Advances in semiconductor materials and device metrology CIICT 2007 - PowerPoint PPT Presentation

Advances in semiconductor materials and device metrology CIICT 2007 DCU, 28 Aug 2007. Patrick J. McNally Nanomaterials & Processing Laboratory School of EE, DCU 1

Co-Workers � Dr. Lisa O’Reilly, Lu Xu, Ken Horan, Jennifer Stopford, Dr. Donnacha Lowney (DCU). � Nick Bennett, Prof. Brian Sealy, University of Surrey. � Dr Jim Greer, Dr. Nicolás Cordero, Yan Lai, Tyndall Natioinal Institute, Cork. � Prof. Nick Cowern, University of Newcastle. � Dr. Gabriela Dilliway, University of Southampton. � Acknowledgements: Science Foundation Ireland, Enterprise Ireland. 2

Metrology Challenges � Length scales � Lateral (X-Y) � Vertical (Z). � Requirements: � Non-destructive. � In situ . � In line. � “Nano” sensitivities: small volumes, areas (nm- scale); impurity sensitivites (e.g. 10 12 - 10 13 cm -3 ); nano-void and nano-pore detection; etc. 3

Low-k metrology challenges � New low-k dielectrics have different mechanical/physical properties compared to SiO 2 . � Pores in the material. � Fragile – delamination; stress-induced fracture. � Back end of the line (BEOL): problems with assembly and packaging. � No convenient and competent metrology tools. Source: ITRS 2005 4

Cu metallisation metrology challenges Source: ITRS 2005 � Measuring barrier layer(s) under seed copper. � Detection of voids in copper lines after CMP and anneal processes. � Thick Cu lines mask this voiding. � Detection through multi-layer structures e.g. individual layer thicknesses. � Delamination of Cu from e.g. low-k layers before and after CMP. � Local stress vs. wafer stress. � Adhesion strength measurements are still done using destructive methods. � Detection of killer pores and voids is not yet possible. 5

A Selection of DCU’s metrology approaches � Gas Cell Photoacoustic Microscopy � Micro-Raman Spectroscopy 6

Cochin Univ. Sci. & DCU proprietary PA cells Source: N. George, 7 Photoacoustic Microscopy (PAM) Tech., India. (incl. sample)

8 Automated PAM system for Si wafer analysis

Analysis of multi-layer structures on Si wafers Cu, SiO 2 layer thicknesses = 500nm SiO 2 Si Si Cu Si Cu SiO 2 Si 9

IC Chip Cracking & Delamination -236.1 -226.4 -216.8 -207.1 -197.4 -187.8 -178.1 -168.5 -158.8 -150.8 PA Phase [Deg.] Delamination Optical Micrograph Photoacoustic Phase Image 10

Wafer bonding defects (Phase Contrast) L. Xu, P McNally, DRIP XII , 9 - 13 September 2007 Berlin (Germany). -87.00 Wafer edge effects -86.10 -85.20 ICPAM Phase image ( f =216 Hz). -84.30 8 -83.40 -82.50 •10,000 pixel image shown is -81.60 obtained in 8 minutes. 6 -80.70 Position [mm] -79.80 •The bonding defects acted as extra -78.90 thermal barrier, shown as extra time -78.00 4 delay in phase images. Phase [Deg.] 2 Inhomogeneous Doppler interference effect at bonded bonding interference interface effects ?? 0 Wafer bonding defects 2 4 6 8 10 Wafer Bonding confirmed by IR and Optical defects confirmed by OM Position [mm] Microscopy 11

12 DELAMINATION After Image Processing

Ongoing PAM Developments � Upgrade to 200 mm & 300 mm wafer capability (end-2007 – Enterprise Ireland Proof of Concept Fund) � Porous dielectric measurements – early results promising. � Wafer edge sub-surface cracks. � Measure nm-scale delamination. � Technology licensing underway. 13

Summary for Photoacoustic Microscopy � Can see through opaque (metallic) layers. � Multi-layer characterisation : thicknesses, delamination, porosity. � Nanometric vertical scale sensitivities. � Whole wafer scanning. 14

Micro-Raman Spectroscopy ( µ RS) � Incident light excites vibrational modes in the sample. Subsequently scatter the light. � Some light is scattered at a different energy (wavelength). � Energy exchange between incident photons and semiconductor phonons (internal vibrational modes). • Probe regions ~ 1 µ m diam. � Raman light intensity is very weak. • 325nm laser � Si � Typically about one photon penetration depth ~ 9nm � out of 10 12 . � E phonon = E incident - E scattered • True nanometric scale depth metrology. 15

• Strained/deformed crystal. • Vibrations of crystal lattice altered. • “Spring constant(s)” between atoms changed. • Shifts frequency of inelastically scattered Raman photons. • A plot this shifted light ouput intensity vs. frequency is a Raman spectrum. 16 Source: www.kosi.com

JY-Horiba LabRam 800 µ RS System Confocal hole 488nm or 325nm Laser source Mirror CCD detector Notch filter Mirror Lens and mirror Microscope XY stage holding sample Sample 17

Strained Silicon CMOS Technology “The industry cannot live without strain-engineering for enhanced mobility” Semiconductor International, 2006 • Strained Si channels improve electron and hole mobilities. • Greater current drive for less power. source drain (Source: Thompson et al., Intel) (Source: Valencia, IBM) How does strain affect heavily doped device regions? 18

Research Motivation Source/drain extensions are one example ITRS identifies 3 key requirements… Single nMOS transistor 1) Increasingly shallow junction depth (x j ) Gate Stack 2) Increasingly steep junction profile Oxide 3) Maintain low resistance (R s ) Channel source drain � I mportantly… Sb has the edge over As 19

1500 Sb in Bulk Si 1400 As in Strained Si Sample Information 1300 Sb in Strained Si Sheet resistance ( Ω /sq.) 1200 � 17 nm strained Si layer 1100 grown on a graded Si 0.83 Ge 0.17 virtual 1000 substrate 900 � 43 nm strained Si layer 800 grown on Si 0.80 Ge 0.20 virtual substrate 700 � Antimony and Arsenic 600 Ion Implantation 500 � 2keV, 4e14cm -2 Sb 600 650 700 750 800 � 2keV, 4e14cm -2 As o C) Anneal temperature ( � Annealed @ 600, 700, 800 0 C in N 2 ambient � Large R s reduction for strain vs bulk for Sb � Comparison to bulk doping unstrained Si � Lower R s for Sb doping compared to As in strained Si. � Sb more highly activated than As in the presence of strain N. S. Bennett et al., Appl. Phys. Lett. 89 , 182122 (2006) 20

Raman Spectra: 325nm laser R eference Si Si-Si � 325 nm HeCd UV Strained Si, 20% G e 2000 Raman Intensity (a.u.) laser. � d p ≈ 9 nm. 1000 UV Strained Si ~ 40nm 0 200 300 400 500 600 Si 0.8 Ge 0.2 ~ 200nm -1 ) W avenum ber (cm 2000 Reference Si Si-Si Strained Si, 20% G e Raman Intensity (a.u.) Si 0.9 Ge 0.1 Tensile ~ 1000nm 1000 Si Substrate UV Laser penetration in SiGe 0 structure 500 510 520 530 RED 21 -1 ) SHIFT W avenum ber (cm

Raman Spectra: UV laser Sb, 2keV, 4e14cm -2 As, 2keV, 4e14cm -2 ε Si Ref Si Si Ref SSi 0.7% SSi-Sb 800 1200 1200 SSi-Sb600 Intensity (a.u.) SSi-Sb700 SSi-Sb800 700 600 600 600 Imp. 0 0 500 510 520 500 510 520 530 -1 ) -1 ) Wavenumber (cm Wavenumber (cm � Red-shift of Si peak indicates the presence of tensile strain in the Si cap layer � Spectra of Sb and As implanted samples show similar behaviour with clear intensity variation with heat treatment. 22

UV Raman Spectra Analysis Si Raman Peak Intensity 2400 SSi-Sb Average Si Intensity (a.u.) SSi-As 2100 1800 1500 1200 900 600 300 0 Ref. Strain. Imp. 600 700 800 Si Si � Si-Si peak intensity variation is consistent with lattice disorder introduced by ion implantation which recovers with heat treatment. � Lattice recovery may not be complete following RTA at 800 0 C for 10 sec. 23

UV Raman Si Peak Shift � Biaxial stress σ xx = σ yy = - ∆ω SiUV /4 GPa Sb implant � Strained Si, 17% Ge: -5.0 � ∆ω SiUV = -5.64 ± 0.3 cm -1 17% Ge -5.5 20% Ge � ⇒ σ = 1.4 ± 0.1 GPa -1 ) -6.0 Raman shift (cm � ε =0.77 ±0.06% -6.5 � Strained Si, 20% Ge: -7.0 � ∆ω SiUV = -6.27 ± 0.2 cm -1 -7.5 � ⇒ σ = 1.57 ± 0.1 GPa -8.0 � ε =0.86 ±0.06% Strained Sb Sb, Sb, Sb, Silicon Implant 600 700 800 � Strain relaxation would be detected by a blue shift (higher wavenumber) of the � Peak shift Relative to Si Raman peak. Unstrained Reference Si � But peak red-shift clearly increases following ion implantation and RTA 24 Ref: I. De Wolf, Semicond. Sci. Technol. 11 , 139(1996)

What is causing the net red shift of the Si peak position observed in the UV Raman spectra? Sb Doping As Doping Linear fit, R=0.85 Linear fit, R=0.80 -0.2 Normalised Raman shift (cm -1 ) -0.4 -0.6 -0.8 -1.0 -1.2 -1.4 -1.6 -1.8 13 14 14 14 13 14 14 14 0.0 0.0 6.0x10 1.2x10 1.8x10 2.4x10 6.0x10 1.2x10 1.8x10 2.4x10 -2 ) -2 ) Sheet Carrier Concentration (cm Sheet Carrier Concentration (cm � Clear linear dependence between the normalised Raman shift and the sheet carrier concentration! 25

Why does doping cause a Raman shift? � Cerdeira and Cardona* observed carrier-concentration related frequency shifts in the Raman spectra of both p- type and n-type Si. � N-doping of Si alters the lattice deformation potential, effectively “softening” the lattice � ⇒ lower phonon vibrational frequencies � ⇒ Raman red-shifts � Usually a very small effect in n-type Si � Only significant when doping concentration is large. � Which it is here!!! 10 20 – 10 21 cm -3 effective doping!!** � Independent of dopant type. • * F. Cerdeira & M. Cardona, Phys. Rev. B. 5 , 1440 (1972). • ** L. O'Reilly et al., INSIGHT-2007, Napa, California, U.S.A., May 6-9, 2007. 26