How Science Works | Quantum Mechanics — States, Interference, Measurement and the Evidence Behind the Theory

Reading route: what entanglement allows. Read sections 19–21 for joint correlations, Bell tests and the limit on faster-than-light messages, then section 23 for experimental checks. For a worked two-qubit state, see Entanglement is not ordinary correlation. Return to the How Science Works master guide to choose another mechanism.

Reading route: Describe the experiment → Understand amplitudes → Follow energy and confinement → Test correlations → Audit evidence → Learn and check understanding.

1. Begin with a preparation and a question, not a mysterious object

A quantum experiment needs a preparation procedure, a specified evolution or interaction, a measurement choice and a record of outcomes. “An electron is in the apparatus” is not a complete description. Its prepared state and the question asked of it matter. Likewise, changing the measurement is not merely looking at the same hidden answer from another angle.

The state is the mathematical object used to assign measurement probabilities within the theory. A wavefunction is one representation for suitable systems; it is not the only possible representation. The operational connection between state and measurement is introduced in OpenStax: Wave Functions. A useful first habit is to describe what the apparatus does before choosing a metaphor for what the system “really is”.

For a learning exercise, write four separate sentences: how the system is prepared, what is changed, which outcome is registered, and what distribution is predicted over repeated trials. If one sentence is missing, the explanation probably contains an unexamined assumption.

2. Why classical probability alone does not reproduce interference

Ordinary alternatives often add as probabilities: if mutually exclusive routes have specified chances and no interference relation is retained, the total chance is their sum. Quantum theory instead assigns amplitudes to coherent alternatives and combines those amplitudes before obtaining a probability. The cross-terms can increase or decrease the result.

This is the central difference behind the double-slit example. It is not enough to replace a classical trajectory with an unknown trajectory and keep the rest unchanged. The calculation rule has changed. Feynman’s Quantum Behavior lecture develops the comparison between particle-like registrations, wave-like interference and quantum amplitudes. The same observations must be explained together, rather than selecting whichever picture is convenient in isolation.

3. A probability amplitude has magnitude and phase

Amplitudes are generally complex numbers. Their magnitudes contribute to probabilities, while relative phases determine how coherent alternatives combine. For amplitudes a and b, the probability associated with their combined alternative is |a + b|², which need not equal |a|² + |b|². Squaring only after addition retains interference.

A global phase multiplying an entire pure state does not change ordinary measurement probabilities. A relative phase between components can change them. This distinction prevents the phase from being treated as either meaningless everywhere or directly observable by itself. The amplitude rules are discussed in Quantum Behavior; the following numerical comparison is an original teaching construction.

4. Worked example: same contributions, different interference

Original hypothetical example. Consider two coherent contributions to one detector outcome. Each has amplitude magnitude 0.30. If their phases agree, the combined amplitude is 0.60 and the outcome probability is 0.36. If their phases differ by half a cycle, the amplitudes are 0.30 and −0.30, giving probability 0.

If information about the alternatives makes them effectively incoherent for the measurement, the corresponding sum of their separate probabilities is 0.09 + 0.09 = 0.18. The values 0.36, 0 and 0.18 come from different physical assumptions about coherence; they are not three arbitrary answers to one fully specified experiment.

This example concerns one selected outcome, not a complete apparatus with only one possible destination. A full model must account for all outcomes so total probability remains one. A dark interference location does not imply the particle’s energy vanished. The distribution of possible registrations has changed. The arithmetic illustrates why a description must say whether amplitudes or probabilities are being combined.

5. Single detections and a wave-like distribution belong to one dataset

A double-slit pattern can be assembled from individual detections. Each event is registered locally, while the distribution accumulated across many events has an interference structure. Neither description should be discarded. Calling the detections “particles” does not explain away the pattern, and calling the pattern “a wave” does not erase the discrete registrations.

Shalm and collaborators reported a time- and space-resolved measurement of a single-photon interference pattern. Its educational importance is that the pattern’s buildup is measured rather than inferred from a familiar diagram alone. The experiment distinguishes predictions about distributions; it does not provide a camera view of an unmeasured classical path between source and detector.

6. Which-path information changes the physical situation

A device that records which alternative occurred becomes correlated with the system. Interference between alternatives can be reduced when the relevant path information is physically available in the apparatus or environment. The issue is not whether a person happens to read a display at that moment. The interaction and retained correlations matter.

Kincaid, McLelland and Zwolak analyse this connection in Measurement-Induced Decoherence and Information in Double-Slit Interference. The useful scientific question is what interaction altered the measurable interference, not whether human attention possesses a special force. Claims about consciousness require evidence beyond the quantum measurement patterns described here.

As an investigation design, compare predictions with and without a specified path-recording interaction while keeping the rest of the arrangement controlled. “Observed” must refer to a defined physical measurement, not a vague change in the narrator’s knowledge.

7. A state is more than the probabilities of one measurement

Two pure states can have identical outcome probabilities in one basis while differing in their relative phases. A measurement in another basis can distinguish them. A list of probabilities for only one measurement is therefore not generally a complete specification of the state.

This is why experimental state preparation and verification usually involve a family of controlled operations or measurement settings. An example of deliberate state construction is the trapped-ion work described in the NIST-hosted report on generating arbitrary superposition states. The paper describes a physical experiment, whereas the next example uses idealised state vectors to expose the logic.

8. Worked example: two states that look the same until the question changes

Original idealised example. Let |0⟩ and |1⟩ be two orthogonal basis states. Define |+⟩ = (|0⟩ + |1⟩)/√2 and |−⟩ = (|0⟩ − |1⟩)/√2. Measuring either preparation in the 0/1 basis gives equal probabilities of 0 and 1. From that measurement alone, the preparations have the same statistics.

Now measure in the +/− basis. The ideal |+⟩ preparation gives the + outcome with probability one, while the ideal |−⟩ preparation gives it with probability zero. The relative sign mattered even though it was invisible in the first measurement’s probabilities.

The lesson is not that the second apparatus uncovers a tiny prewritten label that every other apparatus ignored. It is that the state assigns probabilities to different measurement choices through the amplitude rules. This simple comparison explains why phase control matters and why a single histogram cannot establish every feature of a quantum preparation.

9. The Born rule links a mathematical state to measurement statistics

For a position-space wavefunction, the squared magnitude |ψ(x)|² gives a probability density. A probability for a spatial interval comes from integrating over that interval, not from treating the value at one point as a dimensionless probability. Normalisation ensures that the total probability across all allowed positions is one.

This is a practical distinction. A graph of probability density can have values larger than one in suitable units without violating probability rules, because the integral over an interval is what must remain between zero and one. OpenStax’s wavefunction treatment sets out the interpretation. Always identify whether a graph shows an amplitude, its squared magnitude or an accumulated probability.

10. Worked example: a probability prediction still permits fluctuations

Original hypothetical experiment. Assume an independently repeated preparation has probability 0.80 of producing outcome A on each valid trial. For 1,000 trials, the expected number of A outcomes is 800. Under the independent binomial model, the standard deviation of the count is √[1,000 × 0.80 × 0.20] ≈ 12.65.

A result of 794 does not refute an 80% prediction merely because it is not exactly 800. Conversely, a result near 800 does not prove that the entire apparatus model is correct. The comparison should consider statistical variation, losses, possible drift and the alternatives being tested.

The independence assumption deserves attention. If a source slowly changes during the run, treating every trial as identically prepared may understate uncertainty. A scientific record should therefore include trial order and relevant operating conditions, not only a final count. These numbers are constructed for teaching and do not describe an actual quantum experiment conducted by eduKate.

11. Quantisation does not mean every quantity is always discrete

Bound quantum systems can have discrete allowed energy values because their states must satisfy the relevant dynamical equations and boundary conditions. Other situations have continuous spectra. The word “quantum” therefore does not mean every measurable quantity comes in identical fixed-sized chunks in every context.

A particle confined by ideal boundaries provides a tractable example. Only particular standing-wave forms satisfy those boundaries, and they correspond to particular energies. The quantum particle in a box explains the model. Its value lies in showing how confinement changes allowed states, not in pretending that an ideal infinite wall is an exact description of every real nanoscale object.

12. Worked example: confinement changes the energy scale

Original calculated example. For an electron in an ideal one-dimensional infinite box of length L, the energies are E_n = n²h²/(8mL²), with n = 1, 2, 3 and so on. Taking L = 1.0 nm gives a ground-state energy of approximately 0.376 eV. The second level has four times that energy, approximately 1.504 eV, so the gap between the first two is approximately 1.128 eV.

Double the box length to 2.0 nm, keeping the same particle and ideal boundary model. Because energy scales as 1/L², the ground-state energy becomes approximately 0.0940 eV. The wider confinement has reduced the energy scale by a factor of four.

This is a model comparison, not a measured property of an actual device. Finite barriers, additional dimensions, interactions and material structure can change the spectrum. The valuable result is the explained scaling: a smaller spatial region requires more rapidly varying allowed wavefunctions and changes the kinetic-energy scale. The formula’s domain is given in OpenStax’s box model.

13. The photoelectric effect separates photon energy from beam intensity

In a simple single-photon photoelectric model, a photon carries energy hf. An electron requires a material-dependent work function to escape. When emission is energetically possible, the maximum kinetic energy is hf minus that work function. Increasing intensity at fixed frequency changes the photon supply, not the energy carried by each photon.

This distinction explains why a frequency threshold matters in the simple regime. It is not enough to imagine arbitrarily weak-frequency light steadily accumulating energy on an electron until the familiar classical picture works. OpenStax’s photoelectric-effect discussion connects the prediction to measurements. More intense multiphoton regimes require a different description and are outside this introductory example.

14. Worked example: compare two wavelengths under a stated model

Original hypothetical material. Assume a work function of 2.20 eV and the single-photon model. A 450 nm photon has energy hc/λ ≈ 2.76 eV, leaving a maximum emitted-electron kinetic energy of about 0.56 eV. A 600 nm photon has energy about 2.07 eV, below the assumed work function, so it cannot produce emission by that one-photon mechanism.

The example does not identify a particular real surface or claim its work function is exactly 2.20 eV. Surface condition and experimental context would have to be established. The point is to compare an energy per photon with the required energy, while keeping wavelength separate from intensity.

A diagnostic question follows: why is a brighter beam not automatically a higher-energy-per-photon beam? The two variables describe different properties of the source. That same distinction recurs when interpreting a spectrum, an optical detector or a claim about how radiation interacts with matter.

15. The Schrödinger equation describes state evolution within its domain

For a suitable nonrelativistic closed system, the time-dependent Schrödinger equation is written iħ ∂|ψ⟩/∂t = H|ψ⟩. The Hamiltonian H represents the system’s energy-related dynamics. Given an initial state and an appropriate Hamiltonian, the equation predicts how the state evolves before the chosen measurement is modelled.

The distinction between evolving a state and predicting an individual registered outcome matters. A deterministic evolution equation for amplitudes is compatible with probabilistic measurement predictions. Practical systems also interact with environments and may require an open-system description. The relationship between wave and particle descriptions is developed in Feynman’s second quantum lecture. This equation is not a claim that one elementary wavefunction model covers every relativistic or many-body problem unchanged.

16. Uncertainty is not merely an instrument that needs improvement

The position–momentum uncertainty relation states that their standard deviations in a state obey σ_xσ_p ≥ ħ/2. This is a property of the distributions allowed by the quantum state, not simply a statement about clumsy instruments. Instrumental error and disturbance can exist as well, but they are different parts of an experimental analysis.

A state made narrowly localised in position generally requires a broader distribution of momentum components. The mathematical relation does not say that every quantity is equally uncertain or that scientific prediction has become impossible. It specifies a constraint. OpenStax’s uncertainty-principle treatment explains the relevant quantities. Avoid replacing a precise inequality with the vague phrase “we can never know anything”.

17. Tunnelling does not require borrowing energy

A wavefunction can extend into a finite potential barrier even when a classical particle with the same energy could not cross it. Matching the state across the regions can give a nonzero transmission probability. For a time-independent barrier, energy conservation is not suspended during the process.

The popular phrase “borrow energy briefly” can mislead by replacing the actual boundary-value problem with a story about a conservation-law exception. Barrier shape, width, particle mass and energy determine the transmission. OpenStax’s tunnelling treatment develops the calculation. A strong conceptual answer identifies where the classical restriction differs from the quantum state description, while leaving conservation intact.

18. Identical particles change how states are counted

Identical quantum particles are not treated as distinguishable objects carrying permanent name tags that an experiment merely fails to read. The symmetry properties of their joint states affect allowed occupations and measurement predictions. Fermions and bosons follow different exchange rules, with consequences for the structure and collective behaviour of matter.

For fermions, the exclusion principle concerns occupation of the same complete single-particle quantum state, not the vague statement that two particles can never be “in the same place”. Feynman’s Identical Particles chapter develops the difference. Precise state labels matter because an apparently simple verbal shortcut can erase the conditions that make the rule meaningful.

19. Entanglement belongs to the joint state

For a pure joint state, entanglement means that the state cannot be written as a product of states for its parts. More generally, a mixed state is entangled when it cannot be represented as a probabilistic mixture of product states. The resulting correlations can depend on the measurement settings used on each subsystem. This is more than ordinary ignorance about two independent objects, but it is not automatically established by seeing any correlation whatsoever.

Two boxes containing prearranged matching cards would also give correlated results in one fixed comparison. A stronger quantum test examines correlations across a carefully selected family of settings. Controlled preparation, interaction and detection are central to the trapped-ion research discussed in Wineland’s account of superposition and entanglement. The next distinction concerns what such tests can rule out, rather than what a metaphor suggests.

20. Bell tests turn a foundational dispute into a quantitative comparison

Bell inequalities constrain correlations for specified classes of local hidden-variable explanations under assumptions that include suitably independent measurement settings. Quantum mechanics predicts violations for appropriate states and measurements. Testing the inequality requires more than collecting two matching lists: timing, setting choices, detection and trial selection affect the inference.

In 2015, Shalm and collaborators reported a strong loophole-free test, and Giustina and collaborators reported an independent photonic test. These are specific experiments with specified assumptions, not permission to claim that every imaginable interpretation has been experimentally excluded. The important lesson is how a philosophical-sounding question becomes a discriminating statistical test.

21. Entanglement does not provide a controllable faster-than-light message

Correlated outcomes do not automatically provide a channel on which one party can freely choose the remote result. The correlations become visible when records are compared, and ordinary communication remains necessary for that comparison. A claim about nonclassical correlations must be kept distinct from a claim about controllable signalling.

The assumptions connecting Bell-test data, independent measurement choices and the prohibition of superluminal signalling are explicit in the research on experimentally generated certified random numbers. This provides a useful evidence boundary: the experiment supports a carefully defined statement, not the unrestricted slogan that quantum theory lets information travel instantly on demand.

22. Decoherence explains a loss of accessible interference

When a system becomes correlated with an uncontrolled environment, interference visible in the reduced description of the system can diminish. Different environmental couplings can produce different decoherence behaviour. This is a physical interaction problem that can be investigated, rather than a mysterious transition caused simply by an object becoming “large”.

Myatt and collaborators measured decoherence of superposed motional states of a trapped atom using engineered reservoirs. By controlling the coupling, the experiment linked a changed environment to a changed interference response. Decoherence is important for understanding classical-looking behaviour, but by itself it should not be advertised as settling every interpretive question about why one individual measurement outcome is experienced.

23. A quantum claim must survive ordinary experimental weaknesses

Quantum experiments still face losses, detector inefficiency, background events, calibration error and drift. An apparent effect can be altered by which trials were retained or discarded. A claim about ideal probabilities must therefore be connected to an explicit model of the recorded data and the apparatus response.

For Bell tests, detection and timing are especially important because they determine which alternative explanations remain available. The experimental motivation is discussed in the detection-loophole-free test by Christensen and collaborators. For detectors themselves, Gerrits, Lita and Calkins describe transition-edge sensors for quantum optics. More sophisticated physics does not remove the need for ordinary measurement discipline.

24. A qubit is not a box from which all answers can be read at once

A two-level pure state can be represented by amplitudes α and β with |α|² + |β|² = 1. Measurement in the corresponding basis produces an outcome according to those probabilities. Having a superposition does not permit unrestricted extraction of both amplitudes from one ordinary measurement on a single unknown state.

The earlier +/− example shows why the choice of operation and measurement matters. Controlled interference can make a desired distinction measurable; merely possessing a superposition does not solve every problem. Experimental state manipulation is discussed in the trapped-ion state-generation work. A responsible explanation separates the mathematical space of states from the information that a specified experiment actually returns.

25. Interpretation and prediction should be labelled separately

Quantum mechanics has a highly developed predictive formalism. Questions about the ontological meaning of the wavefunction, measurement outcomes and the relationship between a system and its environment involve additional interpretive commitments. Not every statement made in the language of an interpretation is a separately tested experimental result.

For reading scientific explanations, use three labels. “Measured” identifies the recorded pattern and its uncertainty. “Predicted” identifies what follows from a model and its assumptions. “Interpreted” identifies a proposed account of what the formalism means. This is an editorial discipline, not an assertion that interpretation is unimportant. It makes disagreement clearer by showing which layer is actually under discussion.

A sentence that begins with a measured interference pattern and ends with a broad claim about intention, consciousness or personal destiny has crossed several unsupported steps. A careful explanation stops where its evidence stops, then marks any further hypothesis explicitly.

26. How to read a quantum research result without exaggerating it

Begin with the experimental unit. Is the result about one trapped ion, photons in a particular optical arrangement, a many-body material, or a simulated ideal system? Identify the preparation and measurement settings. Then separate the central observation from the model used to infer an underlying state or parameter.

Next inspect the comparison. Was a classical alternative tested quantitatively? Were device limitations included? Does the result generalise to another platform, or is that an additional proposal? For example, a report of controlled interference is not automatically a demonstration of a practical computer, and a sensitivity gain for one measurement is not an advantage for every task.

This reading procedure is a proposed learning tool. Its purpose is to preserve the actual scientific achievement rather than making it sound smaller or larger. A bounded, well-tested result is more valuable than an inflated universal claim that the original experiment did not investigate.

27. A staged route from observable patterns to formal states

First encounter: begin with a histogram of individually registered events and ask what pattern appears across many trials. Distinguish one outcome from a distribution. Use ordinary probability before introducing amplitudes, so the learner can see exactly which addition rule changes. Do not demand acceptance of a dramatic metaphor before the evidence is understood.

Secondary-to-JC bridge: connect photon energy to frequency, examine a supplied photoelectric dataset, and compare wave interference with discrete registrations. Make students state which quantity was measured and which was inferred. The box example can introduce the connection between confinement, allowed states and energy without requiring a full treatment of Hilbert spaces.

Higher-resolution route: add complex amplitudes, vectors, operators, basis changes, differential equations and mixed states. Derive predictions for several measurements of the same preparation. Then read an experiment with its apparatus model, uncertainty and competing explanations. This is a proposed teaching sequence, not a statement of any current examination specification.

28. Checkpoints with answers

Does a 50/50 outcome distribution prove that a preparation is an incoherent mixture? No. A coherent superposition can produce the same probabilities in that basis. Additional measurement choices are needed to test phase-sensitive differences.

Does an interference minimum mean the particle has lost its energy? No. It concerns the probability of a selected outcome. A complete model accounts for the distribution across all outcomes and the relevant energy transfers.

Does the uncertainty principle merely describe a poor instrument? No. The stated position–momentum relation constrains the distributions allowed by the quantum state. Instrumental uncertainty is an additional issue.

Does tunnelling break energy conservation? Not in the time-independent barrier problem described here. It follows from the quantum state and boundary conditions rather than temporary borrowing of energy.

Do correlated outcomes alone prove a Bell-inequality violation? No. The relevant inequality concerns correlations across specified settings and assumptions. Matching outcomes in one fixed setting can have much simpler explanations.

Can an exactly predicted probability make every repeated count identical? No. Finite samples fluctuate under the probabilistic model. Compare the observation with an appropriate sampling distribution and investigate systematic deviations rather than demanding the expected count every time.

29. The final skill is separating surprise from evidence

Quantum mechanics is unfamiliar because its rules differ from everyday intuitions about trajectories and hidden labels. Unfamiliarity is not a licence to abandon precision. A complete explanation should identify the state, the interaction, the measurement and the predicted distribution, then compare that distribution with the record.

For independent practice, take one worked example from this article and change exactly one condition: relative phase, measurement basis, confinement length, photon wavelength or trial count. Predict which quantities change and which remain unchanged. Then explain why. That exercise turns “quantum is strange” into something more useful: a set of relations that can be calculated, tested and corrected.

Sources, limits and connected subjects

The formal foundations are supported by the linked Feynman lectures and OpenStax University Physics chapters. Experimental examples are identified through the original researchers’ reports hosted by NIST, including single-photon interference, trapped-ion state preparation, engineered decoherence and Bell tests. Dates refer to the cited studies, not claims that they are the latest results. All worked numerical cases and diagnostic scenarios written for this guide are explicitly hypothetical or calculated.

Continue through the existing Quantum Mechanics root, Optics, Physical Chemistry, Materials Science and Electromagnetism. Each retains its own explanatory scope.

Return to How Science Works or the How X Works Hub.