Entanglement Without Faster-Than-Light Myths

Factorability is the dividing line#

Turn the definition into coefficient equations.

Entanglement means the state of a composite system cannot be written as a product of independent subsystem states. The information is in the joint state; the parts, considered separately, carry less than the whole. That is the complete definition, and everything else in this chapter is a consequence of it.

For two qubits, a product state has the form $\lvert a\rangle\otimes\lvert b\rangle$; each qubit has its own state, and the pair is just the two descriptions stapled together. An entangled state permits no such stapling. The standard first example is the Bell state $(\lvert00\rangle+\lvert11\rangle)/\sqrt2$: measured in the computational basis, the pair comes out 00 or 11, never 01 or 10.

Here is the sentence the myths depend on your skipping: correlation is not controllable messaging. The two outcomes match when compared, but each local outcome is random on its own, and neither party can choose which shared result appears. A mailbox requires choosing what to send; entanglement offers no such knob.

Evidence boundary. Entangled pure states are joint states that cannot be expressed as tensor products of subsystem states. [John Watrous] [Michael A. Nielsen]

Build the Bell pair gate by gate#

Trace H and CNOT from |00>.

Two qubits live in a four-dimensional basis; $\lvert00\rangle$, $\lvert01\rangle$, $\lvert10\rangle$, $\lvert11\rangle$; and a general state is $a\lvert00\rangle+b\lvert01\rangle+c\lvert10\rangle+d\lvert11\rangle$. If the Bell state were secretly a product, you could write it as $(x\lvert0\rangle+y\lvert1\rangle)\otimes(u\lvert0\rangle+v\lvert1\rangle)$, which expands to:

Matching the Bell state forces $xv = 0$ and $y u = 0$ while $xu$ and $y v$ stay nonzero. But $xv = 0$ kills either $x$ or $v$: if $x = 0$ then $xu = 0$, and if $v = 0$ then $y v = 0$. Either way a term that must be nonzero vanishes. No choice of four numbers escapes, so the state is genuinely joint; not approximately, not usually, but always.

This is worth doing once by hand, because it converts entanglement from an aura into arithmetic. The "mystery" is a system of four equations with no solution.

Evidence boundary. Bell-state correlations violate constraints on local hidden-variable explanations under suitable measurement settings. [John S. Bell] [Michael A. Nielsen]

Local marginals remain featureless#

Compute each subsystem's distribution without access to the other record.

The headline trap is equating entanglement with faster-than-light messaging. The correlations are real and stronger than any classical shared-variable story can reproduce; Bell tests settle that; but strength of correlation is not control of outcome. Any product, pitch, or plot that routes a chosen message through an entangled pair alone has left the physics.

The quieter trap is treating entanglement as automatically good. Entanglement can be a resource, a symptom, a debugging challenge, or a source of unwanted coupling. When your qubits entangle with the environment instead of each other, the phenomenon has a different name; decoherence; and it is the enemy of everything Part VI of this book builds. The same mathematics, pointed the wrong way, is the main thing hardware teams fight.

Evidence boundary. Quantum theory preserves no-signaling because local marginal statistics do not depend on a spacelike-separated measurement choice. [John Preskill] [John Watrous]

Correlations appear only in comparison#

Show the role of basis choice and classical record matching.

That observation prices everything downstream. Product states are cheap to simulate; highly entangled states are expensive in a straightforward statevector representation, which is both why quantum hardware is hard to imitate classically and why tensor-network methods; classical algorithms that exploit limited entanglement structure; can simulate more than naive counting suggests. Entanglement structure, not qubit count, sets the difficulty.

It also sets the protocol menu. Teleportation, superdense coding, and error correction all spend carefully prepared entanglement alongside classical communication and measurement. In every case the classical leg is load-bearing: remove it and the protocol collapses back to unexplained coincidences.

A signaling test any explanation must pass#

Vary one side's setting and check the other marginal.