Quantum Measurements and Back Action (Spooky and Otherwise) SM Girvin Yale University Thanks to Michel, Rob, Michael, Vijay, Aash, Simon, Dong, Claudia 1 for discussions and comments on Les Houches notes. Quantum Back Action is a Weird Thing CNOT gate control qubit target qubit c t c t truth table 0 0 0 0 0 1 0 1 1 0 1 1 1 1 1 0 Target is affected but control is not Are you sure??? 2 Quantum Back Action is a Weird Thing CNOT gate control qubit 0 −1 0 +1 target qubit 0 −1 0 −1 The control qubit is flipped from ← to 1 + Z1 1 − Z1 CNOT = X2 + I2 2 2 → !! 3 It’s all about the measurement: Stern Gerlach Experiment Silver atom has magnetic moment due to the electron ‘spin’ Silver atom is a qubit. r r V = µ gB r r F = −∇V Magnetic moment (spin) can point in any direction and can be measured by passing 4 the atom through a magnetic field gradient. Stern Gerlach Experiment: Quantum Measurement Silver atom has magnetic moment which can point in any direction, and yet…. (Measured) 5 http://hyperphysics.phy-astr.gsu.edu DISPERSIVE READOUT qubit + readout pulses Qubit + resonator (MHD) AMP dispersive shi8 χ Readout amplitude 1 0 fr f width κ 90 Readout phase (deg) 1 θ -­‐90 fr 0 f MEASUREMENT HISTOGRAM (data from Devoret lab) (unlike Stern-Gerlach, the qubit is in nearly a pure state) 0 1 Counts 2000 -­‐2 0 2 Stern & Gerlach did not measure spin! They entangled spin with position and measured position. (Measured) 8 http://hyperphysics.phy-astr.gsu.edu More precisely, they used a spin-dependent force to entangle spin with momentum and waited for momentum to turn into position. (Measured) 9 http://hyperphysics.phy-astr.gsu.edu This is a measurement of EM modes that were entangled with the qubit ‘spin’ 0 1 Counts 2000 -­‐2 0 2 1D toy model with spin-dependent impulsive force X Bz 2 p H ( p , x, t ) = − hk0 xσ z δ (t ) 2m F ( x, t ) = + hk0 σ z δ (t ) [ H , σ z ] = 0 ⇒ QND ‘Quantum Non-Demolition’ (QND) Measurements are Repeatable Z Z First result is random, rest are repeats. Z Z measurement measurement Z measurement 12 1D toy model with spin-dependent impulsive force X Bz We will measure momentum just after impulse rather than waiting for it to turn into position. 2 p H ( p , x, t ) = − hk0 xσ z δ (t ) 2m F ( x, t ) = + hk0 σ z δ (t ) 2 p H ( p , x, t ) = − hk0 xσ z δ (t ) 2m Input product state ( ) ψ ( x, t = 0 − ) = a ↑ + b ↓ Φ ( x ) Output entangled entangled state ( ψ ( x, t = 0+ ) = aeik x Φ( x) ↑ + be −ik x Φ ( x) ↓ 0 0 ) Output state in momentum basis ( ψ [k , t = 0+ ] = aΦ[k − k0 ] ↑ + bΦ[k + k0 ] ↓ ) 14 0.8 P(k |↓) P(k |↑) 0.6 Strong measurement 0.4 0.2 -4 -2 2 4 k (gaussian input packet) 0.8 P(k |↓) P(k |↑) 0.6 Weak measurement 0.4 0.2 -4 -2 k 2 4 15 0.8 P(k |↓) P(k |↑) 0.6 0.4 0.2 -4 -2 k 2 4 P(k |↑) is easy to understand but what we need is: P(↑| k ) 16 Practice on two continuous variables P ( x , y ) = ψ ( x, y ) y 2 Y x P ( x, y ) = P ( x | y ) P ( y ) ψ ( x, Y ) 2 P ( x, Y ) P( x | Y ) = = 2 P ( Y ) ʹ′ ʹ′ dx ψ ( x , Y ) ∫ 17 Density Matrix Equivalent Reduced density matrix for spin conditioned on measurement of momentum k ρ k k ρ k ρk = = Tr k ρ k P(k ) k ρ k ρk Full density matrix projected onto observed momentum state Reduced density matrix for spin given observed value of momentum 18 Density Matrix Equivalent Full state ψ = a ↑ Φ↑ + b ↓ Φ↓ ⎛ a Φ ↑ ⎞ = ⎜ ⎟ ⎜ b Φ ⎟ ↓ ⎠ ⎝ Full density matrix ⎛ aa * Φ ↑ Φ ↑ ρ = ψ ψ = ⎜ ⎜ a * b Φ Φ ↓ ↑ ⎝ ab * Φ ↑ Φ ↓ ⎞ ⎟ bb * Φ ↓ Φ ↓ ⎟⎠ 19 Reduced density matrix for spin conditioned on measurement of momentum k ρ k k ρ k ρk = = Tr k ρ k P(k ) Full density matrix ⎛ aa * Φ ↑ Φ ↑ ρ = ⎜ ⎜ a * b Φ Φ ↓ ↑ ⎝ ab * Φ ↑ Φ ↓ ⎞ ⎟ bb * Φ ↓ Φ ↓ ⎟⎠ Reduced density matrix for spin conditioned on measurement of momentum 1 ⎛ aa * k Φ ↑ Φ ↑ k ρk = ⎜ P(k ) ⎜⎝ a * b k Φ ↓ Φ ↑ k ab * k Φ ↑ Φ ↓ k ⎞ ⎟ bb * k Φ ↓ Φ ↓ k ⎟⎠ 20 Reduced density matrix for spin conditioned on measurement of momentum 1 ⎛ aa * k Φ ↑ Φ ↑ k ρk = ⎜ P(k ) ⎜⎝ a * b k Φ ↓ Φ ↑ k ab * k Φ ↑ Φ ↓ k ⎞ ⎟ bb * k Φ ↓ Φ ↓ k ⎟⎠ Easy to verify conditional state is pure: Det ρ k = 0 Tr ρ k = 1 ⎛ 1 0 ⎞ ⇒ ∃ exists a basis s.t. ρ k = ⎜ ⎟ ⎝ 0 0 ⎠ If we fully measure the state of the ‘bath’ then the conditional state remains pure! 21 Reduced density matrix for spin conditioned on measurement of momentum 1 ⎛ aa * k Φ ↑ Φ ↑ k ρk = ⎜ P(k ) ⎜⎝ a * b k Φ ↓ Φ ↑ k ab * k Φ ↑ Φ ↓ k ⎞ ⎟ bb * k Φ ↓ Φ ↓ k ⎟⎠ Easy to verify conditional state is pure: Det ρ k = 0 Sanity check: ⎛ aa * ρ = ∫ dkP(k ) ρ k = ⎜ ⎜ a * b Φ Φ ↑ ↓ ⎝ ab * Φ ↓ Φ ↑ ⎞ ⎟ ⎟ bb * ⎠ Averaging over measurement results yields measurement-induced dephasing 22 If the state is pure, there is a corresponding wave function for the spin alone ψk = 1 a ↑ k Φ↑ + b ↓ k Φ↓ P(k ) { ‘Spooky’ back action: [H ,σ z ] = 0 } and yet: 1 P(↑| k ) = aa * k Φ ↑ Φ ↑ k ≠ aa * P(k ) 1 P(↓| k ) = bb * k Φ ↓ Φ ↓ k ≠ bb * P(k ) 23 0.8 P(k |↓) Gaussian packet P(k |↑) 0.6 x2 1/4 ⎛ 1 ⎞ − 4σ 02 Φ ( x) = ⎜ e 2 ⎟ ⎝ 2πσ 0 ⎠ 0.4 0.2 -4 k -2 2 2 |a| P(↑| k ) = e Z + 4 2 0 1/4 ⎛ 2σ ⎞ − ( k ± k0 )2 σ 02 Φ[k ± k0 ] = ⎜ ⎟ e ⎝ π ⎠ k0 k ( Δk )2 1 2 (Δk ) = k0 k | b |2 − ( Δk )2 P(↓| k ) = e Z 2 Z ≡| a | e + 4σ 2 0 k0 k ( Δk )2 2 − k0 k 24 ( Δk ) +|b| e 2 0.8 P(k |↓) P(k |↑) 0.6 S = − Tr ( ρ log ρ ) 0.4 0.2 -4 k -2 2 2 |a| P(↑| k ) = e Z 2 + k0 k ( Δk )2 k0 k | b | − ( Δk )2 P(↓| k ) = e Z 4 (log base e) Average Shannon entropy reduction (information gain) for a weak measurement 2 0 k I= 2(Δk ) 2 25 Summary so far: • • • • Spin-dependent force entangles spin with momentum Measurement of momentum improves knowledge of spin σ z changes even though [ H , σ z ] = 0 ‘Spooky’ back action drives qubit up and down in latitude What happens if we measure x instead of k? The back action changes! Qubit moved in longitude instead. 26 1D toy model with spin-dependent impulsive force X Bz If we measure position just after the impulse, we gain NO information about the momentum change. We DO however learn the value of the magnetic field that acted on the qubit. 2 p H ( p , x, t ) = − hk0 xσ z δ (t ) 2m F ( x, t ) = + hk0 σ z δ (t ) ( ψ ( x, t = 0+ ) = aeik x Φ( x) ↑ + be −ik x Φ ( x) ↓ 0 0 ) − iϕ ⎛ e − iϕ ab * ⎞ aa * e ab * ⎞ 1 ⎛ aa * 2 ρx = ⎜ + iϕ ⎟ | Φ ( x) | = ⎜ + iϕ ⎟ PX ( x) ⎝ e a * b bb * ⎠ bb * ⎠ ⎝ e a * b Measuring position gives no information about momentum or spin but produces rotation of qubit: p2 H ( p, x, t ) = − hk0 xσ z δ (t ) 2m Non-spooky back action! ϕ ≡ 2k0 x Qubit in pure state. 28 − iϕ ⎛ e − iϕ ab * ⎞ aa * e ab * ⎞ 1 ⎛ aa * 2 ρx = ⎜ + iϕ ⎟ | Φ ( x) | = ⎜ + iϕ ⎟ PX ( x) ⎝ e a * b bb * ⎠ bb * ⎠ ⎝ e a * b Measuring position gives no information about momentum or spin but produces rotation of qubit at constant latitude: Non-spooky back action! Qubit in pure state. 29 Same dephasing as before if average over x measurements. DISPERSIVE READOUT is Exactly Analogous qubit + readout pulses Qubit + resonator AMP dispersive shi8 χ Readout amplitude 1 0 fr f width κ 90 Readout phase (deg) 1 θ -­‐90 fr 0 f Y † V = χ a a σ ≈ χ aX σ ↑ X ↓ z a ≡ a +δa (a real) † X ≡ δa +δa Homodyne measurement of Y is analogous to Stern-Gerlach measurement of momentum. Spooky back action. Homodyne measurement of X is analogous to Stern-Gerlach measurement of position. Non-spooky back action due to photon shot noise. z Y † ↑ X ↓ { ψ in = a ↑ + b ↓ + iθ z V = χ a a σ ≈ χ aX σ z X ≡ δ a† + δ a }α − iθ ψ out = a e α ↑ + b e α ↓ { n ψ out = ae + inθ ↑ + be − inθ ↓ } nα Each photon passing through the cavity rotates the qubit by 2θ (but only if we measure n or X!!) Pseudo-heterodyne measurement qubit + readout pulses vacuum port Qubit + resonator X Y Back action has both spooky and non-spooky components. Qubit is now entangled with two independent oscillators (field modes). But still in a PURE state if we measure one quadrature of each. Pseudo-heterodyne measurement qubit + readout pulses vacuum port Qubit + resonator X Y Back action has both spooky and non-spooky components. Measurement efficiency is 50% but ‘added noise’ does not dephase if both modes are fully measured. Thanks Many further details ‘soon’ in my Les Houches notes. Gerlach’s Postcard to Bohr 8 February 1922 ‘Attached [is] the experimental proof of directional quantization. We congratulate [you] on the confirmation of your theory.’ (Historical note: they did not realize this was the discovery of electron spin.) 36 AIP Emilio Segrè Visual Archives.