Two Systems

Opening the second shield changed the room's sound. A deeper pump entered through the deck, and status lights moved across the walls as isolation boundaries transferred from one enclosure to two. The second device had been assembled at Sable rather than shipped from the inner worlds. Its housing carried hand-cut access panels and a maker's mark shared by three station residents. Nothing about it looked like a duplicate.

In the model, however, adding the second qubit doubled the number of systems and quadrupled the number of basis states.

|00>  |01>  |10>  |11>

Each label contained two positions because a complete basis state had to state the basis value of each qubit. Wheeler's canonical measurement integer treated q[0] as the low-order bit.

Iona prepared both qubits independently, applied H to each, and paused before measurement. Each one-qubit amplitude table contained two entries. Combining independent systems multiplied entries pairwise, producing a product state.

( |0> + |1> ) / sqrt(2)
  tensor
( |0> + |1> ) / sqrt(2)

= ( |00> + |01> + |10> + |11> ) / 2

Here tensor named the rule for combining state spaces. In this two-qubit example it created four pairwise products. Each basis amplitude had magnitude 1/2, so every basis outcome had probability 1/4.

Tala covered the probabilities and asked Mara which outcomes could appear. Mara named all four, then objected that she had been asked a question whose answer was printed immediately underneath. Osei pointed out that printed answers had not improved anyone's predictions in the Archive.

A seeded run of 2,048 fresh preparations produced all four outcomes.

0  | ################################################## 505
1  | ################################################### 513
2  | ################################################## 501
3  | ##################################################### 529

Around the equal model, the counts varied. More importantly, no outcome was absent. Independence had spread amplitude across the complete four-row basis.

Osei replaced the second Hadamard with a controlled operation. CNOT(q[0], q[1]) used q[0] as control and flipped q[1] when the control basis value was one.

On basis states the table resembled a classical conditional flip. In a quantum region, however, the control was not measured and converted into an ordinary branch condition. The operation transformed the complete amplitude table without selecting one measured branch.

Such an amplitude transformation was unitary when it preserved normalization and possessed an adjoint that reversed it on every accepted state. Wheeler's unitary method modifier admitted only operations that met that contract. It did not mean that a sampled hardware execution would avoid physical error.

That difference returned them to the Bell program in the mission package. Iona called up the source identity received at Catenary, Sana bound it to the copy carried across the reach, and the two digests matched. The program had not changed while their ability to read it had.

quantum class BellPair {
  state long measured = 0;
  qreg q = new qreg(2);

  unitary void prepareBell() {
    H(q[0]);
    CNOT(q[0], q[1]);
  }

  entry void main() {
    prepare(q, 0);
    prepareBell();
    measured = measure(q);
  }
}

Starting from |00>, the Hadamard created two contributions.

( |00> + |01> ) / sqrt(2)

Because q[0] occupied the right-hand position in the basis label, CNOT left |00> unchanged and carried |01> to |11>.

( |00> + |11> ) / sqrt(2)

Only two rows in the ideal amplitude table remained nonzero.

Before execution, the station entered a scheduled communications shadow behind Sable. Catenary disappeared from the relay panel. For forty-three minutes, no distant authority could approve or dispute what the crew did. Iona treated the interval as ordinary. People who lived there trusted local procedures because light itself enforced the limits of supervision.

Mara predicted that 0 and 3 would appear and refused to predict equal counts. Tala ran 2,048 shots with a fixed seed.

wheeler run tutorial/BellPair.wbc \
  --target ideal \
  --shots 2048 \
  --seed 161803

0  | ################################################## 1009
1  |                                                       0
2  |                                                       0
3  | #################################################### 1039

Every recorded pair matched. The sample established correlation under the declared preparation, simulator, seed, and shot count. It did not establish entanglement from histogram shape alone.

Sana demonstrated the gap with a classical record from the Archive's old two-key doors. The door authority chose one ordinary bit and copied that known value to two audit locations. Repeating the procedure also produced only 00 and 11. With a suitable seed, even the counts could match the Bell histogram exactly.

The Archive had once displayed those matching records under the title Perfect Agreement. Sana had changed the label while she was an apprentice. Edrin changed it back. The preserved revision dispute was probably still larger than the example.

Two preparations had produced the same measurement distribution. One was a classical mixture of two definite records. The other had the ideal amplitude state shown above. Distinguishing them required the state and preparation model, not a more enthusiastic reading of the bars.

Iona asked whether the Bell amplitude table could be separated into independent one-qubit tables. Tala assigned symbolic entries a and b to the first qubit, then c and d to the second. A product state would have joint amplitudes

|00>: a*c
|01>: a*d
|10>: b*c
|11>: b*d

For the Bell table, a*c and b*d had to be nonzero, which made a, b, c, and d nonzero. It also required a*d and b*c to be zero. Those conditions could not all hold together. No pair of independent one-qubit tables produced the Bell state.

A joint pure state that could not be factored into states for its parts was entangled. The term described the complete joint state. It did not imply that the devices communicated outside the operations in their preparation, that measurement sent a useful message across distance, or that any human relation had become a physical law.

Correlated counts remained evidence consistent with that state, not a replacement for its preparation and model. Iona entered the word only after the failed factorization was attached to the record. At Sable, impressive nouns received the same suspicion as unlabeled spare parts.

Mara returned to the CNOT table. When the target began at zero, the output target matched a known control basis value. "It copied the bit."

"A known basis value," Iona said.

For a superposed control, CNOT produced the Bell state rather than two independent copies of the original one-qubit state. If it had cloned the unknown state, the result would have factored into two identical tables. It did not. CNOT could copy classical basis information into a clean basis target without becoming a universal quantum cloning operation.

In the last experiment, measurement disappeared so that the Bell preparation itself could attempt a return. Osei deleted the measurement line, paused over the empty space, and left it visible in the diff. One absent boundary changed what restoration could mean.

quantum class BellReturn {
  state long measured = 0;
  qreg q = new qreg(2);

  unitary void prepareBell() {
    H(q[0]);
    CNOT(q[0], q[1]);
  }

  entry void main() {
    prepare(q, 0);
    prepareBell();
    reverse prepareBell();
    measured = measure(q);
    assert(measured == 0);
  }
}

A generated adjoint reversed the gate order and replaced each gate with its inverse. CNOT inverted itself. Hadamard inverted itself. Applied before measurement, the adjoint restored |00> exactly in the ideal model.

The target scheduler showed the reverse gates as new work with their own duration. Nothing vanished from the station clock. Pumps ran, controls fired, and the crew waited through the operations they had asked the device to perform.

BellReturn (quantum) halted
measured = 0
measurements = [0]

This was not VM rewind. The target would physically execute the adjoint gates as new operations. Nor did it undo the measured Bell run from the opening record. That run had crossed a measurement boundary and produced a classical observation.

Sana placed the successful restoration beside the earlier measured result of 3. The two records shared a preparation body and diverged at the boundary that mattered. One ended with coherent restoration before observation. The other ended with a classical number that could be retained but not fed backward into an unknown prior state.

By the time of the homeward failure, the distinction had become familiar. In this first pass through it, all four crew members remained quiet long enough to hear Sable's shadow end: relay acquisition tones climbed one by one, and Catenary returned to the communications panel as a delayed clock and a queue of messages.

Osei closed the second shield and opened the source for the classical XOR flip from the Archive. If an exact finite permutation already knew how to go forward and backward, perhaps the quantum machine could use it without translating the logic by hand.

In the field manual, that passage was The Bridge.