical Engineering Department University of Texas at Arlington Arlington, Texas, 76019 [email protected] Outline Motivation: Problems Encountered as the Devices Shrink, Frequencies Increase, and Voltages Reduce Improved Model for 1/f Noise in MOSFETs Random Telegraph Signals in MOSFETs Complex RTS Extraction of trapping parameters using RTS Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 2 UTA - Noise Characterization Facilities 6' x 6' x 8' Shielded Room 3 Spectrum and Signal Analyzers, f=1 mHz - 20 GHz. 3 Cryostats, T= 2 K to 350 K. Various Lock-ins, Preamps, System Controllers, Battery Operated Sources etc. Optical Equipment Computer Software for Modeling Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 3 Problems Encountered as the Devices Shrink, Frequencies Increase, and Voltages Reduce Signal-to-noise ratio decreases. Noise models based on large number of electrons break down. Quantum effects become dominant. Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 4 Signal to Noise Ratio Decreases For a MOSFET Start from W=100mm, L=10mm, tox=800Å, NSS=4x1010 eV1cm-2. Assume scaling factor is K. Assume trap and surface state densities remain the same. W W K , L L K , tox tox K in noise level due to the K1/2 law chosen for tox. Unpredictability of noise level for K>20. NSS is actually a two dimensional Poisson variable. Increase Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 5 Large Area Noise Models Break Down Single electron, single trap effects. NSS=4x1010 eV-1 cm-2, W=1mm, L=0.1mm. EC kT=26 meV EF EV Si 1 trap per channel SiO2 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 6 Large Area Noise Models Break Down Break-down of large-area models for sub-micron channel length. kTq 2 I d m eff NO N * 1 2 2 S Id ( f ) A ln B ( N N ) C N N O L O L 2 * 2 fL Cox NL N • • • • A=Nt (cm-3 eV-1) B=ameffNt (cm-1 eV-1) C=a2meff2Nt (cm eV-1) A=B2/(4C) Independent parameters: a and Nt Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 7 Large Area Noise Models Break Down -10 -11 10 Vgs-VT= -1 V Vds= -50 mV S Vd 2 (1Hz) (V /Hz) 10 -12 10 -13 10 -14 10 0.1 1 Channel length (mm) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 10 8 Large Area Noise Models Break Down 20 10 2 channel region model uniform channel model 19 -1 N (cm eV ) 10 18 -3 10 17 t 10 16 10 15 10 0.2 0.4 0.6 0.8 L (mm) 1 1.2 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 1.4 9 Large Area Noise Models Break Down Modified 1/f noise model that takes into account threshold variation along the channel. • For simplicity assume two regions: – DV, DL, VT2,, A2, B2, C2 – Vds-DV, L-DL, VT1, A1, B1, C1 – – – – – DL<<L, VTVT1 A1 = A2, since Nt1 = Nt2 B12/C1 = B22/C2 = 4A I1 = I2 = Id meff1 = meff2, Independent parameters: Nt, a1, a2, VT2, and DV Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 10 Large Area Noise Models Break Down Modified 1/f noise model that takes into account threshold variation along the channel. -17 10 -18 -19 10 Id 2 S (A /Hz) 10 L=0.32mm -20 10 L=0.45mm -21 10 L=1.0mm -22 10 0 0.5 1 1.5 2 |V -V | (V) gs T Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 11 RTS in MOSFETs Random Telegraph Signals: single electron switching. t1 RTS (Arbitrary Units) 0.0002 0.00015 0.0001 5 10 -5 10 DId -5 0 -5 -0.0001 5.2 5.3 5.4 5.5 Time (sec) 5.6 t0 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 5.7 12 RTS in MOSFETs Time Scale seconds PSD Random Telegraph Signals (RTS) with a Lorentzian on 1/f spectum. Frequency Time Scale milliseconds S( f ) (f) 4DI 2 t0 t1 1 t0 1 t1 2 2f 2 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 13 NMOS,W/L(mm)=5/0.23, VDS=175mV, VGS=0.60V 10-9 10-10 Sv (V2/Hz) 10-11 10-12 Sv = 6.11e-12 / ( 1 + f / 1260 ) 2 10-13 1 RTS process 10-14 10-15 100 101 102 103 104 105 Frequency (Hz) DV (10-4 V) 2 0 -2 -4 -6 2 RTS levels -8 -10 0 1 2 3 4 5 6 7 8 Time (ms) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 14 PMOS T1 W/L=5/0.25 VDS=150mV VGS=0.9V 10-9 2 Sv = 1.4e-10 / ( 1 + f / 1.8) 10-10 Sv (V2/Hz) 10-11 2 Sv = 2.3e-14 / ( 1 + f / 23700) 10-12 10-13 2 RTS processes 10-14 10-15 10-1 100 DV (10-4 V) 2 101 102 103 104 105 Frequency (Hz) 1 0 3 RTS levels -1 0 1 2 3 Time (ms) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 15 NMOS ,W/L(mm)=5/0.23 ,VDS=150mV,VGS=0.775V 10-8 Sv = 1.11e-10 / ( 1 + f / 3.12 ) 2 10-9 Sv = 1.18e-11 / ( 1 + f / 48) 2 Sv = 1.08e-12 / ( 1 + f / 1345) 10-11 2 2 Sv (V /Hz) 10-10 4 RTS processes 10-12 Sv = 2.93e-14 / ( 1 + f / 36 780) 2 10-13 10-14 10-15 10-1 100 101 102 103 104 105 3 Frequency (Hz) DV (10-4 V) 2 1 0 -1 5 RTS levels -2 -3 0 1 2 3 4 5 6 7 8 Time (ms) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 16 COMPLEX RTS 9.E-03 (a) Voltage (V) 7.E-03 level 4 5.E-03 level 3 3.E-03 level 2 1.E-03 level 1 -1.E-03 2 2.02 2.04 2.06 2.08 2.1 Time (s) Complex random telegraph signals due to multiple traps SI ( f ) I 2 N traps k 1 DI I 2k t0 t1 k 1 t0 1 t1 2k 2f 2 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 17 RTS in MOSFETs RTS can be used to characterize trapping sites. RTS modeling. EC S( f ) ECox-ET EFp xT 4DI 2 t0 t1 1 t0 1 t1 2 2f 2 qVc EFn qs EFg gate qVgs S( f ) oxide AF I d2 2f 2 KF 2 silicon Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 18 RTS in MOSFETs RTS can be used to characterize trapping sites. • • • • Position of the trap along the channel, yT Position of the trap in the oxide, xT Trap energy, ECox - ET Screened scattering coefficient, a Vc y Vds L DI d DN Dm 1 1 am Id N m Weff Leff N t xT 1 ln c E E E E qV q q V V Cox T C Fp c 0 s gs FB s te kT Tox Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 19 tc / t e Trapping Parameters Through RTS in MOSFETs 10 1 10 0 yT/L=0.6 ECox-ET=3.04 eV (b) 0.01 0.1 Drain Voltage (V) xT=2.7 nm 10 2 10 1 10 0 1 t /t c e 10 -1 Forward Reverse (a) 10 -1 0.25 0.3 0.35 0.4 Gate Voltage (V) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 0.45 0.5 20 Trapping Parameters Through RTS in MOSFETs 10 -1 10 -2 10 -3 10 -4 ds DV / V ds D N/N 10 -1 0.04 Forward Reverse 0.06 0.08 0.1 V -V 10 -2 10 -3 T (V) ds DV / V ds gs 0.3 0.01 0.1 Drain Voltage (V) 1 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 21 Scattering Coefficient (V-s) Trapping Parameters Through RTS in MOSFETs 6 10 -14 5 10 -14 4 10 -14 3 10 -14 2 10 -14 1 10 -14 Forward Reverse 0.04 0.06 0.08 0.1 V -V (V) gs 0.2 T a K1 K 2 ln N Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 22 Effects of Quantization •Increase in effective energy band-gap: change in te and tc • Shift in carrier distribution: change in Cox Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 23 3-D Treatment of RTS 1 1 tc cn n3D n (3D) Vth n3D exp ( E F ET ) / k BT te n (3D) Vth n3D cn n 3D Vth Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 24 2-D Treatment of RTS - tc and te tc 1 cn n2 D 0z p( z ) dz z 1 n (2 D) Vth n2 D 0z p( z ) dz z 1 exp ( EF ET ) / k BT te en (2 D) V n z p( z ) dz n th 2 D 0 z cn n 2D Vth Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 25 2-D Treatment of RTS • From Stern - Howard wave-function: b3 2 p( z ) z exp bz 2 1/ 3 12qml 11 b 2 QB Qinv 32 Si 0 z 3/ b Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 26 2-D Treatment of RTS • Calculate the inversion carrier concentration assuming they are located primarily at E0: 1 1 N n2 D p( z )dz 2k BTmt z exp ECS DE0 EF / k BT 0 p( z )dz 2 1/ 3 2 DE0 2m l 9q 8 Si 0 2/3 1 2 Si 0qN B (VSB 2F )1/ 3 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 27 2-D Treatment of RTS - tc and te te tc exp ( ECS ET DE0 ) / k BT n (2D) Vth (2k BTmt b / 5 2) exp ( ECS EF DE0 ) / k BT n (2D) Vth (2k BTmt b / 5 2) tc 1 zT ln E E E E q q V V T CB F 0 s gs FB s Cox t k T T B ox e • To first order, the ratio is not affected by quantization. Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 28 RTS Measurements • • • • • • MDD n-MOSFETs Weff Leff = 1.37 0.17 mm2 Tox = 4 nm VT = 0.375 V for VSB = 0 V strong inversion, linear region VDS = 100 mV VSB = 0 - 0.4 V, VGS = 0.5 - 0.75 V Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 29 ECox-ET and zT from tc and te tc 1 zT ln Vgs VFB s ECox ET ECB EF 0 q s q t k BT Tox e 3.5 VSB=0 V 3 t ln( t / ) c e ln(t c/te) 2.5 2 1.5 1 0.5 0 0.45 0.5 0.55 0.6 0.65 V (V) 0.7 0.75 0.8 GS Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 30 ECox-ET and zT from tc and te tc 1 zT ln Vgs VFB s ECox ET ECB EF 0 q s q t k BT Tox e 3.5 VSB=0.4 V 3 t t/ ) c )e ln(tln(c/t e 2.5 2 1.5 1 0.5 0 0.45 0.5 0.55 0.65 0.6 0.7 0.75 0.8 V (V) GS Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 31 ECox-ET and zT from tc and te Tox =4 nm VSB (V) VT (V) zT (Å) ECox-ET (eV) 0 0.375 11.22 3.09 0.1 0.382 11.53 3.08 0.2 0.393 11.37 3.08 0.3 0.401 11.64 3.07 0.4 0.408 11.08 3.08 Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 32 Dependence of te on VSB te te(s) te (s) 2 10 exp ( ECS ET DE0 ) / k BT n (2D) Vth (2k BTmt b / 5 2) -3 10 -3 8 10 -4 6 10 -4 VGS=0.75 V VGS=0.55 V 4 10 -4 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 V (V) SB Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 33 Dependence of tc on VSB tc 10 exp ( ECS EF DE0 ) / k BT n (2D) Vth (2k BTmt b / 5 2) -2 VGS=0.55 V (s) tc t(s) c VGS=0.65 V VGS=0.75 V 10 -3 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 V (V) SB Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 34 cn Extracted from tc and te Vth 8k BT / mn n 0 exp DEB k BT -11 cn n 2D Vth n 3 capture coefficient c (cm /s) 10 * 1/ 2 10 -12 10 -13 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 V (V) GS Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 35 2-D Treatment of RTS - Amplitude 1 DN 1 m DI D 1 1 DNt am ID Weff Leff N DN DNt m DNt m 1 1 1 m n mt 1 m n aNt • Question: How does quantization affect number and mobility fluctuations? – Number fluctuation through N – Mobility fluctuations through oxide charge scattering, mt. Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 36 Extraction of Scattering Coefficient • Mobility Fluctuations: – Using Surya’s 2D surface mobility fluctuations model, 2 exp( 4 kz sin ) sin / 2 1 mt dNt E , z dz dE 0 2 c 2 8av E p (sin ) 2k mn*q 3 k 0.82 aSi 2 N 2q 2 d v mn* c 1 exp 2 * 4 si k BTd v mn Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 37 Calculation of Scattering Coefficient • Considering a single trap: Nt(E,z) = Nt(E-ET) (z-zT) 2 sin / 2 1 mt exp( 4kzT sin )dNt 0 2 c 2 8av E p (sin ) 2k mn*q 3 a Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 38 RTS Amplitude -4 DS D V /V DS 3 10 -4 2 10 V =0V SB V =0.1V SB V =0.2V SB V =0.3V SB V =0.4V SB -4 1 10 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 V (V) GS Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 39 Extraction of Scattering Coefficient 2.4 10 -14 Tox =4 nm 2 10 -14 experimental data @V =0V SB fitting with z =0.11nm t a (V-s) fitting with z =0.12nm t 1.6 10 -14 1.2 10 -14 8 10 -15 6 10 a = 2.91x10-13 - 9.93x10-15 ln(N) -11 10 12 -2 N(cm ) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 10 40 13 Extraction of Scattering Coefficient -14 Scattering Coefficient (V-s) 1 10 Tox =8.6 nm -15 8 10 W L = 1.2 0.35 mm2 -15 6 10 -15 4 10 experimental results of Hung et al. 8 fitting of Pacelli et al.11 -15 2 10 theoretical calculation from 2-D mobility fluctuation model 11 10 zT =0.25 nm 12 10 -2 N(cm ) Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 13 10 41 Possible Reasons for Discrepancy • Threshold non-uniformity along the channel is not taken into account. • Location of the trap along the channel • Variation of the channel voltage from source to drain is neglected. • DN/DNt 1 is not valid, even in strong inversion, for very thin oxides. Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 42 ACKNOWLEDGEMENTS • This work has been supported by NSF, THECB-ATP, SRC, TI, Legerity, Motorola and ST-Microelectronics. Noise and Reliability Laboratories, Zeynep Celik-Butler, [email protected] 43
© Copyright 2025 Paperzz