1. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 1 Analysis of Mechanical Stress Effects in Short Channel MOSFETs C. Gallon 1, G. Reimbold 1, G. Ghibaudo 2, R.A. Bianchi 3 and R. Gwoziecki 1,3. 1 CEA-Leti, 38054 Grenoble Cedex 9, France. Tel. : +33 (0)4 38 78 49 93. E-mail: cgallon@cea.fr 2 IMEP, BP257, 38016 Grenoble Cedex 9, France. 3 STMicroelectronics, Central R&D, 38921 Crolles, France.

2. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 2 OUTLINE Introduction Experimental Method – Four point bending technique – Tested devices Experimental Results: –Stress Influence on Long Channel Devices –Stress Influence on Short Channel Devices Conclusion and Perspectives

3. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 3 INTRODUCTION (1) Context: –Generation of mechanical stress at various process steps. –These effects are more important in scaled CMOS devices –A key role of mechanical stress in MOSFETs devices:  Performance improvements (SiGe, SiGe:C… )  or Performance reductions (STI,…) Needs: –Improve extraction methodologies versus mechanical stress. –Better understanding and evaluation of stress effects on MOSFETs devices.

4. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 4 INTRODUCTION (2) Purpose of this work: –Analysis of external mechanical stress effects on MOSFETs from advanced 0.13µm CMOS technology:  Relative variations of mobility with external stress,  Extraction of Piezoresistive Response (PR),  Simple approach proposed to extract PR on short devices accounting for R sd influence. –Objectives: Try to provide data for device simulators and a better analysis of stress effects.

5. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 5 Principle of a four-point bending technique: Interest: Application of an uniform uniaxial stress between the two central fulcrums. a 2a a F F EXPERIMENTAL METHOD: 4-POINT BENDING

6. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 6 y dis Evaluation of stress : Estimated error: - Uncertainty on L, a, y measurements and mainly on E value. - Global accuracy: 7% a 2a a F F EXPERIMENTAL METHOD: 4-POINT BENDING

7. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 7 Micrometer screw EXPERIMENTAL METHOD: 4-POINT BENDING aa 2a L aa L compressive tensile

8. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 8 Specific characteristics: –Rectangular strip are cut from a saw technique. –With an appropriate preparation of the strips, mechanical stress:  longitudinal direction (// to current flow)  or transversal direction (  to current flow). –Most strips fail for 150-200MPa  Keep the applied mechanical stress below 100MPa. EXPERIMENTAL METHOD: 4-POINT BENDING

9. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 9 Bulk and SOI similar technologies - nMOS and pMOS fabricated on (100) substrates - T ox =2nm; W=10µm; Long (L=10µm) or short (L=0.13µm) channel length; - Important point on our short devices: a long distance between STI and gate  limit parasitic internal stress. - Mechanical stress ranging from 0 to 100MPa was applied. DEVICES TESTED ON 4-POINT BENDING Substrate Gate DrainSource Bulk Technology Buried oxide Substrate Gate DrainSource SOI technology

10. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 10 EXPERIMENTAL RESULTS: Long Channel Devices Effects of mechanical stress on transfer characteristics - Linear region characteristics of nMOS BULK: - Use of standard expressions to extract V T and µ. - Note the invariance of V T and mobility variations.

11. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 11 Normalized mobility variations versus applied stress: - Excellent linear dependence for both n and p MOS devices. EXPERIMENTAL RESULTS: Long Channel Devices

12. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 12 Mobility variations and piezoresistance response:  L longitudinal coefficient  T transverse coefficient Majors coefficients of cubic structure for silicon EXPERIMENTAL RESULTS: Long Channel Devices

13. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 13 Piezoresistance coefficients (.10 -12 Pa -1 ): BULKSOIErrors  L nMOS  T nMOS  S nMOS  44 nMO S 622 272 894 349 734 393 1127 341  40  60  L pMOS  T pMOS  S pMOS  44 pMO S -770 491 -280 -1261 -967 531 -435 -1499  40  60 Bulk&SOI results 0.13µm techno. EXPERIMENTAL RESULTS: Long Channel Devices Bradley&al. IEEE2001 0.3µm techno. Ref. Si

14. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 14 EXPERIMENTAL RESULTS: Long&Short Channel Devices 0 2 4 6 8 0255075100 Tensile Stress (MPa)  µ/µ (%) L=0.13µm L=10µm nMOS Longitudinal Stress D’ S’ D S R R G D’ S’ D S R R G Comparison between Long and Short devices: Bradley&al.: “Reduction is only due to the influence of R sd ". (IEEE 2001)

15. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 15 Bradley approach: Problems of this approach: - Extraction of R on is V g dependent - Choice in Vg extraction results in significant variation on R on  Significant uncertainty on piezoresistive coefficients. A novel approach is proposed. EXPERIMENTAL RESULTS: Short Channel Devices

16. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 16 New approach proposed: 1.Correction from R sd influence on Id 0 : 2.Calculation of equivalent Vg shift to get the same Id with and without stress: 3.  Vg is related to Vt and mobility change by: [Roux-dit-Buisson, IEEProceedings-G, 1993] EXPERIMENTAL RESULTS: Short Channel Devices

17. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 17 1. Experimental variations after various stress levels for a 0.13µm pMOS/SOI.  Note excellent linearity. 3. Piezoresistive coefficients extraction. EXPERIMENTAL RESULTS: Short Channel Devices Longitudinal Transversal 0 2 4 6 8 0255075100 0 2 4 6 8 0255075100 No Rsd Correction Rsd correction - 2 - 4 - 6 - 8 Tensile Stress (MPa)  µ/µ (%) 2. Normalized mobility change versus applied uniaxial stress.

18. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 18 Example of calculations including R sd corrections: -Agreement between  coefficients for both short and long L. -Local or 2D stress do not affect significantly short devices, however a slight longitudinal effect may exist. SOI pMOSFET parameters Measured values Corrected values Errors Channel length10µm0.13µm Parasitic Rsd 35  VdVd -0.1V  L  T  S  44 -967 531 -435 -1499 -545 473 71 -1018 -708 567 -141 -1275  100  150 EXPERIMENTAL RESULTS: Short Channel Devices

19. 21/03/2003 ULIS 2003 - Udine (Italy) - C. Gallon 19 CONCLUSIONS & PERSPECTIVES Study of mechanical stress effects on long and short channels. Proposition of a simple approach to determine directly  Vt and  µ –Vt is independent of stress, –Mobility variations dominate the piezoresistive response, –Bulk & SOI: similar piezoresistive response both n and p MOS, slightly higher for SOI. After R sd corrections, comparable results on short and long devices:  2D or local effects are small for a 0.13µm technology. A first step to provide piezoresistive data for device simulators and a better analysis of mechanical stress effects.