Ph.D. Dissertation Proposal First Principles Studies of Defects in HfO2 and at Si:HfO2 Heterojunctions Chunguang Tang (唐春光) (Bachelor Eng.: Univ. Sci. Tech. Beijing) (Master. Sci.: NUS) Chemical, Materials & Biomolecular Engineering Institute of Materials Science University of Connecticut Principal Advisor: Prof. R. Ramprasad Associate Advisor: Prof. L. Shaw Associate Advisor: Prof. P. S. Alpay # of transistors Introduction: device miniaturization First transistor radio: 4 transistors. Quad core processor contains 820 million transistors high k (dielectric constant) transistor K~4 K ~ 30 High k issues • Formation of interface phases (SiOx, silicate, Hf-Si) Wong et al, Microelectronic Eng. (2006) Locquet et al, JAP (2006); Stemmer et al – Effects of oxygen point defects* • Point defects migration may contribute to interfacial phase formation • High oxygen pressure favors silica & low pressure favors Hf silicide * D. Y. Cho et al, APL, 86, 041913 (2005); X. Y. Qiu et al, APL, 88, 072906 (2006); S. Stemmer, JBSTB, 22, 791 (2004) High k issues • High leakage currents and low dielectric constant due to crystallization – As-deposited: amorphous (preferred) – Crystallizes at 400~500 °C amorphous k ~ 30 cubic k ~ 29 tetragonal k ~ 70 – Doped HfO2 with alloying elements. • Si, Y, La, F, N • Increase crystallization temperatures • Stabilize higher k phases monoclinic k ~ 16 Proposed research plan • Undoped HfO2 – The formation and migration of O vacancies, O interstitials and Hf vacancies. – Their contribution to the interfacial phases. • Doped HfO2 – Dopants: Si, Y, La, F, N. – Effects of dopants on relative stabilities of various phases of HfO2. – Effects of dopants on O defect chemistry. Computational Methods • Density Functional Theory (DFT) – Many nuclei-many electron problem one electron problem 2 2 V r eff i r i i r 2m – Supercell approach – Phase and structure information, defect energies • Computational times – Defect formation energies 16 days in one AMD 2.0 GHz processor (Supercell of ~230 atoms). – Migration energy calculation 45 days. 0 E 1 2 Emigr. 3 4 Reaction path Completed Research • Bulk HfO2 results Table I: relative energies (eV) and lattice constants (Å) of bulk HfO2 DE/HfO2 a (calc./expr.) b (calc./expr.) c (calc./expr.) c-HfO2 0.25 5.06/5.08* =a =a t-HfO2 0.16 5.06/5.15** =a 5.14/5.29 m-HfO2 0 5.14/5.12*** 5.19/5.17 5.30/5.29 * J. Wang, H. P. Li, and R. Stivens, J. Mater. Sci. 27, 5397 (1992) ** D. M. Adams, S. Leonard, D. R. Russel, and R. J. Cemik, J. Phys. Chem. Solids 52, 1181 (1991) *** J. Adam and M. D. Rodgers, Acta Crystallogr. 12, 951 (1959) Defects in bulk HfO2 Eform + 1 O atom Eform = Evac+ (EO2)/2 - Eperf Table II: Formation energies (eV) of point defects in bulk HfO2 3-fold site 4-fold site O interstitial 1.7 2.5 O vacancy 6.6 6.5 Hf vacancy 6.1 O interstitial Formation and Migration* Si HfO2 Hf O Interfacial segregation: Thermodynamic driving force (decreasing Eform as interface is approached) Kinetic driving force, and O penetration into Si (decreasing Emigr as interface is approached) Experimental Emigr. of O interstitial in bulk Si: 2.44 eV** (2.26 eV, calculated) O interstitials could lead to the formation of SiOx * C. Tang & R. Ramprasad, Phys. Rev. B 75, 241302 (2007); ** J. C. Mikkelsen, Appl. Phys. Lett. 40, 336 (1981). O Vacancy Formation and Migration* Si HfO2 Hf O Interfacial segregation: Aided by thermodynamic & kinetic driving forces O vacancies could lead to the formation of Hf silicide * C. Tang, B. Tuttle & R. Ramprasad, Phys. Rev. B 76, 073306 (2007) Hf Vacancy Formation and Migration* Hf O Si penetration • Hf vacancies prefer the interface • Si strongly prefers to penetrate into HfO2 Hf vacancies could lead to the formation of Hf silicate * C. Tang & R. Ramprasad, Appl. Phys. Lett., 92, 152911 (2008) Accumulation of O Point Defects* Thermodynamics favors accumulation of point defects at interface, and consequently, the creation of Hf silicide or SiOx Abrupt Interface “SiOx” Interface * C. Tang & R. Ramprasad, Appl. Phys. Lett., 92, 182908 (2008) “Hf-Si” Interface Si doped HfO2 (SDH) C-SDH (1-x) HfO2 + xSiO2 + Ef = Hf1-xSixO2 t-SDH m-SDH 1. If Si > 12% t-HfO2 most stable 2. The local chemistry of Si prefers SiO2 configuration c-SDH m-SDH t-SDH Y doped HfO2 (YDH) (1-x) HfO2+(x/2)Y2O3+Ef=Hf1-xYxO2-x/2 m-YDH c-YDH Charge neutrality 2 Y atoms & 1 O vacancy t-YDH 1. If Y > 12%, t-HfO2 and c-HfO2 more stable. 2. Similar stabilization phenomenon in c-YSZ for fuel cell application. 3. Instead of Y, positively charged O vacancies are identified as the major stabilizing factor. Remaining research • Undoped HfO2 – Amorphous HfO2 and Si heterojunction; • Lower leakage current • High dielectric constant – Various charged states of O defects (VO0, VO+1, VO+2, iO0, iO-1, iO-2); • Formation energies Remaining research • Doped HfO2 – Effects of dopants on HfO2 stabilities (La, F, N); – Formation and migration energies of O defects close to and far away the dopants. • How they influence the behaviors of defects in HfO2 and Si heterojunctions Publication list 1. C. Tang and R. Ramprasad, "Oxygen defect accumulation at Si:HfO2 interfaces" , Appl. Phys. Lett., 92, 182908 (2008). 2. C. Tang and R. Ramprasad, "A study of Hf vacancies at Si:HfO2 heterojunctions" , Appl. Phys. Lett., 92, 152911 (2008). 3. C. Tang and R. Ramprasad, "Oxygen pressure dependence of HfO2 stoichiometry: An ab initio investigation" , Appl. Phys. Lett., 91, 022904 (2007). 4. C. Tang, B. R. Tuttle and R. Ramprasad, "Diffusion of O vacancies near Si:HfO2 interfaces: An ab initio investigation", Phys. Rev. B, 76, 073306 (2007). 5. C. Tang and R. Ramprasad, "Ab initio study of O interstitial diffusion near Si:HfO2 interfaces", Phys. Rev. B, 75, 241302(R) (2007). 6. B. R. Tuttle, C. Tang and R. Ramprasad, "First-principles study of the valence band offset between silicon and hafnia", Phys. Rev. B, 75, 235324 (2007). 7. R. Ramprasad and C. Tang, "Circuit elements at optical frequencies from first principles: a synthesis of electronic structure and circuit theories", J. Appl. Phys. 100, 034305 (2006). 8. Tang CG, Li Y, Zeng KY, Mater. Lett., 59, 3325, (2005). 9. Tang CG, Li Y, Zeng KY, Mater. Sci. Eng. A, 384, 215, (2004). 10. Li Y, Cui LJ, Cao GH, Ma QZ, Tang CG, Wang Y, Wei L, Zhang YZ, Zhao ZX, Baggio-Saitovitch E, Physica C, 314, 55, (1999). 11. Li Y, Wang YB, Tang CG, Ma QZ, Cao GH, SCI CHINA SER A, 40, 978, (1997). 12. Li Y, Tang CG, Ma QZ, Wang YB, Cao GH, Wei T, Wang WH, Zhang TB, Physica C, 282, 2093, (1997). Acknowledgment Committee members: Profs. Rampi Ramprasad, Leon L. Shaw and Pamir S. Alpay Profs. Puxian Gao and George A. Rossetti Group students: Ning, Luke, Tom, Ghanshyam and Hong Computational resources: IMS computation clusters; SGI supercomputer in SoE Funding: NSF & ACS-PRF Backup slides (P, T) dependence of O defects • DFT computations of O vacancy & interstitial formation energies as a function of defect concentration … combined with … thermodynamic model yields (P,T) dependence of stoichiometry formation E defect E defect E perfect 1 0 O kT ln PO2 2 2 Jiang et al Appl. Phys. Lett. 87, 141917 (2005) Pick up T, find P to make formation energy 0, corresponding to equilibrium condition. C. Tang & R. Ramprasad Appl. Phys. Lett. 91, 022904 (2007) (P, T) dependence of interface morphology T = 1200 K T = 400 K formation coverage E N Ecoverage Eabrupt O2 2 P (coverage) e rma tio n Ecofovera g e / kT e rma tio n Ecofovera g e / kT coverage “Hf-Si” Interface Abrupt Interface “SiOx” Interface • Pressure changes could stabilize silicide or SiOx • Increase in T makes abrupt Si:HfO2 interface less stable Si doped HfO2 (SDH) (1-x) HfO2+xSiO2+Ef=Hf1-xSixO2 a1 a2 c-SDH m-SDH The local chemistry of Si prefers SiO2 configuration t-SDH Y doped HfO2 Density Functional Theory Hˆ i (r ) i i (r ) (r ' ) 1 2 ˆ H v pseudopot(r ) d 3r ' XC [ (r )] 2 r r' occ (r ) i (r ) 2 i Initial guess of wave function & electron density Set up Hamiltonian Energy, forces on atoms New electron density no DE < Ebreak yes end Accuracy of DFT • • • • • Structures (bond lengths, bond angles, lattice constants) predicted to within 1 % of experiments Material Expt. DFT error type Ag (FCC) 4.09 4.11 0.6% Metal V (BCC) 3.03 3.02 0.3% Metal LaBi 6.57 6.65 1.2% Alloy Si 5.43 5.43 0.0% Semicon GaAs 5.65 5.66 0.2% Semicon HfO2 5.08 5.06 0.4% Oxide NbO 4.21 4.23 0.6% Oxide CoSi2 5.36 5.30 -1.1% Silicide ZrN 4.62 4.63 0.3% Nitride CaF2 5.46 5.50 0.6% Halide Elastic properties (bulk & shear modulus, etc.) accurate to within 5% of experiments Bond energies, cohesive energies within 10% of experiments Relative energies (energy difference between FCC & BCC, for example) are accurate to within 2% Band gaps are off by about 50% !!! Si:HfO2 heterostructure models Tetragonal HfO2-based Monoclinic HfO2 -based Mater ial k E_gap DEc SiO2 3.9 8.9 3.2 amor Si3N4 7 5.1 2 amor Al2O3 9 8.7 2.1 amor Y2O3 15 5.6 2.3 cubic ZrO2 25 5.8 1.2 m, t, c 25 5.7 1.5 m, t, c La2O3 30 4.3 2.3 Hex, cubic Ta2O5 26 4.5 0.5 orthorhombic unstable TiO2 80 3.5 1.2 t, rutile, anatase unstable E_gap DEc HfO2 Tc 400500 Doped Tc k HfTaO 1000 - HfSiO 1050 ? HfSiO (N) > 5 4.6 (~8) HfYO NA HfAlO > 900 HfLaO > 900 Structure Structure unstable (& hygroscopic) Comments 2.1(2.4VBO) JAP-87-484 data source thinner IL apl-85-2893 amor sharp interface JAP-87-484 various Eg, CBO reported app.surf.sci.253-2770 cubic 5.6 (RPP, 69, 327, Robertson) amor amor 1823 data source Wallace & Wilk 2003 0.5-1.5 (3.0VBO) 27 Comments amor apl-86-102906 reduce mobility IEEE-ele. Dev. Lett-24-556 apl-89-032903 & 85-3205 & 88-202903 Why HfO2 Source: R. M. Wallace and G. D. Wilk, Crit. Rev. Solid State Mater. Sci. 28, 231 Influence of defects on performance • Charges are trapped in defects, shifting threshold voltage and making operation unstable. • Trapped charges scatter carriers in the channel lower carrier mobility • Cause unreliability (oxide breakdown) Effect of F • (APL 90, 112911) – Remove midgap states from Hf dangling bonds at HfO2/SiO2 interface; – Excessive F increase leakage current. • (APL 89, 142914) – Defect passivation