SUPPEMENTAL INFORMATION Tapered Silicon Nanowires for Enhanced Nanomechanical Sensing O. Malvar1, E. Gil-Santos1, J.J. Ruz1, D. Ramos2, V. Pini1, M. Fernandez-Regulez3, M. Calleja1, J. Tamayo1 and A. San Paulo1,3 Tapered Si NW Growth: Si NWs were grown in a CVD reactor at 800 ºC and atmospheric pressure with 10% H2/Ar as both diluent and carrier gas. A fixed flow rate of 270 s.c.c.m. was used for the diluent gas line (direct to growth chamber). The carrier gas was passed through a refrigerated (0ºC) liquid SiCl4 bubbler in order to deliver the precursor gas to the growth chamber. Variations in the carrier gas flow rate in the range 40-90 s.c.c.m. were introduced in different growth sessions in order to obtain NWs with different degree of tapering. Au colloidal nanoparticles (NP) with a diameter of 50 and 80 nm where used as metal catalyst. Theoretical model The Euler-Bernoulli equation for a beam with non-uniform cross section is given by ππ(π₯) ∂2 π€π (π₯,π‘) ∂π‘ 2 ∂2 + ∂π₯ 2 ( πΈπΌ(π₯) ∂2 π€π (π₯,π‘) πΏ4 ∂π₯ 2 )=0 (1) where π€π (π₯, π‘) is the ππ‘β order mode contribution to the nanowire displacement at the normalized longitudinal position π₯ and time t (actual position π is normalized to the NW length, π₯ = π/πΏ), and π(π₯) and πΌ(π₯) are the cross-section area and second moment of area, 1 respectively and both depend on the longitudinal position π₯ in order to account for the tapered geometry. Assuming a harmonic oscillator response for each mode, the solution to eq. (1) can be written as: π€π (π₯, π‘) = π€π (π₯)π ππππ‘ (2) where ππ = 2πππ is the angular resonance frequency of the ππ‘β mode. Then, the temporal and spatial parts of the equation can be separated so that the spatial part of the equation can be writen as: π2 πΈπΌ(π₯) π2 π€π (π₯) −ππ(π₯)ππ2 π€π (π₯) + ππ₯ 2 ( πΏ4 ππ₯ 2 )=0 (3) A general solution for this equation for any tapered beam-like structure can be obtained if π(π₯) and πΌ(π₯) are expressed as: π(π₯) = π0 (1 − πΌπ₯)π πΌ(π₯) = πΌ0 (1 − πΌπ₯)π+2 (4) (5) where π0 and πΌ0 are the area and second moment of area at the clamp. In the case of a tapered nanowire which diameter decreases linearly from the clamp to the free-end, the radius can be written as: π (π₯) = π 0 (1 − πΌπ₯) (6) where πΌ is the tapering parameter given by 2 πΌ= π 0 −π πΉ π 0 (7) and equations (4) and (5) are written using π = 2. After these considerations, the beam equation can be rewritten as: −ππ4 π€π (π₯) + 12πΌ 2 π2 π€π (π₯) ππ₯ 2 + (1 − πΌπ₯) (−8πΌ π3 π€π (π₯) ππ₯ 3 + (1 − πΌπ₯) π4 π€π (π₯) ππ₯ 4 )=0 (8) where ππ is related to the resonance frequency of the ππ‘β order mode by: π 4 πΈπΌ ππ2 = πππΏ4 π0 0 (9) It can be shown that the solution to equation (8) can be expressed as: 4 π€π (π) = πΌ2 π 2 [π΄π πΎ2 (ππ π) + π΅π π½2 (ππ π) + πΆπ π2 (ππ π) + π·π πΌ2 (ππ π)] (10) where π½2 is the Bessel function of first specie and second order, π2 is the modified Bessel function of second order and specie, πΌ2 is the modified Bessel function of first specie and second order and πΎ2 is the modified Bessel function of second specie and second order. The variable π is related to π₯ by: π= 2(1−πΌπ₯)1/2 πΌ (11) 3 and this expression is only valid for πΌ > 0, i.e for a nanowire with a radius that decreases from the clamp to the free-end. The constants π΄π , π΅π , πΆπ , π·π and ππ must be calculated by imposing the boundary conditions for a singly clamped beam which are zero slope and deflection at the clamp and zero bending moment and shear force at the free end. Applying these boundary conditions the following system of equations is obtained: π΄π πΎ2 ( 2ππ −π΄π πΎ3 ( 2ππ ) + π΅π π½2 ( πΌ 2ππ πΌ ) + π΅π π½3 ( 2 4(1−πΌ)ππ π΄π (√1 − πΌ ( π΅π (√1 − πΌ ( πΌ2 2 4(1−πΌ)ππ πΌ2 2 4(1−πΌ)ππ πΆπ (√1 − πΌ ( πΌ2 2 4(1−πΌ)ππ π·π (√1 − πΌ ( π΄π (( πΌ2 2ππ ) + πΆπ π2 ( πΌ 2ππ πΌ ) − πΆπ π3 ( + 24) πΎ0 ( − 24) π½0 ( πΌ − 24) π0 ( πΌ2 π·π ( πΌ2 πΌ )+ −( πΌ )=0 4(1−πΌ)π2 π +6)πΎ (2ππ √1−πΌ) 1 πΌ πΌ2 4πΌ( )+ ππ 4(1−πΌ)π2 π −6)π½ (2ππ √1−πΌ) 1 πΌ πΌ2 4πΌ( )+ ππ 4(1−πΌ)π2 π −6)π (2ππ √1−πΌ) 1 πΌ πΌ2 4πΌ( )+ ππ 4(1−πΌ)π2 π +6)πΌ (2ππ √1−πΌ) 1 πΌ πΌ2 4πΌ( πΌ )=0 ππ 2ππ √1−πΌ )+ − (− + 384) π½1 ( πΌ2 πΌ )+ )+ πΌ 4(1−πΌ)π2 π −72) πΌ2 2ππ √1−πΌ (12) 4(1−πΌ)π2 π +16)πΎ (2ππ √1−πΌ) 24√1−πΌππ ( 0 πΌ πΌ2 2( 4(1−πΌ)ππ + 384) π1 ( 4(1−πΌ)π2 π +16)πΌ (2ππ √1−πΌ) 0 πΌ πΌ2 24√1−πΌππ ( πΌ )− + 384) πΎ1 ( πΌ 4(1−πΌ)π2 π −72) πΌ2 πΌ 2ππ √1−πΌ + 24) πΌ0 ( )+ )+ 2ππ √1−πΌ 4(1−πΌ)π2 π −16)π½ (2ππ √1−πΌ) 0 πΌ πΌ2 πΆπ (( πΌ )=0 πΌ 2ππ ) − π·π πΌ3 ( 2ππ √1−πΌ 2 2 (4(1−πΌ)ππ +72) 4(1−πΌ)ππ πΌ2 2( 4(1−πΌ)ππ πΌ 2ππ √1−πΌ 24√1−πΌππ ( π΅π ( 2ππ ) + π·π πΌ2 ( πΌ 2ππ 2ππ √1−πΌ πΌ )) + 4(1−πΌ)π2 π −16)π (2ππ √1−πΌ) 0 πΌ πΌ2 24√1−πΌππ ( )+ πΌ 4(1−πΌ)π2 π +72) πΌ2 πΌ2 2( 4(1−πΌ)ππ 2ππ √1−πΌ − 384) πΌ1 ( πΌ )) = 0 Since this system of equations is homogeneous, in order to have a non-trivial solution the determinant of the matrix of the system must be zero, condition from which the eigenvalue equation is obtained. It is impontant to notice that every eigenvalue ππ obtained by 4 numerically solving the eigenvalue equation depends on πΌ. This dependence is plotted on figure 1 for the first two modes of vibration. We can provide a simple analytical expression for the resonance frequencies by fitting the numerical solutions for ππ to polynomial functions of πΌ. The two polynomial functions obtained from such fitting are: π1 (πΌ) = 1.150πΌ 4 − 1.115πΌ 3 + 0.697πΌ 2 + 0.336πΌ + 1.875 π2 (πΌ) = 21.337πΌ 7 − 59.437πΌ 6 + 66.091πΌ 5 − 36.654πΌ 4 + 10.501πΌ 3 − 1.512πΌ 2 − 0.440πΌ + 4.694 (13) Interestingly, for 0<α<0.9, the ratio of the square of these polynomial functions, i.e f2/f1 , gives a linear dependence on α. Figure 2 shows the linear fitting of this ratio with α. The ratio f2/f1 can thus be approximated by the simple expression: π2 π1 ≈ 6.267 − 4.103πΌ (11) After imposing the determinant to be zero, the system of equations 12 is undetermined. This means that the coefficients π΅π , πΆπ and π·π will be proportional to π΄π and thus our new system will consists of just three equations, but this time the system will not be homogeneous. Once we solve the new system of equations we have the values of π΅π , πΆπ and π·π and thus we have the complete expressions for the vibration mode shapes given by eq (10). 5 Figure 1. Numerical solution (red symbols) and polynomial fitting (blue line) for the eigenvalues kn for the first (left) and second (right) modes. The fitting are used to provide simple analytical expressions for the resonance frequencies in the range. 6 Figure 2. Numerical solution (red symbols), analytical approach (red line) from eq (13) and linear fitting for the ratio f2/f1 in the range 0<ο‘<0.9. 7