One incoming wave, several outgoing paths

A path is a mode of propagation

A splitter creates more than one outgoing propagation channel from an incoming field. A partially reflecting optical element sends amplitude into reflected and transmitted modes; an aperture arrangement can provide alternative paths to a screen. The channels may differ in direction, position, polarization or another distinguishable mode label.

The word split therefore refers first to the field organization. It does not automatically mean that the physical constituents of the wave have been cut apart. A classical field can occupy both outputs continuously, while a quantum state requires a description of amplitudes and detection statistics.

A useful diagram labels modes rather than drawing a small object literally sawn in half. It also includes every relevant output, including loss channels when necessary. This makes the meaning of the splitter operational: it is a transformation from a specified input state to a set of output states. The observable might be optical power, photon counts or interference after recombination, and each must be calculated with the appropriate description.

Amplitude fractions and power fractions differ

For a simple lossless splitter with compatible mode normalization, transmission and reflection amplitudes t and r satisfy |t|² + |r|² = 1 for one input. The corresponding power fractions are T and R. A balanced splitter has T = R = 1/2, so the amplitude magnitudes are 1/√2 rather than one half.

This square-root distinction is essential. Assigning amplitude one half to each output would give total normalized power one half, leaving an unexplained loss. Conversely, assigning equal full input amplitudes to both outputs would double the power unless another input or active source were included.

At interfaces between different media, field amplitudes can require impedance or flux factors before their squared magnitudes become power ratios. The compact T–R notation assumes those factors are already incorporated. A clear calculation states the normalization rather than applying a familiar formula to any pair of electric-field values. The conservation rule concerns physical flux, not an arbitrary coordinate amplitude.

Both input ports matter to a reversible device

An ideal two-port optical splitter has two input modes even when only one is deliberately illuminated. Its transformation maps both inputs to both outputs. Including the second input is necessary to write a unitary, reversible mode transformation and to predict interference when both inputs are occupied.

One convenient real convention uses the matrix with rows (√T, √R) and (√R, −√T). The minus sign ensures orthogonal columns when T + R = 1. Other conventions place phase factors on reflection, but they must give consistent measurable predictions when the optical paths are tracked.

The apparently unused input is not an excuse to discard half the transformation. In quantum optics it is typically assigned a vacuum state, and the full operator mapping preserves the required relations. In classical optics it determines how another coherent input would redistribute the output power. A one-input cartoon is useful for a first picture, but the complete two-input map is what makes recombination and conservation transparent.

A single photon has path amplitudes, not divided energy parcels

For an ideal single-photon input and vacuum in the other port, the output can be written t|1,0〉 + r|0,1〉, with the phase convention included in t and r. The state has one total photon distributed coherently across two mode alternatives.

An ideal direct detection gives one photon event in one output or the other, with probabilities |t|² and |r|². It does not give two simultaneous detections each carrying half the original photon energy. Real inefficiency and background complicate the record, so the ideal statement must be tested with appropriate source and detector characterization.

The classic Grangier, Roger and Aspect experiment used triggered photon measurements to study anticorrelation at a splitter together with interference behavior. The verified paper abstract supports that foundational distinction; this article does not claim to reproduce its full apparatus or numerical analysis. The important lesson is that path division and energy quantization coexist. The state transformation is wave-like, while individual absorption records need not look like divided classical power.

Direct path counts do not reveal the relative phase

A balanced path superposition and an equal incoherent mixture can give the same fifty-fifty direct detector counts. The difference lies in off-diagonal coherence, which becomes visible when the paths are recombined in a suitable basis.

For the coherent state (|L〉 + exp(iφ)|R〉)/√2, changing φ changes later interference. For the mixture with half the trials in each path and no stable phase relation, the same phase scan does not produce the corresponding fringe. Thus equal output populations alone do not establish coherent splitting.

This distinction guides the experiment. Measuring path probabilities characterizes the splitter ratio, while recombination tests the retained phase relationship. Both are useful, but they answer different questions. A claim that the particle “chose a path” cannot be inferred solely from direct counts when a coherent superposition predicts the same counts. Conversely, a bright classical interference pattern does not by itself establish a single-photon source. State preparation and measurement must be connected carefully.

Recombination can undo an ideal split

If the two outgoing modes remain coherent and controllable, an inverse optical transformation can return the amplitude to a chosen original mode. This is the basic logic of an interferometer. A relative phase inserted between splitting and recombination changes which final output is populated.

Using balanced real splitters, a single input followed by a phase φ in one arm and a second identical transformation gives output probabilities cos²(φ/2) and sin²(φ/2). Their sum is one. At one phase, all ideal signal exits one port; shifting by π exchanges the ports.

The result shows why splitting is not automatically an irreversible measurement. A coherent transformation can be reversed or redirected through phase control. A detector or environmental interaction may change that situation by recording path information or introducing loss. The distinction depends on the actual operations, not on the visual fact that two paths were drawn. A complete experiment records the recombination response as well as the initial division.

Loss adds channels rather than invalidating conservation

A real splitter can absorb, scatter or send light into uncollected modes. The measured transmission and reflection fractions may then satisfy T + R < 1. The deficit should be assigned to calibrated loss or additional outputs, not interpreted as a failure of energy conservation.

For an illustrative device with T = 0.45 and R = 0.45, ten percent of incident power is outside the two measured outputs. That fraction might be absorption, but a measurement of only two ports cannot distinguish absorption from uncollected scattering. Additional diagnostics or a validated model are needed.

In a quantum description, the environment can be included as extra modes so that the larger transformation remains consistent. Conditioning on detected events changes the normalized statistics of the surviving subset and must be labeled. A fifty-fifty ratio among detected photons does not prove a lossless splitter. Absolute detection efficiency and total input normalization are separate parts of the characterization.

Path splitting differs from fission or frequency conversion

Nuclear fission changes a nucleus into reaction products with different identities and an energy release determined by nuclear masses and states. Ordinary optical path splitting changes mode amplitudes without that nuclear transformation. The shared word splitting should not hide the different objects being divided.

Nonlinear optical processes can also change photon frequencies or create correlated output pairs under suitable conditions. Those are distinct interactions with their own energy and momentum or phase-matching requirements. They are not what an ideal passive beam splitter does merely by reflecting and transmitting a field.

A good explanation therefore asks what remains invariant. In the ideal optical path example, the input frequency and total photon number are preserved while mode occupation changes. In a particle-breakup reaction, species and particle counts can change while conserved totals are carried by the products. The appropriate measurement follows that difference: interference and mode powers for path splitting, product spectroscopy and reaction kinematics for breakup.

Choose the measurement that answers the actual question

To characterize a splitter, measure its output fractions, phases, losses and dependence on wavelength or polarization where relevant. To test coherence, recombine the paths and scan phase. To test single-photon behavior, characterize counting statistics and backgrounds with a suitable source. No one of these measurements automatically supplies the others.

The central interpretation is a transformation between modes. That view keeps the amplitude square roots, complementary ports and environmental channels visible. It also explains why a split path can later recombine and why an individual photon detection remains a whole-energy event.

Splitting is therefore a precise physical process when the input state, output modes and measurement are specified. It is not a universal mechanism of material fragmentation or quantum collapse. Following the full transformation from preparation through propagation to detection reveals exactly what was divided, what remained conserved and which information the experiment can recover.

Go to the original work.

These links identify the sources behind specific claims. Verification scope varies by source and is stated below.

1. 1 / University teachingBarton Zwiebach, MIT 8.04: Chapter 2, Experiments with photons ↗

   Beam-splitter phase conventions, two-path amplitude transformations and Mach–Zehnder recombination.

   Verification scope: MIT-hosted lecture-note text and beam-splitter equations inspected.

2. 2 / Primary experimentGrangier, Roger and Aspect (1986): Experimental Evidence for a Photon Anticorrelation Effect on a Beam Splitter ↗

   Triggered antibunching/anticorrelation at a splitter and interference in the same experimental program.

   Verification scope: Publisher abstract and bibliographic record inspected; full paper not claimed reviewed.

3. 3 / University teachingOpenStax University Physics: Fission ↗

   Fission changes nuclear identities; example U-235 neutron-induced channels balance charge and nucleon number and have channel-dependent products.

   Verification scope: Official textbook equations, examples and assumptions inspected.

4. 4 / Official researchNIST: Sources of Nonclassical Light for Quantum Networks ↗

   Spontaneous parametric down-conversion is a nonlinear pump-to-photon-pair process, distinct from a passive linear beam splitter.

   Verification scope: Official research-program explanation inspected; no performance numbers transferred to the illustrative comparison.