Quantum Mechanics ECE 663-1, Fall ‘08 Why do we need it? QM interference creates bandgaps and separates metals from insulators and semiconductors ECE 663-1, Fall ‘08 But how can electrons (particles) create interference? Lessons on http://hyperphysics.phy-astr.gsu.edu/hbase/quacon.html#quacon or http://www.colorado.edu/physics/2000/quantumzone/ECE 663-1, Fall ‘08 Solar system model of atom mv2/r = Zq2/4pe0r2 Centripetal force Electrostatic force Continuous radiation from orbiting electron Pb1: Atom would be unstable! (expect nanoseconds observe billion years!) Spectrum of Helium Transitions E0(1/n2 – 1/m2) (n,m: integers) Pb2: Spectra of atoms are discrete! ECE 663-1, Fall ‘08 Why waves? Only certain modes allowed (like a plucked string) nl = 2pr (fit waves on circle) Momentum ~ 1/wavelength (DeBroglie) p = mv = h/l (massive classical particles vanishing l) This means angular momentum is quantized mvr = nh/2p = nħ From 2 equations, rn = (n2/Z) a0 a0 = h2e0/pq2m = 0.529 Å (Bohr radius) ECE 663-1, Fall ‘08 Bohr’s suggestion E = mv2/2 – Zq2/4pe0r Using previous two equations En = (Z2/n2)E0 E0 = -mq4/8ħ2e0 = -13.6 eV = 1 Rydberg Transitions E0(1/n2 – 1/m2) (n,m: integers) Explains discrete atomic spectra So need a suitable Wave equation so that imposing boundary conditions will yield the correct quantized solutions ECE 663-1, Fall ‘08 What should our wave equation look like? ∂2y/∂t2 = v2(∂2y/∂x2) String y w x k Solution: y(x,t) = y0ei(kx-wt) w2 = v2k2 What is the dispersion (w-k) for a particle? ECE 663-1, Fall ‘08 What should our wave equation look like? Quantum theory: E=hf = ħw (Planck’s Law) p = h/l = ħk (de Broglie Law) and E = p2/2m + U (energy of a particle) w Thus, dispersion we are looking for is w k2 + U So we need one time-derivative and two spatial derivatives k 2y/∂tX 2 = v2(∂2y/∂x2) ∂X ECE 663-1, Fall ‘08 Wave equation (Schrodinger) iħ∂Y/∂t = (-ħ22/2m + U)Y Kinetic Energy Potential Energy Makes sense in context of waves Eg. free particle U=0 Solution Y = Aei(kx-wt) = Aei(px-Et)/ħ w k We then get E = p2/2m = ħ2k2/2m ECE 663-1, Fall ‘08 For all time-independent problems iħ∂Y/∂t = (-ħ22/2m + U)Y = ĤY Separation of variables for static potentials Y(x,t) = y(x)e-iEt/ħ Oscillating solution in time Ĥy = Ey, Ĥ = -ħ22/2m + U BCs : Ĥyn = Enyn (n = 1,2,3...) En : eigenvalues (usually fixed by BCs) yn(x): eigenvectors/stationary states ECE 663-1, Fall ‘08 What does it all mean? ECE 663-1, Fall ‘08 What does Y(x,t) represent? • Probability amplitude of finding particle at x at time t (Like electric field in phasor notation E, a complex #) Y itself hard to measure so overall phase irrelevant • Probability density P(x,t) = y*(x,t)y(x,t) (Like intensity E*E easier to detect) • Must have ∫P(x,t)dx = 1 at all times • Charge density r(x,t) = qP(x,t) • Current density J(x,t) = qvP(x,t) (“v” to be defined later) ECE 663-1, Fall ‘08 Averages O = ∫|y(x,t)|2O(x)dx O = ∫y*(x,t)Ô(x)y(x,t)dx Symmetrized! One can show that averages follow ‘classical rules’ Dx.Dp ≥ ħ/2 (Uncertainty Principle) d<x>/dt = <v> md<v>/dt = <F> l ECE 663-1, Fall ‘08 Examples of meaningful averages n = Y*Y (Electron Density) ˆ v=ˆ p/m = -iħ/m (check KE) J = q<v> = q∫y*(x)[-iħ/m]y(x)dx (Symmetrize !!!) ˆ J = iqħ/2m(YY*/x – Y*Y/x) For plane waves Y = Y0eikx J = n(ħk/m) = nqv Also check: n/t + J/x = 0 (charge conservation) ECE 663-1, Fall ‘08 Relation between Y and yn ? yn s are allowed solutions (like mode shapes of a fixed string) Their energies (‘frequencies’) are the eigenvalues En Aside: They are orthogonal (independent), like modes of a string ∫y*n(x)ym(x)dx = 0 if n ≠ m and normalized ∫y*n(x)yn(x)dx = 1 Y shows the actual solution (superposition of allowed ones) In general, Y(x,t) = Sn an yn(x)e-iEnt/ħ ECE 663-1, Fall ‘08 Why are levels quantized as Bohr suggested? Due to confinement, like acoustic waves on a string Same for H-atom Vacuum U(r) = -Zq2/4pe0r -0.85 eV -1.51 eV -3.4 eV -13.6 eV ECE 663-1, Fall ‘08 Simplest eg of a confined system U= U= Particle in a Box U=0 y eikx doesn’t satisfy boundary conditions (should be zero at both ends) But can superpose allowed solutions y = Asin(kx) is zero at x = 0 What about x = L ? ECE 663-1, Fall ‘08 Particle in a box U= U= y = Asin(kx) is zero at x = L only for special values of k knL = np (n = 1, 2, 3, …) U=0 Quantization condition knL = np (n = 1, 2, 3, …) ie, L = nln/2 (exactly like acoustic waves) ECE 663-1, Fall ‘08 Particle in a box U= w yn = Asin(knx) knL = np (n = 1, 2, 3, …) U= E3 E2 E1 k k 1 k2 k3 U=0 Fixed k’s give fixed E’s En = ħ2kn2/2m = ħ2n2p2/2mL2 Coeff A fixed by normalization A=√2/L Full Solution Yn = 2/L sin(npx/L) exp(-iħn2p2/2mL2 t) ECE 663-1, Fall ‘08 Particle in a box Yn=` 2/L sin(npx/L) exp(-iħn2p2/2mL2 t) w En = ħ2kn2/2m = ħ2n2p2/2mL2 E3 E2 E1 k k1 k2 k3 Find J J = iqħ/2m(YY*/x – Y*Y/x) ECE 663-1, Fall ‘08 Extracting Physics from Pictures !! U= U= Ground State y1(x) Smoothest curve with no kinks Next mode y2(x) should also minimize energy but be orthogonal to the first mode (since modes must be independent!) U=0 Hence the single kink! Next one y3(x) must be orthogonal to the other two Since Kinetic energy ~ ∂2y/∂x2 Lowest energy wavefunction must have smallest curvature It must also vanish at ends and be normalized to unity ECE 663-1, Fall ‘08 Extracting Physics from Pictures !! U =U= UU== U=0 If we decrease box size, but keep area under modes same, then each mode must peak more This increases curvature and thus energy levels and their separation Notice this from exact results, DEn 1/L2 U=0 (Uncertainty: Localizing particle increases its energy!) Small boxes (atoms), energies discrete and well separated Large boxes (metal contacts), they are bunched up ECE 663-1, Fall ‘08 Constant, non-zero potential U= knL = np still U= U = U0 But dispersion kn = √2m(En-U0)/ħ2 ECE 663-1, Fall ‘08 Finite potential walls: thinking ‘outside the box’ U = U0 U = U0 22 exp(±ik’x),kk’= =2m(U exp(±kx), 2m(E-U 0-E)/ħ 0)/ħ U=0 Asinkx + Bcoskx, k = 2mE/ħ2 Solve piece-by-piece, and match boundary condns (Match y, dy/dx) Wavefunction penetrates out (“Tunneling”) ECE 663-1, Fall ‘08 Particle at a Step U0 x 0 y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 Boundary Conditions: y(0-) = y(0+) dy/dx|0- = dy/dx|0+ k = 2mE/ħ2 k’ = 2m(E-U0)/ħ2 E > U0 k,k’ real ECE 663-1, Fall ‘08 Particle at a Step U0 0 x y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 Boundary Conditions: 1+r=t k(1-r) = k’t t = 2k/(k+k’) r = (k-k’)/(k+k’) ECE 663-1, Fall ‘08 Particle at a Step U0 x 0 y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 Transmission = curr transmitted/curr incident Reflection Coeff = curr reflected/curr incident ECE 663-1, Fall ‘08 Particle at a Step U0 0 x y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 Free particle J = ħk/m|y0|2, y0: amplitude (More correctly, J = Re(ħk/m|y0|2), in case k is complex) Thus, T = Re(k’|t|2/k), R = |r|2 ECE 663-1, Fall ‘08 Particle at a Step U0 0 x Y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 T = Re(k’|t|2/k) = 4kk’/(k+k’)2 R = |r|2 = (k-k’)2/(k+k’)2 T+R=1 t = 2k/(k+k’) r = (k-k’)/(k+k’) ECE 663-1, Fall ‘08 Repeat for E < U0 U0 x 0 y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 k = 2mE/ħ2 k’ = 2m(E-U0)/ħ2 E < U0 k’ imaginary = ih ECE 663-1, Fall ‘08 Repeat for E < U0 U0 0 y(x) = eikx + re-ikx , x < 0 = teik’x, x > 0 T = Re(ih|t|2/k) = 0 R = |r|2 = |k-ih|2/|k+ih|2 = 1 Expected? x k = 2mE/ħ2 h = 2m(U0-E)/ħ2 t = 2k/(k+ih) r = (k-ih)/(k+ih) ECE 663-1, Fall ‘08 Summary: Particle at a step U0 x 0 T Classical 1 Quantum 0 U0 E $$$ Question: How do we make the curves approach each other? ECE 663-1, Fall ‘08 Particle at a Barrier eikx + re-ikx Aeik’x + Be-ik’x 0 teikx U0 L x Boundary Conditions: y(0-) = y(0+) dy/dx|0- = dy/dx|0+ k = 2mE/ħ2 y(L-) = y(L+) dy/dx|L- = dy/dx|L+ E > U0 k,k’ real k’ = 2m(E-U0)/ħ2 ECE 663-1, Fall ‘08 Particle at a Barrier eikx + re-ikx Aeik’x + Be-ik’x 0 1+ r = A + B k(1-r) = k’(A-B) Aeik’L + Be-ik’L = teikL k’(Aeik’L – Be-ik’L) = kteikL U0 L teikx x 2 = (1+k’/k)A + (1-k’/k)B 0 = (1-k’/k)Aeik’L + (1+k’/k)Be-ik’L ECE 663-1, Fall ‘08 Particle at a Barrier eikx + re-ikx Aeik’x + Be-ik’x 0 teikx U0 L x A = 2e-ik’L(1+k’/k)/[(1+k’/k)2e-ik’L– (1-k’/k)2eik’L] B = 2eik’L(1-k’/k)[(1-k’/k)2eik’L-(1+k’/k)2e-ik’L] 2 = (1+k’/k)A + (1-k’/k)B 0 = (1-k’/k)Aeik’L + (1+k’/k)Be-ik’L ECE 663-1, Fall ‘08 Particle at a Barrier eikx + re-ikx Aeik’x + Be-ik’x 0 teikx U0 L x A = 2e-ik’L(1+k’/k)/[(1+k’/k)2e-ik’L– (1-k’/k)2eik’L] B = 2eik’L(1-k’/k)[(1-k’/k)2eik’L-(1+k’/k)2e-ik’L] teikL = 2(2k’/k)/ [(1+k’/k)2e-ik’L– (1-k’/k)2eik’L] = 2kk’/[-i(k2+k’2)sink’L + 2kk’cosk’L] ECE 663-1, Fall ‘08 Particle at a Barrier eikx + re-ikx Aeik’x + Be-ik’x 0 teikx U0 L x T = 4k2k’2/[(k2+k’2)2sin2k’L + 4k2k’2cos2k’L] Resonances: Maximum if cosk’L = 1, sink’L = 0 ie, k’L = 0, p, 2p, ... L = 0, l/2, 3l/2, .... teikL = 2(2k’/k)/ [(1+k’/k)2e-ik’L– (1-k’/k)2eik’L] = 2kk’/[-i(k2+k’2)sink’L + 2kk’cosk’L] ECE 663-1, Fall ‘08 More obvious if it’s a well eikx + re-ikx Aeik’x + Be-ik’x 0 teikx U0 L x ie, k’L = 0, p, 2p, ... L = 0, l/2, 3l/2, .... Represent Resonances, with E > U0 (They would be bound states if E < U0) Here k = 2m(E-U0)/ħ2 k’ = 2mE/ħ2 ECE 663-1, Fall ‘08 Back to Barrier but lower Energy k’ = ik eikx + re-ikx Ae-kx + Bekx 0 U0 L teikx x T = 4k2k2/[(k2-k2)2sinh2kL + 4k2k2cosh2kL] Large or wide barriers: kL >> 1, sinh(kL) ~ cosh(kL) ~ ekL/2 T ≈ 16 k2k2e-2kL/(k2+k2)2 ~ [16E(U0-E)/U02]e-2kL teikL = 2ikk/[(k2-k2)sinhkL + 2ikhcoshkL] ECE 663-1, Fall ‘08 Back to Barrier but lower Energy eikx + re-ikx Ae-kx + Bekx 0 U0 L teikx x T ≈ [16E(U0-E)/U02]e-2kL Even though E < V0, T > 0 (tunneling) U(x) E x1 x2 More generally, WKB approximation x2 T ~ exp[-2∫dx 2m[U(x)-E]/ħ2] x1 ECE 663-1, Fall ‘08 Example: Tunneling x2 T ~ exp[-2∫dx 2m[U(x)-E]/ħ2] x1 Well Barrier • • • • Alpha particle decay from nucleus Source-Drain tunneling in MOSFETs Single Electron Tunneling Devices (SETs) Resonant Tunneling Devices (RTDs) ECE 663-1, Fall ‘08 Example: Tunneling Quantum states (Speer et al, Science ’06) Needed for designing Heterojunctions/superlattices/ Photonic devices, etc Upswing in Current due To Tunneling Gloos ECE 663-1, Fall ‘08 Barrier problem: Summary U0 L 0 T 1 x Classical Quantum 0 Tunneling T ~ e-2kL, k ~ (U0-E) U0 E Resonances k’L = np, E = U0 + ħ2k’2/2m ECE 663-1, Fall ‘08 Matlab plots As barrier width increases, we recover particle on a step ECE 663-1, Fall ‘08 Matlab code • • • • • • • • • • • • • • • • subplot(2,2,4); % vary this from plot window to plot window m=9.1e-31;hbar=1.05e-34;q=1.6e-19; L=1e-9; %m, vary this from plot window to plot window! U0=1; %Volts Ne=511;E=linspace(0,5,Ne); k=sqrt(2*m*E*q/hbar^2);%/m eta=sqrt(2*m*(E-U0)*q/hbar^2);%/m T=4.*k.^2.*eta.^2./((k.^2+eta.^2).^2.*sin(eta.*L).*sin(eta.*L) + 4.*k.^2.*eta.^2.*cos(eta.*L).*cos(eta.*L)); plot(E,T,'r','linewidth',3) title('L = 15 nm','fontsize',15) % vary this from plot window to plot window! grid on hold on tcl=2.*k./(k+eta);tcl=tcl.*conj(tcl); Tcl=real((eta./k).*tcl); plot(E,Tcl,'k--','linewidth',3) gtext('step','fontsize',15) ECE 663-1, Fall ‘08 Can we solve for arbitrary Potentials? Approximation Techniques Graphical solutions (e.g. particle in a finite box) Special functions (harmonic oscillator, tilted well, H-atom) Perturbation theory (Taylor expansion about known solution) Variational Principle (assume functional form of solution and fix parameters to get minimum energy) Numerical Techniques (next) ECE 663-1, Fall ‘08 Finite Difference Method y yn-1 xn-1 y= yn-1 yn yn+1 Uy = One particular mode yn y n+1 xn xn+1 Un-1 Un-1yn-1 Unyn Un+1yn+1 = yn-1 Un Un+1 = [U][y] yn yn+1 ECE 663-1, Fall ‘08 What about kinetic energy? y yn-1 xn-1 y= yn-1 yn yn y n+1 xn xn+1 (dy/dx)n = (yn+1/2 – yn-1/2)/a (d2y/dx2)n = (yn+1 + yn-1 -2yn)/a2 yn+1 ECE 663-1, Fall ‘08 What about kinetic energy? y yn-1 xn-1 y= yn-1 yn yn y n+1 xn xn+1 -ħ2/2m(d2y/dx2)n = t(2yn - yn+1 - yn-1) yn+1 t = ħ2/2ma2 ECE 663-1, Fall ‘08 What about kinetic energy? y yn-1 xn-1 yn y n+1 xn xn+1 -ħ2/2m(d2y/dx2)n = t(2yn - yn+1 - yn-1) y= yn-1 -t 2t -t yn yn+1 Ty = -t 2t -t -t 2t -t yn-1 yn yn+1 ECE 663-1, Fall ‘08 What about kinetic energy? y yn-1 xn-1 y= yn-1 yn yn y n+1 xn xn+1 [H] = [T + U] yn+1 ECE 663-1, Fall ‘08 What next? y yn-1 xn-1 yn y n+1 xn xn+1 Now that we’ve got H matrix, we can calculate its eigenspectrum >> [V,D]=eig(H); % Find eigenspectrum >> [D,ind]=sort(real(diag(D))); % Replace eigenvalues D by sorting, with index ind >> V=V(:,ind); % Keep all rows (:) same, interchange columns acc. to sorting index (nth column of matrix V is the nth eigenvector yn plotted along the x axis) ECE 663-1, Fall ‘08 Particle in a Box Results agree with analytical results E ~ n2 Finite wall heights, so waves seep out ECE 663-1, Fall ‘08 Add a field ECE 663-1, Fall ‘08 Or asymmetry Incorrect, since we need open BCs which we didn’t discuss ECE 663-1, Fall ‘08 Harmonic Oscillator Shapes change from box: sin(px/L) exp(-x2/2a2) Need polynomial prefactor to incorporate nodes (Hermite) E~n2 for box, but box width increases as we go higher up Energies equispaced E = (n+1/2)ħw, n = 0, 1, 2... ECE 663-1, Fall ‘08 Add asymmetry ECE 663-1, Fall ‘08 Matlab code • • • • • t=1; Nx=101;x=linspace(-5,5,Nx); %U=[100*ones(1,11) zeros(1,79) 100*ones(1,11)];% Particle in a box U=x.^2;U=U;%Oscillator %U=[100*ones(1,11) linspace(0,5,79) 100*ones(1,11)];%Tilted box • • • • • • • %Write matrices T=2*t*eye(Nx)-t*diag(ones(1,Nx-1),1)-t*diag(ones(1,Nx-1),-1); %Kinetic Energy U=diag(U); %Potential Energy H=T+U; [V,D]=eig(H); [D,ind]=sort(real(diag(D))); V=V(:,ind); • • • • • • • • % Plot for k=1:5 plot(x,V(:,k)+10*D(k),'r','linewidth',3) hold on grid on end plot(x,U,'k','linewidth',3); % Zoom if needed axis([-5 5 -2 10]) ECE 663-1, Fall ‘08 Grid issues For Small energies, finite diff. matches exact result Deviation at large energy, where y varies rapidly Grid needs to be fine enough to sample variations ECE 663-1, Fall ‘08 Summary • Electron dynamics is inherently uncertain. Averages of observables can be computed by associating the electron with a probability wave whose amplitude satisfies the Schrodinger equation. • Boundary conditions imposed on the waves create quantized modes at specific energies. This can cause electrons to exhibit transmission ‘resonances’ and also to tunnel through thin barriers. • Only a few problems can be solved analytically. Numerically, however, many problems can be handled relatively easily. ECE 663-1, Fall ‘08