IBM C1000-179 — Fundamentals of Quantum Computing Using Qiskit v2.X Developer Passing target: 24/34. All code outputs, ASCII drawings, and QASM programs were verified by execution with qiskit 2.5.1 / qiskit-ibm-runtime 0.48.0.

Answer Key

Q	Ans	Q	Ans	Q	Ans	Q	Ans
1	B	10	D	19	C	28	C
2	C	11	A	20	B	29	B
3	A	12	B	21	D	30	C
4	B	13	A	22	A	31	A
5	C	14	C	23	C	32	D
6	D	15	B	24	B	33	B
7	B	16	D	25	A, D	34	C
8	A	17	A	26	A
9	C	18	B	27	D

Explanations

Q1 — B. The drawer packs gates as far left as they fit: x(2) shares the first column with h(0) because q2 is idle there, while cx(1, 2) must wait for cx(0, 1). Drawing verified by execution. A places the x(2) after the second CX (wrong order); C flips the wire layout; D turns cx(0, 1) into two one-qubit gates on q0.

Q2 — C (3 1). AncillaRegister qubits are real qubits: num_qubits counts all of them (2 + 1 = 3) and num_ancillas reports 1. Verified by execution.

Q3 — A. Reading left→right: X on the top wire (q0), then a SWAP (the two ×'s joined by a vertical line), then H on q1. B would show a CX (dot + ⊕), not two ×'s. C starts with X on q1 and ends with H on q0 — the wrong wires. D applies the gates in the wrong order. Drawing verified by execution.

Q4 — B (2 2). The barrier forces x(1) into a layer after h(0), so depth is 2 even though the gates touch different qubits. But size() counts only operations, not directives — the barrier itself adds nothing: h + x = 2. Verified by execution.

Q5 — C. The v2.x classical-feedforward syntax is the if_test context manager, whose as else_ handle opens the else branch. c_if (A) is the removed pre-2.x style; B and D are invented signatures.

Q6 — D. a.tensor(b) builds a ⊗ b with the argument circuit b occupying the lower-indexed qubits: the CX sits on q0/q1 and the H from a lands on q2. Widths need not match. Verified by execution (drawing shows h on q_2).

Q7 — B (+y). H sends |0⟩ to |+⟩ (+x axis); S then rotates the equator by 90°, giving (|0⟩ + i|1⟩)/√2 = |i⟩, which points along +y. Statevector [0.707, 0.707j] verified by execution.

Q8 — A (False True). Single X and Z anticommute. But XX vs ZZ: each factor pair anticommutes, and two minus signs cancel — the tensor products commute. Verified by execution.

Q9 — C (0.5). Fidelity between pure states is |⟨0|+⟩|² = (1/√2)² = 0.5. Option A (0.71) is the overlap amplitude, forgetting to square. Verified by execution.

Q11 — A (1.0 0.0). For GHZ, ZZI (Z on q2 and q1) has value +1 on both |000⟩ and |111⟩ branches (two −1's multiply to +1). A single Z (IIZ, on q0) averages +1 and −1 across the branches → 0. Verified by execution.

Q12 — B. rz(π) = diag(e^{−iπ/2}, e^{iπ/2}) = −i·Z: equal to Z only up to a global phase, so the matrices differ (C false) but every measurement statistic is identical (A false).

Q13 — A. Passing a backend object as mode selects job mode: every run() is its own independently queued job. Sessions and batches must be created explicitly.

Q14 — C. Fake backends live in qiskit_ibm_runtime.fake_provider in the v2.x stack. They mimic real-device noise/coupling locally with no account. The old qiskit.providers.fake_provider V1 location and an Aer import are the distractors.

Q15 — B. Independent jobs + shared scheduling = Batch. Session would reserve the device unnecessarily; a plain loop of job-mode submissions pays the queue repeatedly.

Q16 — D ({'1': 100}). The PUB (qc, [np.pi]) binds θ=π at execution time — no assign_parameters needed (C is V1-era thinking). rx(π)|0⟩ = −i|1⟩, so every one of the 100 shots reads 1. Verified by execution.

Q17 — A. Shot precedence is per-PUB > run(shots=...) > options.default_shots. The most specific setting wins; values never accumulate.

Q18 — B (100 '01'). get_bitstrings() returns one string per shot → length 100. x(0) sets qubit 0, the rightmost bit of the 2-bit string → '01'. Verified by execution.

Q19 — C. The Sampler samples classical output; a circuit with no measurements has none, and V2 raises an error rather than appending measurements or returning probabilities.

Q20 — B. Dynamical decoupling has its own options group: options.dynamical_decoupling.enable plus sequence_type ("XX", "XpXm", "XY4"). It is not under resilience (A), not a run() kwarg (C), and unrelated to twirling (D).

Q21 — D. measure_all() creates a register named 'meas', so the field is r.data.meas; r.data.c raises AttributeError: 'DataBin' object has no attribute 'c'. Verified by execution.

Q22 — A ([ 1. 0. -1.]). ry(θ)|0⟩ has ⟨Z⟩ = cos θ: cos 0 = 1, cos π/2 = 0, cos π = −1. The three parameter sets broadcast against the single observable. Verified by execution.

Q23 — C. The Estimator computes ⟨ψ|O|ψ⟩; the observable defines the measurement basis internally. Explicit measurement instructions make the input invalid for expectation-value estimation.

Q24 — B. Precision is the second positional lever of EstimatorV2.run(pubs, precision=...) (or the 4th PUB element). Shots is an integer count and belongs to the Sampler world; default_shots = 0.01 and resilience_level = 0.01 are type-nonsense.

Q25 — A and D. ZNE and measurement-error mitigation (TREX) act on the results after execution — mitigation. Dynamical decoupling and twirling modify the circuit before/during execution — suppression.

Q26 — A. The drawing shows the ░ barrier column and a classical register labeled meas: 2/ — both fingerprints of measure_all() (fresh 'meas' register + automatic barrier). B would show c: 2/ and no barrier; C draws H on q1 with the CX pointing up; D's measure_active() only measures qubits with activity (and would also name the register differently). Drawing verified by execution.

Q27 — D. A Bell pair's reduced single-qubit states are maximally mixed, so both per-qubit Bloch vectors have zero length. Any option with definite arrows describes a product state.

Q28 — C. reverse_bits=True puts the highest-index qubit on top, so q1 is the top (empty) wire and q0 — still the qubit carrying the X — moves to the bottom. B relabels the gate onto q1, which reverse_bits never does: it reorders wires, it doesn't move gates. Verified by execution.

Q29 — B. On the qsphere, color encodes the amplitude's phase. For (|0⟩ − |1⟩)/√2 the |1⟩ amplitude carries phase π, so its blob is rendered in the opposite color; both blobs are the same size (equal magnitudes), ruling out A.

Q30 — C (0.488). Probability = counts/shots = 488/1000. You report what was measured, not the ideal value (D).

Q31 — A. V2 primitive run() calls are asynchronous and return a job handle immediately; job.status() polls it, and the job persists server-side, retrievable later with service.job(job_id).

Q32 — D. job.usage() reports consumed QPU seconds (job.metrics() gives the broader timing/metadata dictionary). The other calls don't exist.

Q33 — B ({'10': 100}). Trace it: x q[0] → q0=1; cx q[0], q[1] fires → q1=1; the second x q[0] returns q0 to 0. Final state |q1 q0⟩ = |10⟩, so the little-endian bitstring is '10'. Verified by execution via qasm3.loads + Sampler.

Q34 — C. h q[0] then cx q[0], q[1] entangles q0 and q1; q2 is never touched and stays 0. The state is (|000⟩ + |011⟩)/√2 in q2 q1 q0 order → bitstrings '000' and '011'. Verified by execution.

Distractor Analysis — 10 Trickiest

Q1 (drawing columns). Option A is the near-miss: same gates, but the x(2) pushed to the end. The drawer left-packs independent gates, so x(2) shares column 1 with the H — misjudging gate packing is the most common circuit-reading error.
Q4 (barrier accounting). The two halves pull in opposite directions: the barrier does affect depth() (it blocks layer merging) but does not count in size(). A assumes it affects neither; C assumes it affects both. Only B splits them correctly.
Q6 (tensor order). C is the trap for people who assume the calling circuit keeps the low qubit indices. a.tensor(b) = a ⊗ b with b on the lower-indexed wires — mirroring how Statevector.from_label('01') puts the rightmost label character on q0.
Q8 (commutation). B (False False) is intuitive but wrong: anticommutation is multiplicative across tensor factors, so an even number of anticommuting pairs yields overall commutation — the algebra behind stabilizer measurements.
Q9 (fidelity). A (0.71) is the classic amplitude/probability confusion — quoting |⟨ψ|φ⟩| instead of |⟨ψ|φ⟩|². The exam reuses this square-vs-no-square trap across fidelity, amplitudes, and histogram questions.
Q16 (PUB binding). C is pure V1 muscle memory: in V2, parameter values ride in the PUB and are bound server-side. A tempts those who half-remember rx(π) as "a rotation, so it must be random" — it is a full flip.
Q17 (shots precedence). B is plausible because run(shots=...) does beat the options default — but the per-PUB value is more specific still. Precedence: PUB > run > options.
Q21 (register name). A is the "it just works" answer people expect after measure_all(). The result field is named after the classical register, and measure_all() names it 'meas' — data.c only exists for the default register of QuantumCircuit(n, m).
Q28 (reverse_bits). B keeps q1 on top and moves the X onto it — the misreading that reverse_bits renames wires. It only reverses the display order; the gate stays with its qubit, which therefore changes rows.
Q33 (QASM trace). A ('01') is the big-endian read of the correct final state. C forgets the second x q[0] undoes the first. Tracing QASM line by line and remembering the bitstring prints c[n−1]…c[0] is exactly what this section tests.