---
title: Photoreactivity Score in Photochemical Systems
url: https://www.emergentmind.com/topics/photoreactivity-score
type: topic
---

# Photoreactivity Score in Photochemical Systems

Searching arXiv for the cited papers and topic.
“Photoreactivity score” denotes a scalar quantity used to rank or compare light-driven chemical performance, but the literature assigns the term to different observables depending on the physical scale and application. In semiconductors, it is an orientation-resolved fraction of optical transition probability associated with a surface normal; in cancer-targeted photodynamic therapy, it is a composite score combining normalized absorption in the 700–850 nm therapeutic window with a normalized intersystem-crossing proxy; in a benchmark H\(_2\)-photoproduction reactor, it is the overall quantum yield defined as the ratio of hydrogen-production rate to absorbed-photon rate [2202.01073] [2512.15889] [2011.03792].

## 1. Terminological scope and domain-specific definitions

The term is used for quantitatively different constructs that share a common purpose: condensing a photophysical or photochemical workflow into a single ranking variable. The cited literature does not present a single universal definition; instead, it presents three domain-specific operationalizations.

| Domain | Score definition | Interpretation |
|---|---|---|
| Semiconductor photochemistry | $S(\hat n)=\dfrac{\int_{\Omega(\hat n)}P(\hat k)\,d\Omega}{\int_{4\pi}P(\hat k)\,d\Omega}$ | Fraction of all photo-excited carriers generated within a chosen angular cap around surface normal $\hat n$ |
| PDT photosensitizers | $S_i=(w_A\bar A_i+w_k\bar k_i)$ or $S_i=(\bar A_i)^{w_A}(\bar k_i)^{w_k}$ | Composite ranking from normalized therapeutic-window absorption and normalized ISC proxy |
| Photoreactor H\(_2\) production | $\Phi=\dfrac{\langle r_{H_2}\rangle}{\bar G_{abs}}$ | Overall quantum yield, i.e. mol H\(_2\) per mol photon absorbed |

In all three cases, the score is dimensionless after normalization or ratio formation, and each formulation is intended to support comparison across candidates: crystal facets, molecular photosensitizers, or photocatalytic operating conditions. A plausible implication is that the phrase “photoreactivity score” should be interpreted contextually rather than as a standardized metric across subfields.

## 2. Orientation-resolved score in semiconductor photochemistry

For semiconductors, the score is constructed from the Fermi–Golden-Rule rate of vertical optical transitions at crystal momentum $\mathbf k$ under monochromatic illumination [2202.01073]. The directional optical transition probability is written as
\[
P(\mathbf k;\nu)\propto \sum_{v\in \mathrm{VB}}\sum_{c\in \mathrm{CB}} |\mu_{cv}(\mathbf k)|^2\,
\delta(E_{c,\mathbf k}-E_{v,\mathbf k}-\hbar\nu),
\]
with transition-dipole matrix element
\[
\mu_{cv}(\mathbf k)=\left\langle c,\mathbf k\left|\tfrac{e}{m}\mathbf p\right|v,\mathbf k\right\rangle.
\]
The $\delta$-function enforces energy conservation and encodes the joint density of states at $\mathbf k$. An equivalent notation introduces
\[
J(\mathbf k;\nu)=\sum_{v,c}\delta(E_{c,\mathbf k}-E_{v,\mathbf k}-\hbar\nu),
\]
so that
\[
P(\mathbf k;\nu)=\sum_{v,c}|\mu_{cv}(\mathbf k)|^2 J(\mathbf k;\nu).
\]

The practical quantity is obtained by coarse-graining over a finite energy window, typically from the band gap $E_{\rm g}$ up to $E_{\rm g}+\Delta E$:
\[
P_{\mathbf k}=\int_{\nu_{\min}}^{\nu_{\max}}P(\mathbf k;\nu)\,d\nu
\approx
\sum_{\substack{v,c\\ E_{\min}<E_{c,\mathbf k}-E_{v,\mathbf k}<E_{\max}}}
|\mu_{cv}(\mathbf k)|^2.
\]
Each $\mathbf k_i$ is normalized to $\hat k_i=\mathbf k_i/|\mathbf k_i|$, and the weights $P_{\mathbf k_i}$ are interpolated onto a regular $(\theta,\varphi)$ grid on the unit sphere to form a continuous spherical heat map
\[
P(\hat k):S^2\to \mathbb R^+.
\]

To isolate carriers generated toward a surface with normal $\hat n$, the construction restricts to a spherical cap
\[
\Omega(\hat n)=\{\hat k\in S^2:\hat k\cdot \hat n\ge \cos\theta_c\},
\]
where $\theta_c$ is a small half-angle, e.g. \(10^\circ\ldots 20^\circ\). The raw score is
\[
S_{\rm raw}(\hat n)=\int_{\Omega(\hat n)}P(\hat k)\,d\Omega,
\]
and the normalized photoreactivity score is
\[
S(\hat n)=\frac{\int_{\Omega(\hat n)}P(\hat k)\,d\Omega}{\int_{4\pi}P(\hat k)\,d\Omega}\in[0,1].
\]
In discrete form,
\[
S(\hat n)\approx
\frac{\sum_{i:\,\hat k_i\cdot \hat n\ge \cos\theta_c}P_i w_i}
{\sum_{i=1}^{N}P_i w_i},
\]
with $w_i$ the integration weight. For a regular Monkhorst–Pack mesh one may take $w_i\equiv 1/N$.

This formulation is designed to identify the surfaces associated with the largest number of photo-generated carriers. The underlying significance is that photochemical reactions on semiconductors are anisotropic, and the score supplies a facet-resolved ranking that uses all possible transitions weighted by their transition dipole moments rather than band dispersion alone.

## 3. Hybrid-DFT implementation and the rutile TiO\(_2\) example

The semiconductor implementation is specified within hybrid-DFT in the independent-particle approximation [2202.01073]. The exchange–correlation functional is HSE06 with 25% exact exchange; the plane-wave cutoff is 500 eV; PAW potentials are used as in Table S1 of the paper; and transition matrix elements are computed via the PAW $\mathbf p$-operator. For rutile TiO\(_2\), the Brillouin zone is sampled with a \(4\times 4\times 6\) $\Gamma$-centered mesh, giving \(N=96\) k-points. The energy window extends from the PBE-HSE gap \(E_{\rm g}\) up to \(E_{\rm g}+0.8\,\mathrm eV\), intended to mimic visible/near-UV excitation.

The worked example uses rutile TiO\(_2\) with \(E_{\rm g}=3.48\) eV and \(\Delta E=0.8\) eV. Choosing \(\theta_c=15^\circ\), the discrete evaluation yields
\[
S(001)\approx 0.42,\qquad S(101)\approx 0.22.
\]
Under this definition, roughly 42% of all photo-excited carriers are generated within \(15^\circ\) of the [001] direction, compared to 22% within \(15^\circ\) of [101]. The paper interprets this as quantifying the stronger photoreactivity expected for the TiO\(_2\)(001) facet under near-band-gap illumination.

The methodological significance lies in the contrast with conventional band-structure inspection. The results indicate that it is generally possible to correlate the heat maps with anisotropy visible in conventional band-structure plots, but they also demonstrate that band-structure plots do not always provide all the informations. Taking into account the contribution of all possible transitions weighted by their transition dipole moments is therefore presented as crucial for a complete picture.

## 4. Composite score for cancer-targeted photosensitizers

For photodynamic therapy, the score is built from two observables that the study identifies as central to photosensitizer performance: cumulative absorption in the therapeutic window and the efficiency of singlet–triplet intersystem crossing [2512.15889]. The one-photon absorption cross-section is
\[
\sigma_A(\omega)=\frac{4\pi}{3\hbar c}\,\omega
\sum_f \frac{\eta\,|\langle E_f|D|E_0\rangle|^2}{(E_f-E_0)^2+\eta^2},
\]
and the total absorption in the 700–850 nm window is
\[
A_{\rm window}=\int_{\omega_{\rm lo}}^{\omega_{\rm hi}} d\omega\,\sigma_A(\omega).
\]
In the stick-spectrum limit \(\eta\to 0\),
\[
A_{\rm window}=\sum_{\Delta E_f\in [E_{\rm lo},E_{\rm hi}]}
|\langle E_f|D|E_0\rangle|^2.
\]

The second ingredient is an intersystem-crossing proxy. In the weak-coupling regime, the true ISC rate out of the first singlet into triplets is, to leading order, proportional to \(|\langle T|H_{\rm SOC}|S\rangle|^2\). Instead of computing every spin–orbit matrix element, the study defines
\[
\tilde k_{\rm ISC}(t)\equiv
|\langle E_{1,S=1,M}|e^{-itH_{\rm SOC}}|E_{1,S=0}\rangle|^2
\]
and states that for \(t\ll \hbar/\|H_{\rm SOC}\|\),
\[
\tilde k_{\rm ISC}(t)\simeq t^2 |\langle T|H_{\rm SOC}|S\rangle|^2
\Rightarrow
|\langle T|H_{\rm SOC}|S\rangle|^2\simeq \tilde k_{\rm ISC}(t)/t^2.
\]

The composite photoreactivity score is then formed by combining normalized absorption and normalized ISC proxy. For candidate \(i\),
\[
\bar A_i\equiv A_{{\rm window},i}/\max_j A_{{\rm window},j},
\qquad
\bar k_i\equiv (\tilde k_{{\rm ISC},i}/t^2)/\max_j (\tilde k_{{\rm ISC},j}/t^2),
\]
and one may define
\[
S_i=(w_A\bar A_i+w_k\bar k_i).
\]
In its simplest form, equal weights \(w_A=w_k=\tfrac12\) may be used. An alternative geometric mean is also given:
\[
S_i=(\bar A_i)^{w_A}(\bar k_i)^{w_k}.
\]
The text states that the geometric mean can be used if one wants to penalize poor performance in either channel more sharply.

By ranking molecules by \(S_i\), the study frames the score as a physically meaningful Pareto-optimal trade-off between deep-tissue absorption and spin-flip efficiency. This differs structurally from the semiconductor score: there the score is a directional fraction of transition probability, whereas here it is an explicit multi-objective aggregation over normalized molecular descriptors.

## 5. Fault-tolerant quantum workflow and resource estimates

The PDT formulation is explicitly tied to a fault-tolerant quantum workflow for BODIPY derivatives, including heavy-atom and transition-metal-substituted systems that are described as challenging for classical methods [2512.15889]. To compute \(A_{\rm window}\) without resolving every spectral peak, the method prepares the normalized dipole-acted ground state
\[
|\tilde\psi\rangle = D|E_0\rangle/\sqrt{\mathcal N_D},
\qquad
\mathcal N_D=\sum_f |\langle E_f|D|E_0\rangle|^2,
\]
block-encodes the electronic Hamiltonian \(H\) in a low-rank THC form, and then uses qubitization plus generalized Quantum Signal Processing to build two Heaviside-style projectors, one onto energies \(\le E_{\rm hi}\) and one onto energies \(\ge E_{\rm lo}\). In each shot, both filters are applied and a single ancilla qubit is measured; if its value is \(b\in\{0,1\}\), then repeating \(S\) shots yields
\[
P_{\rm window}=\langle b\rangle \Rightarrow A_{\rm window}=\mathcal N_D\cdot P_{\rm window}.
\]

To achieve a sampling error \(\epsilon_{\rm samp}\) with confidence \(1-\delta_{\rm samp}\), the shot count is
\[
S\approx \left\lceil (2\epsilon_{\rm samp}^2)^{-1}\ln(2/\delta_{\rm samp})\right\rceil.
\]
The specific example uses \(\epsilon_{\rm samp}=0.1\), \(\delta_{\rm samp}=10^{-2}\), and therefore \(S\approx 1.3\times 10^2\). Each shot costs two QSP projections, each of degree \(d=O((\lambda'/\Delta)\,{\rm poly}\log \epsilon_H^{-1})\) calls to the walk operator \(W\).

For ISC, the method prepares a two-branch superposition of singlet and triplet reference states via the sum-of-Slaters method and a single-sided QSP-based low-energy projector, evolves under the one-body \(H_{\rm SOC}\), and then performs a modified Hadamard test to read off \(\operatorname{Re}\langle T|U(t)|S\rangle\) and \(\operatorname{Im}\langle T|U(t)|S\rangle\) from ancilla \(X\) and \(Y\) measurements. Repetition to precision \(\epsilon\) requires \(S_{\rm Had}=O(1/\epsilon^2)\) shots; for \(\epsilon=0.1\), \(S_{\rm Had}\approx 4.8\times 10^2\). The workflow also describes a vibronic alternative based on the spin-vibronic Koppel–Domcke–Cederbaum Hamiltonian
\[
H=T_{\rm nuc}+V_0+W'(Q),
\]
with rate extraction from
\[
k_{\rm ISC;vib}\simeq (d/dt)\,P_T(t)\rvert_{t\to 0}.
\]

The resource estimates state that active spaces ranging from 11 to 45 spatial orbitals can be simulated using 180–350 logical qubits and Toffoli gate depths between \(10^7\) and \(10^9\). The detailed workflow specifies active-space construction for four BODIPY derivatives, Hamiltonian factorization with PySCF followed by THC or compressed double factorization, and sum-of-Slaters state preparation with depth \(O(D\log D)\) for \(D\sim 10^4\) determinants. A width–depth trade-off in the QROM-based SELECT gives cumulative-absorption costs up to \(\simeq 2.5\times 10^9\) Toffolis per shot and ISC-proxy costs up to \(\simeq 5.8\times 10^9\) per shot; multiplying by the required \(10^2\)–\(10^3\) shots yields total Toffoli budgets \(\sim 10^{11}\)–\(10^{12}\) in the worst case, but \(\sim 10^9\)–\(10^{10}\) for more moderate accuracy targets or smaller spaces.

## 6. Reactor-scale quantum-yield score in photocatalytic H\(_2\) production

In the benchmark photoreactor study, the photoreactivity score is the overall quantum yield \(\Phi\), defined from the mean volumetric H\(_2\)-production rate and the mean volumetric rate of radiant light absorbed [2011.03792]. In the flat-torus reactor, neglecting reflections at the walls, the mean volumetric rate of photon absorption is
\[
\bar G_{abs}=(A)=
a_{\rm light}
\int_{\lambda_{\min}}^{\lambda_{\max}}
[q_{in,\lambda}-q_{out,\lambda}]\,p_\lambda\,d\lambda,
\]
with illuminated specific surface \(a_{\rm light}=36.8\ {\rm m}^{-1}\). The incoming monochromatic hemispherical photon flux density is \(q_{in,\lambda}=q\,f_{a,LED}(\lambda)\), and the outgoing flux is \(q_{out,\lambda}=q_{out}\,f_{a,rear}(\lambda)\). The integration limits are \(\lambda_{\min}=400\) nm and \(\lambda_{\max}=490\) nm. The probability that an absorbed photon is absorbed by eosin Y rather than by the catalyst is
\[
p_\lambda=
\frac{\varepsilon_{a,EY}(\lambda)\,C_{EY}}
{\varepsilon_{a,EY}(\lambda)\,C_{EY}+\varepsilon_{a,Cat}(\lambda)\,C_{Cat}}.
\]

Hydrogen production is measured from pressure rise in the sealed headspace. Under continuous stirring at 1000 rpm and isothermal conditions \(25.0\pm 0.1^\circ{\rm C}\), the steady-regime balance gives
\[
r_{H_2}=\frac{dn_{\rm tot}}{dt}
=
\left(\frac{V_G}{RT}+\frac{V_L}{H_{H_2}(T)}\right)\frac{dP}{dt},
\]
with \(V_L=160\) mL, \(V_G=30\) mL, \(H_{H_2}(25^\circ{\rm C})=1.3077\times 10^5\ {\rm m}^3\cdot {\rm Pa}\cdot {\rm mol}^{-1}\), \(R=8.314\ {\rm J}\cdot {\rm mol}^{-1}\cdot {\rm K}^{-1}\), and \(T=298.15\) K. The pressure rise is recorded by a Keller PA 33X transducer and Read 30 software, while online GC confirms that the only evolving gas is H\(_2\).

Plotting \(\langle r_{H_2}\rangle\) versus \(\bar G_{abs}\) gives a straight line through the origin, yielding the linear coupling law
\[
\langle r_{H_2}\rangle = \Phi\,\bar G_{abs}.
\]
The score is then defined as
\[
\Phi=\frac{\langle r_{H_2}\rangle}{\bar G_{abs}},
\]
which is dimensionless and reported as mol H\(_2\) per mol photon. An equivalent expression uses total absorbed photons per second, \(N_{ph,abs}=\bar G_{abs}V_L\):
\[
\Phi=\frac{r_{H_2}\;(\text{mol}\cdot {\rm s}^{-1})}
{N_{ph,abs}\;(\text{mol hv}\cdot {\rm s}^{-1})}
=
\frac{r_{H_2}}{\bar G_{abs}\,V_L}.
\]

Typical values at \([\mathrm{Cat}]=0.044\) mM and \([\mathrm{EY}]=0.20\) mM give
\[
\Phi=(4.7\pm 0.2)\times 10^{-3}\ {\rm mol\ H_2\cdot mol\ hv^{-1}},
\]
while other runs varying \([\mathrm{EY}]/[\mathrm{Cat}]\) give \(\Phi\) in the range \(2.2\)–\(6.1\times 10^{-3}\) with uncertainties of \(0.2\)–\(0.9\times 10^{-3}\). The protocol is explicitly generalized to other photocatalytic systems by characterizing incident and transmitted spectra, computing \(\bar G_{abs}\), measuring the reaction rate, testing linear or non-linear coupling, and extracting the slope as \(\Phi\) or a more complex function if non-linear.

## 7. Comparative interpretation, methodological cautions, and recurring misconceptions

Across the three formulations, the score always compresses a high-dimensional photophysical problem into a scalar, but the compressed object differs substantially. In the semiconductor case, the score is an angularly restricted share of total transition probability. In PDT molecular screening, it is a weighted combination of two normalized observables. In the photoreactor study, it is a process-level ratio between chemical productivity and absorbed radiant input [2202.01073] [2512.15889] [2011.03792].

A recurrent misconception is to treat band-structure anisotropy by itself as a sufficient predictor of facet photoreactivity. The semiconductor study explicitly states that conventional band-structure plots do not always provide all the informations, and that a complete picture requires all possible transitions weighted by their transition dipole moments. Another potential misconception is to regard any photoreactivity score as intrinsically universal. The cited literature instead shows that normalization conventions, observables, and intended use cases are domain dependent. This suggests that cross-study comparison is meaningful only after verifying what has been normalized, over which energy or wavelength window, and whether the score measures directionality, molecular trade-offs, or reactor-scale quantum yield.

The three definitions also differ in what they omit. The semiconductor score is constructed within hybrid-DFT in the independent-particle approximation. The PDT score may use an ISC proxy, with vibronic dynamics introduced only where appropriate. The reactor score is valid within the measured radiative balance and can support either linear or non-linear coupling analysis, depending on the observed behavior. These are not contradictions; they are differences in modeling target and experimental or computational granularity.

Taken together, the literature supports using “photoreactivity score” as a family of operational metrics for ranking light-driven systems. The common structure is not a shared formula but a shared function: converting absorption, transition probability, spin conversion, or chemical output into a dimensionless or normalized ranking variable that is specific to the photochemical question being asked.

Source: https://www.emergentmind.com/topics/photoreactivity-score