Quadratures The quadrature operators Q,P (the BIG ones, Mandel 1034): πΜ = πΜ = πΜ† + πΜ √2 π(πΜ† − πΜ) √2 I apologize for using Q,P – once again, nobody asked me before re-using the same letters – it is a convention. They are a scaled version of the canonical variables q,p, and they should not be confused with the P representation and the Q representation. Some cal them X1 X2. The normalization varies between 1 (Mandel&Wolf) to ½ (the green book). I like the symmetric 1/√2, but sometimes I neglect the normalization for simplicity. [πΜ , πΜ] = π And their uncertainty: 2 2 √〈(ΔπΜ ) 〉 √〈(ΔπΜ) 〉 ≥ 1/2 The electric field can be presented as: βππ † πΈΜ (π, π‘) = π ∑ √ πΊ {πΜ π π(πβπ−πππ‘) − πΜπ,π π −π(πβπ−ππ π‘) } = 2π0 π π,π π,π π,π βππ = π ∑√ πΊ {πΜ (cos(π β π − ππ π‘) + π π ππ(π β π − ππ π‘)) 2π0 π π,π π,π π,π † − πΜπ,π (cos(π β π − ππ π‘) − π π ππ(π β π − ππ π‘))} βππ = −∑√ πΊ {πΜ cos(π β π − ππ π‘) +πΜπ,π π ππ(π β π − ππ π‘)} π0 π π,π π,π π,π i.e. they are the cosine and sine parts of the field – hence the Q-P Wigner presentation We saw that coherent has minimal uncertainty of 1 in each The angles=phases are arbitrary – we can define πΜ (π) = πΜπ† π ππ + πΜπ π −ππ πΜ(π) = πΜπ† π π(π+π/2) + πΜπ π −π(π+π/2) = πΜ (π + π/2) That will give the same with βππ πΈΜ (π, π‘) = − ∑ √ πΊ {πΜ cos(π β π − ππ π‘ + π) +πΜπ,π π ππ(π β π − ππ π‘ + π)} π0 π π,π π,π π,π Coherent state (reminder): β¨πΌ|πΜ |πΌβ© = β¨πΌ|πΜ† + πΜ|πΌβ© = √2 π π(πΌ) β¨πΌ|πΜ|πΌβ© = β¨πΌ|π(πΜ† − πΜ)|πΌβ© = √2 πΌπ(πΌ) Which means πΌ= πΜ + ππΜ √2 THE DRAWING OF THE QUADRATURES, THE MEANING OF THE AREA IS NOISE Let’s start DOING things to light. The Beam-Splitter Formalism: 1 2 0 3 Classically, we view it as a 3-port problem: incoming wave v1 gives rise to two waves v2=rv1, v3=tv1, that conserve energy thanks to the 3rd rule above. Now suppose we want to implement this to our quantum mechanical creation and annihilation operators, which of course obey: [ πΜπ πΜπ† ] = πΏπ,π But if we write a2=ra1, a3=ta1 we get (except for the trivial cases where t or r are 1 and the other 0, so this is not a BS at all): [ πΜπ πΜπ† ] = |π|2 [ πΜπ πΜπ† ] = |π|2 ≠ 1 [ πΜπ πΜπ† ] = |π‘|2 [ πΜπ πΜπ† ] = |π‘|2 ≠ 1 [ πΜπ πΜπ† ] = ππ‘ ∗ [ πΜπ πΜπ† ] = ππ‘ ∗ ≠ 0 First of all – you can’t turn one mode into two or the other way (but yes a linear combination) so there is no 1X2 device, only 2X2, so it is wrong to ignore the other input. Specifically, the problem is the Vacuum – H=Sum(hw(n+1/2)). Even for VAC we have not only Energy, but INFINITE energy !! These are the vacuum fluctuations (VF), and what is called the zero-point energy (ZPE). One cannot overestimate the fundamental importance of this in Nature, and it is part of the answer to what is quantum in quantum optics. So it is like in Gollum’s riddle to Bilbo: “It cannot be seen (a|0>=0), cannot be felt, cannot be heard, cannot be smelt. It lies behind stars, and under hills, and empty holes (modes) it fills” The answer is “Dark”, and to see that vacuum=dark indeed creeps everywhere in a way that is fundamentally important to the quantum description of Nature, including the most simple nontrivial optical elements – the Beam Splitter (BS): So the reason is the fact that we ignored the NOTHING – Gollum’s Dark.. – the vacuum that fills the empty modes and comes in through the empty port: r,t is from one side, and r’,t’ is from the other – assuming there are no losses, it’s matrix is unitary, which means that: |π| = |π ′ | |π‘| = |π‘ ′ | |π|2 + |π‘|2 = 1 ππ‘ ∗ + π′∗ π‘ ′ = 0 It is SO easy, HW ο So the correct transformation is: πΜπ = π πΜπ + π‘′ πΜπ πΜ3 = π‘ πΜπ + π′ πΜπ And now indeed [ πΜπ πΜπ† ] = |π|2 [ πΜπ πΜπ† ] + |π‘|2 [ πΜπ πΜπ† ] = 1 [ πΜπ πΜ3† ] = |π‘|2 [ πΜπ πΜπ† ] + |π|2 [ πΜπ πΜπ† ] = 1 [ πΜπ πΜπ† ] = ππ‘ ∗ [ πΜπ πΜπ† ] + π′∗ π‘′[ πΜ0 πΜ0† ] = 0 So the vacuum field plays a fundamental role and required for internal consistency, and it has consequences in quantum electrodynamics that have no classical counterpart in classical electrodynamics and cannot be ignored. It is from here trivial to show that this is the correct transformation and that, for example, the number of photons is conserved n0+n1=n2+n3. A set of r,r’,t,t’ that satisfies the above conditions and is suitable for DIELECTRIC BS is: π′ = π = π|π| π‘′ = π‘ π€βπππ π, π‘ ππ ππππ Alternatively you can have: π ′ = −π = |π| In practice, it usually doesn’t matter (I never saw a case where it mattered), as long as you are consistent throughout your calculation What happens to quadratures in a BS: 2 1 πΜπ = ππ πΜπ + π‘ πΜπ 0 3 πΜ3 = π‘ πΜπ + ππ πΜπ It is easy to see that in “reverse” we get: πΜπ = π‘ πΜπ − ππ πΜπ πΜ0 = π‘ πΜπ − ππ πΜπ So πΜπ = πΜπ† + πΜπ = −ππ πΜπ† + π‘πΜπ† + ππ πΜπ + π‘ πΜπ = π‘ πΜπ − π πΜπ πΜπ = ππΜπ† − π πΜπ = π πΜπ† + ππ‘πΜπ† + π πΜπ − ππ‘ πΜπ = π‘ πΜπ + π πΜπ πΜπ = πΜπ† + πΜπ = −ππ πΜπ† + π‘πΜπ† + ππ πΜπ + π‘ πΜπ = π‘ πΜπ − π πΜπ πΜ3 = ππΜπ† − π πΜπ = π πΜπ† + ππ‘πΜπ† + π πΜπ − ππ‘ πΜπ = π‘ πΜπ + π πΜ0 SO What happens to COHERENT STATES in a BS: πΜπ + π πΜπ = π‘(π0 + ππ0) + ππ(π1 + ππ1) we can also see that by: π·0 (πΌ)π·1 (π½)|π£ππ〉 = ππ₯π(πΌπΜπ† − πΌ ∗ πΜπ + π½πΜπ† − π½ ∗ πΜπ )|π£ππ〉 = ππ₯π(πΌπ‘ πΜπ† + πΌ ππ πΜπ† − πΌ ∗ π‘ πΜπ + πππΌ ∗ πΜπ + π½π‘ πΜπ† + ππ½π πΜπ† − π½ ∗ π‘ πΜπ + ππ½ ∗ π πΜπ )|π£ππ〉 = ππ₯π ((π‘πΌ) πΜπ† − (π‘πΌ)∗ πΜπ + (πππΌ) πΜ3† − (πππΌ)∗ πΜπ + (π‘π½) πΜπ† − (π‘π½)∗ πΜπ + (πππ½) πΜ2† − (πππ½)∗ πΜπ ) |π£ππ〉 = π·2 (π‘πΌ + πππ½)π·3 (π‘π½ + πππΌ) QED So how do you do displacement in the lab? Displacement by a beam-splitter and a strong Local Oscillator (LO) The previous derivation shows that by having a strong LO , using abeam-splitter with very small π and π ≈ 1 you get that the Q,P of your state will be practically the Q0,P0 of the original state, with - rP1 added to the Q quadrature, and rQ1 added to the P quadrature, but the noise, which is the same for large LO as for vacuum, will be decreased by r – i.e. a displacement without adding noise. Homodyne To measure the quadrature of a state (which is Hermitian and so an observable) what you do is Homodyne, which is mixing your state with a strong LO using 50:50 beamsplitter, and SUBTRACT the currents of the two output detectors. This “isolates” the interference term between the two beams (since the average intensities are identical in both detectors, but the interference term is opposite in sign), and in classical optical detection it is used for detection of weak signals (since it is proportional to the LO amplitude). It is typically better to have a difference between the frequency of the LO and the signal, in which case the interference term oscillates at the difference frequency (beating) and so is detected by lock-in amplifier (avoiding the noises around DC, the 1/f in particular). In this case the measurement is called Heterodyne. 1 2 0 3 1 1 〈πΌ3 − πΌ2 〉 ∝ 〈 πΜπ† πΜπ − πΜ2† πΜ2 〉 = ( πΜπ† − π πΜπ† )( πΜπ + π πΜπ ) − ( πΜπ† − π πΜπ† )( πΜπ + π πΜπ ) 2 2 1 = 〈 πΜπ† πΜπ − π πΜπ† πΜπ + π πΜπ† πΜπ + πΜπ† πΜπ − πΜπ† πΜπ + π πΜπ† πΜπ − π πΜπ† πΜπ − πΜπ† πΜπ 〉 2 = π〈 πΜπ† πΜπ − πΜπ† πΜπ 〉 Assume a1 is a strong (=classical, stronger than πΜπ ) local oscillator with amplitude = |πΏπ|π π(π+π/2) : 〈πΌ3 − πΌ2 〉 ∝ |πΏπ|〈 πΜπ† π ππ + πΜπ π −ππ 〉 = |πΏπ| πΜπ (π) And then the noise comes only from the measured state and not from the LO HW! (hint - since we can approximate the expectation value: 〈 πΜπ πΜπ† 〉 = 〈1 + πΜπ† πΜπ 〉 ≈ 〈 πΜπ† πΜπ 〉 Leading to: 2 〈(πΌ3 − πΌ2 )2 〉 ∝ |πΏπ|2 〈( πΜπ (π)) 〉 〈Δ(πΌ3 − πΌ2 )2 〉 = 〈(πΌ3 − πΌ2 )2 〉 − 〈πΌ3 − πΌ2 〉2 2 ∝ |πΏπ|2 〈( πΜπ (π)) 〉 − |πΏπ|2 〈 πΜπ (π)〉2 = |πΏπ|2 〈Δ πΜπ (π)2 〉 DRAW THE DRAWING – while varying the phase of the LO the state rotates around the origin and we measure the projection on one axis so we see the shape – like tomography.