Uncertainty Relation Revisited from Quantum Estimation Theory Yu Watanabe Collaborators: Takahiro Sagawa, Masahito Ueda The University of Tokyo

arXiv:1010.3571 (2010).

2/17/2011

Nagoya Winter Workshop 2011 (NWW2011)

Outline 

Two types of uncertainty relation  Our setup of estimation process  Our results  Relation to information geometry

2

Two types of uncertainty relation 

Indeterminacy --- property of quantum states --Trade-off relation on state preparation Non-commutable observables cannot have definite value simultaneously. 

Kennard-Robertson’s inequality (1927, 1929) : quantum fluctuation



Schrodinger’s inequality (1930)

: symmetrized quantum corelation function 3

Two types of uncertainty relation 

Complementarity --- property of measurements (operations) --Trade-off relation between measurement errors We cannot perform measurement without back-action. We cannot perform precise measurement of non-commutable observables simultaneously.

We cannot know

and

precisely, if

.

There must exist trade-off relations obtained information about non-commutable observables. However, the optimal trade-off relation is not found so far. optimal: the equality is achievable for all states and observables. What is the optimal bound of complementarity? 4

Estimation Theory -- Essential Part of Complementarity

: estimator (estimated value) of

5

Example

6

Example

7

Example

8

Example

9

Example

10

Cramer-Rao inequality 

For all consistent estimators such that

for all

and

: Fisher information matrix

: parameter of  11

Quantum Cramer-Rao inequality 

For all POVM

for all : Quantum Fisher information matrix Ex.) SLD, RLD, etc… 

Combining with classical Cramer-Rao inequality,

Projection measurement of

12

Measurement Error 

Measurement error of observable

in POVM

Projection measurement of 

and

cannot vanish simultaneously, if

There must exist trade-off relations between and ..

13

Theorem 1 

For all • POVM measurements • observables • quantum states



Proof From quantum Cramer-Rao inequality, for all

: RLD Fisher information matrix

then the discriminant of quadratic polynomial is always negative. 14

Arthurs and Goodman’s inequality Unbiased measurement Target

Probe : projection measurement of for all : unbiasedness condition For all unbiased measurements,

15

Estimation Scheme on Unbiased Measurement 

One of the simplest (may not be optimal)way to estimate and is : eigenvalue of : number of times we obtain outcome In this case,

then

16

Theorem 1 

For all • POVM measurements • observables • quantum states

However, the bound is not optimal. Example. For for all

17

Conjecture •

The optimal bound of the trade-off relation is

: “intrinsic” quantum fluctuation

: “intrinsic” quantum correlation function

• •

2-dimensional Hilbert space : proved Higher dimension : not proved but numerical evidences exist.

18

What is intrinsic quantum fluctuation and correlation ? : simultaneous irreducible invariant subspace of

and

: projection on

19

Example 

and

cannot be simultaneous block diagonalizable



and

are commutable



In general

20

Theorem 2 --- achievability of the optimal bound --For all observables and quantum states, there exist a measurement

such that

The measurement that attain the inequality is performing projection measurements of with probability

.

Condition for attaining the bound is the off-diagonal element of

equals to zero. 21

Not Optimal Scheme

In this case,

22

Optimal Scheme

23

Example 

Simultaneous measurement of two spin observables.

24

Theorem 3 For the case of

,

is satisfied.

Lemma

: SLD Fisher information matrix

From Gill and Massar bound,

25

Proof of Theorem 3 If

,

26

Numerical Evidence of the Conjecture POVMs are plotted.

Nagaoka’s inequality

27

Relation to Information Geometry •

Cramer-Rao inequality • Fisher metric: monotone metric on probability distribution manifold • Quantum Fisher metric: monotone metric on quantum state space • Quantum Cramer-Rao inequality • Complementarity on quantum information geometry 28

Fisher metric and Cencov’s theorem 

Monotonicity of statistical model Markov mapping



Monotone metric on statistical model is uniquely determined to Fisher metric .

29

Quantum Fisher metric and Petz’s theorem 

Monotonicity of quantum statistical manifold Quantum process



Monotone metric on quantum statistical manifold has minimum and maximum. •

SLD is the minimum and RLD is the maximum. 30

Quantum Cramer-Rao inequality 

What is quantum measurement? Quantum measurement



Quantum Cramer-Rao inequality for all

31

Complementarity from geometrical point of view 

Equality of

cannot be achievable.

Equality of with real vector attained by the projection measurement of

can be .



Complementarity on quantum information geometry Trade-off relation between and  Our conjecture

is important to understand the effect on quantum state manifold by the measurement

. 32

Summary and future studies 

Summary • We conjecture that the optimal bound of measurement errors is

•



This is important for understanding quantum information geometry.

Future studies • Prove the conjecture. It may come soon!

33

Thank you!

34