SwiftQiskit is a lightweight quantum computing simulator written entirely in Swift. It brings a Qiskit-like experience to the Apple ecosystem, with a strong focus on clarity, correctness, and GUI integration.
This project is experimental and educational, but grounded in real quantum mechanics principles
Differences between this forked repository ("fork") and its parent:
- The usage of Xcode playgrounds.
- Playgrounds 10 to 23 contain many different quantum computing examples; pages 40+ walk
through the SwiftQiskitApp
INTRODUCTION.mdbook chapter by chapter. - Showing of Bloch spheres (in live playgrounds).
- Using Swift Testing.
- A separate app, SwiftQiskitApp, targeting macOS, iOS and iPadOS, is the SwiftUI front-end for this package. SwiftQiskitGUI has been dropped.
- A template system is available to generate Swift apps, e.g. see the SwiftQiskitWalkDemo application.
- A second library product,
SwiftQiskitViews(import SwiftQiskitViews), holds presentational SwiftUI types (BlochVector,CHSHChartView,TensorNetworkView) shared by the playground and bySwiftQiskitApp, keeping the coreSwiftQiskitmodule UI-free.
- ✅ Complex number arithmetic
- ✅ Matrix operations (
+ - *, scalar multiply, Kronecker products) - ✅ Tensor products:
tensor(_:)/⊗onMatrixandStateVector - ✅ Dirac (bra–ket) notation:
Ket/Bra, postfix†(dagger), inner & outer products - ✅ State vector simulation
- ✅ Quantum gates (see the gate tables below):
- Hadamard (H)
- Pauli-X (X)
- Pauli-Y (Y)
- Pauli-Z (Z)
- Phase gates: S, S†, T, T†, and the general P(θ)
- Rotations: RX(θ), RY(θ), RZ(θ)
- CNOT (Controlled-NOT), Toffoli (CCX), and general multi-controlled X (MCX)
- Two-qubit rotations RZZ(θ), RXX(θ), RYY(θ), and the general
pauliRotation
- ✅ Single-qubit gate embedding
- ✅ Quantum circuit abstraction
- ✅ Measurement & state collapse
- ✅ Bell State (Entanglement) example
- ✅ Pauli strings & Hamiltonians (
PauliString/Hamiltonian), Pauli-basis measurement and shot-based expectation values (measureExpectation) - ✅ Hamiltonian simulation via Trotterized time evolution (
evolve/trotterCircuit) - ✅ Variational algorithms: exact parameter-shift gradients and a minimal gradient-descent
optimizer (
ParameterShift,GradientDescent) - ✅ State tomography from measurement statistics (
StateTomography) - ✅ Open-system simulation: density matrices, Kraus channels, and noise models
(
DensityMatrix,KrausChannel,NoiseModel,runDensityMatrix/runTrajectories) - ✅ Register arithmetic (
increment/decrement) and marginal/parity readout helpers - ✅
SwiftQiskitViews— a small, UI-only companion library (Bloch-vector math, a 2D chart view) shared by the playground and bySwiftQiskitApp
Named single-qubit basis kets, defined as Ket (= StateVector) constants in
Sources/SwiftQiskit/Quantum/Dirac.swift:
| Constant | State | Definition | Bloch sphere |
|---|---|---|---|
.zero |
|0⟩ | (1, 0) | +z (north pole) |
.one |
|1⟩ | (0, 1) | −z (south pole) |
.plus |
|+⟩ | (|0⟩ + |1⟩)/√2 | +x |
.minus |
|−⟩ | (|0⟩ − |1⟩)/√2 | −x |
.plusI |
|i⟩ | (|0⟩ + i|1⟩)/√2 | +y |
.minusI |
|−i⟩ | (|0⟩ − i|1⟩)/√2 | −y |
Multi-qubit basis kets come from the binary-label initializer, e.g. Ket("01") = |01⟩
(qubit 0 is the most-significant bit), with Bra("01") as the matching bra.
Built-in gates — each is a public enum in Sources/SwiftQiskit/Gates/ exposing
static let matrix: Matrix (parameterized gates expose static func matrix(theta:)),
with a matching convenience method on QuantumCircuit:
| Gate | Circuit API | Type |
|---|---|---|
| Hadamard (H) | h(qubit) |
HadamardGate |
| Pauli-X (X) | x(qubit) |
PauliXGate |
| Pauli-Y (Y) | y(qubit) |
PauliYGate |
| Pauli-Z (Z) | z(qubit) |
PauliZGate |
| S / S† | s(qubit) / sdg(qubit) |
SGate / SDaggerGate |
| T / T† | t(qubit) / tdg(qubit) |
TGate / TDaggerGate |
| Phase P(θ) | p(theta, qubit) |
PhaseGate |
| RX/RY/RZ (θ) | rx/ry/rz(theta, qubit) |
RXGate / RYGate / RZGate |
| CNOT (CX) | cx(control, target) |
CNOTGate |
| Toffoli (CCX) | ccx(control1, control2, target) |
ToffoliGate |
| CZ / multi-controlled Z | cz(a, b) / mcz(controls, target) |
ControlledZGate / MultiControlledZGate |
| SWAP | swap(a, b) |
SwapGate |
| MCX | mcx(controls, target) |
MultiControlledXGate |
| RZZ/RXX/RYY (θ) | rzz/rxx/ryy(theta, q0, q1) |
RZZGate / RXXGate / RYYGate |
Used in:
- Hadamard (H) — Bell example; all five test suites; playground pages 01, 02, 05–16, 20; page 19 (
HadamardGate.matrixapplied directly viaStateVector.apply(_:)— Bell-pair prep, basis rotations — rather than the circuit method, since page 19 doesn't useQuantumCircuitat all); page 22 (HadamardGate.matrixas the walk's coin flip) - Pauli-X (X) —
TensorProductTests,AdditionalGatesTests; pages 02, 05, 09–14, 17, 18; page 19 (PauliXGate.matrixas a Kraus operator) - Pauli-Y (Y) —
AdditionalGatesTests; pages 05 (y(0)), 08 (Y† == Y, ⟨ψ|Y|ψ⟩); pages 19–20 (PauliYGate.matrixas a Monte-Carlo error gate / an exact-expectation check) - Pauli-Z (Z) —
DiracNotationTests,AdditionalGatesTests; pages 01, 02, 05, 08, 13, 14; pages 19–21 (PauliZGate.matrixas a Monte-Carlo error gate and the exact ZZ-rotation identity's core piece) - S / S† —
AdditionalGatesTests; pages 02 (the |±i⟩ states), 05; page 20 (SDaggerGate.matrixapplied raw for the Y-basis measurement rotation, sign-checked against a known Y-eigenstate) - T / T† —
AdditionalGatesTests; page 05 (tapplied twice, T² == S —tdgisn't used on any page) - Phase P(θ) —
AdditionalGatesTests; pages 01, 05; page 16 (the building block of the hand-built controlled-phase CP(θ), five gates deep) - RX/RY/RZ (θ) —
AdditionalGatesTests; page 05; page 13 (ry/rzprepare the teleported payload); page 14 (rx(θ)as a partial bit-flip error); page 15 (ry(-θ)rotates into a tilted measurement basis); page 18 (ry(θ)is the VQE ansatz's only parameter); page 20 (RYGate/RZGate.matrix(θ)build a generic tilted state, raw); page 21 (RXGate.matrix(θ)self-checksMatrix.expm();RZGate.matrix(θ)is the core of the hand-derived ZZ-rotation identity — the actual Ising evolution now runs throughHamiltonian/evolveinstead) - CNOT (CX) — any distinct control/target pair; also
CNOTGate.matrix(qubits:control:target:). Bell example;BellStateTests,CNOTTests; pages 05, 07–11, 13–18; pages 19–21 (CNOTGate.matrix(qubits:control:target:)applied directly viaStateVector.apply(_:)for Bell-pair prep and the ZZ-rotation identity) - Toffoli (CCX) — any distinct control/control/target triple; also
ToffoliGate.matrix(qubits:control1:control2:target:).ToffoliTests; page 14 (three X-conjugated Toffolis build the syndrome-driven bit-flip correction) - MCX (Multi-controlled X) — any number of distinct controls; also
MultiControlledXGate.matrix(qubits:controls:target:)(0 controls = X, 1 = CNOT, 2 = Toffoli). page 14 (the "control on 0" halves of the syndrome correction);increment/decrementbelow are built on it - RZZ/RXX/RYY (θ) — any distinct qubit pair; fixed 2-qubit
exp(-iθ·P⊗P/2).TwoQubitRotationTests; page 21 (RZZGate'scx;rz;cxidentity is the exact building block of every Trotter step; also the generalpauliRotationcircuit methods below are built from the same basis-change + CNOT-staircase idiom)
QuantumCircuit also has a few operations that aren't tied to a single fixed-size gate enum:
pauliRotation(_ pauli:theta:) (the general Pauli-string rotation rzz/rxx/ryy are built
from), rotateToZ(_:_:) / measure(shots:basis:) (Pauli-basis measurement), evolve(_:time: steps:order:) (Trotterized Hamiltonian simulation), and increment/decrement(register: controlledBy:) (ripple-carry ±1 on a register, built from mcx). See API.md for
full signatures.
Hand-built gates — constructed in tests/playgrounds from raw Matrix values or gate
compositions and applied with circuit.apply(_:) (or, on page 19, which doesn't build a
QuantumCircuit at all, directly via StateVector/DensityMatrix); not (yet) part of Core.
A few entries that used to live here have since moved into Core — Matrix.expm(), the
RZZGate/rzz ZZ-rotation identity, Hamiltonian/PauliString, and the KrausChannel
factories — pages 18/19/21 now use those Core types directly (see the features list above),
though a couple of pages keep the original hand-built version around as a cross-check:
| Gate | Built from |
|---|---|
| Pauli-Y (Y) | raw 2×2 Matrix |
| CZ | h(1); cx(0,1); h(1) |
| Bell-basis projector | (Ket * Bra) ⊗ I₂ |
| 3-qubit code correction | one 32×32 Matrix |
| Tilted observable A(θ) | cos θ·Z + sin θ·X, built entrywise |
| CCZ | Matrix.identity(size: 8) with the |111⟩ entry set to −1 |
| Modular multiplication U_a (mod 15) | 16×16 / 128×128 permutations |
| QFT† (3-qubit inverse Fourier) | entrywise 8×8 inverse DFT ⊗ I₁₆ |
| Controlled phase CP(θ) | p(θ/2,c); cx(c,t); p(-θ/2,t); cx(c,t); p(θ/2,t) |
| ZZ-rotation exp(−iθ·Z⊗Z/2), by hand | cx(0,1); rz(θ,1); cx(0,1) |
| Coined-walk shift S, by hand | 32×32 permutation on (coin ⊗ 16-site position) |
Where:
- Pauli-Y (Y) —
DiracNotationTests(adjoint of a non-symmetric matrix — the test predatesPauliYGate) - CZ — page 11 (phase oracles and diffusion operator); page 13 (
h(2); cx(0,2); h(2)— the deferred Z^a correction) - Bell-basis projector — built as
(Ket("ab") * Bra("ab")) ⊗ Matrix.identity(size: 2). page 13 (recovering one measurement branch withoutmeasure()) - 3-qubit code correction — three X-conjugated
ToffoliGate.matrix/MultiControlledXGate.matrixcalls, combined into one 32×32Matrix. page 14 (syndrome-driven error correction viaapply(_:)) - Tilted observable A(θ) — page 15 (CHSH correlators; measured via
ry(-θ)) - CCZ — page 11 (3-qubit Grover finale)
- Modular multiplication U_a (mod 15) — one
.oneper column; the controlled versions key on a counting bit. page 12 (Shor order finding) - QFT† (3-qubit inverse Fourier) — built entrywise from
cos/sin. page 12 (phase-estimation readout) - Controlled phase CP(θ) — page 16 (the QFT ladder and standalone phase estimation; reduces to CZ at θ=π)
- ZZ-rotation exp(−iθ·Z⊗Z/2), by hand — page 21 (re-derives the identity and checks it against
Matrix.expm(), even thoughRZZGate/rzznow build the same thing in Core) - Coined-walk shift S, by hand — one
.oneper column. page 22 (kept only as a cross-check baseline; the walk itself now usesincrement/decrement, confirmed to agree exactly)
Custom operators on the quantum types (Ket = StateVector): the postfix dagger † is
declared in Sources/SwiftQiskit/Quantum/Dirac.swift, and the infix tensor product ⊗
(at MultiplicationPrecedence) in Sources/SwiftQiskit/Math/Matrix.swift:
| Operator | Expression | Result | Meaning | Defined in |
|---|---|---|---|---|
† |
Ket† |
Bra |
⟨ψ| = (|ψ⟩)† | Quantum/Dirac.swift |
† |
Bra† |
Ket |
|ψ⟩ = (⟨ψ|)† | Quantum/Dirac.swift |
† |
Matrix† |
Matrix |
adjoint U† (also Matrix.adjoint) |
Quantum/Dirac.swift |
⊗ |
Matrix ⊗ Matrix |
Matrix |
Kronecker product A ⊗ B (also tensor(_:)) |
Math/Matrix.swift |
⊗ |
Ket ⊗ Ket |
Ket |
|a⟩ ⊗ |b⟩ — combines registers, lhs in the high-order bits (also tensor(_:)) |
Quantum/StateVector.swift |
⊗ |
Bra ⊗ Bra |
Bra |
⟨a| ⊗ ⟨b| (also tensor(_:)) |
Quantum/Dirac.swift |
⊗ |
Ket ⊗ Bra |
Matrix |
mixed product = the outer product |a⟩⟨b| | Quantum/Dirac.swift |
⊗ |
Bra ⊗ Ket |
Matrix |
mixed product ⟨a| ⊗ |b⟩ = |b⟩⟨a| | Quantum/Dirac.swift |
* |
Bra * Ket |
Complex |
inner product ⟨φ|ψ⟩ | Quantum/Dirac.swift |
* |
Ket * Bra |
Matrix |
outer product |ψ⟩⟨φ| | Quantum/Dirac.swift |
* |
Bra * Matrix |
Bra |
⟨ψ|U — enables expectation values ψ† * U * ψ |
Quantum/Dirac.swift |
* |
Matrix * Matrix |
Matrix |
matrix product AB | Math/Matrix.swift |
+ |
Matrix + Matrix |
Matrix |
entrywise sum A + B | Math/Matrix.swift |
- |
Matrix - Matrix |
Matrix |
entrywise difference A − B | Math/Matrix.swift |
* |
Matrix * Double / Double * Matrix |
Matrix |
scalar multiple c·M | Math/Matrix.swift |
* |
Matrix * Complex / Complex * Matrix |
Matrix |
scalar multiple c·M | Math/Matrix.swift |
Scalar Complex arithmetic (+ - * / and Double scaling) lives in Math/Complex.swift
and is not listed here — it acts on numbers, not on qubit states or gates.
- No hidden magic — everything is explicit and readable
- Mathematical correctness over shortcuts
- Modular architecture (Core / Examples / GUI-ready)
- Designed for learning, experimentation, and extension
SwiftQiskit is not just a simulator — it’s an attempt to make quantum computing accessible, visual, and native on Apple platforms.
Enjoy exploring the quantum world
For a full listing of every public type and member in the core library, see API.md.
SwiftQiskit/
├── Sources/
│ ├── SwiftQiskit/
│ │ ├── Math/
│ │ │ ├── Complex.swift
│ │ │ └── Matrix.swift
│ │ ├── Quantum/
│ │ │ ├── StateVector.swift
│ │ │ ├── Dirac.swift
│ │ │ ├── SimulationResult.swift
│ │ │ ├── PauliBasis.swift
│ │ │ ├── PauliString.swift
│ │ │ ├── Hamiltonian.swift
│ │ │ ├── ParameterShift.swift
│ │ │ ├── StateTomography.swift
│ │ │ ├── DensityMatrix.swift
│ │ │ ├── KrausChannel.swift
│ │ │ └── NoiseModel.swift
│ │ ├── Gates/
│ │ │ ├── Hadamard.swift
│ │ │ ├── PauliX.swift
│ │ │ ├── PauliY.swift
│ │ │ ├── PauliZ.swift
│ │ │ ├── Phase.swift
│ │ │ ├── Rotation.swift
│ │ │ ├── CNOT.swift
│ │ │ ├── Toffoli.swift
│ │ │ ├── MultiControlledX.swift
│ │ │ └── TwoQubitRotation.swift
│ │ ├── Circuit/
│ │ │ ├── QuantumCircuit.swift
│ │ │ └── TensorNetwork.swift
│ │ └── Utils/
│ │ └── String+Padding.swift
│ └── SwiftQiskitViews/
│ ├── BlochVector.swift
│ ├── CHSHChartView.swift
│ └── TensorNetworkView.swift
├── Examples/
│ └── main.swift
├── Tests/
│ ├── SwiftQiskitTests/
│ │ ├── BellStateTests.swift
│ │ ├── TensorProductTests.swift
│ │ ├── DiracNotationTests.swift
│ │ ├── CNOTTests.swift
│ │ ├── AdditionalGatesTests.swift
│ │ ├── MatrixArithmeticTests.swift
│ │ ├── MeasurementTests.swift
│ │ ├── MatrixExponentialTests.swift
│ │ ├── TwoQubitRotationTests.swift
│ │ ├── ToffoliTests.swift
│ │ ├── ReadoutTests.swift
│ │ ├── PauliBasisTests.swift
│ │ ├── PauliStringTests.swift
│ │ ├── MeasureExpectationTests.swift
│ │ ├── StateTomographyTests.swift
│ │ ├── TrotterTests.swift
│ │ ├── ParameterShiftTests.swift
│ │ ├── RegisterArithmeticTests.swift
│ │ ├── DensityMatrixTests.swift
│ │ ├── KrausChannelTests.swift
│ │ ├── NoiseModelTests.swift
│ │ ├── CommutingGroupsTests.swift
│ │ ├── PermutationFastPathTests.swift
│ │ └── TensorNetworkTests.swift
│ └── SwiftQiskitViewsTests/
│ ├── BlochVectorTests.swift
│ ├── CHSHChartViewTests.swift
│ └── TensorNetworkViewTests.swift
├── PlaygroundDocs/
│ ├── 01QUBITSHELP.md
│ ├── 02BLOCH2DHELP.md
│ ├── 03BLOCH2DPROJECTIONHELP.md
│ ├── 04BLOCH3DHELP.md
│ ├── 05GATESHELP.md
│ ├── 06SUPERPOSITIONHELP.md
│ ├── 07ENTANGLEMENTHELP.md
│ ├── 08DIRACHELP.md
│ ├── 09TENSORPLAN.md
│ ├── 09TENSORHELP.md
│ ├── 10DEUTSCHPLAN.md
│ ├── 10DEUTSCHHELP.md
│ ├── 11GROVERPLAN.md
│ ├── 11GROVERHELP.md
│ ├── 12SHORPLAN.md
│ ├── 12SHORHELP.md
│ ├── 13TELEPORTATIONPLAN.md
│ ├── 13TELEPORTATIONHELP.md
│ ├── 14ERRORCORRECTIONPLAN.md
│ ├── 14ERRORCORRECTIONHELP.md
│ ├── 15CHSHPLAN.md
│ ├── 15CHSHHELP.md
│ ├── 16QFTPLAN.md
│ ├── 16QFTHELP.md
│ ├── 17DEUTSCHJOZSAPLAN.md
│ ├── 17DEUTSCHJOZSAHELP.md
│ ├── 18VQEPLAN.md
│ ├── 18VQEHELP.md
│ ├── 19NOISEPLAN.md
│ ├── 19NOISEHELP.md
│ ├── 20TOMOGRAPHYPLAN.md
│ ├── 20TOMOGRAPHYHELP.md
│ ├── 21TROTTERPLAN.md
│ ├── 21TROTTERHELP.md
│ ├── 22WALKPLAN.md
│ ├── 22WALKHELP.md
│ ├── 23TENSORNETWORKPLAN.md
│ ├── 23TENSORNETWORKHELP.md
│ ├── 41BASISTRANSFORMATIONSHELP.md
│ └── 90LIVEVIEWHELP.md (not page-numbered — sorts last on purpose; the shared-code/live-view guide)
├── Playgrounds.playground/
│ ├── Sources/ (code shared by all pages — see PLAYGROUNDSUPPORT.md)
│ └── Pages/
│ ├── 00TOC
│ ├── 01Qubits
│ ├── 02Bloch2d
│ ├── 03Bloch2dProjection
│ ├── 04Bloch3d
│ ├── 05Gates
│ ├── 06Superposition
│ ├── 07Entanglement
│ ├── 08Dirac
│ ├── 09Tensor
│ ├── 10DeutschExample
│ ├── 11GroverExample
│ ├── 12ShorExample
│ ├── 13Teleportation
│ ├── 14ErrorCorrection
│ ├── 15CHSH
│ ├── 16QFT
│ ├── 17DeutschJozsa
│ ├── 18VQE
│ ├── 19Noise
│ ├── 20Tomography
│ ├── 21Trotter
│ ├── 22Walk
│ ├── 23TensorNetwork
│ ├── 40ComplexAndMatrices
│ └── 41BasisTransformations
├── Package.swift
├── API.md (full public API reference for the core library)
├── CLAUDE.md (guidance for Claude Code working in this repo)
├── PLAYGROUNDSUPPORT.md (playground implementation reference)
├── STATUSandTODO.md (project status, roadmap, working TODO list)
└── References (tbd)
The package itself (Package.swift) declares swift-tools-version: 5.9 and targets
macOS 27+ / iOS 27+, matching the deployment target of its consumer apps
(SwiftQiskitApp, SwiftQiskitWalkDemo).
This fork's playground pages, however, are developed and tested against Xcode 27.0 beta and macOS 27 beta — some SwiftUI live-view pages need the beta-specific workarounds in PLAYGROUNDSUPPORT.md on Xcode 27 betas (confirmed still needed on beta 5, 27A5237l).
Open Xcode, go to Integrate and clone "https://github.com/SwiftProjectOrganization/SwiftQiskit".
swift run SwiftQiskitExamplesThe Bell state |Φ⁺⟩ is defined as:
|Φ⁺⟩ = (|00⟩ + |11⟩) / √2
import SwiftQiskit
let circuit = QuantumCircuit(qubits: 2)
circuit.h(0)
circuit.cx(0, 1)
let finalState = circuit.run()
print(finalState)
let result = circuit.measure(shots: 1000)
for (state, count) in result.sortedCounts {
let probability = Double(count) / Double(1000)
print("\(state): \(count) (\(String(format: "%.2f", probability)))")
}Note: This is the same code as
Examples/main.swift, run viaswift run SwiftQiskitExamples.
00: 498 (0.50)
11: 502 (0.50)
States 01 and 10 never appear — this confirms quantum entanglement. Measurement outputs are probabilistic and may vary per run.
Playgrounds.playground (at the repo root, macOS target) contains interactive, lecture-style
explorations of the library. Open it in Xcode — pages import SwiftQiskit and, on the pages
that use BlochVector/CHSHChartView, import SwiftQiskitViews too, so the active scheme
must build both products: use the SwiftQiskit-Package scheme (or the autogenerated
SwiftQiskitViews scheme), not the per-product SwiftQiskit scheme, which won't build
SwiftQiskitViews. Pages are linked sequentially with Previous/Next markers.
Code shared by multiple pages (the Bloch-sphere views) lives in the playground's
Sources/ folder — BlochVector and CHSHChartView themselves live in the SwiftQiskitViews
package product instead, so Core stays UI-free. See
PlaygroundDocs/90LIVEVIEWHELP.md for a user-facing
guide to that shared code and to putting a live view on a page, and
PLAYGROUNDSUPPORT.md for the terse implementation reference.
Clickable table of contents (markdown only): links to every page with a one-line
description, plus pointers to the guides in PlaygroundDocs/.
First look at qubit states through the Dirac API, shown in the results sidebar (no
console output): building Kets from amplitudes, the dagger †, inner and outer
products, probabilities, and tensoring a ket with itself. Content provisional.
User guide in PlaygroundDocs/01QUBITSHELP.md.
Visualizes single-qubit states on the Bloch sphere using a SwiftUI Canvas live view.
- Bloch vector math — maps a state |ψ⟩ = α|0⟩ + β|1⟩ to sphere coordinates
(x = 2·Re(ᾱβ), y = 2·Im(ᾱβ), z = |α|² − |β|²) plus the spherical angles θ and φ,
reusing the
Complexarithmetic fromSwiftQiskit. - Rendering — a 2D orthographic projection of the sphere with axes, drawn by the
shared
BlochSphereView, each sphere accompanied by a numeric readout. - Gallery — six canonical states built with real circuits and shown side by side: |0⟩ (north pole), |1⟩ via Pauli-X (south pole), |+⟩ via Hadamard (+x axis), |−⟩ via Hadamard + Pauli-Z (−x axis), |+i⟩ via Hadamard + S (+y axis), and |−i⟩ via Hadamard + S† (−y axis). The same vectors are also printed to the console.
User guide in PlaygroundDocs/02BLOCH2DHELP.md; the general recipe for putting a SwiftUI live
view on a playground page is in PlaygroundDocs/90LIVEVIEWHELP.md.
A general single-qubit state, tilted off the equator of the Bloch sphere (45° from x, 60° from y and z), explored in depth.
- Ket definition — derives |ψ⟩ = cos(θ/2)|0⟩ + e^{iφ}·sin(θ/2)|1⟩ from direction
cosines and builds the state directly from its amplitudes with
StateVector. - Console readout — amplitudes, magnitudes, probabilities, and a round-trip check recovering the Bloch vector from the amplitudes.
- Live view — the state on a large Bloch sphere plus two plane projections
(x–y seen from +z, z–y seen from +x) drawn by the shared
BlochProjectionView.
User guide in PlaygroundDocs/03BLOCH2DPROJECTIONHELP.md.
An interactive 3D Bloch sphere: a rotatable wireframe rendered with a pure SwiftUI
Canvas (no SceneKit/RealityKit), plus live sliders for the spherical angles.
- 3D rendering — latitude/longitude circles are perspective-projected through an orbit camera; drag the canvas to rotate. The far hemisphere is drawn dimmer as a depth cue, and dashed drop lines connect the state vector to the equator plane.
- θ/φ sliders — rebuild |ψ⟩ = cos(θ/2)|0⟩ + e^{iφ}·sin(θ/2)|1⟩ on every change. The two sliders are independent because the parametrization keeps |α|² + |β|² = cos²(θ/2) + sin²(θ/2) = 1 identically — every slider position is a valid normalized state, shown live in the numeric readout.
- Xcode 27 beta note — running SwiftUI playground pages on the Xcode 27 beta
currently needs two workarounds, described in
PLAYGROUNDSUPPORT.md: a shim
libcups.dylibin DerivedData, and keeping@State-based views in the playground'sSources/folder (which is why the slider viewBlochExplorerViewlives there). Both are confirmed still present on beta 5 (27A5237l).
User guide: PlaygroundDocs/04BLOCH3DHELP.md.
A gentle, gate-by-gate tour of the built-in gate set in the results sidebar (no live
view, no prints): x, h, z (with the interference reveal that makes its phase flip
visible), y, s/sdg, t (applied twice to show T² == S), the general phase gate
p(theta:), and the rotations rx/ry/rz, each shown individually on a 1-qubit
QuantumCircuit, plus a one-line h+cx Bell-state teaser pointing to 07Entanglement.
User guide: PlaygroundDocs/05GATESHELP.md.
A 4-qubit console walkthrough: every qubit put into superposition via h, inspecting the
resulting 16-state amplitudes/probabilities and a 1600-shot measurement, plus a
partial-superposition (2-qubit) contrast. User guide:
PlaygroundDocs/06SUPERPOSITIONHELP.md.
Annotated walkthrough of the Bell state |Φ⁺⟩: builds the circuit (h + cx), inspects the
resulting state vector and its amplitudes/probabilities, and runs a 1000-shot measurement.
A GHZ section extends the recipe to 3 qubits — cx(0, 2) spans non-adjacent qubits — and
the page closes by rebuilding the Bell state via apply(CNOTGate.matrix) to show the
matrix form agrees with the fluent cx API.
User guide in PlaygroundDocs/07ENTANGLEMENTHELP.md.
Dirac-notation walkthrough of Quantum/Dirac.swift:
- Bras and kets — basis kets via
Ket("01")and the named states.zero/.one/.plus/.minus/.plusI/.minusI; the postfix dagger†turns aKetinto aBra(and givesMatrix.adjoint). - Products — inner products
Bra * Ket(orthonormality checks) and outer productsKet * Bra(projectors, completeness). - Expectation values — recovers the page-04 initial qubit's Bloch coordinates
as the Pauli expectation values ⟨ψ|X|ψ⟩, ⟨ψ|Y|ψ⟩, ⟨ψ|Z|ψ⟩, shown on a static
Bloch3DView.
User guide in PlaygroundDocs/08DIRACHELP.md.
Tensor-product walkthrough (console only), mirroring
Tests/SwiftQiskitTests/TensorProductTests.swift section by section:
tensor(_:)/⊗onMatrixandStateVector, and the mixed-product identity (A ⊗ B)(C ⊗ D) = (AC) ⊗ (BD).- Gate embedding — building H ⊗ I by hand and checking it matches what
circuit.h(0)applies across a 2-qubit register. - Entanglement — why the Bell state cannot be factored as a tensor product of single-qubit states.
Design notes in PlaygroundDocs/09TENSORPLAN.md; user guide in PlaygroundDocs/09TENSORHELP.md.
Deutsch's algorithm (console only) — deciding whether a black-box function f: {0,1} → {0,1} is constant or balanced with a single oracle query:
- The four oracles — every 1-bit function's oracle U_f built from gates the
library already has: identity,
x(1),cx(0,1), andcx(0,1)+x(1). - Phase kickback — a stage-by-stage state-vector walkthrough showing how the |−⟩ ancilla turns the oracle into a phase (−1)^f(x) on the input qubit.
- Deterministic verdict — the final Hadamard maps the phase to qubit 0, so one measurement reads off constant (0) vs balanced (1) with certainty, confirmed for all four oracles and backed by shot statistics.
Design notes in PlaygroundDocs/10DEUTSCHPLAN.md; user guide in PlaygroundDocs/10DEUTSCHHELP.md.
Grover's search (console only) — finding a marked basis state with quadratically fewer oracle queries:
- CZ from existing gates — the controlled-Z built as
h(1); cx(0,1); h(1), then conjugated by X gates to make a phase oracle for any marked state |w⟩. - Inversion about the mean — an amplitude-by-amplitude walkthrough of one Grover iteration, with exact 1-iteration success on 2 qubits and what happens when you over-rotate by iterating further.
- The diffusion operator in Dirac notation — 2|s⟩⟨s| − I assembled directly from
the outer product in
Quantum/Dirac.swiftand checked against the gate construction. - 3-qubit finale — Grover on 8 states using a hand-built CCZ matrix applied via
apply(_:), with the theoretical success probability after each iteration.
Design notes in PlaygroundDocs/11GROVERPLAN.md; user guide in PlaygroundDocs/11GROVERHELP.md.
Compiled Shor's algorithm (console only) — factoring 15 by quantum order finding, with a 3-qubit counting register and a 4-qubit work register:
- Factoring reduces to order finding — the classical gcd reduction, plus the "lucky guess" cases where no quantum computer is needed at all.
- Modular multiplication as permutation matrices — U_a |w⟩ = |a·w mod 15⟩ and its
controlled powers hand-built (one
.oneper column) and applied viaapply(_:), with the orbit |1⟩ → |7⟩ → |4⟩ → |13⟩ → |1⟩ exposing the order geometrically. - A hand-built QFT† — the 8×8 inverse DFT constructed entrywise on the register's integer index (no bit-reversal bookkeeping), checked against Hadamard and unitarity.
- Phase estimation stage by stage — superposed counts, the entangled orbit, then
exact peaks at y = 8·s/r; shot statistics sampled from one
run()(measure(shots:)now does the same thing internally, but the page keeps its own sampler visible since it also produces the marginal-over-y summary this section wants). - Classical post-processing — measured phase → lowest terms → verified order → gcd factors, then a sweep of every coprime base including the instructive a = 14 failure (a^(r/2) ≡ −1).
Design notes in PlaygroundDocs/12SHORPLAN.md; user guide in PlaygroundDocs/12SHORHELP.md.
Quantum teleportation and its dual, superdense coding — entanglement used as a communication resource, with a Bloch-sphere live view:
- Teleportation on 3 qubits — Alice's payload, a shared Bell pair, her Bell-basis
rotation (
cx(0,1); h(0)), and Bob's X^b Z^a correction. - Measurement branches without measuring — the four outcomes recovered with Dirac projectors (|ab⟩⟨ab|) ⊗ I₂, showing P(ab) = ¼ regardless of |ψ⟩ (no signalling) and fidelity 1 once each branch gets its own correction.
- Deferred measurement — classical feedback replaced by
cx(1,2)and CZ(0,2), after which the register factors exactly as |+⟩ ⊗ |+⟩ ⊗ |ψ⟩ (~8e-17). - No cloning, concretely — Bob's marginal reproduces |ψ|² while Alice's qubit is left in |+⟩: the state moved rather than copied.
- Superdense coding — two classical bits carried by one qubit, decoded with certainty, with the Bell basis's Gram matrix printed as the identity to show why.
Design notes in PlaygroundDocs/13TELEPORTATIONPLAN.md; user guide in PlaygroundDocs/13TELEPORTATIONHELP.md.
The 3-qubit bit-flip/phase-flip repetition code — how a quantum computer protects one fragile qubit without ever looking at it directly, with a Bloch-sphere live view:
- Encode and extract a syndrome —
cx-based encoding and two ancilla parities that name the flipped qubit (or "none") without touching α or β. - A hand-built correction — a 32×32 permutation, built from three X-conjugated
ToffoliGate/MultiControlledXGatecalls (one per syndrome branch), that flips whichever qubit the syndrome accuses, applied viaapply(_:)as a single combined operation. - Continuous errors, digitized exactly — an
rx(θ)sweep shows the coherent correction restoring fidelity 1.0000 at every θ, while the syndrome ancillas alone carry the cos²(θ/2)/sin²(θ/2) branch weights. - Where distance 3 breaks — two simultaneous errors alias to the wrong syndrome,
producing a silent, fully "corrected" logical X; the exact logical error rate
p_L = 3p² − 2p³ is confirmed by enumeration, then reproduced a second way by driving the
page's own circuit through a
KrausChannel.bitFlipNoiseModel, exactly (runDensityMatrix) and by Monte Carlo (runTrajectories). - Phase flips for free — Hadamard-conjugating the same code (H Z H = X) turns a Z error into the X error the rest of the page already fixes.
Design notes in PlaygroundDocs/14ERRORCORRECTIONPLAN.md; user guide in PlaygroundDocs/14ERRORCORRECTIONHELP.md.
The CHSH inequality — whether a Bell pair's correlations could ever come from a shared classical instruction list — with a live chart of the violation:
- The classical bound, exhaustively — all 16 deterministic ±1 strategies checked by brute force (max |S| = 2), plus a shared-direction hidden-variable model that saturates the bound and doubles as the chart's classical comparison curve.
- A pinned measurement convention — the tilted observable A(θ) = cos θ·Z + sin θ·X,
built entrywise, measured via
ry(-θ), with the sign checked against the exact expectation value (and against page 04/08's ⟨Z⟩/⟨X⟩ for the same qubit) before it's trusted. - Correlators two ways — exact via
psi† * (A(a) ⊗ A(b)) * psiand sampled viameasure(shots:), agreeing with cos(a−b). - The violation and its limits — a Bell pair's S = 2√2 against a product-state control (S = √2) and a fine angle sweep confirming the Tsirelson ceiling of 2√2, never higher.
- A live chart —
CHSHChartView(sharedSwiftQiskitViewstype) plots the exact cos θ curve, sampled points, and the classical line together.
Design notes in PlaygroundDocs/15CHSHPLAN.md; user guide in PlaygroundDocs/15CHSHHELP.md.
The quantum Fourier transform as a gate circuit (console only) — closing the gap page 12 left open when it built the QFT as a single entrywise matrix:
- The missing gate — controlled phase CP(θ) derived from
p+cxalone (p(θ/2,c); cx(c,t); p(-θ/2,t); cx(c,t); p(θ/2,t)), checked against CZ at θ = π. - The QFT ladder — Hadamards and CP's per qubit, plus a swap network, checked against page 12's entrywise DFT to ~1e-15 on every basis state.
- Why the swaps — dropping them reproduces the exact bit-reversal of the correct output.
- The inverse QFT and unitarity — QFT then QFT† returns every basis state to itself.
- Standalone phase estimation — exact recovery of dyadic phases, a spread for phases that aren't, and a precision comparison at 3 vs. 6 counting qubits.
Design notes in PlaygroundDocs/16QFTPLAN.md; user guide in PlaygroundDocs/16QFTHELP.md.
Deutsch–Jozsa and Bernstein–Vazirani (console only) — page 10's algorithm generalized from 1 bit to n:
- The n-qubit circuit — page 10's shape widened to n input qubits + 1 ancilla.
- Oracles from
cx— constant and balanced functions built the same way page 10 did. - The verdict — P(all-zero input) is exactly 1 or 0, from a single query, for any n.
- A shot-sampling gotcha — the ancilla's bit is a free coin flip; only the input bits are
deterministic in
measure(shots:)output. - Bernstein–Vazirani — the identical circuit recovers an entire hidden n-bit string in one query.
- The query-count gap — quantum stays at 1 while classical Deutsch–Jozsa's worst case grows exponentially and classical Bernstein–Vazirani grows linearly.
Design notes in PlaygroundDocs/17DEUTSCHJOZSAPLAN.md; user guide in PlaygroundDocs/17DEUTSCHJOZSAHELP.md.
The variational quantum eigensolver — the one page where the circuit isn't fixed in advance, with a live chart of the optimization:
- The Hamiltonian — the qubit Hamiltonian for H₂ (Jordan–Wigner, minimal basis), six
PauliStringterms summed into aHamiltonian. - A one-parameter ansatz —
x(0); ry(θ,1); cx(1,0), provably confined to the {|01⟩,|10⟩} subspace. - The energy —
Hamiltonian.expectation(_:), page 08's Dirac expectation-value idiom summed over terms. - The exact answer — a closed-form 2×2 eigenvalue, used only to grade the optimizer.
- Parameter-shift gradients — exact, not approximate, for a single-rotation ansatz, via
ParameterShift.gradient(at:_:); pinned against a finite difference. - Gradient descent —
GradientDescent.minimizeconverges to the exact ground energy (error 0.00e+00) in ~10 steps. - A live chart — the E(θ) landscape and the optimizer's own visited points, on the shared
CHSHChartView.
Design notes in PlaygroundDocs/18VQEPLAN.md; user guide in PlaygroundDocs/18VQEHELP.md.
Open systems, built on Core's DensityMatrix/KrausChannel types, plus a live Bloch gallery:
- ρ and coherence — the density matrix ρ = |ψ⟩⟨ψ| via
DensityMatrix(_:)(itself the existingKet * Braouter product); a classical mixture ½|0⟩⟨0| + ½|1⟩⟨1| (viaDensityMatrix(mixture:)) contrasted against the superposition |+⟩⟨+| — identical Z-statistics, different off-diagonals. - Kraus channels — the four
KrausChannelfactories (bit-flip, phase-flip, depolarizing, amplitude damping), each checked for trace preservation viaisTracePreserving(). - Decoherence, exactly — coherence decaying as (1−2p)ⁿ under repeated dephasing, and amplitude damping pulling the Bloch vector inside the sphere — the picture no pure state can draw.
- A Monte-Carlo unraveling — the exact channel reproduced from ordinary pure-state code: flip a coin per shot, apply the error gate or not, then measure.
- Entanglement via a reduced state —
DensityMatrix.partialTrace(keeping:)on one qubit of a Bell pair gives entropy exactly 1 bit (.vonNeumannEntropy), against 0 for a product state — the explanation page 13's marginals were owed.
Design notes in PlaygroundDocs/19NOISEPLAN.md; user guide in PlaygroundDocs/19NOISEHELP.md.
Reconstructing a state from measure(shots:) statistics alone — the honest version of "what a
real device gives you," depending on page 19's mixed states for its sharpest result:
- Basis rotations, pinned by hand —
hfor X,sdg+hfor Y (matching Core'srotateToZ(_:_:)), checked against a known Y-eigenstate rather than assumed. - The estimator and its 1/√N error —
StateTomography.estimate/estimateBlochVector's RMS error against the exact expectation value falls by roughly √10 each time the shot count grows tenfold. - Pure vs. mixed unphysical estimates — a pure state's per-axis reconstruction lands outside the Bloch ball about half the time at any N (it sits exactly on the boundary); only a genuinely mixed state's frequency shrinks toward zero.
- An entangled qubit's marginal, from shots — a Bell pair's qubit-0 Bloch vector reconstructs to the origin, restating page 13's no-cloning result statistically.
- Why full tomography doesn't scale — a 3ⁿ-settings cost table, motivating page 18's per-term Pauli measurements.
Design notes in PlaygroundDocs/20TOMOGRAPHYPLAN.md; user guide in PlaygroundDocs/20TOMOGRAPHYHELP.md.
Hamiltonian simulation — evolving a state in time under a Hamiltonian too large for a single gate, the original motivation for quantum computers, with a live chart:
Matrix.expm(), self-checked — Core's matrix exponential (scaling-and-squaring Taylor series) validated against Core's exactRXGatebefore being trusted as ground truth.- An exact gate identity — exp(−iθ·Z⊗Z/2) =
cx(0,1); rz(θ,1); cx(0,1), re-derived from Core'sRZGateand checked againstexpm, not assumed — the same identity Core'sRZZGate/rzzbuild directly. - Trotter error scaling — the target is now a
Hamiltonian(threePauliStringterms) evolved viaQuantumCircuit.evolve(_:time:steps:order:); first-order error shrinking as O(1/n), second-order (Suzuki) as O(1/n²), at the observable level (⟨Z₀⟩(t)) as well as the operator level. - Why the error exists — the non-zero commutator [Z⊗Z, X⊗I] identified as the cause; a commuting-only Hamiltonian is exact at n=1.
Design notes in PlaygroundDocs/21TROTTERPLAN.md; user guide in PlaygroundDocs/21TROTTERHELP.md.
The discrete-time quantum walk — interference producing a distribution, rather than answering an oracle question or amplifying a marked item, with a live chart:
- The shift, as a permutation — a conditional-shift permutation on a 16-site cycle built
from Core's
increment/decrement(register:controlledBy:), cross-checked for exact equality against the page's original hand-built permutation and confirmed unitary (S†S = I). - Ballistic vs. diffusive spreading — the quantum walk's spread grows roughly linearly in t; a classical random walk's grows as exactly √t, at every step.
- A cyclic-coordinate gotcha, caught and documented — computing spread from raw site indices breaks near the cycle's wraparound boundary; the fix is a signed offset from the start.
- Interference, not asymmetry — an |0⟩ coin gives a lopsided distribution; Core's existing
|+i⟩basis ket restores left-right symmetry exactly.
Design notes in PlaygroundDocs/22WALKPLAN.md; user guide in PlaygroundDocs/22WALKHELP.md.
Building and drawing a tensor network from a circuit — a second, independent way to read a
circuit, alongside run()'s full-matrix replay, with a live gallery:
- Wires as edges, gates as tensors — every qubit wire is a bond-dimension-2 edge, a
k-qubit gate is a rank-2k tensor (Core's own un-embedded local matrix for that gate), and
each qubit's
|0⟩start is a cap on the left of its wire. TensorNetwork(circuit).contract(), checked againstrun()— evaluated from nothing but those small local tensors and the wiring, never the full embedded matricesrun()uses, so agreement between the two is a genuine cross-check.- Four worked networks — a Bell pair, a GHZ state with a non-adjacent
cx(page 07, drawn with its skipped wire crossing the gate's box dashed),rzzunfolding into exactly thecx;rz;cxidentity (page 21), and a small 2-qubit QFT ladder (page 16). - A bug found along the way —
t()was tagging every qubit for noise purposes instead of just its own; fixed, with a regression test.
Design notes in PlaygroundDocs/23TENSORNETWORKPLAN.md; user guide in
PlaygroundDocs/23TENSORNETWORKHELP.md.
The first of the 40+ pages, numbered separately from 01–22 because they accompany chapters
of the SwiftQiskitApp INTRODUCTION.md book rather than continuing that sequence: every code
fragment from the book's Chapter 2, in order — Complex construction/*///conjugate/
magnitude, the Born rule, phase via Euler's formula, plain [Complex] vector helpers, inner
products and normalization, Matrix as a transformation and why multiplication order matters,
unitarity and why exact == lies on H, and the tensor product ⊗ (console only, no live
view, no companion PlaygroundDocs/ guide).
Standalone, not tied to a book chapter (console only): expressing |ψ⟩ in a different basis by
building the change-of-basis matrix T from the new basis kets as columns, constructing T†
by hand as the stack of the new bras and checking it against Matrix.adjoint, reading off the
new amplitudes as inner products ⟨bⱼ|ψ⟩, the {|+⟩, |−⟩} case reducing exactly to H, why the
complex {|+i⟩, |−i⟩} case needs the conjugate transpose (a plain transpose silently swaps
|+i⟩'s probabilities and fails the unitarity check), and measuring in the new basis via
h/sdg;h before an ordinary measure(shots:).
User guide in PlaygroundDocs/41BASISTRANSFORMATIONSHELP.md.
The Bloch views (BlochSphereView, BlochProjectionView, Bloch3DView, BlochExplorerView)
are shared between these pages via the playground's Sources/ folder (not part of Core).
BlochVector, the shared 2D chart (CHSHChartView, used by pages 15, 18, 21, and 22), and the
circuit-aligned TensorNetworkView (page 23) live instead in the SwiftQiskitViews package
product — a small SwiftUI library depending only on SwiftQiskit, also used by the sibling
SwiftQiskitApp repo, so Core itself stays UI-free —
see PlaygroundDocs/90LIVEVIEWHELP.md for a user guide to
each type and PLAYGROUNDSUPPORT.md for the implementation reference.
BlochVector gained an additive init(x:y:z:) for page 19's mixed-state (sub-unit-length)
vectors, used by pages 19 and 20, and DensityMatrix-driven initializers when it moved into
SwiftQiskitViews; every earlier call site is unaffected.
Contributions, ideas, and discussions are welcome. This project is built step by step and open for exploration.
Project status, what works in v0.2, and the roadmap live in STATUSandTODO.md, together with this fork's working TODO list.
MIT License © 2025 Ali Nasser