Luigi Colombo. Texas Instruments Dallas, TX, USA. Jozef Stefan Institute Ljubljana, Slovenia April 9, 2013

Transcription

1 Luigi Colombo Texas Instruments Dallas, TX, USA Jozef Stefan Institute Ljubljana, Slovenia April 9, 2013

2 UT Austin Ruoff group UT Austin - Banerjee group UT Austin - Tutuc group UT Dallas - Wallace group UT Dallas - J. Kim group GIT/UT Dallas Vogel group Nanoelectronic Research Initiative and NIST 2

3 Introduction Graphene based devices Graphene crystal growth Graphene device integration Summary 3

4 10 mm 1 mm Modern CMOS Beginning of Submicron CMOS Deep UV Litho 100 nm 10 nm >40 Years of Scaling History Every generation Feature size shrinks by 70% Transistor density doubles Wafer cost increases by 20% Chip cost comes down by 40% Generations occur regularly On average every 2.9 years over the past 40 years Recently every 2 years 90 nm in nm ? 32 nm in 2010 Presumed Limit Need a New Switch 4

5 Performance per power density vs. gate length that the slowing of voltage scaling causes a reversal of the trend beyond 130-nm-node technology Power density vs. gate length: active and passive power density W. Haensch et al IBM J. RES. & DEV. VOL. 50 NO. 4/5 JULY/SEPTEMBER

6 6

7 Spin based devices Spin Wave Spin torque Spin FETs All spin logic Nano magnetic logic devices Tunnel FETs III-V, graphene Graphene PN Junction Devices Bilayer Pseudospin FETs (BiSFET) Lateral graphene tunneling devices 7

8 One Atom Thin A single layer of carbon atoms arranged in a hexagonal lattice as in graphite (a 0 =1.42Å) Linear spectrum Strength Energy dispersion is linear as a function of k, Ambipolar High Fermi velocity v F =10 8 m/s; high (~10 3 K) Graphene Fermi energy for typical (10 12 cm -2 ) densities High (m > 10 5 cm 2 /Vs) mobility at room temperature High mobility Highly flexible Thermal conductivity ~3000 W/m-K in plane and highly anisotropic; ~ 2 W/m-K out of plane mobility ~1100 GPa modulus, fracture strength ~125 GPa Unique optical properties Exceptionally high specific surface area (2630 m 2 /g) High temperature base (support) material (in reducing or neutral conditions) Barrier material impermeable? Density ~2.2 g/cm 3 (lightweight yet strong) 8

9 Graphene /CNT Interconnect? Graphene Switch? PCB Graphene Thermal spreader 9

10 BiSFET Bose-Einstein Condensate High Quality SC Graphene J.J. Su and A.H. McDonald, Nat. Phys., 2008 Banerjee et al, EDL 2009 Tunneling FET: Low SS Bilayer graphene? GNR? Q. Zhang et al, EDL, 2008 P-N Junction Veselago lens switch High Quality SC Graphene V. V. Cheianov et al, Science,

11 Very intriguing device, will require significant process development to realize as Gr/h-BCN There could be other options, easier than lateral composition control for implementation G. Fiori et al., ACS Nano

12 Elements of a BiSFET Device I hole ( ) I electron ( ) Gate dielectric: perhaps high-k for increased gate control; perhaps ferroelectric to induce required sheet charge layers in graphene under zero gate bias. h e V p V n graphene leads Metal contacts Top Gate: used to control charge densities in graphene layers; perhaps work-function engineered to provide required sheet charge densities under zero bias. V G,p V G,p Graphene bilayer Tunnel barrier: perhaps dielectric or just misaligned bilayers? Bottom Gate: Perhaps one gate used as only a backgate with fixed voltage/workfunction to provide required sheet charge densities under zero bias. S. K. Banerjee et al., UT Austin (2009) 12

13 13

14 ~ 120 ~ 1000C ~ 20 RT! Zazula, JM, On Graphite Transformations at High Temperature and Pressure Induced by Absorption of the LHC Beam, LHC Project Note 78 / 97,, CERN-SL/BT(TA) 14

15 Bonaccorso et al., Materials Today 15(12) 584,

16 Chemical vapor deposition: Cu, Ir, Ru, Pt, etc Plasma enhanced CVD processes: Cu, Precipitation: Ni, Ru, Co, Pt, Pd, etc. Growth by desorption of Si from SiC X. Li et al, Science (2009): Copper P. Sutter et al., Physical Review B 80 (24) (2009): Platinum N.A. Kholin et al., Surface Science (1984): Iridium Karu and Beer, JAP (1966). Nickel J. Sanchez-Barriga et al., Diamond and Related Materials (2010). Cobalt J. Lee, et al, in IEDM - Technical Digest, (ICP-CVD) D. V. Badami, Nature (1962). SiC 16

17 T (K) T (K) T (K) T (K) Substrate selection Metals Dielectrics Process type: LPCVD APCVD PECVD etc Ti/(C+Ti) Cu/(C+Cu) Precursor Sources: Gases Liquids Solids Solutions Ni/(C+Ni) Ir/(C+Ir) Okamoto H., Phase Diagrams for Binary Alloys, Desk Handbook, Vol. 1,

18 T g ~ 1000 C; CH 4 :H 2 = 1:100; P ~ 100 mtorr Surface catalyzed and limited growth SEM TEM graphene 6-fold Cu (111) 100 mm 4-fold Cu (100) 200 mm 20 mm Grain boundary Li et al. Science 324 (2009) Li et al. Nano Lett mm Y. Hao et al. in preparation 18

19 Photoelectron intensity (cts/s) C 1s HOPG Cu + CVD graphene Binding energy (ev) Intensity (counts) x G HOPG Graphene on Cu 2D Raman shift (cm -1 ) CVD graphene exhibits asymmetric C 1s XPS spectrum, expected for sp 2 -bonded graphene Comparable FWHM to HOPG) indicates that CVD graphene is well-ordered CVD graphene: 0.71 ev FWHM HOPG: 0.52 ev FWHM CVD graphene on Cu shows a low Raman D-band intensity similar to HOPG A. Pirkle, UTD PhD Thesis,

20 Cu substrate Pt substrate 400 mm Yan Z, ACS Nano mm Y.Hao Unpublished results UT Austin L. Gao et al Nat. Comm. (2012) 20

21 Mobility 16,400 25,000 cm 2 /V - for Gr/SiO 2 Mobility 27,000 45,000 cm 2 /V - for Gr/h-BN X. Li et al., Nano Letters (2010) Petrone et al., Nano Letters (2012) 21

22 Single Crystal Graphene Exfoliated CVD on Cu CVD on Pt CVD on Cu Manchester UT Austin Shenyang UT Austin 1 mm 1 mm Microns millimeters Centimeters meters? Polycrystalline CVD Graphene UT Austin SKKU IBM Graphene Quartz 4 cm 22

23 Graphene surface is chemically inert Need to functionalize the surface to deposit dielectrics E-beam metals, O 3, NO 2, organics,. Scaling of dielectrics down to ~ 1 nm needed for some devices Enable a variety of dielectrics high-k, low-k and 2D dielectrics (TMDs and h-bn) 23

24 Kim et al. PRB 83, (R) (2011) Dean et al. SCC 152 (2012)

25 CVD graphene transferred to SiO 2 using PMMA method Sample received in-situ 300 C / 3 hr vacuum anneal (P ~ 1x10-9 mbar) Height (nm) a) Transferred b) Annealed a) b) x ( m) x ( m) Pirkle, A, PhD Thesis, UT Dallas 2011 Height (nm) Photoelectron intensity (cts/s) Binding energy (ev) C 1s XPS states corresponding to PMMA are largely removed AFM shows a much smoother surface RMS roughness drops from 4.6 nm to 0.6 nm upon annealing 300 C / 3 hr vacuum anneal is effective for significant removal of PMMA residue from graphene x10 3 (f) (e) (g) (i) (h) (i) (h) (g) (e,f) (d) C 1s (c) (b) (a) c) Transferred CVD graphene on SiO C / 3 hr UHV anneal b) Transferred CVD graphene on SiO 2 a) CVD graphene on Cu 25

26 Standard ALD Al 2 O 3 processes (TMA/H 2 O) lead to non-uniform deposition at step edges Nucleation strategies for ALD on graphene Thin polymer (NFC CP) layer D. B. Farmer, et. al., Nano Lett. 9(12), 4474 (2009) 1 nm e-beam Al / oxidation in air S. Kim, et. al., APL 94, 6, (2009) Evaporated PTCDA M. Hersam et. al. NO 2 noncovalent functionalization D. B. Farmer and R. G. Gordon, Nano Lett. 6(4), 699 (2006) Y.-M. Lin, et. al., Nano Lett. 9(1), 422 (2009) Y. Xuan, et al., APL 92, (2008) O 3 functionalization B. Lee, et. al., APL 92(20), (2008) G. Lee, et. al., J. Phys.Chem. C 113(32), (2009) B. Lee, et. al., APL 97(4), (2010) LC April 11,

27 Standard ALD Al 2 O 3 processes (TMA/H 2 O) lead to non-uniform deposition of the dielectric on graphene Need to use a nucleation layer or functionalize the graphene surface E-beam Al and Ti provide nucleation centers for the ALD growth uniform coverage critical No detrimental effect on gate capacitance ultra-thin interfacial layer desirable 2.2nm Oxidized Al 1.1nm 0 2.2nm 0 150nm 150nm 300nm 300nm0 Oxidized Ti 1.1nm nm 300nm 0 150nm 300nm 27

28 Graphene/Pt (700C) followed by PVD Y in O 2 at RT Y 2 O 3 : hp10 a = nm Graphene: a = nm Yittrium oxide ML interacts weakly with graphene and is stable up to high temperatures. STM reveals that the yittria layer has a two-dimensional hexagonal lattice rotated by 30 degrees relative to the hexagonal graphene lattice. R. Addou et al. Nat Nanotech. 18, 41 (2013) 28

29 Large-band-gap semiconductor (~5-6 ev) Lattice parameter ~ 0.25 and interlayer distance of ~ 0.34nm Nearly lattice matched with graphene Dielectric constant ~ 4.1 Mechanical properties (Young s Modulus ~ GPa) Excellent lubricating properties High dielectric breakdown strength (~35kV/mm) Chemical and thermal stability Excellent thermal properties Atomically flat Flexoelectricity [Wittkowski et al. Thin Solid Films 353 (1999)] [Kudin et al, PRB 64, , (2002)] [Li et al, Nanotechnology 20 (2009)] [Naumov et al, PRL 102, (2009)] 29

30 MG grows on a monolayer h-bn layer in an incommensurate manner. The Intermediate h-bn layer possesses the bulk-like insulating properties, normal phonon sptectrum, and bulk lattice constants compared to h-bn/ni(111). h-bn: B 3 N 3 H 6 at 800 C Graphite/(ene): C 2 H 2 or Benzene at ~ 600 C Oshima et al SSC 116, (2000) Nagashima et al PRB 54 (19), 491 (1996) 30

31 Number of Layers [from TEM] Ni BN Precursors B 2 H 6 (5% in H 2 ):NH 3 = 1:18 T ~ 1025 C P ~ 135 mtorr 100 mm Growth Time [mins] Ismach et al. ACS Nano, 2012, 6 (7), pp

32 Ismach et al. ACS Nano, 2012, 6 (7), pp

33 ) (a) Ni/Au Ni/Au Parylene Parylene i/auni/au Graphene Graphene Ni/Au Ni/Au SiO 2 (90 SiOnm) 2 (90 nm) (b) (b) (a) (b) Nickel Parylene ~7 nm 5nm Graphene n++ Si n++ substrate Si substrate 1 um 5 nm SiO 2 Intensity cps (a) Without Parylene With Parylene Raman Shift (cm -1 ) Mordi et al., Appl. Phys. Lett. 100, (2012) Resistance [K] (b) 3 Mobility ~ 5000 cm 2 /V s nm V BG [V] V D = 10 mv 33

34 PI (a.u.) C 1s HOPG + 1 nm Ni+ 500 C / 10 min anneal (UHV) HOPG + 1 nm Ni HOPG BE ( ev ) XPS analysis indicates absence of carbide formation ( ~ 282 ev) at the Ni graphene interface I d - V d plots for Ni on graphene indicative of ohmic behavior higher total resistance (R) and hence higher R c 34

35 C 1s 1.0 PI (a.u.) HOPG + 1 nm Ni+ 500 C / 10 min anneal (UHV) HOPG + 1 nm Ni HOPG I d (ma) μm V bg (V) BE ( ev ) V d (V) XPS analysis indicates absence of carbide formation ( ~ 282 ev) at the Ni graphene interface I d - V d plots for Ni on graphene indicative of ohmic behavior higher total resistance (R) and hence higher R c 34

36 Exfoliated 2 CVD grown CVD Graphene R c (k) 1 old transfer new transfer TiO x process L c = 3mm W c (mm) ~ 5X reduction in R c observed Residue at interface significantly contributes to R c 35

37 The effect of plasma treatment of Gr/SiC on contact resistance 36

38 Large area polycrystalline graphene is now common place Millimeter size single crystal graphene can be routinely grown Large area h-bn thickness control is still challenging progress is being made Integration of thin dielectrics on graphene making progress high-k and low-k Progress is being made on contact resistance reduction still a lot of work to be done to meet the real device requirements Integration of h-bn/graphene shows excellent results Devices using exfoliated films show excellent transport properties Integrated processes at large scale still a big challenge 37

39