PadhaikoAll subjects

Quantum Hardware and NISQ Era

Quantum Computing (ENCT 328)

Open in the Padhaiko app

5.1 Physical realization (Ion trap, superconducting, photonic)

Unlike classical bits implemented reliably on silicon CMOS transistors, physical qubits are delicate two-level quantum systems that must remain isolated from environmental noise while still permitting fast, accurate external control and readout.

DiVincenzo's Criteria for Quantum Hardware

Definition

DiVincenzo's Five Criteria for Quantum Computation

In 2000, David DiVincenzo formulated five necessary requirements for any physical hardware platform to realize a scalable quantum computer:

  1. Scalable Physical System with Well-Characterized Qubits: A physical array of two-level systems whose internal Hamiltonian, couplings, and unwanted stray interactions are thoroughly understood.
  2. Qubit Initialization: The ability to reset every qubit to a simple, known fiducial state (typically ) prior to computation.
  3. Long Decoherence Times: Coherence times () must be much longer than the elementary gate execution time () so many operations complete before quantum information decays ().
  4. Universal Set of Quantum Gates: Hardware mechanisms to implement arbitrary single-qubit rotations and at least one two-qubit entangling gate (e.g., or ).
  5. Qubit-Specific Measurement: The ability to measure individual qubits cleanly in the computational basis with high readout fidelity.

(For distributed quantum networks, DiVincenzo added two communication criteria: interconversion between stationary and flying qubits, and faithful transmission of flying qubits over distance.)

Major Physical Qubit Platforms

  1. Superconducting Qubits (e.g., Transmon):
    • Physical Encoding: Artificial atoms formed by lithographically fabricated superconducting circuits (aluminum/niobium) containing a non-linear Josephson junction shunted by a capacitor. The junction creates an anharmonic energy spectrum so and (transition frequency ) can be addressed without exciting .
    • Operating Environment: Cooled to inside dilution refrigerators to suppress thermal excitations ().
    • Control & Connectivity: Single- and two-qubit gates use shaped microwave pulses (). Typically arranged on planar 2D grids or heavy-hexagonal lattices with nearest-neighbor coupling only.
    • Strengths & Trade-offs: Ultra-fast gates and compatibility with semiconductor nanofabrication (IBM, Google), but short coherence times () and sparse connectivity.
  2. Trapped Ion Qubits:
    • Physical Encoding: Individual charged atoms (such as or ) suspended in ultra-high vacuum using oscillating RF electric fields (Paul trap). The and states correspond to stable hyperfine or optical electronic energy levels.
    • Control & Connectivity: Manipulated via precision laser pulses. Two-qubit entangling gates (e.g., Mølmer–Sørensen or Cirac–Zoller gates) couple internal qubit states through shared quantized vibrational modes (phonons) of the ion chain, providing all-to-all connectivity within a trap.
    • Strengths & Trade-offs: Longest coherence times (milliseconds to seconds) and very high two-qubit gate fidelities (), but relatively slow two-qubit gates () and scaling bottlenecks as ion chains grow large.
  3. Photonic Qubits:
    • Physical Encoding: Single photons encoded via polarization (), dual-rail spatial paths ( vs. ), or time-bin modes.
    • Control & Connectivity: Single-qubit gates are executed passively at room temperature using linear optical elements (beam splitters, phase shifters, waveplates).
    • Strengths & Trade-offs: Virtually zero thermal decoherence and native compatibility with fiber-optic quantum communication; however, photons do not interact directly with each other, making deterministic two-qubit entangling gates difficult (requiring measurement-induced nonlinearity as in the KLM scheme or fusion-based architectures) and demanding high-efficiency single-photon sources and superconducting nanowire detectors.
Comparison of Physical Qubit Hardware Platforms
FeatureSuperconducting (Transmon)Trapped IonsPhotonic Qubits
Physical QubitAnharmonic LC circuit with Josephson junctionHyperfine/optical states of confined ionsPhoton polarization, path, or time-bin
Operating Temp.Cryogenic ()Room temp. vacuum chamber (laser-cooled ions)Room temperature (cryogenic detectors)
Gate ControlMicrowave pulses ()Precision laser beams ()Beam splitters, phase shifters ()
Coherence TimeShort ()Very long ()Extremely long (photon loss is main error)
ConnectivityNearest-neighbor (2D grid / heavy-hex)All-to-all within a trap chainProgrammable via optical routing / fusion
Primary ChallengeCryogenic wiring, crosstalk, short Slow gate speeds, multi-zone ion shuttlingPhoton loss, probabilistic two-qubit gates

5.2 Noise, decoherence, and qubit metrics

Decoherence is the unwanted coupling of a quantum processor to its surrounding electromagnetic, thermal, or vibrational environment, which leaks quantum phase information and destroys superposition and entanglement.

Definition

Relaxation Time (T1) and Coherence Time (T2)

Two primary timescales characterize single-qubit noise:

  1. Energy Relaxation Time (, Longitudinal Relaxation): Measures how quickly a qubit initialized in the excited state spontaneously decays to the ground state by dissipating energy into the environment (amplitude damping):
  1. Dephasing / Coherence Time (, Transverse Relaxation): Measures how quickly the relative phase between and (the off-diagonal density matrix elements ) scrambles in a superposition state such as :

Because energy relaxation itself destroys phase coherence, is governed by both and a pure dephasing time :

Consequently, always holds. ( denotes raw free-induction dephasing time before dynamical decoupling / spin-echo pulses cancel low-frequency drift.)

Standard Qubit and Processor Metrics

Key Hardware Performance Metrics
MetricMeaningMathematical / Practical Definition
Single-Qubit Gate Fidelity ()Accuracy of single-qubit rotationsTypical: ()
Two-Qubit Gate Fidelity ()Accuracy of entangling gates ()Typical NISQ: ()
Readout Fidelity ()Accuracy of classifying vs. (typ. )
Gate-to-Coherence RatioCoherent operation budget before decay
Quantum Volume ()Holistic processor capability metric, where is max square circuit ( qubits depth) passing heavy-output test ()
CLOPSExecution speed throughputCircuit Layer Operations Per Second (includes classical control loop latency)

5.3 NISQ model and limitations

In 2018, John Preskill coined the term Noisy Intermediate-Scale Quantum (NISQ) to describe the current era of quantum hardware.

Definition

The NISQ Computational Model

A NISQ processor consists of roughly to several thousand physical qubits operating without full fault-tolerant quantum error correction:

Fundamental Limitations of NISQ Devices

  1. Exponential Decay of Circuit Fidelity (Depth Wall): If a circuit executes single-qubit gates of fidelity and two-qubit gates of fidelity , the probability of error-free execution drops exponentially:

Deep fault-tolerant algorithms such as Shor's factoring (requiring millions of gates) collapse into uniform random noise on NISQ hardware.

  1. Sparse Connectivity and SWAP Routing Overhead: Two-qubit gates can only act between physically coupled qubits. To entangle two distant qubits separated by edges, the compiler inserts gates. Since each decomposes into gates, routing sharply increases gate count and error accumulation.
  2. Lack of Fault-Tolerant Logical Qubits: Full quantum error correction requires large physical-to-logical qubit overhead ( physical qubits per logical qubit) and sub-microsecond real-time classical decoding. NISQ workflows instead rely on Quantum Error Mitigation (QEM)—techniques such as Zero-Noise Extrapolation (ZNE) and Readout Error Mitigation—which reduce bias in expectation values via extra classical post-processing and increased sampling shots.
  3. Barren Plateaus in Variational Algorithms: NISQ algorithms frequently use hybrid Variational Quantum Algorithms (e.g., VQE, QAOA) with parameterized circuits. In deep random circuits, the gradient of the cost function vanishes exponentially with qubit count, stalling classical optimizers.
AspectNISQ DevicesFault-Tolerant Quantum Computers (FTQC)
Qubit TypeRaw noisy physical qubitsError-corrected logical qubits
Maximum Circuit DepthShallow ( two-qubit layers)Arbitrarily deep (polynomial/exponentially long)
Error HandlingError mitigation (ZNE, PEC, measurement calibration)Active quantum error correction (Surface code, stabilizer decoding)
Target AlgorithmsVQE, QAOA, quantum simulation, QML heuristicsShor's algorithm, QPE, Grover's search, HHL

5.4 Quantum circuit execution on IBM quantum simulators

IBM Quantum and its open-source SDK Qiskit provide a standard software workflow for designing, transpiling, simulating, and running quantum circuits on both classical simulators and cloud-hosted superconducting QPUs.

Definition

Qiskit Circuit Execution Pipeline

Executing a quantum circuit in Qiskit follows four sequential stages:

  1. Circuit Construction: Instantiate QuantumCircuit(n, c) with qubits and classical bits, append abstract gates (h, x, cx, rz), and add measure instructions.
  2. Transpilation: Compile the abstract circuit for a target backend's native basis gate set (e.g., {ecr, id, rz, sx, x} on modern IBM processors) and physical coupling map (routing non-adjacent two-qubit gates via insertion and cancelling redundant gates).
  3. Simulator / Backend Execution:
    • Statevector Simulation: Computes the exact -component complex vector deterministically with zero shot noise (feasible up to qubits in RAM).
    • Shot-Based / Noisy Simulation (AerSimulator): Samples discrete measurement bitstrings over shots repetitions (e.g., or shots), optionally attaching a NoiseModel calibrated from a real IBM QPU's , gate errors, and readout confusion matrices.
  4. Primitives & Post-Processing: Modern Qiskit Runtime exposes two execution primitives:
    • Sampler: Returns quasi-probability distributions or bitstring counts from repeated circuit shots.
    • Estimator: Evaluates expectation values of Pauli observables , integrating error suppression and mitigation options.
Qiskit Bell-State Simulation (Ideal vs. Noisy Shot Execution)Python
from qiskit import QuantumCircuit, transpile
from qiskit_aer import AerSimulator
from qiskit_aer.noise import NoiseModel, depolarizing_error, ReadoutError

# 1. Build a 2-qubit Bell state circuit |Phi+> = (|00> + |11>) / sqrt(2)
qc = QuantumCircuit(2, 2)
qc.h(0)
qc.cx(0, 1)
qc.measure([0, 1], [0, 1])

# 2. Ideal shot-based simulation
ideal_sim = AerSimulator()
ideal_tqc = transpile(qc, ideal_sim)
ideal_counts = ideal_sim.run(ideal_tqc, shots=1024).result().get_counts()
print("Ideal counts:", ideal_counts)

# 3. Build a simple NISQ noise model (1Q/2Q depolarizing + readout error)
noise_model = NoiseModel()
noise_model.add_all_qubit_quantum_error(depolarizing_error(0.002, 1), ["h"])
noise_model.add_all_qubit_quantum_error(depolarizing_error(0.02, 2), ["cx"])
noise_model.add_all_qubit_readout_error(ReadoutError([[0.98, 0.02], [0.03, 0.97]]))

noisy_sim = AerSimulator(noise_model=noise_model)
noisy_counts = noisy_sim.run(transpile(qc, noisy_sim), shots=1024).result().get_counts()
print("Noisy counts:", noisy_counts)

Worked Examples

Example 5.1

5.1

A superconducting transmon qubit has energy relaxation time , pure dephasing time , and a two-qubit gate duration of ().

  1. Calculate the qubit coherence time .
  2. If the qubit is prepared in and left idle for , what is the probability of still measuring ?
  3. Estimate the maximum number of sequential gates executable within one coherence time .

Calculating Coherence Time : Using the relaxation-dephasing relation :

Notice that .

Excited-State Survival Probability After : By longitudinal amplitude relaxation with :

Coherent Gate Budget: Dividing by :

Answer:
  1. Coherence time:
  2. Survival probability in :
  3. Sequential budget: gates
Example 5.2

5.2

A NISQ circuit compiled for a -qubit superconducting processor contains native single-qubit gates (each with fidelity ), two-qubit gates (each with fidelity ), and qubit measurements (each with readout fidelity ).

  1. Estimate the overall circuit execution fidelity .
  2. If a separate processor reliably passes the heavy-output Quantum Volume test up to width and depth , what is its Quantum Volume ()?

Estimating Overall Circuit Fidelity: Assuming independent gate and readout errors across operations:

Substituting the given fidelities (, , ):

  • Single-qubit contribution:
  • Two-qubit contribution:
  • Readout contribution:

Multiplying all factors:

Notice that the two-qubit gates account for the largest fidelity loss ().

Computing Quantum Volume: By definition, Quantum Volume for maximum passing square circuit dimension is:

Answer:
  1. Overall circuit fidelity: ()
  2. Quantum Volume:
Example 5.3

5.3

A Bell state circuit preparing is executed for shots on an IBM Quantum noisy simulator, producing the measurement counts: {'00': 465, '01': 25, '10': 30, '11': 480}.

  1. Calculate the empirical probabilities of all four bitstrings and the total corrupted-state probability .
  2. For an ideal noiseless Bell state (), calculate the standard deviation of the estimated probability due purely to finite-shot sampling over shots.

Empirical Bitstring Probabilities and Error Fraction: Dividing each count by :

  • ()
  • ()
  • ()
  • ()

Ideal Bell-state population: (). Corrupted-state population caused by gate/readout noise:

Finite-Shot Sampling Standard Deviation (): Each measurement shot is a Bernoulli trial with ideal probability of yielding 00. The binomial standard deviation of the sample proportion over shots is:

Thus, even on a noiseless simulator, shots yield .

Answer:
  1. Probabilities: , , , ; corrupted fraction ().
  2. Ideal shot-noise standard deviation: ().

Practice Exercises

Practice Exercises — Quantum Hardware and NISQ Era

Question 1

List DiVincenzo's five criteria for quantum computation and explain one key engineering trade-off between superconducting transmon qubits and trapped ion qubits.

DiVincenzo's Five Criteria:

  1. A scalable physical system with well-characterized qubits.
  2. Ability to initialize qubits to a simple fiducial state (e.g., ).
  3. Long decoherence times relative to gate execution times ().
  4. A universal set of quantum gates (single-qubit rotations + a two-qubit entangling gate).
  5. Qubit-specific measurement capability.

Key Engineering Trade-Off (Speed & Fabrication vs. Coherence & Connectivity):

  • Superconducting transmons achieve very fast gate speeds () using microwave pulses and leverage mature lithographic microchip fabrication, but suffer from shorter coherence times (), cryogenic dilution cooling (), and restricted nearest-neighbor connectivity.
  • Trapped ions offer dramatically longer coherence times (milliseconds to seconds), higher two-qubit gate fidelities, and all-to-all connectivity via shared motional phonon modes, at the expense of much slower two-qubit laser gates () and complex optical/shuttling control as qubit counts scale.
Answer:

DiVincenzo's criteria require (1) scalable well-characterized qubits, (2) initialization, (3) long coherence (), (4) universal gates, and (5) qubit-specific readout. Superconducting qubits excel in gate speed and planar chip fabrication but have shorter coherence and sparse connectivity, whereas trapped ions offer all-to-all connectivity and long coherence at the cost of slower gates.

Question 2

A physical qubit has measured relaxation time and total coherence time .

  1. Verify that these values satisfy the fundamental physical bound .
  2. Calculate the pure dephasing time .

Verifying the Coherence Bound: With :

Since , the bound is satisfied.

Computing Pure Dephasing Time : From :

Answer:
  1. Bound verified: .
  2. Pure dephasing time: .
Question 3

Consider a -qubit NISQ chip whose coupling map is a linear nearest-neighbor chain . A quantum algorithm requires executing a gate between (control) and (target).

  1. How many gates are needed to bring adjacent to (without uncomputing back), and how many total physical gates does this transpiled operation require?
  2. If each physical gate has fidelity (ignoring single-qubit gate errors), compare the fidelity of a direct with the routed operation.

Counting and Physical Gates: In the chain , and are separated by hops ( intermediate qubits ). To move the logical state of adjacent to (onto ):

  • Apply and : gates.
  • Each gate decomposes into gates, contributing physical s.
  • Adding the actual target gives:

Comparing Direct vs. Routed Fidelity:

  • Direct (all-to-all hardware): ().
  • Routed on linear chain:

Routing increases the error rate nearly sevenfold (from to ).

Answer:
  1. Requires gates, totaling physical gates.
  2. Direct fidelity is , whereas routed fidelity drops to .
Question 4

What is the Quantum Volume () metric, and why is it preferred over raw physical qubit count alone for benchmarking NISQ processors? If an IBM Quantum processor achieves , what is the width and depth of the largest square random circuit it reliably executes?

Definition and Motivation for Quantum Volume: Raw qubit count does not reflect useful computational power: noisy or disconnected qubits may fail circuits that high-fidelity qubits complete easily. Quantum Volume is a single holistic benchmark measuring the largest -qubit, depth- random square circuit whose heavy outputs are sampled with probability (with statistical confidence), defined as:

It simultaneously captures qubit count, gate fidelity, coherence time, connectivity, crosstalk, and compiler efficiency.

Determining Square Circuit Dimension for : Solving :

Thus, the processor reliably executes square circuits of width qubits and depth layers.

Answer:

Quantum Volume () captures holistic processor quality (fidelity, connectivity, and compiler efficiency) rather than qubit count alone. A processor with reliably executes square circuits of qubits depth layers.

Question 5

Explain the difference between Qiskit's Statevector simulator and noisy shot-based (AerSimulator with NoiseModel) simulator, and explain the role of transpilation before executing a circuit on an IBM Quantum device.

Statevector Simulator vs. Noisy Shot-Based Simulator:

  • Statevector Simulator: Computes the exact -dimensional complex state vector by linear-algebraic matrix multiplication. It has zero statistical shot noise and zero hardware decoherence, making it ideal for verifying theoretical circuit math, though RAM grows exponentially as .
  • Noisy Shot-Based Simulator: Simulates discrete circuit execution for a finite number of shots (e.g., ), sampling bitstrings randomly from the output distribution while injecting Kraus/depolarizing gate errors, thermal relaxation, and SPAM readout errors via a NoiseModel. This predicts realistic NISQ hardware behavior.

Role of Transpilation: Real IBM quantum hardware only executes a restricted set of native basis gates (e.g., ecr, rz, sx, x) across a sparse coupling map. Transpilation translates abstract gates (like H, CCX, SWAP) into native basis gates, inserts SWAP routing for non-adjacent two-qubit gates, and optimizes adjacent gates to minimize total depth and two-qubit gate count.

Answer:

The Statevector simulator computes exact noiseless amplitudes without sampling noise, whereas a noisy shot-based simulator samples finite measurement bitstrings under realistic gate, decoherence, and readout noise models. Transpilation compiles abstract circuits into the processor's native basis gates and physical coupling topology while minimizing circuit depth.

← Quantum Algorithms (part 2: 4.4–4.5)Quantum Error and Correction →