Physical vs Logical Qubits and Layout

qc/hardware

Logical qubits are the abstract q[0], q[1], … you design a circuit with. Physical qubits are the actual hardware elements, each sitting at a specific location on the chip with its own error rates and connectivity. A layout is the mapping between the two.

By default, logical qubit $i$ maps to physical qubit $i$ — but you can control this explicitly via initial_layout during transpilation, which matters a lot because backend properties (gate errors, readout errors, $T_1$/$T_2$ — see Backend Properties) are not uniform across a chip. A good layout puts your circuit’s qubits on the hardware’s best-performing, best-connected physical qubits.

from qiskit import transpile
 
initial_layout = list(range(qc.num_qubits))  # ascending logical->physical mapping
 
transpiled = transpile(
    qc,
    backend=backend,
    initial_layout=initial_layout,
    basis_gates=['h', 'cx', 'swap'],
    optimization_level=3,
)

Why list(range(n)) and not just range(n): Qiskit’s transpiler internals distinguish list from other iterables when disambiguating “one layout” vs “a list of layouts, one per circuit” — passing a bare range can get silently misinterpreted. See Qiskit API Gotchas.

Layout choice interacts directly with Coupling Map and Topology: if the layout puts two qubits that need to interact onto physical qubits that aren’t directly connected, the transpiler has to insert extra SWAP gates to route around it.

Related

• Backend Properties
• Coupling Map and Topology
• SWAP Overhead and Routing
• Qiskit API Gotchas

Self-Check

• Could you explain the difference between a logical qubit and a physical qubit to someone new to this?
• Why does layout choice matter if every physical qubit can run the same gates?
• Why list(range(n)) and not just range(n) for initial_layout?