Analysis of Strain Effect in Ballistic Carbon Nanotube FETs Nov. 30, 2006 Youngki Yoon Dept. of Electrical & Computer Engineering University of Florida
Outline Carbon nanotube field-effect transistor Uniaxial strain on CNTs Material properties of strained CNTs Strain effect on Eg Strain effect on band-structure-limited velocity Simulated device structure & approach Simulation results I-V characteristics Strain effect on Imin Strain effect on Ion Strain effect on intrinsic delay Concluding remarks
What is CNTFET? G S D CNTFET Conventional MOSFET CNTFET with doped source drain extentions CNTFET with metal source drain contacts
Why strained CNTs? J. Cao et al., PRL (2003) (a) Tensile uniaxial strain and (b) compressive uniaxial strain on the channel of a CNTFET. Conductance is change by several orders of magnitude Sentitivity change is available. T. Tombler et al., Nature (2000)
Let’s apply uniaxial strain! (16,0) CNT Band gap is increased (Egh=0.33eV to 0.44eV). Slope (band-structrue-limited velocity) is decreased.
Strain effect on CNTs (Variation of Eg and band-structure-limited velocity) Tensile strain Eg of (16,0) CNT ↑ (n=3q+1 group) Eg of (17,0) CNT ↓ (n=3q+2 group) Compressive strain Eg of (16,0) CNT ↓ (n=3q+1 group) Eg of (17,0) CNT ↑ (n=3q+2 group) Eg vs. uniaxial strain strength Band-structure-limited velocity: Tensile strain B.S.L. vel. of (16,0) CNT ↓ (n=3q+1 group) B.S.L. vel. of of (17,0) CNT ↑ (n=3q+2 group) Compressive strain B.S.L. vel. of of (16,0) CNT ↑ (n=3q+1 group) B.S.L. vel. of of (17,0) CNT ↓ (n=3q+2 group) The lowest subbands of (16,0) CNTs. Solid lines: unstrained (16,0) CNT. Dashed lines: 2% strained CNT.
Device structure & approach Coaxially gated Schottky Barrier CNTFET ( ) 3nm HfO2 gate oxide with a dielectric constant of 16 40nm strained (16,0) and (17,0) CNT channel 0.4V power supply Approach Self-consistent NEGF formalism with Poisson equation Mode space approach M Gate Strained CNT Device structure
Mode space approach Real space approach Mode space approach A part of (n,0) zigzag nanotube lattice in real space Mode space approach (n,0) ZNT is decoupled into n one-dimensional mode space lattice. Mode space lattice
ID-VG characteristics (16,0) CNTFET w/ uniaxial strain (17,0) CNTFET w/ uniaxial strain Device characteristics strongly depend on the band gap of the channel material. ID-VG characteristics change significantly with even a small strain.
Strain effect on Imin Imin ≡ minimum current delivered (VG=0.2V) Main figure Solid line: (16,0) CNTFET Dashed line: (17,0) CNTFET Subset: band profile vs. channl position at VG=0.2V Solid line: unstrained (16,0) CNTFET Dashed line: 2% strained (16,0) CNTFET Imin ≡ minimum current delivered (VG=0.2V) A simple estimation for Imin
Strain effect on Ion Ion ≡ current at VG=Von=Voff+VDD , Ioff VDD=0.4V Solid line: (16,0) CNTFET Dashed line: (17,0) CNTFET (16,0) CNTFET w/ uniaxial strain Ion ≡ current at VG=Von=Voff+VDD , where Voff is the voltage at Ioff=10-7A. unstrained 2% unstrained 2% uniaxial (16,0) CNTFET at on-state
Strain effect on intrinsic delay (16,0) CNTFET w/ uniaxial strain (17,0) CNTFET w/ uniaxial strain ON OFF Quantum reflection 0% 2% 0% 2% Lowest conduction band of (16,0) CNT Ec vs. X for the same Ion/Ioff
Summary Two important material property changes after applying uniaxial strains: Eg Band-structure-limited velocity Nominal device & approach Coaxially gated CNT SBFET with half band gap SB height Self-consistent NEGF with Poisson’s eq. Mode space approach Results I-V characteristics are changed a lot with even a small strain strength. Imin , Ion , and intrinsic delay are affected by Eg and B.S.L velocity changes. Strain engineering can be effectively used to tune up the device performance, but trade-off should be carefully considered. Thank you