In standard quantum mechanics, probability is the bedrock of physical interpretation. For the probabilities of all possible outcomes to sum to exactly one, the state space of a quantum system must possess a positive-definite inner product. This geometric guarantee ensures that every physical state has a positive "length" or norm.
However, when we attempt to merge quantum mechanics with special relativity to describe fundamental forces, this foundational assumption runs into a wall. Building local, covariant gauge theories—or modeling quantum fields in the curved spacetimes of general relativity—frequently forces physicists to abandon positive-definite metrics in favor of indefinite metrics, where states can have zero or even negative norms.
This introduces a profound tension: how can a theory with an indefinite metric remain physically meaningful if it risks producing negative probabilities and unstable particles?
The Locality-Positivity Conflict
The shift toward indefinite metrics is not a mathematical whim; it is a necessity driven by the requirements of relativity. In quantized gauge theories (such as quantum electrodynamics or Yang-Mills theories), there is a fundamental conflict between maintaining local interactions and enforcing a positive-definite state space.
To preserve local causality and Lorentz covariance, the fields must contain auxiliary degrees of freedom—often called ghost states. As explained by Albeverio et al., relaxing the positivity requirement allows researchers to preserve locality in gauge interactions, avoiding the mathematical deadlocks that plague standard formulations.
By transitioning from a strict Hilbert space to a non-positive definite state space, fields can be constructed cleanly across all spacetime dimensions. In Models of Local Relativistic Quantum Fields with Indefinite Metric (in All Dimensions), the authors establish that relativistic quantum field theories can maintain locality and covariance by adopting an indefinite metric, provided they satisfy a modified set of Wightman axioms.
The Price of Interaction
While an indefinite metric solves the problem of locality, it introduces a steep physical price: it alters the stability of matter and complicates how particles interact.
In a standard Hilbert space, the positivity of the metric acts as a barrier that prevents unphysical behavior. For instance, Gottschalk demonstrated that a positive-definite metric inherently stabilizes one-particle states, preventing them from spontaneously decaying into other configurations even in the absence of a mass gap. When this constraint is lifted to accommodate an indefinite metric, this geometric protection vanishes, allowing one-particle states to decay into auxiliary configurations.
More surprisingly, an indefinite metric may be an unavoidable consequence of particle scattering itself. In Quantum Fields Obtained from Convoluted Generalized White Noise Never Have Positive Metric, Albeverio proved a strict no-go theorem for quantum fields constructed via stochastic white noise: introducing the non-Gaussian elements necessary for particle interactions automatically destroys the positivity of the metric.
[Positive-Definite Metric] ───► Stable Particles, No Interactions (Free Fields Only)
[Indefinite Metric] ───► Non-Trivial Scattering, Local Gauge Symmetries, Particle Decay
This creates an explicit interaction-positivity tradeoff. To build a mathematically rigorous model with non-trivial scattering, one must accept an unphysical indefinite metric; to keep a standard probability interpretation, the model forces the system back into a non-interacting free field.
Vacuum Ambiguity and Canonical Non-Invariance
When the geometry of the state space is no longer anchored by a positive-definite metric, defining a unique physical reality becomes significantly harder.
In standard flat Minkowski space, the positive-definite metric uniquely picks out a stable vacuum state. But in curved spacetimes or globally hyperbolic manifolds, global symmetries disappear. As Thaler and Sanders point out, the lack of a preferred Hilbert space or unique vacuum state complicates the very definition of what a "particle" is. To maintain locality and covariance on these curved manifolds, the inner product space must be relaxed to a non-degenerate, indefinite inner product space.
This lack of uniqueness manifests as a failure of canonical invariance. When researchers attempt to define vacuum states using traditional methods like Hamiltonian diagonalization, the results often depend entirely on the arbitrary choice of coordinate systems or variable definitions.
Agocs et al. demonstrated that this sensitivity to mathematical representation creates an infinite family of possible vacuum states for a single physical model. To resolve this ambiguity, one must look for invariant physical quantities—such as minimizing the renormalized stress-energy tensor—rather than relying on coordinate-dependent state definitions.
Restoring Physical Reality
If indefinite metrics are mathematically required but physically uninterpretable on their own, how do we extract real-world predictions from them? Two primary mechanisms bridge this gap: Krein space representations and metric operators.
Krein Spaces and LSZ Scattering
To handle negative norms rigorously, mathematicians embed the state space into a Krein space. A Krein space decomposes an indefinite inner product space into two parts: a positive-definite piece and a negative-definite piece.
Using this structure, Albeverio successfully derived the Lehmann-Symanzik-Zimmermann (LSZ) asymptotic condition for theories with indefinite metrics. This derivation provides the formal justification needed to map field operators to actual particle creation and annihilation processes, proving that valid scattering matrices (S-matrices) can still be extracted from non-positive definite spaces.
Metric Operators ($\eta_+$)
In non-Hermitian or open quantum systems, an alternative approach modifies the inner product itself to restore positive probabilities. If a Hamiltonian is non-Hermitian under a standard inner product, its eigenvalues can become complex, and time evolution can lose unitarity.
To fix this, Li et al. and Lee et al. utilize a positive-definite metric operator, $\eta_+$, to redefine the inner product:
$$\langle \psi | \phi \rangle_{\text{new}} = \langle \psi | \eta_+ | \phi \rangle_{\text{standard}}$$
This operator accounts for the non-orthogonal nature of the system's eigenstates. By bending the geometry of the state space, the modified inner product ensures that physical observables yield real-valued energy spectra and preserve unitary time evolution, successfully mapping exotic or dissipative systems back into the realm of viable quantum mechanics.
Key Foundational Work
Models of Local Relativistic Quantum Fields with Indefinite Metric (in All Dimensions) (1997) — contributed 6 concept elements to this direction.
Remarks on some new models of interacting quantum fields with indefinite metric (1997) — contributed 6 concept elements to this direction.
On the stability of one-particle states generated by quantum fields fulfilling Yang—Feldman equations (2001) — contributed 6 concept elements to this direction.