Higher-Order Interference in Quantum Mechanics

• Higher-order interference is defined as interference patterns that cannot be fully reconstructed from lower-order (pairwise) contributions, challenging the Born rule's quadratic structure.
• Experimental methodologies use multi-slit setups with alternating inclusion–exclusion sums to isolate nonzero third-order terms while controlling systematics like detector nonlinearity and phase drift.
• A positive Sorkin signal would have pivotal implications, potentially requiring modifications to the Born rule or quantum measurement postulates to accommodate post-quantum phenomena.

Searching arXiv for recent and foundational papers on higher-order interference. Higher-order interference denotes interference structure that cannot be reduced to lower-order contributions. In its foundational sense, formalized by Sorkin, it asks whether an $n$-path experiment contains genuinely irreducible $k$-path terms with $k \geq 3$, beyond the pairwise terms generated by the Born rule. In standard quantum mechanics, two-path interference is generically nonzero, but all higher Sorkin functionals vanish, so quantum probability is “grade-2.” At the same time, the literature uses the same phrase in adjacent but distinct senses, including higher-order optical correlation functions, many-particle interference, and nonlinear or topological interference phenomena that remain entirely within standard quantum theory. Distinguishing these meanings is essential, because nonzero higher-order interference in the Sorkin sense would challenge core quantum postulates, whereas nonzero higher-order correlations in quantum optics need not do so (Bolotin, 2016, Rotari et al., 20 Feb 2026).

1. Sorkin hierarchy and operational definitions

For an $n$-slit or $n$-path experiment, let $P_S$ denote the probability or intensity at a fixed detector outcome when exactly the subset $S \subseteq \{1,\dots,n\}$ is open, and let $P_{\emptyset}$ denote the background with all paths closed. Sorkin’s hierarchy defines alternating inclusion–exclusion functionals that isolate genuinely $k$-body contributions. The basic cases are

$I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$

and

$k$0

More generally,

$k$1

with $k$2 appearing with sign $k$3. A nonzero $k$4 would mean that the three-path pattern cannot be reconstructed from pairwise and single-path data alone. Grade-2 additivity is the statement that $k$5 for all $k$6 (Bolotin, 2016).

Experiments often report a normalized third-order quantity rather than the raw alternating sum. One convention is $k$7, where $k$8 is a characteristic intensity scale such as $k$9 or $k \geq 3$0. In the atomic triple-path proposal, the experiment-friendly notation is

$k \geq 3$1

with

$k \geq 3$2

where $k \geq 3$3 and $k \geq 3$4. This normalization compares any residual third-order term to the dominant pairwise interference scale (Lee et al., 2019).

Sorkin also showed that if $k \geq 3$5 for some $k \geq 3$6, then all $k \geq 3$7 for $k \geq 3$8. In that sense, a theory’s position in the hierarchy is determined by the highest order at which irreducible interference can occur. Classical probability sits at level 1; standard quantum theory sits at level 2 (Lee et al., 2019).

2. Why standard quantum mechanics is grade-2

The standard argument is algebraic. If $k \geq 3$9 denotes the complex amplitude for propagation through slit $n$0, then the Born rule gives

$n$1

Only pairwise cross terms appear. There are no cubic terms such as $n$2, and therefore every alternating sum designed to isolate genuinely three-way or higher contributions vanishes: $n$3, $n$4, and so on. In this sense, the quantum measure induced by the Born rule is quadratic in amplitudes and hence grade-2 (Bolotin, 2016).

In generalized probabilistic and measure-theoretic language, this restriction can be expressed as a projector identity or as grade-2 additivity of a decoherence functional. In the GPT formulation used by Barnum, Müller, and Ududec, absence of third-order interference is equivalent to

$n$5

for all multi-slit masks, while in quantum measure theory one writes

$n$6

for disjoint alternatives $n$7. These formulations are operationally equivalent to $n$8 (Barnum et al., 2014).

Bolotin’s analysis adds a logical interpretation. Using exclusive disjunction over $n$9 alternatives, Bolotin argues that the proposition “exactly one outcome occurs” must be supplemented by a factor

$n$0

which explicitly excludes the simultaneous truth of three-or-more outcomes. Once this factor is imposed, the multi-alternative XOR reduces to pairwise structure only, yielding

$n$1

The paper’s conclusion is that quantum theory predicts only second-order interference because of the assumption, underlying the Born rule, that in each measurement in the $n$2-event space exactly one outcome may only occur; nonzero third- or higher-order interference would contradict the proposition that measurements always yield definite results (Bolotin, 2016).

A positive Sorkin signal would therefore not be a small perturbation of textbook interference. It would force a modification of the Born rule, or of linearity, or of the measurement postulates, and proposed changes of that kind are often entangled with no-signaling, tomographic locality, convexity, and exclusivity constraints (Bolotin, 2016).

3. Experimental searches and the control of systematics

Direct tests cycle through all open/closed path configurations and compute the relevant alternating sums. The first prominent three-slit photon experiment measured

$n$3

normalized it by the total pairwise interference scale

$n$4

and reported $n$5. In that experiment, the magnitude of three-path interference was bounded to less than $n$6 of the expected two-path interference, with central-maximum values $n$7 for laser plus power meter, $n$8 for laser plus photon counting, and $n$9 for heralded single photons (Sinha et al., 2010).

Later interferometric tests pushed the null bounds much lower. A stabilized 5-path interferometer at $P_S$0 measured all $P_S$1 configurations, corrected independently calibrated detector nonlinearities, and reported nonlinearity-corrected values $P_S$2 in the classical regime, together with consistent null results for $P_S$3 and $P_S$4. The paper describes this as ruling out higher-order interference terms to an extent more than four orders of magnitude smaller than the expected pairwise interference (Kauten et al., 2015).

Across optical triple-slits, multi-path interferometers, NMR, microwave networks, and related platforms, the overall empirical pattern is null results consistent with $P_S$5 within uncertainty. The literature summarized in 2016 describes sensitivities progressing from $P_S$6 down to $P_S$7 and, in some controlled interferometric implementations, further, with the overwhelming majority of experiments delivering results in accordance with second-order interference only (Bolotin, 2016).

The central practical difficulty is not the combinatorics of the Sorkin sums but the suppression of spurious contributions. Detector nonlinearity is especially dangerous: if the measured signal is $P_S$8 with $P_S$9, then even when the true $S \subseteq \{1,\dots,n\}$0 vanishes, the $S \subseteq \{1,\dots,n\}$1 and $S \subseteq \{1,\dots,n\}$2 terms generally do not cancel in the alternating sum, producing a false $S \subseteq \{1,\dots,n\}$3. Other major systematics include background drift, phase drift, unequal intensities, finite slit thickness, edge scattering, and the fact that physically blocking a slit changes electromagnetic boundary conditions, so the measured $S \subseteq \{1,\dots,n\}$4 may not represent the pair contribution that would have occurred had the third slit been open but unpopulated. Interferometers that vary path amplitudes without changing apertures mitigate this back-action (Bolotin, 2016).

An instructive alternative is the atomic internal-state proposal based on a tripod configuration in $S \subseteq \{1,\dots,n\}$5. It replaces spatial paths by three long-lived internal states, implements a tritter by Raman couplings on the $S \subseteq \{1,\dots,n\}$6 intercombination line, and studies several realizations of “slit blockers,” including state transfer out of the tripod, dephasing maps, and a single-cycle spontaneous-emission blocker. With $S \subseteq \{1,\dots,n\}$7, $S \subseteq \{1,\dots,n\}$8, and tritter duration $S \subseteq \{1,\dots,n\}$9, the projected statistical precision is at the $P_{\emptyset}$0 level for $P_{\emptyset}$1, with dominant systematic effects identified as AC-Stark-induced phase errors, detector background drift, crosstalk, and repeated spontaneous-emission cycles. Because this architecture works entirely in the internal-state domain, it avoids near-field boundary-condition artifacts present in spatial slit experiments (Lee et al., 2019).

4. Distinct meanings of “higher-order interference” in contemporary literature

A persistent source of confusion is that not every “higher-order interference” signal is a Sorkin-type violation of grade-2 quantum probability. In many-particle interferometry, standard quantum mechanics already allows higher Sorkin order for higher-order correlation functions. For $P_{\emptyset}$2 detected particles, Born’s rule permits interference among up to $P_{\emptyset}$3 single-particle paths contributing to the $P_{\emptyset}$4th-order correlation function. In a two-photon five-slit experiment, nonzero two-particle interference was observed up to fourth order, while fifth-order interference was bounded to $P_{\emptyset}$5 in the intensity-correlation regime and to $P_{\emptyset}$6 in the photon-correlation regime. This is not a violation of quantum mechanics; it is the structure predicted for two-particle correlations (Pleinert et al., 2020).

A second distinction concerns nonlinear evolution and many-body interactions. In second-quantized treatments of generalized multi-slit interferometers, a single particle or any linear multiport still gives $P_{\emptyset}$7. However, when multiple quantum particles interact nonlinearly, Sorkin functionals such as $P_{\emptyset}$8 can become nonzero while remaining fully inside standard quantum theory. Explicit examples include photons in Kerr-type media and interfering Bose–Einstein condensates with contact interactions. The key point is that nonlinearity introduces multi-mode correlators, such as $P_{\emptyset}$9, that are not reducible to pairwise linear processes (Rozema et al., 2020).

This theoretical possibility has been demonstrated experimentally in a nonlinear optical “triple slit” implemented by three co-propagating beams overlapping in a $k$0 crystal. In that setup, the third-order Sorkin parameter

$k$1

was measured as a function of crystal position, and at a fringe maximum the experiment reported

$k$2

The signal vanished when the crystal was translated out of focus, confirming that the nonlinearity, not a Born-rule failure, generated the higher-order term (Namdar et al., 2021).

Quantum optics also uses the phrase for higher-order spatial or temporal correlation effects. In a model of three dipole-dipole coupled emitters in an equilateral triangle, the relevant observables are $k$3, $k$4, $k$5, and $k$6; the paper explicitly states that it studies higher-order interference in the sense of Glauber correlation functions and “does not address Sorkin-type higher-order interference in multi-path experiments beyond standard quantum mechanics” (Rotari et al., 20 Feb 2026). Likewise, fourth-order optical interference in unbalanced interferometers can persist even when path delays are much larger than the coherence time, provided one measures time-resolved intensity correlations rather than single-detector fringes (Ou et al., 2021). Higher-order optical interference has also been used as a protocol primitive for visibility-based hypothesis testing without a shared phase reference, where the relevant resource is the full joint photocount statistics and optimal operation is typically reached in the few-photon regime (Jachura et al., 2017).

The common misconception is therefore twofold. First, “quantum theory only exhibits second-order interference” is correct only for single-particle, linear, Born-rule Sorkin tests. Second, nonzero higher-order optical correlations or nonlinear many-particle Sorkin terms do not by themselves indicate post-quantum physics.

5. Generalized probabilistic theories and post-quantum models

Higher-order interference occupies a central role in GPT reconstructions. One line of work elevates “no higher-order interference” to an axiom. In the single-system reconstruction of Barnum, Müller, and Ududec, three postulates—no higher-order interference, classical decomposability of states, and strong symmetry—characterize finite-dimensional systems as either classical simplices or Euclidean Jordan algebras, including real, complex, and quaternionic quantum theories, spin factors, and the $k$7-level octonionic case. Adding observability of energy as a fourth postulate singles out standard complex quantum theory (Barnum et al., 2014).

A complementary result derives the same interference restriction from purity principles. In sharp theories with purification, the four principles Causality, Purity Preservation, Pure Sharpness, and Purification imply that there can be no $k$8th-order interference for $k$9. The proof proceeds by constructing a self-dual inner product and pure orthogonal projectors onto faces associated with slit subsets, then invoking the Barnum–Müller–Ududec projector-based criterion. The same framework shows that single-system state and effect cones are isomorphic to cones of squares in Euclidean Jordan algebras (Barnum et al., 2017).

Not all GPT models obey grade-2 additivity. Dakić, Paterek, and Brukner proposed “density cubes,” rank-3 tensors intended to generalize density matrices by adding genuine triple-coherence components. In their constructed three-level theory, the triple-slit-like interferometer yields

$I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$0

and the model also violates the Leggett–Garg inequality beyond the quantum Tsirelson bound for temporal correlations (Dakic et al., 2013). A categorical variant based on “density hypercubes” uses a double-dilation construction to obtain a probabilistic theory exhibiting higher-order interference up to fourth order. In the multi-slit experiment analyzed there, the theory predicts

$I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$1

hence nonzero $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$2 and $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$3, while all $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$4 vanish for $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$5; hyper-decoherence maps recover quantum theory in the Karoubi envelope (Gogioso et al., 2018).

These models are not uncontroversial. A detailed comparison of Density Cubes and Quartic Quantum Theory argues that the Density Cubes axioms are insufficient to uniquely characterize the theory, while Quartic Quantum Theory appears to exhibit irreducible interference to all orders only because of an ambiguity in the current operational definition of higher-order interference. The conclusion is that a new operational definition is needed if one wants the hierarchy to discriminate genuine post-quantum coherence rather than artifacts of underconstrained effect spaces (Lee et al., 2015).

6. Foundational significance, misconceptions, and open problems

The foundational significance of higher-order interference lies in what a positive result would force one to abandon. In the standard formalism, a statistically significant nonzero $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$6 would indicate failure of the Born rule’s quadratic dependence on amplitudes, or of linearity, or of the assumption that exactly one outcome occurs in a measurement. Bolotin’s formulation is especially explicit: ongoing experiments are operationally testing the truthfulness of the proposition that measurements always yield definite results (Bolotin, 2016).

This does not mean that every experimental nonzero $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$7 has such significance. The literature repeatedly identifies detector nonlinearity, slit blocking back-action, phase drift, and nonlinear medium effects as mechanisms that can generate apparent or genuine higher-order terms without any departure from ordinary quantum mechanics (Bolotin, 2016, Namdar et al., 2021). A careful encyclopedia-level treatment must therefore separate three questions: whether the quantity being measured is a Sorkin functional or a Glauber correlation; whether the evolution is linear or nonlinear; and whether the observed nonzero term survives all known systematics.

Several open problems recur across the literature. One is to construct explicit, consistent post-quantum models with higher-order interference that also satisfy no-signaling and operational coherence (Bolotin, 2016). Another is to refine the operational definition of higher-order interference so that theories with enlarged effect spaces are not misclassified (Lee et al., 2015). A third is experimental: devise protocols that isolate higher-order terms while guaranteeing that lower-order and background contributions are fully accounted for, ideally in phase-stable interferometric architectures that avoid aperture-blocking artifacts (Lee et al., 2019).

Taken together, the present state of the subject is sharply defined. In the Sorkin sense, standard quantum mechanics is grade-2: $I_2 = P_{12} - P_1 - P_2 + P_{\emptyset},$8. Extensive experiments remain consistent with that prediction, while increasingly constraining deviations. Yet the broader literature also shows that higher-order interference, in other technically precise senses, is abundant inside standard quantum theory: in many-particle correlations, in nonlinear media, in unbalanced fourth-order interferometry, and in spatial photon correlation functions. The topic therefore sits at a junction of foundations, quantum optics, and GPT theory, where terminological precision is itself part of the physics.