The spy detector
★★☆In a balanced Mach-Zehnder (\(\varphi = 0\)), a perfect detector is placed on one of the arms. It signals the passage of the photon without absorbing it. What probabilities are observed at the outputs?
Chapter 1 · Lectures 4 h + Tutorials 2 h
A single photon, two paths, and an interference pattern that builds up dot by dot. The rest of the course follows from this observation.
Send photons one at a time towards two slits. Each one lands at a single point on the screen, like a particle. Yet after a few thousand impacts, fringes appear, as they would for a wave.
To describe this, we do not add probabilities. We give each path an amplitude, which is a complex number. When the paths are indistinguishable, we add the amplitudes, then take the squared modulus. This is the Born rule. Two opposite amplitudes cancel out, and the fringe is dark.
As soon as a physical trace reveals which path was taken, the fringes fade. This is also what happens to Schrödinger's cat: its surroundings observe it all the time. We call this decoherence.
The qubit puts this idea to work for information. Its state is written alpha zero plus beta one, where the modulus of alpha squared plus the modulus of beta squared equals one. We represent it as a point on the Bloch sphere. Be careful, though. A measurement gives only one outcome, and a qubit never yields more than one bit of classical information.
Einstein“God does not play dice. Every particle must have a well-defined position and velocity, even if we do not know them.”
Bohr“Albert, stop telling God what to do. A quantum object is neither a wave nor a particle: it shows us one face or the other depending on the question we ask it.”
In 1801, Thomas Young sends light through two slits: bright and dark fringes appear, the signature of a wave. Let us repeat the experiment, sending the photons one at a time. Classical intuition predicts two spots, one behind each slit. That is not what we observe.
Before running the lab: write down your prediction, run it, compare, then explain the gap in one sentence. Where will the first ten hits land: at random or in fringes? What will we see after several thousand?
counted impactsprobability distribution
Push D towards 1: a detector placed behind the slits indicates the path taken with growing certainty, and the fringes fade. The complementarity relation \(D^2 + V^2 \le 1\) (Englert, 1996) forbids knowing the path perfectly while keeping high-contrast fringes.
Interference appears when several indistinguishable paths lead to the same detection. The more a physical trace makes the path knowable, the more the fringes fade; they vanish completely when the path can be determined perfectly.
To control the paths precisely, we use two 50/50 beam splitters (BS), two mirrors (M) and a phase plate that adds a phase shift \(\varphi = 2\pi\,\Delta\ell/\lambda\) on one arm. The lecture notes describe four set-ups:
measured frequencypredicted probability
Without a which-path detector, the photon is represented on both arms at once: only the recombination at BS₂ fixes the probabilities. Add the detector, or remove BS₂, and you get 50/50 again, whatever the value of φ.
Interference of a single photon in a Mach-Zehnder was cleanly demonstrated by Grangier, Roger and Aspect in 1986, with a source that truly emits only one photon at a time. In 1999, Arndt and Zeilinger’s team in Vienna observed interference with fullerene molecules C60: 60 carbon atoms, about 1,080 nucleons and electrons.
In a balanced Mach-Zehnder (\(\varphi = 0\)), a perfect detector is placed on one of the arms. It signals the passage of the photon without absorbing it. What probabilities are observed at the outputs?
A probability is a positive real number: two probabilities cannot cancel each other. To describe a dark fringe, we need an object that adds or subtracts, like a wave. So we associate a complex number with each path: the amplitude.
The state of a system is described by a ket \(\ket{\psi}\). The transition from a state \(\ket{i}\) to a state \(\ket{f}\) is described (in the absence of any intermediate evolution) by the complex amplitude
Born rule: \(P(f \leftarrow i) = |\braket{f|i}|^2\).
Paths in parallel (indistinguishable): add the amplitudes. Steps in series: multiply the amplitudes. Probability: take the squared modulus of the total amplitude.
Drag the tips of \(\mathcal A_1\) and \(\mathcal A_2\), or start from a case in the lecture notes:
The amplitudes are not normalised here: read the squares as relative intensities. The difference between the “sum of the probabilities” and the “square of the sum” is the interference term \(2\,\mathrm{Re}(\mathcal A_1 \mathcal A_2^*)\). It is positive when the phases agree and negative when they oppose each other. That is the whole secret of the fringes.
A beam splitter transmits with amplitude \(1/\sqrt2\) and reflects with amplitude \(i/\sqrt2\) (a phase shift of \(\pi/2\) on reflection). Following the four paths of MZ4, the lecture notes obtain
The phase plate introduces \(\varphi = \pi/3\). What is the probability of detection at \(D_x\)? (decimal or fraction)
Set φ to 60° in lab 1.2 and send 500 photons: the red mark shows the expected value.
An electron crosses a unit cell by two paths with equal amplitudes \(\mathcal A_0\). A defect shifts the phase of one of them: \(\mathcal A_\text{tot} = \mathcal A_0(1 + e^{i\varphi})\). Show that \(P = 4|\mathcal A_0|^2\cos^2(\varphi/2)\), then give the smallest value \(\varphi > 0\) (in radians) that cancels the transmission.
When random disorder makes these cancellations systematic, the electron stays trapped: this is Anderson localisation (Nobel Prize 1977).
In 1935, Schrödinger imagines a cat shut in a box with a radioactive atom, a Geiger counter and a vial of poison. If we apply the theory literally, the cat ends up described by
More precisely, the cat is entangled with the atom: the overall state is \(\tfrac{1}{\sqrt2}\big(\ket{\text{decayed},\text{dead}} + \ket{\text{intact},\text{alive}}\big)\) (see chapter 4). Schrödinger wanted to show that the Copenhagen interpretation was absurd. The irony is that his cat became the standard illustration of superposition.
| Superposition (the AND) | Statistical mixture (the OR) | |
|---|---|---|
| Nature | A single state that combines A and B | A or B, but we do not know which |
| Example | \(\tfrac{1}{\sqrt2}(\ket{0} + \ket{1})\) | Hidden coin: heads or tails |
| Interference | Yes, the amplitudes combine | No, the probabilities add |
A cat is never isolated: air molecules and infrared photons “observe” it constantly. Each interaction carries a little which-path information away into the environment. This is exactly what the slider D in lab 1.1 does: when the information leaks out, the fringe visibility drops to zero and the state behaves like a statistical mixture. For a macroscopic object, this takes a tiny fraction of a second. It is also the first enemy of quantum computers.
Each qubit rotates about the \(z\) axis at a slightly different rate (standard deviation σ), because the environment is not known. Individually the state stays pure; for the ensemble, the mean Bloch vector shrinks as \(r(t)=e^{-\sigma^2t^2/2}\) and the purity falls from 1 towards ½: coherent superposition → mixture. The fringe visibility of an interferometer follows \(r\). This is decoherence, with no qubit being “measured”.
We prepare 1,000 qubits in \(\tfrac{1}{\sqrt2}(\ket0 + \ket1)\), and 1,000 others by tossing a coin for each one to choose between \(\ket0\) and \(\ket1\). Which experiment can tell the two batches apart?
We “mark” the path of each photon, and the fringes disappear. If we then erase this marking and sort the impacts according to the result of the erasure, the fringes reappear in each subset. This does not allow a message to be sent into the past: without the sorting, done afterwards with information sent by classical means, the screen shows only a blur without fringes.
A qubit is a two-level quantum system with basis states \(\ket0\) and \(\ket1\). A pure state is written
Up to a global phase (which has no physical effect), we can always write \(\alpha = \cos\frac\theta2\) and \(\beta = e^{i\varphi}\sin\frac\theta2\): a point \((\theta, \varphi)\) on the Bloch sphere. Statistical mixtures (chapter 4) occupy the interior of the sphere.
One sometimes reads that a qubit “stores an infinite amount of information” because \(\alpha\) and \(\beta\) are continuous. This is misleading: no more than one bit of classical information can be extracted from a qubit (the Holevo bound), and a measurement projects the state onto one of the possible outcomes. The power of quantum computing comes from interference between amplitudes, not from reading out “all the results in parallel”.
The states \(\ket+\) and \(\ket{+i}\) both give 50/50 in the Z basis, but not in the X or Y basis: the relative phase φ is invisible in one basis and decisive in another.
Let \(\ket{\psi} = \frac{i}{2}\ket0 + \frac{\sqrt3}{2}\ket1\). What is the probability of obtaining \(0\)?
We want a qubit that gives \(0\) one time in four in the Z basis. What angle θ (in degrees) is needed? Then check in lab 1.4 with 1,000 copies.
The gate \(H\) acts as \(H\ket k = \frac{1}{\sqrt2}\big((-1)^k\ket k + \ket{1-k}\big)\). What does \(\ket0\) become?
In a cold-atom interferometer, \(\varphi = kB\) and \(P_1(B) = \sin^2(kB/2)\). For the best sensitivity, we work at the point where \(|\mathrm dP_1/\mathrm dB|\) is largest. Which bias field \(B_0\) should we choose?
In the basis \(\{\ket x, \ket y\}\), the beam splitter is written \(B = \frac{1}{\sqrt2}\begin{pmatrix}1 & i\\ i & 1\end{pmatrix}\), the pair of mirrors \(M = \begin{pmatrix}0 & i\\ i & 0\end{pmatrix}\) and the phase plate \(\Phi = \mathrm{diag}(e^{i\varphi}, 1)\). Calculate \(\ket{\psi_\text{out}} = B\,M\,\Phi\,B\ket x\) and recover \(P(x) = \cos^2(\varphi/2)\) (up to the labelling of the outputs).
Start with \(B\ket x = \frac{1}{\sqrt2}\binom{1}{i}\). Keep \(e^{i\varphi}\) factored out until the end, then use \(1 + e^{i\varphi} = 2e^{i\varphi/2}\cos(\varphi/2)\).
\(\Phi B\ket x = \frac{1}{\sqrt2}\binom{e^{i\varphi}}{i}\), then \(M\Phi B\ket x = \frac{1}{\sqrt2}\binom{-1}{ie^{i\varphi}}\), and finally
Hence \(P(x) = \cos^2(\varphi/2)\) and \(P(y) = \sin^2(\varphi/2)\), as in the lecture notes.
A message encrypted with RSA-2048 is intercepted today and stored (“harvest now, decrypt later”). Which statement is correct?
Analogy: “in a Ndolé, the bitterness of the leaves and the sweetness of the peanut offset or reinforce each other depending on quantity and timing”. Which assessment is correct?
A single photon at a time crosses the interferometer, and fringes appear after thousands of photons. Which conclusion is correct? Discuss in pairs before answering.
Which observation distinguishes the state \((\ket0+\ket1)/\sqrt2\) from a statistical mixture of half \(\ket0\), half \(\ket1\)? Discuss in pairs before answering.
Choose one analogy from this chapter. With the person’s consent, who may be an elder, a practitioner, a craftsperson or a musician, ask: (1) how they present this notion, in what language and with what words; (2) under what conditions this knowledge is passed on; (3) what, in their view, does not translate into physics. Write down the words in the local language, return your file to the person, then present both systems, each coherent by its own criteria, without concluding that one proves the other.
Main ideas.
Formulas.
Pitfalls. Confusing superposition with classical ignorance (only a superposition interferes); forgetting Holevo’s bound: despite two continuous complex numbers, a measurement on one qubit yields at most 1 classical bit; believing a measurement reveals a pre-existing value, whereas an incompatible quantity has no definite value before the measurement.
| Term | Terme français | Definition |
|---|---|---|
| Probability amplitude | Amplitude de probabilité | Complex number \(c_k\) with \(P(k)=|c_k|^2\). |
| Decoherence | Décohérence | Leak of information to the environment: a superposition becomes a mixture. |
| Wave-particle duality | Dualité onde-corpuscule | Interference or a point-like impact depending on the measurement; \(\lambda=h/p\). |
| Hilbert space | Espace de Hilbert | Complete complex vector space with an inner product. |
| Qubit | Qubit | Two-level system, \(\alpha\ket0+\beta\ket1\). |
| Bloch sphere | Sphère de Bloch | Representation of the pure states of a qubit on \(S^2\). |
| Superposition | Superposition | Linear combination of basis states, \(\sum_ic_i\ket{\phi_i}\). |
The cold-atom interferometers of this chapter measure gravity. In 2022 a quantum gravity gradiometer surveyed an 8.5 m line at 0.5 m spatial resolution and detected a 2 m tunnel (signal-to-noise ratio 8): mapping the subsurface (cavities, aquifers, ores) is an intended use.
Stray et al., Nature 602, 590 (2022)John Clarke, Michel Devoret and John Martinis were awarded the prize “for the discovery of macroscopic quantum mechanical tunnelling and energy quantisation in an electric circuit”. Their superconducting circuits of the 1980s are the direct ancestors of the Google and IBM qubits.
Nobel Committee press release, 7 October 2025 — nobelprize.orgIn 2024, Google showed on its Willow chip that the error of a logical qubit decreases as the error-correcting code is enlarged: it is divided by about 2.1 each time the code distance is increased. In November 2025, Quantinuum presented Helios (98 barium ions, two-qubit gates with 99.92% fidelity). Take care not to confuse the hundreds of physical qubits in these machines with the few logical qubits that are actually protected.
Nature 638, 920 (2025); Nature 655, 81 (2026)The Mach-Zehnder formula \(P = \cos^2(\varphi/2)\) lies at the heart of commercially available cold-atom gravimeters, which measure \(g\) to within about \(10^{-8}\ \mathrm{m/s^2}\), and of atomic magnetometers for brain imaging without cryogenics.
The International Year of Quantum Science and Technology (2025) closed in Accra (Ghana) in February 2026. The UNESCO review notes that activity was more than seven times less dense in Africa than in Europe and North America. Closing this gap also depends on courses like this one.
UNESCO, The Quantum moment: a global report, May 2026