Germanium Channels for CMOS ... Facts and Fantasies (pdf) - IM2NP

Germanium Channels for CMOS... Facts and FantasiesTony Peaker, Vladimir Markevich, Eddy Simoen*,Alexandra Satta* and Bruce HamiltonUniversity of Manchester, M60 1QD, UK* IMEC, Kapeldreef 75, B-3001 Leuven, Belgium• why go back to germanium !• comparisons with silicon• activation and doping ceiling• implantation defects

Scaling silicon CMOS has made ICs cheaper faster andreduces power per functionBeyond 90nm scaling has given diminishing returns so newapproaches ie novel device structures, SOI, new gatematerials and possibly new channel materials• oxide can’t be scaled beyond 1.2nm because of tunnellingcurrents• SiON probably inadequate beyond 45nm node so need hi kdielectrics to improve short channel effects and gateleakage• mobility in thin channels compromised need higher μmaterials• strained Si, III-Vs. Ge?

The mobility problem … use strain or throw away Si?INTEL 90nm node process:tensile strain in NMOS vianitride cap & uniaxialcompressive via local SiGe10% drive current increasefor NMOS 20% for PMOS~50% increase in “bulk”hole mobility available inprincipleThompson et al (2004) IEEEED Letters 25 p191Thompson et al (2004) IEEETrans ED 51 p 1790

Is Ge a contender for ICs beyond the 45nm node?Si Ge GaAsElectron mobility(cm 2 /Vs)1350 3900 8500Hole mobility (cm 2 /Vs) 450 1900 400Band gap (eV) 1.12 0.67 1.42Electron saturationvelocity (x10 7 cm 2 /s)1 0.6 2Thermal conductivity at300K (W/cm/?C)1.3 0.58 0.55‣ 400% increase in hole mobility with Ge but… there are a few problems!• GOI or localised growth essential• interface state density high and uncontrollable• little fundamental data on implantation

Interface states … conflicting reportsin the literature and on the street• Ge diffuses into HfO 2 and degrades it as a dielectric… hafnium aluminates better but rare earth oxidesfavoured by some ….• nitrided germanium oxide good compromise as firststack layer• interface state density >10 12 cm -2 … not controllable• forming gas anneal doesn’t work for most people• reasonable PMOS devices from Chui (Stamford nowat INTEL)• no interface states seen in EPR … why?

until recently not much basic data on implantationinto germanium had been determined•implantation defects•defect removal•activation of dopants•doping ceilings•transient enhanced diffusion

Is germanium fundamentally different to silicon inrelation to implantation? …‣ YES because the energy needed to create a vacancyin Ge is much less than in Si (neutral vacancy 1.9eV inGe but 3.3eV in Si) and so:•diffusion of all technologically important dopants isthought to be vacancy mediated …•transient enhanced diffusion may result from metastablevacancy clusters (not interstitials)?•the doping limitation mechanisms in n-type may bemore dramatic than in silicon (vacancy driven)

Diffusion in Ge compared to Si (normalised to melting point)T m (Ge)= 1210 K, T m (Si) = 1685 Ksubstitutional atomshybrid atomsinterstitials atomsself diffusion10 -6FeFeCuCucontinuous lines Gebroken lines Siboron is a slowdiffuser in GeD X(cm 2 s -1 )10 -910 -1210 -15from Harmut BrachtMünster EMRS 2006 10 -18Zn↓B10 -3 1.0 1.2 1.4 1.6AsSiAsGeZnSi:XGe:XBT m/T

dopants in germaniumMaximum equilibriumsolid solubility (600ºC) ofdopants in germaniumis much lower thanin silicon. For equilibriumB a 10nm implant will beexpected to give 30,000 Ω-sq.ITRS (Si) requirement forthe 30nm node is 300 Ω-sq.for a 12nm implant… a long way off (~2orders) but Ga seems OKand P is nearMaximum Equil. Solid Solubility (cm -3 )Trumbore Bell Sys Tech J 39 (1960) 20510 2010 1910 18B Ga P As Sb10 21Doping Element

B implants in Ge with and without Ge pre-amorphisationBoron concentration (at cm -3 )10 2210 2110 2010 19(a)eqilibrium concentration at 600°CB as implantedB after 400°C annealB after 600°C annealBoron concentration (at cm -3 )10 2210 2110 2010 19(b)as implantedafter 400°C annealafter 600°C annealeqilibrium concentration at 600°C10 180 20 40 60 80 100 120Depth (nm)10 180 20 40 60 80 100 120Depth (nm)• Pre-amorphization affects the as-implanted B profile: (channeling suppression)• No significant diffusion in c-Ge or a-Ge up to 600 o C.

p-type dopants: B activation with & without pre-amorphization1500BB+100keV GeR s(Ω/square)100050080 Ω/square0300 400 500 600Anneal temperature (ºC)• B highest activation (20%) takes place at low T in c-Ge (300 o C)• However, 60% activation achieved from a-Ge phase.• Increasing the B doping dose improves Rs (down to ~80 Ω-sq).

p-type dopants Re-crystallization of pre-amorphized and B-doped GeSiO 2SiO 2SiO 220 nmα-Ge100 nmc-Ge20 nmAs-implantedDepth of amorphous layer: 100 nmSiO 2Annealed at 400 o C, 60s, N 2Fully re-grownNo detectable EOR defects20 nm

ut beware of sputtering and vacancy clustering whichcause surface roughness and form voids with heavy ionsvoids are 100nm deep after 10 15 Sb +at 70keVsurface roughness increases with ionmass … SiO 2 capping reducessputteringJanssens et al J. Vac Sci & Tech B 24p510 (2006)

n-type dopants … a very different storyP activation and diffusionP-implantations suffer from:- low activation ≤6x10 19 cm -3- P-precipitation at highimplantation doses ≥5x10 15at/cm 2- Concentration-enhanceddiffusion above 2x10 19carriers/cm 3Concentration (at/cm 3 )As-imp500 o C, 30s10 21 500 o C, 60s500 o C, 120s10 2010 19Arsenic also has major problems 10 18(b)0 20 40 60 80 100Depth (nm)

What about defects and junction leakage?summary of recombination-generation states characterized in germaniumafter annealing for 30mins at the temperatures indicated theconcentration of the defect has reduced to half its original value

Irradiation damage removalΔC, pF0.50.40.30.20.1U b= -5.0 V; U p= -0.5 Vt p= 1 ms; e n= 80 s -1Ge:Sbas irradiated100 o C140 o C200 o C400 o C0.050 100 150 200 250Temperature, KDLTS of Ge:Sb as irradiated andannealed for 30mins at theindicated temperature. Similarmeasurements have been done onP, As and Bi doped material.MESSAGE:All irradiation inducedgeneration-recombinationcentres can be removedwith a very low thermalbudget

DLTS of Ge with low dose Si implant and annealΔC, pF43210Ge 0.9992Si 0.0008:P3 MeV Si ionsΦ = 1x10 10 cm -2U b= -5.0 VU p= 0 Vt p= 1 mse n= 80 s -1as implanted100 o C200 o C300 o Co400 CMESSAGE:much more complexspectrum and evolutionduring annealing thanirradiation case. Verysubstantial concentrationof generation centresremaining after 400 o Canneal50 100 150 200 250Temperature, K

DLTS of Si implanted Ge after annealing2.0Ge 0.9992Si 0.0008:PΔC, pF1.51.00.5U b= -4.0 VU p= 0 Vt p= 1 mse n= 80 s -110 12 Si 400 0 Cincreased doseresults in morepersistent generationcentres (substantialleakage current after500 0 C anneal)10 10 Si 400 0 C0.010 12 Si 500 0 C50 100 150 200 250 300Temperature, K

Diode leakage currentsI diode(A cm -2 )10 310 210 110 010 -110 -210 -310 -410 -510 -610 -7300°C anneal400°C anneal500°C annealintrinsic leakage at room temperature for 10 16 cm -3 material-1.0 -0.5 0.0 0.5 1.0p + n diode afterannealing for 5 minutesThe diode was made byimplanting 13keV B +ions into n-typegermaniumafter 500ºC anneal thedominant current is dueto surface leakageresidual deep statescontribute 8x10 -5 A cm -2(calculated from DLTSdata)V diode(V)

Overview• major interface state problem still to be resolved• boron seems a well behaved dopant in Ge … goodactivation no significant diffusion with adequateactivation thermal budgets.• n-type dopants do not appear to provide adequateactivation with simple techniques.• surface leakage two orders more significant than in Si• Implant damage difficult to remove … a 500 0 C anneal ofa 10 12 cm -2 3MeV Si leaves a substantial concentrationof generation centres and junction leakage.

Acknowledgements• UK Engineering and Science Research Council andEuropean Commission (INTAS) for funding• Marc Meuris, W. Vandervorst and Cor Claeys of IMECfor many valuable discussions• Arne Larsen and Knud Bonde Nielsen (University ofÅrhus, Denmark) for collaboration on the electron andgamma irradiation work• Russel Gwilliam (UK National Ion Implantation Facility atSurrey University)• Umicore Belgium for germanium materials