deepdive

The Spectral–Angular Structure of the Flux

Why Φ is resolved in energy and direction, why an average intensity misrepresents it, and how useful power is the overlap integral of Φ(E,θ) and σ(E) in P(t).

Last updated: 2026-07-26

A single number — a total intensity, a mean flux — is a poor description of an incident field, and treating it as adequate is one of the most common ways to misestimate a converter’s output. The master formula does not multiply a scalar intensity by a scalar cross-section; it integrates a flux $\Phi(E,\theta,t)$ that is resolved in energy $E$ and direction $\theta$ against a coupling $\sigma(E)$ that is itself energy-dependent. The useful power is therefore an overlap integral, and overlap is not captured by averages. This chapter explains why the spectral and angular structure of the flux is indispensable, why an averaged intensity misrepresents it, and how to think about the overlap-weighted quantities that actually govern output.

Why resolve in energy at all

The reason the flux must be spectrally resolved is that the coupling is spectrally selective. If $\sigma(E)$ were flat — the same at every energy — then only the total, energy-integrated flux would matter, and a mean intensity would suffice. But real cross-sections are structured: broad in some ranges, sharply peaked at resonances in others, negligible elsewhere, as detailed in sigma analysis. When $\sigma(E)$ varies strongly with energy, the same total flux distributed differently across the spectrum produces wildly different outputs. Flux piled up where $\sigma$ is large yields much; the identical total flux piled up where $\sigma$ vanishes yields nothing. The spectral shape of $\Phi$ is thus not a refinement but a determinant of the result, and collapsing it to a total discards exactly the information that decides the outcome.

Why the average misrepresents the flux

The failure of the averaged description can be stated as a mathematical inequality, which makes the point unambiguous. The quantity that governs output is the spectral overlap $\int \Phi(E)\,\sigma(E)\,dE$ . The naive averaged estimate replaces this with the product of the mean flux and the mean cross-section, $\langle\Phi\rangle\,\langle\sigma\rangle$ times the bandwidth. These two are equal only if $\Phi$ and $\sigma$ are uncorrelated across the spectrum; in general they differ by exactly the covariance,

\int \Phi(E)\,\sigma(E)\,dE \;=\; \big[\langle\Phi\rangle\langle\sigma\rangle + \operatorname{Cov}(\Phi,\sigma)\big]\,\Delta E ,

where the covariance is positive when flux and coupling are large in the same energy band and negative when they are anti-aligned. An averaged intensity throws the covariance away and keeps only the first term, so it systematically misestimates the output — understating it when flux and cross-section overlap constructively, overstating it when they do not. Since a converter is typically engineered to align its $\sigma(E)$ with the ambient spectrum, the discarded covariance is usually the dominant, favourable term, and the averaged estimate is not merely imprecise but biased in the wrong direction. The definition and units of the flux density that this argument rests on are fixed in radiation flux density.

The overlap integral and the overlap-weighted mean

The energy-dependent part of the master formula is, at its core, an overlap integral, and it is worth isolating it. Holding the other factors aside, the spectral content of the output is

I \;=\; \int \Phi(E)\,\sigma(E)\;dE ,

which is maximised, for a fixed total flux, when the flux is concentrated where $\sigma(E)$ is largest — the statement that spectral overlap is what one engineers. It is often useful to summarise the effective coupling actually experienced by the flux as an overlap-weighted mean cross-section,

\bar{\sigma}_{\text{eff}} \;=\; \frac{\displaystyle\int \Phi(E)\,\sigma(E)\;dE}{\displaystyle\int \Phi(E)\;dE} ,

the average of $\sigma$ weighted by the flux that samples it. This $\bar{\sigma}_{\text{eff}}$ is generally very different from the unweighted mean of $\sigma(E)$ : a resonance that is narrow in energy contributes strongly to $\bar{\sigma}_{\text{eff}}$ only if the flux actually populates its band, and a large cross-section in an unpopulated band contributes nothing. Reporting $\bar{\sigma}_{\text{eff}}$ rather than a bare $\langle\sigma\rangle$ is the honest way to compress the spectral overlap into a single figure of merit, because it carries the alignment information the plain average discards. The role of $\bar{\sigma}_{\text{eff}}$ within the full integrand is set out in mathematical architecture.

The angular dimension: anisotropy and acceptance

Direction enters for a reason parallel to energy but with its own physics. The flux is resolved in $\theta$ because both the incident field and the converter’s response can be anisotropic. On the field side, an ambient flux is rarely perfectly isotropic; it may arrive preferentially from one hemisphere, carrying a net directional bias that is essential to the momentum-transfer argument, since a perfectly isotropic flux deposits no net momentum. On the converter side, the device accepts incoming quanta only within some angular acceptance — a range of $\theta$ over which it couples effectively — beyond which the coupling falls off. The angular integral

\int \Phi(E,\theta)\,A(\theta)\;d\theta ,

with $A(\theta)$ the acceptance profile, is again an overlap: output is gathered only where the flux’s angular distribution and the converter’s acceptance coincide. A device with a narrow acceptance pointed away from the flux’s preferred direction collects little, however intense the flux; one whose acceptance is aligned with an anisotropic flux collects much. As with energy, the average over angle discards this alignment, and an isotropic-flux approximation applied to an anisotropic reality misstates both the magnitude and — through the momentum term — the very existence of a directed output. The connection between angular structure and the directed transfer $\Delta p$ is developed in non-equilibrium energy conversion.

Reading Φ(E,θ,t) in the master formula

All of this is why the flux appears in the master formula as a fully resolved field rather than a scalar,

P(t) = \iint \Phi(E,\theta,t)\,\sigma(E)\,\Delta p\,v_{\text{ph}}\,\eta_{\text{conv}}\;dE\,d\theta ,

with the double integral performing the spectral and angular overlaps simultaneously. The time argument $t$ carries the further truth that the flux can vary — diurnally, seasonally, or with source conditions — so that $P(t)$ inherits a time dependence from $\Phi$ even when the material factors are static. The correct summary of the chapter is that the useful power is a double overlap: of $\Phi(E,\theta)$ with $\sigma(E)$ in energy, and of $\Phi(E,\theta)$ with the angular acceptance in direction. Neither overlap survives being replaced by an average, because averaging destroys the correlations that overlap is made of. Any estimate that begins from a single mean intensity should be regarded as a lower-fidelity approximation whose error is the discarded covariance, and any illustrative number quoted for $P(t)$ inherits whatever spectral and angular assumptions went into $\Phi$ ; those assumptions must be stated, not buried. The precise symbols, units, and ranges for every factor are catalogued in variable definitions.