08.

The Kochen-Specker Theorem and the Measurement Problem

How Should Superpositions be Interpreted? Part 2. (A) Literally (QM description is complete): Options: (1) Standard Claim: The properties of a quantum system in a superposed state are indeterminate (do not possess values).

Topics I. KS Theorem II. Measurement

EPR say: non-local!

(B) Non-literally (QM description is incomplete): Options: (1) Local Hidden Variables with VD (2) Non-local Hidden Variables with VD The properties of a quantum system are determinate (possess values) at all times, even when the system is in a superposed state. Bell says: No! Conflicts with experiment.

Value Definiteness (VD)

We now know that non-locality isnt all that spooky. So maybe Option (B2) is okay. It would let us off the hook if we want to maintain VD (which we would if we want to keep as close to classical notions about properties as possible). BUT: The Kochen-Specker Theorem says NO! to VD.

Briefly: The Kochen-Specker (KS) Theorem is a mathematical result about the nature of Hilbert spaces (the special type of vector spaces that are the most general representation of the state space for a quantum system). It says that, if properties are represented as operators on a Hilbert space in a 1-1 fashion (i.e., each property is represented by a unique operator), then these properties cannot all be said to simultaneously have values. As we shall see, this is due specifically to the structure of Hilbert spaces. This is not a problem for classical systems with state spaces represented by point sets. (So, in a sense, the KS Theorem just reconfirms that our original choice of using Hilbert spaces to represent quantum state spaces is correct, if we want to be able to represent quantum properties that are fundamentally different from classical properties.)

I. The Kochen-Specker (KS) Theorem

Let v(A), v(B), v(C), ... be the values, in a given state, of properties represented by operators A, B, C, ... . Kochen-Specker Theorem: For Hilbert spaces H of dimension 3, the following are contradictory: (Value Definiteness) Any set of properties represented by operators A, B, C, ... on H simultaneously have values v(A), v(B), v(C), ... (Value Constraints)
(a) If A, B, C are compatible and C = A + B, then v(C) = v(A) + v(B). (b) If A, B, C are compatible and C = AB, then v(C) = v(A)v(B).
Compatibility means A, B, C all have a set of eigenvectors in common.

(Sum Rule) (Product Rule)

Value Constraints is a consequence (in part) of requiring a 1-1 correspondence between properties and operators (this is sometimes called non-contexuality).

To see how the KS Theorem is proved, we first need the notion of a projection operator (these will also be important in later lectures):

A projection operator P| associated with a unit vector | maps any vector | to another one P|| defined by, P|| || that is the result of projecting | along the direction of |.
P| is sometimes written as || to make this definition explicit. Note again: | is a number and | is a vector, so || is a vector.

example:
| | P||

2 Important Characteristics of Projection Operators: (P1) (P|)2 = P| (P2) P|a1 + P|a2 + ... + P|aN = IN (idempotency) (resolution of the identity)
Proof: (P|)2| = P|(P||) = P|(||) = ||| = ||

where |a1, |a2, ... , |aN form an orthonormal basis for an Ndimensional Hilbert space H with identity operator IN.
Proof: Suppose |B is any vector in H. Then: (P|a + P|a + ... + P|a )|B = (P|a + P|a + ... + P|a )(b1|a1 + b2|a2 + ... + bN|aN)
1 2 N 1 2 N

= b1|a1a1|a1 + b2|a2a2|a2 + ... + bN|aNaN|aN = b1|a1 + b2|a2 + ... + bN|aN = |B

since ai|aj = 0, unless i = j

The significance of projection operators: They encode yes/no properties about whether or not a system in a given state possesses a given value of a given property. Recall: Any orthonormal basis |a1, |a2, ... , |aN is a set of eigenvectors of some (complete) operator A. For these eigenvectors, (a) A|ai = ai|ai (b) P|ai|ai = |aiai|ai = |ai
|ai is an eigenvector of A with eigenvalue ai |ai is an eigenvector of P|a with eigenvalue +1
i

(c) P|ai|aj = |aiai|aj = 0|ai = 0 = 0|aj , for i j

really the zero vector |I

|aj is an eigenvector of P|a with eigenvalue 0

i

SO: (i)

Any eigenvector of P|ai with eigenvalue +1, is also an eigenvector of A with eigenvalue ai, which means it represents a state that possesses the value ai of the property represented by A.

(ii) And any eigenvector of P|ai with eigenvalue 0, is also an eigenvector of A that does not have the eigenvalue ai, which means it represents a state that possesses some value, other than ai, of the property represented by A. Thus: Let P|a be a projection operator, where |ai is an eigenvector of the operator A. Then P|a represents the property "The value of A is ai".
i i

There are only two values of this property in a given state: (a) +1, which means the state has the value ai of A. (b) 0, which means the state does not have the value ai of A. 3

The Significance of projection operators to the KS Theorem It turns out that the (Value Constraints) part of the KS Theorem entails a particular value constraint on projection operators, call it (PVC): Projection Operator Value Constraint (PVC) Let |a1, |a2, ... , |aN form an orthonormal basis for an N-dimensional Hilbert space. Then, v(P|a1) + v(P|a2) + ... + v(P|aN) = 1, where v(P|ai) = 1 or 0, for i = 1...N.

Proof: Suppose |a1, |a2, ... , |aN form an orthonormal basis for an N-dimensional Hilbert space with identity operator IN. Then (P2) says: P|a + P|a + ... + P|a = IN. The P|a are all compatible since they share the |ai as eigenvectors. So the Sum Rule says: v(P|a ) + v(P|a ) + ... + v(P|a ) = v(IN). This yields (PVC) by means of the following two results: v(P|a )v(P|a ) = v(P|a 2) Product Rule
i i i 1 2 N 1 2 N i

v(P|a ) = 1 or 0
i

= v(P|a ) (P1)
i

Since values are nonnegative numbers

For any operator R, v(R) = v(INR) = v(IN)v(R) Product Rule

v(IN) = 1

Since values are nonnegative numbers

So the KS Theorem can be recast as the claim that (VD) and (PVC) are contradictory. Lets see what this really means: (PVC) says: The operator A with eigenvectors |ai has a definite value (i.e., just the value ai for which v(P|ai) = 1). (VD) says: All operators have definite values (not just A, but even those incompatible with A). So (VD) requires that (PVC) holds, not only for the |ai orthonormal basis, but for all orthonormal bases. The proof of the KS Theorem is just a demonstration that this cannot be: We cant have (PVC) for all orthonormal bases of Hilbert spaces with dimension greater than or equal to 3.

Now: Consider how we might implement (PVC) by the following: Label each P|ai either black or white, depending on whether v(P|ai) = 1 or 0. (PVC) then says: For the set of bases corresponding to P|ai, one is labeled white and all the others are labeled black. (VD) now requires us to do this for all sets of bases. SO: To prove the KS Theorem, we just need to demonstrate that you cant color all the bases of a Hilbert space of dim 3 such that one member of each set is white and all the other members of the set are black. Lets see how this works for dimension 2, but doesnt work for dimension 3.
Note First: P|a doesnt correspond exactly to the unit vector |ai. Rather P|a corresponds to the ray through |ai. P|a projects other vectors along this ray (i.e., the direction that |ai is pointing in). So what were doing below is coloring basis rays, and not basis vectors. P|a P|a
2 i i i

dim = 2

There are an infinite number of pairs of orthogonal P|a

1

rays {P|a , P|a }, {P|a , P|a }, etc. Each one obtained by rotating {P|a , P|a } by some angle , 0 <
1 2 1 2

P|a

90. For each set, can consistently color one white and the other black. If we continue coloring as in the diagram, well color the entire circle such that each alternating quadrant is black or white.

P|a

dim = 3

Rotating about P|a colors the equator q, as in the 2-dim case. Rotating about P|a colors the circle r,
3

P|a

and rotating about P|a colors the circle s. But can

1

also rotate about any other ray through the origin. P|a
2

Claim: Cant consistently color the entire surface of the sphere in this manner. At some point, youll run over a previously colored portion!

(PVC) doesnt work for dim = 3 due to the extra degree of freedom in the 3rd dimension. Intuitively, it wont work for higher dimension spaces with even more degrees of freedom. (This isnt all that obvious, but its true.) 5

How Should Superpositions be Interpreted? Part 3. (A) Literally (QM description is complete): Options: (1) Standard Claim: The properties of a quantum system in a superposed state are indeterminate (do not possess values). (B) Non-literally (QM description is incomplete): Options: (1) Local Hidden Variables with VD (2) Non-local Hidden Variables with VD Bell says: No! Conflicts with experiment. KS Theorem says: No!
Value Definiteness (VD) The properties of a quantum system are determinate (possess values) at all times, even when the system is in a superposed state.

EPR say: non-local!

We still have options under (B). For instance, we can claim that, for any quantum system, some of its properties are determinate at all times, others are not (well see later that modal interpretations of QM take this route). Maybe we should seriously consider Option (A1): Why not simply adopt it?

BUT:

New problem arises: Dynamics!

II. Albert Chap. 4: The Measurement Problem

According to the Principles of QM, there are two ways a state can change: 1. In general, states change via the Schrdinger Equation: |(t1) Schrdinger |(t2)
evolution

t1 t2

Deterministic: For any initial state |(t1), there is exactly one final state |(t2) that it Schrdinger-evolves into. Temporal. 2. When a measurement occurs, states change according to the Projection Postulate:
When a measurement of B is made on a state | = a1|b1 + a2|b2 + ... + aN|bN expanded in the eigenvector basis of B with result bi, then | collapses to |bi:

| |bi collapse Indeterministic: | collapses to |bi with probability = |ai|2. Instantaneous! Inconsistent? On the surface, yes. In fact, given Option (A1), we can show explicitly that they lead to very different predictions. Consider physical system consisting of Hardness-measuring device m and electron e:

ready

electron enters

hard

soft

electron
exits

Device + electron can be considered a 2-particle system (2 subsystems). The Schrdinger equation then tells us how the states of this system evolve in time.

initial state |readym|harde |readym|softe Schrdinger

evolution

final state |hardm|harde

|softm|softe
Schrdinger evolution

pointer set to ready

pointer set to soft

Now: Suppose a black electron is measured for Hardness. Initial state is: |readym|blacke = |readym{|harde + |softe} = |readym|harde + |readym|softe SO: Since Schrdinger evolution is linear, final state must be: |hardm|harde + |softm|softe BUT: According to the Projection Postulate, the final state is: either or |hardm|harde |softm|softe with prob = with prob =

If we interpret superpositions literally via Option (A1), then these represent totally different states! (In fact, well see later that they are empirically different!) Initial response: But, according the Principles of QM, the Projection Postulate is supposed to take over during a measurement. So shouldnt we simply ignore what the Schrdinger dynamics predicts for measurement scenarios? Problem: What exactly is a measurement? When is the Projection Postulate supposed to take over from the Schrdinger dynamics?

SO: If we want to adopt Option (A1), were going to have to answer this question.