Quantum Information Theory Entanglement and Measurement Pouya Bastani pbastani@math.sfu.ca CMPT 881 - Quantum Computing Quantum Information Theory – p.1/20 Information Information - classical or quantum - is measured by the minimum amount of communication needed to convey it How much is there to know? Shannon: the amount of information from a source is the memory required to represent its output and is given by its ‘entropy’ How much do we already know? the amount of information we already have is given by ‘mutual entropy How large is our ignorance? Slepian & Wolf: the amount of information needed to obtained full knowledge is given by a quantity called ‘conditional entropy’ Quantum Information Theory – p.2/20 Shannon Entropy Entropy: amount of uncertainty in the state of a physical system Shannon entropy of a random variable X: average amount of uncertainty about X before we learn its value average amount of information gain after we learn the value of X Less probable events convey more information Entropy of independent events is the sum of entropies Entropy of an even with probability p is H(p) ≡ − lg p Entropy associated with probability distribution p(x) of X: H(X) ≡ − X p(x) lg p(x) 0 lg 0 ≡ 0 x Entropy of a bit is 2 Quantum Information Theory – p.3/20 Entropy Types Joint Entropy: total uncertainty about the pair (X, Y ) P H(X, Y ) ≡ − x,y p(x, y) lg p(x, y) Conditional Entropy: average uncertainty about X given Y H(X|Y ) ≡ H(X, Y ) − H(Y ) Mutual Entropy: common information between X and Y H(X:Y ) ≡ H(X) + H(Y ) − H(X, Y ) Conditional entropy is always non-negative Quantum Information Theory – p.4/20 Definitions The density operator for quantum system {|ψi i , pi } is defined by: ρ≡ X pi |ψi i hψi | i A quantum system is said to be in a pure state if its state is known exactly. A pure state can be written as ρ = |ψi hψ| Let ρAB be the density operator of a quantum system AB. The reduced density operator for system A is defined by ρA ≡ TrB (ρAB ) The partial trace TrB over system B is defined by TrB (|a1 i ha2 | ⊗ |b1 i hb2 | = |a1 i ha2 | hb2 |b1 i where |ai i and |bi i are any two vectors in the state space of A and B Quantum Information Theory – p.5/20 Von Neumann Entropy Qubits exist in superposition state, unlike classical bits QIT must reflect this fact by including phase information Von Neumann defined the entropy of a quantum state as S(ρ) ≡ Tr(ρ lg ρ) where ρ is the density operator of the system P Ifρ is expressed in a diagonal basis, ρ = x λx |xi hx|, then the von Neumann entropy reduces to the classical Shannon entropy S(ρ) = − X λx lg λx 0 lg 0 ≡ 0 x An n × n matrix ρ is diagonalizable iff ρ has n independent eigenvectors Quantum Information Theory – p.6/20 Entropy Types Joint Entropy: S(AB) ≡ −Tr(ρAB lg(ρAB )) Conditional Entropy: S(A|B) ≡ S(AB) − A(B) Mutual Entropy: S(A:B) ≡ S(A) + S(B) − S(AB) Quantum Information Theory – p.7/20 Examples 1 Independent particles: ρA = ρB = 21 (|0i h0| + |1i h1|) ρAB = ρA ⊗ ρB S(A) = 1 S(AB) = 2 S(A:B) = 0 S(A|B) = 1 Quantum Information Theory – p.8/20 Example 2 Fully correlated Particles: ρAB = 21 (|00i h00| + |11i h11|) ρA = TrB (ρAB ) = 21 (|0i h0| + |1i h1|) S(A) = 1 S(AB) = 1 S(A:B) = 1 S(A|B) = 0 Quantum Information Theory – p.9/20 Example 3 Entangled (super-correlated) Particles: |ψAB i = √1 (|00i 2 + |11i) ρAB = |ψAB i hψAB | ρA = TrB (ρAB ) = 12 (|0i h0| + |1i h1|) S(A) = 1 S(AB) = 0 S(A:B) = 2 S(A|B) = −1 Quantum Information Theory – p.10/20 Observation In QIT, conditional entropies can be negative. In fact, isolated entangled systems have negative conditional entropy Case III is forbidden is classical (Shannon) information theory the entropy of the entire system AB cannot be smaller than the entropy of one of its subparts: S(AB) ≮ S(B) S(AB) = 0 in case III because the EPR-pair AB is created via a unitary transformation of a zero entropy pure state. In fact, any unitary transformation conserves entropy in a quantum system CN OT H |00i = 1 (|00i + |11i) = |EP Ri 2 Quantum Information Theory – p.11/20 Example Greenberger-Horne-Zeilinger (GHZ) state (EPR-triplet): |ψABC i = √1 (|000i 2 + |111i) ρABC = |ψABC i hψABC | ρAB = 12 (|00i h00| + |11i h11|) S(ABC) = 0 S(C|AB) = −1 Quantum entanglement between any part and the rest of the system Ignoring one part creates classical correlation between the rest Quantum Information Theory – p.12/20 Von Neumann Measurement Measurement: correlation between the observer and the observed Copenhagen interpretation: measurement is the interaction of a classical apparatus with a quantum system, that collapses the wave-function Von Neumann: the measuring ancilla must be quantum mechanical The qantum system Q and ancilla A are initially in state |Q, 0i = X αi |ai , 0i i Measurement introduces correlation between quantum system Q ancilla A U |Q, 0i −→ |Q, Ai = X αi |ai , ii i Observation collapses this wavefunction to |ai , ii with probability αi Quantum Information Theory – p.13/20 Flaw No-cloning theorem: it is possible to copy a quantum state only if it belongs to a set of orthogonal states Attempt to copy an arbitrary quantum state generates entanglement between quantum system Q and ancilla A An arbitrary state cannot be measured without creating entanglement In general |Q, Ai is entangled not just correlated Entangled states have a description distinct from correlated states Quantum Information Theory – p.14/20 So Far Classical information corresponds to correlation between measured system and measuring device (only differences are measurable) Von Neumann measurement creates entanglement between quantum system Q and measurement ancilla A rather than correlations Entangled system QA cannot any reveal classical correlation Quantum Information Theory – p.15/20 Resolution Interact another ancilla A′ with entangled system QA: ′ |QAA i = X αi |ai , i, ii i State of the entire ancilla AA′ unconditional on Q yields: ρAA′ = TrQ [ρQAA′ ] = X |αi |2 |i, ii hi, i| i A and A′ are maximally correlated: S(A:A′ ) = S(A) = S(A′ ) = S(AA′ ) Measure the correlation between A and A′ Correlations between A and Q are unobservable Quantum Information Theory – p.16/20 Pure vs Mixed State No (Shannon) information about Q can be obtained from the ancilla: S(Q:A:A′ ) = S(A:A′ ) − S(A:A′ |Q) = 0 Consequence of the fact that Q is initially in a pure state, S(Q) = 0 No information can be obtained in any measurement of a pure state Pure state has vanishing von Neumann entropy, thus there is no uncertainty about it, so nothing can be learned from it Only mixed-state measurements can yield information (mutual entropy) If a system is prepared in a mixed state described by {ρi , pi }, information I gained by a measurement is limited by the Kholevo bound: I≤S X i pi ρi ! − X pi S(ρi ) i Quantum Information Theory – p.17/20 “... We must abandon at least one cherished notion of orthodox measurement theory: that the apparatus necessarily reflects the state of the system it was built to measure, by being correlated with it in the sense of Shannon. Rather, it is the correlations between the ancillae that create the illusion of measurement.” “...Intuition dictates that performing a measurement means somehow observing the state of Q. Rather, a measurement is constructed such as to infer the state of Q from that of the ancilla – by ignoring Q itself. The correlations (in AA′ ) which emerge from the fact that a part (Q) of an entangled state (QAA′ ) is ignored give rise to the classical idea of a measurement. ” C. Adami and N. J. Cerf Quantum Information Theory – p.18/20 Negative Entropy Interpretation Alice and Bob hold some information X and Y , respectively Bob has some information about X due to correlation between X and Y Additional information Alice needs to send Bob for him to fully learn X: S(X|Y ) = S(XY ) − S(Y ) What if S(X|Y ) is negative? Alice can transfer X using only classical communication Alice and Bob have the potential to transfer additional quantum information in the future at no further cost* to them * Communicating classical information is far easier that quantum information Quantum Information Theory – p.19/20 References 1. N.J. Cerf, C. Adami. Quantum mechanics of measurement, eprint quant-ph/9605002 2. N.J. Cerf, C. Adami. Negative entropy and information in quantum mechanics, PRL 79 (1997), pp. 5194-519 3. N.J. Cerf, C. Adami. Negative entropy in quantum information theory, Proceedings of the 2nd International Symposium on Fundamental Problems in Quantum Physics, Oviedo, 1996 4. N.J. Cerf, C. Adami. What Information Theory Can Tell Us About Quantum Reality, Lecture Notes in Computer Science 1509, 1999 5. M.A. Nielsen, I.L. Chuang. Quantum Computation and Quantum Information. Cambridge University Press, 2000 6. M. Horodecki, Oppenheim, A. Winter. Partial Quantum Information, Nature 436 (August 2005), pp. 673-676 Quantum Information Theory – p.20/20