1. MEIS Mtg University of Huddersfield, 8 Dec 2011 MEIS studies in the EU ANNA I3 project Jaap van den Berg International Institute for Accelerator Applications, University of Huddersfield Michael Reading Centre for Materials and Physics, University of Salford Acknowledgement: EU FP6 I3 project “ANNA” (contract no. 026134 RII3) Paul Bailey, Tim Noakes, Daresbury MEIS facility

2. MEIS mtg U of Huddersfield, 08/12/2011 Outline ANNA I3 project MEIS experimental aspects MEIS quantification MEIS analysis of: –Ultra thin STO/TiN high-k MIM cap nanolayer structures: layer thickness & composition the effect of processing steps, e.g. segregation, layer interdiffusion –Post annealing of shallow Sb implants into Si following SPER Sb precipitation and pile-up under the oxide

3. MEIS mtg U of Huddersfield, 08/12/2011 ANNA I3 project Analytical Network for Nanotechnology FP 6 project completed in February 2011 Details: www.i3-anna.net Networking - formation of a Joint (distributed) Analytical Laboratory with individual ISO 9000 certification Joint Research Activities – 6 themes, e.g. nanolayer characterisation Transnational Access EU funded user access (Uni’s, Research institutes, SME’s) to analytical facilities not available in home country, e.g. MEIS Integrated Infrastructure Initiative - 3 strands:

4. MEIS mtg U of Huddersfield, 08/12/2011 MEIS Experimental aspects 100 – 200 keV He+ ions incident along the [-1-11] channel scattered ions detected along the [111] & [211] blocking directions. ( i.e. 70.5º and 90º scattering angle) Double alignment conditions to minimize the dechanneling background Sub-nm depth resolution (near-surface) Energy spectra to depth profile conversion: Elastic E loss yields the mass of scattering atom Inelastic E loss yields the depth of scattering event Quantification of yield & depth (energy spectrum simulation) He + ions: - higher dE/dz, better depth resolution - higher scattering cross section (move target vertically during analysis) - any assymmetry in inel. energy loss function reduced compared to H + [211] Si (001) surface [001] Detector [111][332]

5. MEIS mtg U of Huddersfield, 08/12/2011 MEIS - quantitative? Stopping powers: from SRIM 2003 onwards - consistent results for 50 - 200 keV ions on SiO 2 layers of different thicknesses - checked against other techniques Ion Yields affected by Scattering X-section and Neutralisation Scattering X-section: Not Rutherford (electron screening), less repulsion- effectively higher energy & reduced cross section. Use Andersen correction for dσ/dΩ using the BZ potential Neutralisation: FOM 50 -100 measurements on a variety of targets:  neutralisation state depends on energy (PB) or velocity (M&Y) Charge fractions: CF (H) = 1-exp(-0.019*E) CF (He) = 1-exp(-0.0061*E) (P Bailey, Daresbury Lab. UK) Depth

6. MEIS mtg U of Huddersfield, 08/12/2011 DRAM MIMcaps Ongoing scale reduction in microelectronics: 40 nm node for DRAM (2011) SiO 2 oxide thickness < 1 nm - serious tunneling leakage current Need for high k & small d in DRAM but leakage current < 10 -7 cm -2 @ 1V Accurate materials characterisation of these nanolayer structures is vital for understanding their properties Materials solution search – range of high-k oxides SrTiO 3 the most promising candidate: high dielectric constant (bulk) ≥ 200, band gap ~ 3.3eV ITRS roadmap for DRAM: Equivalent Oxide Thickness in SiO 2 (EOT) for C=25 fF/cell TiN electrodes, low cost, manufacturing - friendly 2010-11 EOT 0.6 nm 2012-13 EOT 0.5 nm Collaboration with IMEC, Leuven: Christoph Adelmann, Michaela Popovici Planar and high aspect ratio structures

7. MEIS mtg U of Huddersfield, 08/12/2011 Results D02 IGOR spectrum simulation: excellent fit: TiN layer 2.6 nm Ti & N profiles coincide No evidence of SiO 2 layer underneath Near surface reoxidation of TiN disordering of Si lattice to 5 nm depth (consistent with XRR results) Nominal layer structure Ti, Si, O and N peaks well separated (90º) Narrow O peak due to surface reoxidation D02: TiN SiO 2 Si bulk 3 nm ~1nm SiO 2 IMEC clean TiN deposited by PVD (Anelva)

8. MEIS mtg U of Huddersfield, 08/12/2011 Results D05, D06 - annealing Sr/(Sr +Ti) = 0.65; cf. RBS 0.62 (20 nm) D06 (annealed): more uniform Sr / Ti ratio Sr loss ~20% (where?); Ti 20% gain in STO Ti outdiffusion - TiN layer thickness reduced Nominal layer structure: D05: near surface Sr enrichment (ARXPS) Thickness layer (nm) Based onD05D06 STOSr,Ti,O, N3.33.2 TiNSr,Ti,O, N2.92.6 D05: STO Sr rich TiN SiO 2 Si bulk 3 nm 3 nm ~1nm D06: D05 + RTA 650 ºC, 15 s in N 2 (after crystallisation of STO) D06: Sr reduction, Ti increase in STO:10-15%

9. MEIS mtg U of Huddersfield, 08/12/2011 Results D11 D12 - annealing D11: TiN STO stoich TiN SiO 2 Si 2 nm 3 nm 3 nm D12: D11 + RTA 650 ºC, 15 s in N 2 Nominal layer structure (full MIMcap): Thickness layer (nm) Based onD09D10 TiN topSr,Ti,O, N2.0 ±.1 STOSr,Ti,O, N2.8 TiN bottom2.92.8 Clear interdiffusion of TiN/ STO at i/f Increased Ti fraction in STO Surface segregated Sr reduced post annealing Thin layer surface reoxidation Surface segregation of Sr on top of TiN !

10. MEIS mtg U of Huddersfield, 08/12/2011 Sb shallow implants in Si Current generation CMOS transistors require ultra shallow S/D extension junctions; N dopant typically As Collaboration with Fraunhofer IIS-b, Erlangen Stephane Koffel, Peter Pichler 10 S/D Extension junction depths < 15 nm High doping levels to obtain the required sheet resistance (R s ) Sb potential replacement for As as n- type implant for S / D extensions Larger stopping, less straggle leading to shallower implant and steeper profile Higher activation? Lower sheet resistance observed Active Sb concentration up to 10 21 cm -3 measured Diffusion only via V’s, less “ transient” diffusion BUT problems: Sb pile up at SiO 2 /Si interface on annealing Cause?Snowploughing during SPER ? Combined SIMS, MEIS and XTEM study

11. MEIS mtg U of Huddersfield, 08/12/2011 20 keV Sb in Si - SIMS 20 keV Sb @ 1x10 15 cm -2 implant R p ~18 nm (to separate bulk and surface effects) RTP @ 650 ºC 20 s in N 2 SPER & activation Post annealing SIMS depth profiling @ 800ºC 120 s to 1 hr No visible broadening Reduction of peak concentration Sb pile-up at SiO 2 /Si interface @ 900ºC 120 s to 1200 s broadening at 1200 s increased Sb pile-up at SiO 2 /Si i/f SIMS problems: sputter profiling, ion beam mixing, sputter rate changes

12. MEIS mtg U of Huddersfield, 08/12/2011 Sb implants in Si - MEIS MEIS energy spectrum 200 keV He + (Probe sample depth up to ~30 nm) Peaks due to scattering off O, Si and Sb Depth scales added (approx) After RTP (crystal regrown by SPER): Little Sb visible (non-substitutional) No pile-up under oxide after SPER! Si disorder at projected Sb range After RTP + 20 mins anneal: Sb becomes non-substitutional in implant range clear Sb pile-up peak under oxide; (hint @ +10 mins)

13. MEIS mtg U of Huddersfield, 08/12/2011 Sb depth profiles Concentration depth profiles (of non-substitutional Sb) After RTP (SPER): > 80% Sb substitutional No pile-up under oxide MEIS can quantify amounts, location and movement of Sb After RTP + 20 mins post activation anneal: 50 - 60% of Sb becomes visible (= non- substitutional) around projected range Beginning of movement to and pile-up under oxide (hint @ +10 mins)

14. MEIS mtg U of Huddersfield, 08/12/2011 Sb implants in Si - TEM > RTP (SPER) Band of EoR defects Si was amorphised X-tal regrown by SPER No Sb precipitates ! TEM Micrographs Sb precipitation ! MEIS Sb depth profiles SPER+10 min anneal: Precipitates at ~15 nm No Sb defects (precipitates) under Si oxide i/f SPER+1 hr anneal: Precipitates at R p still visible + Sb precipitates at Si oxide i/f Excellent correspondence between MEIS depth profiles and TEM

15. MEIS mtg U of Huddersfield, 08/12/2011 Analysis - Sb movement Sheet concentrations (0-8 nm) & (10-25 nm) as function of anneal time (after RTP) Most Sb is substitutional after RTP; following post anneal rapid depopulation of substitutional Sb sites up to 20 mins then slow decay Pile-up peak increases linearly with anneal time Pile-up is not due to snowploughing during SPER Growth of near-surface precipitates is “diffusion limited” Calculated diffusion coefficients unrealistic for Fermi-level dependent diffusivity (Pichler) Percolation process?

16. MEIS mtg U of Huddersfield, 08/12/2011 Conclusions MEIS (with energy spectrum simulation) provides high depth resolution, quantitative compositional and structural information on nano-layers: Layer thickness (precision of ± 0.1- 0.2 nm) (accuracy depends on e.g. density) Layer stoichiometry close to the nominal parameters / HR RBS data Dopant concentration and location (Sb implants) MEIS shows effect of processing steps on materials in MIMcap nano layers: TiN PVD (sputtering) process removes SiO 2 and causes deeper Si disorder: Clear layer interdiffusion at the STO/ TiN interfaces upon annealing Small but distinct Sr segregation on top of the TiN top electrode in shallow Sb implants into Si: the depopulation of Sb in substitutional sites and Sb movement quantitatively indicates the operation of an unusual Sb diffusion process

17. MEIS mtg U of Huddersfield, 08/12/2011

19. 1 m MEIS facility Daresbury Lab

20. MEIS mtg U of Huddersfield, 08/12/2011 MEIS - Yield corrections Au O ± 5% Mass 20 - 160 product of X-section * neutralisation correction factor ± 5% MEIS is “quantitative” Included in spectrum simulation Combined effect of X-section correction & neutralisation factor Ti Hf N Conditions: [-1-11] in [111 out] Θ  = 70.5 o Sr Si Andersen * neutralisation correction factor Atomic number Sb

21. MEIS mtg U of Huddersfield, 08/12/2011 DRAM MIMcaps Ongoing microelectronic scale reduction: 40 nm node for DRAM (2011) SiO 2 oxide thickness < 1 nm - serious tunneling leakage current In DRAM, the scale reduction forces high k & small d but leakage current < 10 -7 cm -2 @ 1V Accurate materials characterisation of such nanolayer structures is vital to understand their properties Materials solution search SrTiO 3 the most promising candidate: dielectric constant (bulk) ≥ 200, band gap ~ 3.3eV ITRS roadmap for DRAM: Equivalent Oxide Thickness in SiO 2 (EOT) for C=25 fF/cell TiN electrodes, low cost, manufacturing- friendly EOT 0.65 nmEOT 0.5 nm Collaboration with IMEC, Leuven: Christoph Adelmann, Michaela Popovici

22. MEIS mtg U of Huddersfield, 08/12/2011 MIM cap layers analysed STO / TiN layers grown by ALD / PVD @ IMEC (STO = SrTiO 3 ) Systematic change of variable D02: TiN SiO 2 Si bulk 3 nm D03: STO Sr rich SiO 2 Si bulk 3 nm D05: STO Sr rich TiN SiO 2 Si bulk 3 nm 3 nm D06: + RTA 650 ºC, 15 s in N 2 (crystallisation of STO) D04: STO stoich SiO 2 Si bulk 3 nm ~1nm D07: STO stoich TiN SiO 2 Si bulk 3 nm 3nm D08: + RTA 650 ºC, 15 s in N 2 D09: TiN STO Sr rich TiN SiO 2 Si 2 nm 3 nm 3 nm D10: + RTA 650 ºC, 15 s in N 2 D11: TiN STO stoich TiN SiO 2 Si 2 nm 3 nm 3 nm TiN PVD (Anelva) SiO 2 ~1nm IMEC clean STO Stoichiom. (4:3 recipe) ALD STO Sr rich (3:1 recipe) ALD D12: + RTA 650 ºC, 15 s in N 2

23. MEIS mtg U of Huddersfield, 08/12/2011 Spectrum simulation Dechannelling background subtracted from the spectrum A trial sample layer structure based on available information is sliced up in layers of 0.1 nm thick These layers are transformed into Gaussians, to account for energy resolution and depth dependent straggling The Z 2 2 dependence of X-section determines the backscattering yield. All Gaussians are summed. Energy loss rates obtained from SRIM (for up to two regions of different stopping powers) The model is optimized until a best fit (  min 2 ) with the spectrum is obtained. Energy spectra are simulated using a program developed at Daresbury Laboratory that runs as a macro within the IGOR © graphics software. (Paul Bailey, Daresbury Lab.)

24. MEIS Mtg University of Huddersfield, 8 Dec 2011 Fractional comp. Counts Any dechannelling background subtracted from the spectrum Energy spectra are simulated using a program developed at Daresbury Lab that runs as a macro within the IGOR © graphics software. (Paul Bailey, Daresbury Lab.) Example: Energy spectrum from SiO 2 on Si SiO 2 Si (disordered) Si (100) crystal The Z 2 2 dependence of X-section determines the backscattering yield. All Gaussians are summed Energy loss rates obtained from SRIM for up to two regions of different stopping powers The model is optimized until a best fit (  min 2 ) with the spectrum is obtained. These layers are transformed into Gaussians, to account for energy resolution and depth dependent straggling A trial sample layer structure based on available information is sliced up in layers of 0.1 nm thick O Si O Spectrum simulation

25. MEIS mtg U of Huddersfield, 08/12/2011 MEIS - Analysis Energy spectra converted into damage/dopant depth profiles (conc. of Si / dopant atoms vs. depth) Ion yields are referenced to the random level Depth scales obtained from inelastic energy loss data (SRIM) Mass & Depth profile analysis Elastic energy loss gives the mass of scattering atom. Inelastic energy loss within sample enables depth analysis Quantification issues Only scattering from near-surface Si atoms and displaced Si or dopant atoms (due to shadow cones). For 100-200 keV H or He ions sub-nm near surface depth resolution due to: Beam energy near max in inelastic energy loss curve; Large  1 and  2 : long pathways in target; Hi res electrostatic energy analyser. (0.4% En. Res.) Inel. En. loss (eV/Å)

26. MEIS mtg U of Huddersfield, 08/12/2011 2-D spectra 2 -D Yield vs. E -  spectra Scattered ions detected using a toroidal electrostatic analyzer and 2-d detector. Thus a 2D spectrum of yield vs scattering angle and energy is obtained. A cut taken along the [111] blocking direction provides the energy spectrum. Si (surface) N (surface) Hf (buried) Si (buried) O (surface) 100 90 80 70 60 556575 Energy (keV) Scattering angle (deg)

27. MEIS mtg U of Huddersfield, 08/12/2011 Results D03 & D04 thickness taken as half height Ti & Si 20 & 40 % under nominal values: Sr diffusion into SiO 2 during growth? D03 Sr/(Sr +Ti) = 0.65 cf. RBS = 0.62 D04 Sr/(Sr +Ti) = 0.42 cf. RBS 0.5 Nominal layer structure: Thickness layer (nm) Based on D03D04 STOSr, O2.82.0 Ti, Si2.41.7 Clear difference in Sr/Ti ratios & peak widths D03: STO Sr rich SiO 2 Si bulk 3 nm ~1nm D04: STO stoich SiO 2 Si bulk 3 nm ~1nm

28. MEIS mtg U of Huddersfield, 08/12/2011 Results D07 D08 - annealing * Nominal layer structure: Minor changes only due to annealing Stoich. STO on TiN stable upon annealing Thickness layer (nm) Based onD07D08 STOSr,Ti,O, N2.82.6 TiNSr,Ti,O, N2.8 Sr/(Sr +Ti) ratio D07 = 0.53 D08 = 0.5 cf. RBS 0.5 D07 & D08: N peak >1 nm deeper into SiO 2 Some interdiffusion at STO/TiN i/f D07: STO Stoich TiN SiO 2 Si bulk 3 nm 3nm ~1nm D08: D07 + RTA 650 ºC, 15 s in N 2 N O S i Ti Sr

29. MEIS mtg U of Huddersfield, 08/12/2011 Results D09 D10 annealing Nominal layer structure: D09: TiN STO Sr rich TiN SiO 2 Si 2 nm 3 nm 3 nm D10: D09 + RTA 650 ºC, 15 s in N 2 Thickness layer (nm) Based onD09D10 TiN topSr,Ti,O, N1.9 ±.1 STOSr,Ti,O, N3.43.1 TiN bottom2.92.8 D09 single scan, high noise level Interdiffusion at TiN/STO i/f post annealing Some Sr segregation at surface (≤0.4 nm) Surface reoxidation