Explaining the magnitude of Chirality-Induced Spin Selectivity via electron-electron exchange
Extracted source text ยท arXiv 2607.06858
%% ABABAB !TEX TS-program = LuaLaTeX
\documentclass[%
reprint,
%superscriptaddress,
%groupedaddress,
%unsortedaddress,
%runinaddress,
%frontmatterverbose,
%preprint,
%preprintnumbers,
%nofootinbib,
%nobibnotes,
%bibnotes,
amsmath,amssymb,
aps,
prl,
%pra,
%prb,
%rmp,
%prstab,
%prstper,
%floatfix,
]{revtex4-2}
\usepackage{dcolumn}% Align table columns on decimal point
\usepackage{amsmath,amsthm,amssymb}
%\usepackage[numbers,sort&compress]{natbib} % \citenum
\usepackage[dvipsnames]{xcolor} % More colours
\usepackage{lmodern} %Font
\usepackage{graphicx}
\usepackage{float} %H in figure environment
\usepackage{bm} %\v
\usepackage{braket}
\usepackage{slashed} %feynman slash notation
\usepackage{tikz}
\usepackage[compat=1.1.0]{tikz-feynman}
\usepackage[utf8]{inputenc}
\usepackage{xr}
\newcommand*{\red}[1]{\textcolor{red}{#1}}
\newcommand*{\green}[1]{\textbf{\textcolor{Green}{#1}}}
\renewcommand{\t}[1]{\text{#1}}
\renewcommand{\Re}[1]{\t{Re}\{{#1}\}}
\newcommand*{\iu}{{\mathrm{i}\mkern1mu}}
\def\bfr{{\bf r}}
\renewcommand*{\v}[1]{\bm{#1}}
\newcommand{\Angstrom}{\AA\ }
\newcommand{\DbyD}[1]{\frac{\partial}{\partial {#1}}}
\newcommand{\td}{\t{d}}
\newcommand{\qdot}{\Dot{q}}
\newcommand{\pdot}{\Dot{p}}
\newcommand{\LorentzFwd}[2]{\Lambda_{\ #2\ }^{#1}}
\newcommand{\LorentzBwd}[2]{\Lambda_{#2}^{\ #1\ }}
\newcommand{\vdot}[1]{\Dot{\v{#1}}}
\newcommand{\half}{\frac{1}{2}}
\renewcommand{\vector}[1]{\begin{pmatrix}#1\end{pmatrix}}
\newcommand{\vsigma}{\v{\sigma}}
\newcommand{\vhat}[1]{\hat{\v{#1}}}
\newcommand{\TotDeriv}[2]{\frac{\td #1}{\td #2}}
\newcommand{\SqBraket}[1]{\left[#1\right]}
\newcommand{\create}[1]{\psi^\dagger(\v{r}#1,t)}
\newcommand{\annihilate}[1]{\psi(\v{r}#1,t)}
\newcommand{\GHat}{\Hat{G}}
\newcommand{\Eh}{\t{E}_{\t{h}}}
\usepackage{filecontents}
\begin{filecontents}{UHFCISSSI.aux}
\relax
\newlabel{FirstPage}{{}{1}{}{}{}}
\@writefile{toc}{\contentsline {title}{Supplementary Information for `Explaining the magnitude of Chirality-Induced Spin Selectivity via electron-electron exchange'}{1}{}}
\@writefile{toc}{\contentsline {section}{\numberline {I}Symmetry considerations}{2}{}}
\newlabel{sec:SymmetryConsiderations}{{I}{2}{}{}{}}
\citation{altmann_rotations_1986,altmann_point-group_1994}
\@writefile{toc}{\contentsline {subsection}{\numberline {A}Point group Symmetry}{3}{}}
\@writefile{lot}{\contentsline {table}{\numberline {I}{\ignorespaces Spinor representation of the point group of $C_{2v}$. The rest of the elements of the double group can be found as the negatives of these elements. This representation is in the Cartan gauge. Note that $g^{-1} = g^\dagger $. }}{3}{}}
\newlabel{table:RepsofC2v}{{I}{3}{}{}{}}
\citation{barron_time_nodate,burton_parity-time_2019}
\citation{barron_true_1986}
\@writefile{toc}{\contentsline {subsection}{\numberline {B}Time reversal symmetry}{5}{}}
\newlabel{eq:TimeReversalAntiUnitary}{{14}{5}{}{}{}}
\newlabel{eq:SigmaUnderT}{{17}{5}{}{}{}}
\@writefile{toc}{\contentsline {subsection}{\numberline {C}Wavefunctions}{6}{}}
\newlabel{SI:Wavefunctions}{{I\,C}{6}{}{}{}}
\newlabel{eq:pointGroupEigenFunction}{{24}{6}{}{}{}}
\citation{barron_true_1986}
\@writefile{toc}{\contentsline {subsection}{\numberline {D}Photoelectrons}{8}{}}
\newlabel{SI:photoelectrons}{{I\,D}{8}{}{}{}}
\@writefile{toc}{\contentsline {subsection}{\numberline {E}Hartree--Fock}{8}{}}
\citation{artacho_nonorthogonal_1991}
\@writefile{toc}{\contentsline {section}{\numberline {II}Green's functions in a non-orthogonal basis}{9}{}}
\newlabel{SI:NonOrthGreensFuncSec}{{II}{9}{}{}{}}
\citation{datta_electronic_1995,rammer_quantum_2007}
\@writefile{toc}{\contentsline {section}{\numberline {III}Self Energy Terms}{10}{}}
\newlabel{SI:SelfEnergyTerms}{{III}{10}{}{}{}}
\newlabel{eq:DysonMatrixEquations}{{40}{10}{}{}{}}
\citation{datta_electronic_1995}
\newlabel{SI:eq:GnDefinition}{{49}{11}{}{}{}}
\citation{pokhilko_iterative_2022}
\citation{rammer_quantum_2007}
\@writefile{lot}{\contentsline {table}{\numberline {II}{\ignorespaces Parameters used to setup the calculation unless otherwise stated. $\gamma $ terms are the coupling strength to the 3 different leads. $\mu $ are the chemical potentials of the different leads. Parameters are described in more detail in section \ref {SI:SelfEnergyTerms}}}{13}{}}
\newlabel{table:Parameters_used}{{II}{13}{}{}{}}
\@writefile{toc}{\contentsline {section}{\numberline {IV}Calculation Specifics}{13}{}}
\@writefile{toc}{\contentsline {section}{\numberline {V}Expression for the current}{13}{}}
\newlabel{SI:sec:Current}{{V}{13}{}{}{}}
\citation{datta_electronic_1995}
\@writefile{toc}{\contentsline {section}{\numberline {VI}Supplementary Figures}{14}{}}
\newlabel{SI:sec:DifferentParameters}{{VI}{14}{}{}{}}
\newlabel{SI:eq:SpinDensityDiff}{{62}{14}{}{}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces Spin difference in number of electrons (equation \ref {SI:eq:SpinDensityDiff}). The $X$, $Y$ component of the spin density difference is zero. This is because spin-orbit coupling was not included in the calculation, therefore each spin `chooses' to align parallel, or anti-parallel. Parameters are as shown in table \ref {table:Parameters_used}. This is using the full $G^n$ not the non-equilibrium component, equation Main Text-\ref {MT-eq:nonEqSpinDensityComponent}. }}{15}{}}
\newlabel{SI:fig:SpinDensityDiffXYZ}{{1}{15}{}{}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces Energy resolved spin polarised current along $Z$ direction. Left: the parameters were modified such that there is no spin bias from the bulk copper, only the photoelectrons. Since this leads to a much smaller spin-symmetry breaking the CISS effect is much smaller. Right: No spin polarisation of photoelectrons, but spin polarisation of bulk copper. Parameters are as shown in table \ref {table:Parameters_used} except for. Left: $\gamma _b^{\uparrow } = \gamma _b^{\downarrow } = 10^{-2}\text {E}_{\text {h}}$. Right: $\gamma _p^{\uparrow } = \gamma _p^{\downarrow } = 10^{-2}\text {E}_{\text {h}}$}}{15}{}}
\newlabel{SI:fig:SpinPolarisedCurrentwithNoSpinBias}{{2}{15}{}{}{}}
\bibdata{UHFCISSSINotes,CISS}
\bibcite{altmann_rotations_1986}{{1}{}{{}}{{}}}
\@writefile{toc}{\contentsline {subsection}{\numberline {A}Stability of effect with variation of parameters}{16}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces Energy resolved spin polarised current along $Z$ direction. Parameters are as shown in table \ref {table:Parameters_used} except for. Left: $\gamma _p^{\uparrow } = 1.04 \times 10^{-5}\text {E}_{\text {h}}$, $\gamma _p^{\downarrow } = 0.96 \times 10^{-5}\text {E}_{\text {h}}$ . Right: $\gamma _s^{\uparrow } = \gamma _s^{\downarrow } = 10^{-3}\text {E}_{\text {h}}$ }}{16}{}}
\newlabel{SI:fig:SpinPolarisedCurrentSmallerGamma}{{3}{16}{}{}{}}
\bibcite{altmann_point-group_1994}{{2}{}{{}}{{}}}
\bibcite{barron_time_nodate}{{3}{}{{}}{{}}}
\bibcite{burton_parity-time_2019}{{4}{}{{}}{{}}}
\bibcite{barron_true_1986}{{5}{}{{}}{{}}}
\bibcite{artacho_nonorthogonal_1991}{{6}{}{{}}{{}}}
\bibcite{datta_electronic_1995}{{7}{}{{}}{{}}}
\bibcite{rammer_quantum_2007}{{8}{}{{}}{{}}}
\bibcite{pokhilko_iterative_2022}{{9}{}{{}}{{}}}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces Energy resolved spin polarised current along $Z$ direction. Parameters are as shown in table \ref {table:Parameters_used} except for. Factor $1.04 \rightarrow 1.01$, Factor $0.96 \rightarrow 0.99$. Right: $1.04 \rightarrow 1$, Factor $0.96 \rightarrow 1$}}{17}{}}
\newlabel{SI:fig:SpinPolarisedCurrentSmallerBias}{{4}{17}{}{}{}}
\@writefile{toc}{\contentsline {section}{\numberline {}References}{17}{}}
\bibstyle{unsrt}
\providecommand\NAT@force@numbers{}\NAT@force@numbers
\newlabel{LastBibItem}{{9}{18}{}{}{}}
\newlabel{LastPage}{{}{18}{}{}{}}
\gdef \@abspage@last{18}
\end{filecontents}
\externaldocument[SI-]{UHFCISSSI}
\DeclareUnicodeCharacter{2212}{-}
\begin{document}
\preprint{APS/123-QED}
\title{Explaining the magnitude of Chirality-Induced Spin Selectivity via electron-electron exchange}
%\thanks{A footnote to the article title}%
\author{Bence Csakany}
\email{bc528@cam.ac.uk}
\author{Alex J. W. Thom}
\email{ajwt3@cam.ac.uk}
\affiliation{Yusuf Hamied Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, UK. CB1 2EW}
\begin{abstract}
Chiral molecular structure can couple to electron spin, leading to unexpected spin polarization effects. This Chirality-Induced Spin Selectivity (CISS) was first reported for DNA molecules adsorbed on gold but its microscopic origin remains unclear. We demonstrated though simulation that the exchange arising from electron-electron Coulomb interactions within the self-consistent mean field (Hartree--Fock) approximation can yield significant ($\sim 2\%$) spin polarisation for $3$-methylcyclohexanone adsorbed on Cu(111) amplifying a much smaller ($\sim 0.0014\%$) initial bias, consistent with experiment. Symmetry considerations ensure the result is physically meaningful, while its ab-initio nature ensures all parameters are physically realistic. This amplification is connected to existing studies on spin-symmetry breaking in Hartree--Fock, providing a new pathway for understanding the magnitude of CISS as an emergent phenomenon of interacting electrons.
%\begin{description}
%\item[Usage]
%Secondary publications and information retrieval purposes.
%\item[Structure]
%You may use the \texttt{description} environment to structure your abstract;
%use the optional argument of the \verb+\item+ command to give the category of each item.
%\end{description}
\end{abstract}
\maketitle
\section{Introduction}
Chirality-Induced Spin Selectivity (CISS) is a phenomenon in which structural chirality leads to strong spin-dependent electronic behaviour. This is in spite of the weakness of the relativistic spin-orbit coupling which is believed to induce it. CISS has been observed using a diverse range of molecules such as DNA, helicenes, oligopeptides, and chiral self-assembled monolayers\cite{gohler_spin_2011,safari_enantioselective_2024,giaconi_efficient_2023}. The effect is typically measured by electron transmission experiments; however, CISS also has an effect on rates of reaction\cite{lu_enantiospecificity_2021}. Beyond being of intrinsic interest, CISS has also been suggested as an explanation of the homochirality of life\cite{ozturk_origin_2023} as well as having applications in spintronics\cite{michaeli_new_2017}. \\
Here we focus on CISS as observed in UV photo-emission spectroscopy experiments, where the spin polarisation (which can range from 0\% (unpolarized) to 100\% (fully spin polarized)) of the resulting photoelectron current is observed in the 1--60\% range, depending on the system.
Despite its experimental reporting over a decade ago\cite{gohler_spin_2011}, the mechanism underlying the CISS effect is still uncertain\cite{bloom_chiral_2024,evers_theory_2022}. Various microscopic mechanisms have been proposed\cite{fransson_chiral_2024,dalum_theory_2019,shitade_geometric_2020,zheng_chirality-driven_2025,di_ventra_chirality-induced_2025}, though none have yet produced a predictive and generally accepted explanation for the magnitude of the effect. Though many explanations reproduce a large ($\sim 10\%$) photoelectron CISS effect, this is typically caused by invoking unphysical parameters\cite{fransson_chiral_2025,upadhyay_weak_2026} or accidental degeneracy\cite{dalum_theory_2019}. \\
A leading explanation for the CISS effect originates in electron--electron exchange\cite{ozturk_origin_2023,fransson_chirality-induced_2019}, which modifies the Coulombic repulsion between electrons of the same spin. However, these Coulomb interactions do not, by themselves, connect \textit{chirality} to spin. In this work, we show that Coulombic electron-electron exchange interactions create an amplification of an existing spin bias and this correctly predicts the magnitude of the observed spin polarisation seen in the the CISS effect. It requires no unphysical parameters, accidental degeneracy or fine-tuning. The exchange interaction provides the amplification necessary for relativistic and chiral phonon effects to create the experimentally observed magnitudes of the CISS effect in a wide range of structures. In contrast to other ab-initio studies\cite{naskar_chiral-induced_2023} incorporating exchange interactions as well as other effects, we show that the exchange interaction is the primary cause of the magnification.
\section{Background}
Spin-Orbit Coupling (SOC) is responsible for coupling the chirality of molecular geometry to spin. SOC results from the coupling of the momentum and the spin of electrons in relativistic (Dirac) theory, along with the conversion of an electric field into a magnetic field via Lorentz transformations. Even with SOC, point group and time reversal symmetries can prevent a spin polarisation being observed (see SI) and so CISS is always observed in non-equilibrium scenarios, with molecules that have no significant point group symmetry. A detailed understanding of SOC is not necessary here; it is sufficient to understand that for electrons with a net (possibly angular) momentum in an electric field, typically there is a small energetic difference between up and down spin. This energy difference is shown as an `external influence' in figure \ref{fig:MoDiagram}. \\
Dalum\cite{dalum_theory_2019} showed that in the case of orbitals close in energy, a significant ($\sim$10--20\%) spin polarisation can arise using SOC. However when the orbitals were further in energy only a much smaller polarisation was obtained. This is not an issue here however, as we show only a small spin polarisation is needed for the exchange effects (`amplification' in figure \ref{fig:MoDiagram}) to cause amplification.
%What appears to be a $O(1/c^0)$ SOC term can be obtained using only non-relativistic Schr\"{o}dinger theory\cite{di_ventra_chirality-induced_2025}. This is a local $U(1)$ gauge transformation of the Schr\"{o}dinger equation into a rotating reference frame \cite{anandan_quantum_2003} any such effects are therefore automatically included here.
%It is not the only relativistic correction to $O(1/c^2)$ in a multi electron picture (section 13.2.3 M. Reiher + A. Wolf). It is however the effect that has been studied the most in regards to CISS. In atomic units this reduces its effect by a factor of $\sim 10000$ compared to the Coulomb interaction. For light atoms it is on the order of meV as compared to the Coulomb interaction on the order of $2\t{eV}$. It has been suggested\cite{shitade_geometric_2020} that by constraining an electron to a helix within Dirac theory, the non-relativistic limit SOC should scale as $O(c^0)$. Shitade calculates this puts the effect on the order of $160\t{meV}$ for a typical DNA helix. For typical molecular parameters this is still an order of magnitude below that of the Coulomb interaction between nearest neighbours. It would however be sufficient to create an initial spin bias that can be amplified by exchange and spin-symmetry breaking. As stated by Shitade\cite{shitade_geometric_2020}, the non-relativistic limit is different depending on if the 1D helical constraint is applied before or after the non-relativistic limit is taken. This issue stems from constraining the Dirac equation in curvilinear coordinates\cite{brandt_dirac_2016}. As is also stated, spin orbit coupling in a 1D system can be removed via a unitary transformation, in this case via returning to the non-rotating coordinate system. It is therefore important to include something that breaks the local nature of a free electron Hamiltonian, such as Coulomb interactions, before returning to the non-relativistic limit. Further work is needed to resolve the nature and effect of the sublattice degrees of freedom.\\
\begin{figure}[h!]
\centering
\includegraphics[clip, trim=3.5cm 4cm 1.5cm 4cm, width=\linewidth]{Plots/SpinMoEnergyShiftAndAmplify.pdf}
\caption{\textbf{Schematic diagram showing the change in energy of each electron in a molecule due to an external influence.} Each electron is in an approximate single particle state, a molecular orbital. The energy of these orbitals is perturbed slightly by an external effect such as chiral phonons or spin-orbit coupling, along with breaking of time-reversal symmetry. Once the symmetry is broken, electron--electron exchange effects amplify the symmetry breaking leading to a much larger effect.}
\label{fig:MoDiagram}
\end{figure}
Another source of small energetic differences between up and down spin electrons is from chiral phonons. Chiral phonons are molecular/crystal nuclear vibrations that have angular momentum. It has been demonstrated in both experiment and theory that chiral phonons can couple to electronic degrees of freedom in a spin-selective way \cite{fransson_vibrational_2020,luo_large_2023}. When out of thermal equilibrium, these chiral phonons act as a spin-pump creating a buildup of spin on an interface\cite{kim_chiral-phonon-activated_2023}. This causes the small difference in the energy of up/down spin as shown in figure \ref{fig:MoDiagram}, which is then amplified by exchange effects. In thermal equilibrium, however, no such effect is observed due to time-reversal symmetry\cite{dalum_theory_2019}.
Point group symmetries can also prevent a spin polarisation as shown in SI-\ref{SI-sec:SymmetryConsiderations}.
Extensive theoretical studies have attempted to use chiral phonons to predict a macroscopic spin difference\cite{fransson_charge_2021,fransson_charge_2022,fransson_chiral_2024}. However, to achieve this in current models purely via phonons, it appears that unphysically small phonon frequencies must be assumed.\\
Ab-initio\cite{naskar_chiral-induced_2023,zheng_chirality-driven_2025} and model\cite{zollner_insight_2020,dalum_theory_2019,shitade_geometric_2020,medina_continuum_2015} Hamiltonians have both been used to investigate the effect of electron-electron and electron-nucleus spin-orbit coupling. Both within the fully relativistic and perturbative relativistic pictures, these calculations have shown the existence of solutions with a spin texture. Because these were complicated relativistic calculations, the effects of exchange, if included, and spin-orbit coupling were not explicitly disentangled and how the spin texture came about was not the focus. It is this disentanglement of effects that is performed here.\\
%\subsection{Symmetry considerations}
%\label{sec:SymmetryConsiderations}
%It is shown in the SI that for \textbf{any} experiment, a non-zero spin polarisation of the form of equation \ref{eq:SpinPolEq} can only be obtained if certain symmetries are broken. In the case of the photoelectron experiment these are time-reversal $\mathcal{T}$ and a mirror plane in the vertical direction, $\sigma_v$. If the experimental setup is not symmetric under these operations, then a non-zero spin polarisation can be observed. Indeed for the experiment under consideration here this is the case. Note that the experimental setup obtained by applying an operation such as a mirror plane will, naturally, have the related spin-polarisation. \\
%As an example consider a water molecule. It is known that water has a dipole, yet if the Schr\"{o}dinger equation including the nuclear degrees of freedom were to be solved exactly for water, it would have no dipole. This is due to the inversion symmetry of the Schr\"{o}dinger equation. If however the dipole of water were to be measured, it is likely that an external electric field would be applied. This breaks the inversion symmetry of the system and allows a dipole to be measured. The experimental and the spatially inverted experimental system are distinct geometrically, but have identical energies and symmetry related dynamics. The same argument follows for CISS and spin polarisation. If the experiment does not break time reversal symmetry along with the other necessary ones, a non-zero spin polarisation cannot be observed. This is discussed further in reference \citenum{m_theiler_non-hermitian_nodate}.\\
%It is important to choose an electronic structure method that respects the symmetry of the problem at hand. For the Hartree--Fock family these symmetries are collected in figure \ref{fig:HFTREE}. Each flavour of Hartree--Fock breaks certain symmetries and is best suited when that symmetry is already broken in the experiment.
%Therefore in our case complex generalised Hartree--Fock\cite{burton_parity-time_2019} is best suited to the experiment.
%\begin{figure}[h!]
%\centering
%\includegraphics[width=\linewidth]{Plots/HFTREE-HUGH.png}
%\caption{The family of Hartree--Fock methods and what symmetries they conserve. The method employed here best aligns with c-GHF. $S_\t{z}$ and $S^2$ symmetry is broken via the explict spin doping. $\mathcal{T}$ symmetry is broken via the experimental setup, the photoelectrons. $\mathcal{K}$ symmetry is broken due to the open boundary conditions. A plane wave single electron solution $\psi = \t{e}^{\iu k x}$ does not have $\mathcal{K}$ or $\mathcal{T}$ symmetry. \green{TODO make this figure myself or cut it.} }
%\label{fig:HFTREE}
%\end{figure}
% CISS experiments frequently incorporate magnets\cite{safari_enantioselective_2024}. A distinction should be drawn between the exchange interaction responsible for a ferromagnet and the magnetic field created by the magnet. The interaction of a spin with a magnetic field is much smaller\footnote{The $B$ field is reduced by a factor $\mu_0$ compared to $H$. In atomic units $\mu_0 = 1/c^2$} than that of two electrons interacting via the Coulomb force. The exchange interaction (Coulomb based) is responsible for aligning the spins in a ferromagnet, not the interaction of the magnetic dipoles of electron. The magnetic interaction leads to the much larger scale effect of magnetic domains. This distinction is often neglected and can lead to misinterpretations. Usually, a magnetic surface is modelled purely by an excess of spin. The magnetic dipole of the electron is not explicitly modelled and there are \textbf{no magnetic dipole-dipole interactions} taking place.
\section{Model Experiments}
To computationally investigate the CISS effect in photoelectron spectroscopy we have created two simulated experiements based upon a chiral molecule, $3$-methylcyclohexanone ($3$-MCHO), adsorbed to a Cu(111) surface as shown in figure \ref{fig:NonEqDensityPIAndPZ}. This system is a simplified model of the experiment conducted in reference \citenum{badala_viswanatha_vectorial_2022} where the spin- and energy-resolved photoelectron spectrum for $3$-MCHO adsorbed on Cu(643) was obtained.
The simulations are performed in a Green's function framework consisting of semi-infinite modelled leads and an explicitly represented \textbf{system}.
The first `equilibrium' experiment is partitioned into two sections: i) a \textbf{bulk} copper lead; and ii) the \textbf{system} consisting of a single layer of Cu(111) with 3-MCHO adsorbed onto it.
The \textbf{bulk} copper lead is modelled by a electron sea via its Green's function. This couples to the \textbf{system} whose is explicitly represented in a localised Gaussian basis and described by a conventional quantum chemical Hamiltonian.
The second `photo-electron' experiment contains the same components as the `equilibrium', but with two additional leads: iii) the \textbf{photoelectron} lead is a semi-infinite lead, physically superimposed on the \textbf{bulk} lead, but does not interact with it; and iv) a \textbf{vacuum} lead lying beyond the \textbf{system} to act as a sink of photoelectrons. The \textbf{photoelectron} lead represents a sea of excited photoelectrons with a lower occupied density of states and higher Fermi energy than the bulk, via a different Green's function, which also coupled to the \textbf{system}. The \textbf{vacuum} lead is represented by a further Green's function with a lower density of states and a lower Fermi energy to act as a sink of outgoing photoelectrons.
These are depicted in Figure \ref{fig:ModelExpFigPhotoelectron}.
\begin{figure}[h!]
\centering
\includegraphics[width=\linewidth]{ModelExperimentFig.pdf}
\caption{Schematic of the Green's-function models used to study photoelectron transport through 3-methylcyclohexanone (3-MCHO) adsorbed on Cu(111). Top: `equilibrium' setup comprising a semi-infinite bulk Cu lead coupled to the explicit Cu(111)$+$3-MCHO system. Bottom: `photoelectron' setup, where additional photoelectron and vacuum leads are introduced to model excited-electron injection and photoelectron emission, respectively. Shaded regions indicate occupied states below the corresponding Fermi levels.}
\label{fig:ModelExpFigPhotoelectron}
\end{figure}
\section{Modelling Non-Equilibrium Transport}
%\label{sec:SpinSymmetryBreaking}
The time evolution of a non-relativistic quantum system is given by the time-dependent Schr\"{o}dinger equation.
\begin{equation}
\hat{H} \psi = \iu \partial_t \psi
\end{equation}
To model the \textbf{system} we will work in second quantisation in a discrete basis and use the Born--Oppenheimer approximation. Creation/annihilation operators on site $p$ are denoted $\hat a_p^\dagger/\hat a_p$ respectively. $\hat{\mathcal{H}}$ denotes the electronic Hamiltonian and is:
\begin{equation}
\hat{\mathcal{H}} = h_{pq}\hat a^\dagger_p \hat a_q + h_{pqrs} \hat a^\dagger_p \hat a^\dagger_q \hat a_r \hat a_s
\end{equation}
$h_{pq}$ are the one-electron interactions such as momentum and nuclear attraction. $h_{pqrs}$ are the two electron interactions --- the Coulomb interaction of electrons with each other. \\
To describe time-dependent non-equilibrium phenomena, it is helpful to use the Green's function, a formally exact encoding of the one particle properties of the system described by this Hamiltonian. Within many-body perturbation theory, the Hartree--Fock (HF) approximation of the Green's function leads to the effective one-particle Hamiltonian:
\begin{equation}
\hat{H} = h_{pq}\hat a^\dagger_p \hat a_q + \frac{1}{2\pi}(h_{pqrs}-h_{qprs}) G^n_{rp} \hat a^\dagger_q \hat a_s\label{eq:HFApproximation}
\end{equation}
where $G^n$ is the (possibly non-)equilibrium density matrix (equation \ref{eq:GnDefinition}) and the two-electron Coulomb integrals are given by:
\begin{equation}
h_{pqrs} = \braket{q(\bfr_1)p(\bfr_2)|\frac1{\lvert \bfr_1-\bfr_2\rvert} |r(\bfr_2)s(\bfr_1)}.
\end{equation}
Within electronic structure theory, the effective one particle Hamiltonian $\hat H$ is called the Fock Hamiltonian.
The framework of Green's functions has the additional advantage that it can also describe an infinite sea of electrons, allowing easy modelling of the leads in our experiment.
The details of non-equilibrium Green's functions(NEGF) are found in standard texts\cite{datta_electronic_1995,rammer_quantum_2007}, and further details are given in the supplementary information (see SI-\ref{SI-SI:NonOrthGreensFuncSec}). The system Hamiltonian, $\hat H$ defines the Dyson equations below, which are solved to produce Green's functions ($G$) and self-energies ($\Sigma$) describing the system. The properties of the leads are specified by their self energies ($\Sigma$) which are given later.
For completeness, the relevant Dyson equations are:
\begin{align}
(EI + \iu \eta - &\hat H - \Sigma^R ) G^R(E) = I \t{, } \eta > 0\\
(EI - \iu \eta - &\hat H -\Sigma^A) G^A(E) = I\t{, } \eta > 0\\
G^K &= G^R(E) \Sigma^K G^A(E)\\
G^n &= \int_{-\infty}^{\infty} -\iu G^<(E) \td E \label{eq:GnDefinition}\\
G^<(E) &= G^K(E) - G^A(E) - G^R(E)
\end{align}
These are solved to yield the spectral Green's functions ($G^{R/A}$), which describe the possible one particle states of the system, and the Keldysh $G^K$ which describe the occupied one particle states. The self energies $\Sigma^{R/A/K}$ are used in the electron transport calculation to describe reservoirs that can source/sink electrons. The molecular orbitals (MOs) of electronic structure theory correspond to the residue of the divergences of $G^{R/A}$ (i.e. the long lived single particle states). \\
\section{Model Parameters}
As discussed above, for a transport calculation reservoirs (leads) from which electrons can enter/leave the system of interest must be defined. Each reservoir is modelled as an infinite band with a uniform density of states -- often termed the wide-band limit, (see SI-\ref{SI-SI:SelfEnergyTerms}) as it approximates the situation when the range of energy levels of interest are much closer together than the band they are connected to.
To this effect each lead self energy is given by:
\begin{align}
\Sigma^R_{\t{lead } i} &= -\iu \frac{\gamma_i}{2} C_i\\
\Sigma^A_{\t{lead } i} &= \iu \frac{\gamma_i}{2} C_i\\
\Sigma^K_{\t{lead } i} &= -\iu \left(1-2f_i(E)\right)\frac{\gamma_i}{2}C_i
\end{align}
$C_i$ is the coupling matrix that dictates the coupling to each explicit basis function (denoted $i$) in the system. For the self energy representing the \textbf{bulk} copper, $C$ is chosen such that the it couples only to the copper atoms. The \textbf{vacuum} consists of outgoing photoelectrons plane wave states that couple only to the basis functions on the adsorbed molecule. This is discussed in more detail in SI-\ref{SI-SI:SelfEnergyTerms} and shown schematically in figure \ref{fig:ModelExpFigPhotoelectron}.
$f_i$ is the Fermi function with chemical potential $\mu_i$. $\gamma_i$ is a parameter for each lead that encompasses both the density of states and the strength of the coupling of the system to the lead. An approximation for $\gamma$ is
\begin{equation}
\gamma = 2\pi t^2 \left(\t{D.O.S}\right)
\end{equation}
Where $t$ is the interaction energy between the two systems, $\t{D.O.S}$ is the density of states of the copper surface.
The $\gamma$ parameters are chosen to be physically meaningful. The value for the bulk copper lead is:
\begin{equation}
\gamma_{\t{bulk}} = \gamma_{\t{vacuum}} = 10^{-2} \t{E}_{\t{h}}
\end{equation}
To represent the reduced occupation number of the photoelectrons the $\gamma$ term there is reduced:
\begin{equation}
\gamma_{\t{photoelectron}} = 10^{-4}\t{E}_{\t{h}}.
\end{equation}
The results are qualitatively the same (section SI-\ref{SI-SI:sec:DifferentParameters}) independent of the exact value chosen.\\
The chemical potential of the bulk copper was set at $\mu_{\t{bulk}} = -0.15 \t{E}_{\t{h}} = -4.08 \t{eV}$ in order to create a neutral system. The experimental\cite{anderson_work_1949} work function for copper is $-4.46\t{eV}$. This discrepancy of $0.38\t{eV}$ is explained by the approximations made. \\
\section{Results}
In the wavefunction found in ground state electronic structure theory, if the MOs of $\alpha/\beta$ electrons are different, then $\alpha/\beta$ electrons will have different transport properties. This is exactly the CISS effect: a difference in transport properties of $\alpha/\beta$ electrons.
In this non-equilibrium system, the Green's function takes the place of the wavefunction describing the \textbf{system}. Here we investigate the effect that a small spin perturbation in the bulk copper has on the Hartree--Fock Green's function of the system and the resulting photoelectron spin current.
We are interested in measuring net spin polarisation. For quantity $Q$, the spin polarisation is given by
\begin{equation}
\t{Spin Polarisation} = \frac{Q_{\uparrow} - Q_{\downarrow}}{Q_{\uparrow} + Q_{\downarrow}} \label{eq:SpinPolEq}
\end{equation}
where $Q_{\uparrow}$ is the quantity of interest in the `up' spin direction and $Q_{\downarrow}$ in the `down' spin direction.
In order induce a spin polarisation in the copper surface as seen in the experiment \cite{badala_viswanatha_vectorial_2022}, $\gamma_{\t{bulk}}$ and $\gamma_{\t{photoelectron}}$ were made spin dependent by increasing the $\gamma$ for $\alpha$ electrons by a factor $1.04$ and decreasing $\gamma$ for $\beta$ electrons by a factor $0.96$\footnote{This choice is arbitrary, reversing the $\alpha$ and $\beta$ electrons would yield related results as seen in CISS experiments where the $L$ enantiomer is exchanged for $R$, SI-\ref{SI-sec:SymmetryConsiderations} }. This method has been used elsewhere to represent a magnetised lead\cite{fransson_charge_2021,dalum_theory_2019}. We do not agree with this interpretation however, as a magnetised lead would have a significant exchange effect on the system which this simple method neglects. This perturbation has been introduced to model the spin symmetry breaking as would be caused by spin-orbit coupling or chiral phonons from the molecule--surface system. \\
No such perturbation was applied to the outgoing photoelectron lead, $\gamma_{\t{vacuum}}$. It is stressed that modifying $\gamma_{\t{photoelectron}}$ in this way does not result in a $4\%$ polarisation of the electron density. Indeed the net spin polarisation of the density (equation \ref{eq:SpinPolEq}) in the copper in `equilibrium' is only $0.0014\%$. This is because the change in $\gamma$ only effects those states that are within $10 k_{\t{b}} T \approx 10\t{mE}_{\t{h}}$ of the Fermi level. All lower states are fully occupied and a change in broadening has no effect.\\
To fit with experiment the chemical potential of the incoming photoelectrons was set at $-0.15\Eh + 5.9\t{eV}$. This corresponds to a $210\t{nm}$ UV light source. \\
\begin{figure}[h!]
\centering
\includegraphics[clip, trim=1.5cm 1.5cm 1.5cm 1.5cm, width=0.3\linewidth]{Plots/Molecules/CU111_Opt_VMD.png}
\includegraphics[clip, trim=3.5cm 3.5cm 3.5cm 3.5cm,width=0.3\linewidth]{Plots/NonEqSpinPI.PNG}
\includegraphics[clip, trim=3.5cm 3.5cm 3.5cm 3.5cm,width=0.3\linewidth]{Plots/NonEqSpinPZ.PNG}
\caption{\textbf{The photelectron (spin) density of the system.} Left: The system of interest, 3-R-MCHO adsorbed into Cu(111). Middle: Photoelectron density (equation \ref{eq:nonEqSpinDensityComponent}). Right: Photoelectron spin density resolved along the direction normal to the plane. Cyan/Orange is spin parallel/antiparallel to the surface normal respectively. The surface parallel components are zero, Figure SI-\ref{SI-SI:fig:SpinDensityDiffXYZ}.}
\label{fig:NonEqDensityPIAndPZ}
\end{figure}
\begin{figure}[h!]
\centering
\includegraphics[width=0.55\linewidth]{Plots/CHO_CuSmall_PE0.04_Current_Z_PE.png}
\includegraphics[width=0.42\linewidth]{Plots/CHO_CuSmall_PE0.04_Current_Z_PESPINPOL.png}
\caption{\textbf{The magnitude and polarisation of the resultant photoelectrons.} Left: Incoming photoelectron and outgoing (scattering) current. The variation in the current is due to the location of the virtual orbitals of the molecule. Most of the current does not make it across the interface. This is due to the limited delocalisation across the interface, as is typical for physisorbed molecules. The bulk copper was biased by $4\%$ as described in the main text. Right: The spin polarisation of the resultant current. The bias creates an inherent $4\%$ polarisation in the incoming electrons. The outgoing electrons do not have this and all variation is due to the spin-symmetry broken molecular orbitals. Spin polarisation is as defined by equation \ref{eq:SpinPolEq}. In both cases the spin is resolved along the surface-normal direction.}
\label{fig:SpinPolAndCurrent0.04}
\end{figure}
Figure \ref{fig:NonEqDensityPIAndPZ} shows the distribution of the non-equilibrium density and spin density. The non-equilibrium component of the density matrix is defined by:
\begin{equation}
G^n_{\t{Non-equilibrium Comp.}} = \int_{\mu_{\t{bulk}}}^{\infty} -\iu G^<(E) \td E
\label{eq:nonEqSpinDensityComponent}
\end{equation}
The spin density has a complicated texture. Far away form the chiral molecule the spin texture is approximately symmetric as the surface alone has a $\sigma_v$ mirror plane and so we would expect an approximately symmetric solution. Close to the molecule however, the spin texture is not symmetric due to the influence of the molecules chirality and it is only here that the transport can occur in our calculation. It can be seen that the surface mainly interacts with the CO $\pi$ bond. This suggests that large delocalised structures, such as Helicenes, would exhibit a larger spin textures. There have been multiple reports of CISS with Helicenes\cite{safari_deposition_2022,kettner_chirality-dependent_2018,safari_enantioselective_2024,rodriguez_weakly_2023}\\
%Indeed, investigations into benzene reveal that the Hartree--Fock equations have symmetry broken solutions. Curiously these solutions vanish in DFT calculations\cite{pillai_effect_2025}. This is not to say that benzene (a delocalised molecule) alone should display a CISS effect as it has a mirror plane. However the existence of these solutions shows the possibility of a CISS effect in delocalised molecules if the required symmetries are broken. See section \ref{sec:SymmetryConsiderations}. Due to the buildup of $\alpha$ spin on the copper surface, a diamagnetic response of the molecule can be seen. This response is not due to an external effect, it is the true response of the molecule interacting with a slightly polarised surface.
Figure \ref{fig:SpinPolAndCurrent0.04} shows the incoming photoelectron and outgoing (scattering) photoelectron current (section SI-\ref{SI-SI:sec:Current}). Most of the electrons do not make it across the interface and are instead scattered back into the copper. This is due to the limited delocalisation of transport channels across the interface. There is an oscillation in both the incoming and scattered current. These oscillations correspond to the changing of the electron transport channels between different MOs. The oscillations are coincident with the fluctuations in the spin polarisation. This is due to each transport channel having its own spin texture. As the electron transport channels change so does the spin of the transport, this leads to oscillations in transmitted spin. This is qualitatively what is seen in the experiment of reference \citenum{badala_viswanatha_vectorial_2022}. It is not expected that the theory here reproduce the experiment perfectly, it is too simple and neglects important features such as the electronic structure of the bulk copper and the interactions with the scattering plane wave states. Nevertheless a surprising agreement in both the magnitude (Expt. $\sim 4\%$) and shape of the photoelectron spectrum is found.\\
It is stressed that the average $4\%$ polarisation of the incoming photoelectrons (blue line in figure \ref{fig:SpinPolAndCurrent0.04}) is due to the way $\gamma$ was chosen. The CISS signature found is the \textit{oscillations} in the transmitted spin polarisation spectrum, along with the spin texture in figure \ref{fig:NonEqDensityPIAndPZ}. \\
We therefore conclude that the exchange interaction caused by a small ($0.0014\%$) amount of spin buildup greatly influences the resultant photoelectron spectrum leading to a large ($\sim 2\%$) polarisation in transmitted electrons.
\section{Links to Spin Symmetry breaking in Hartree--Fock}
%\label{sec:SSBreakInHF}
First we must define symmetry breaking. The symmetries we are interested in are those of the Hamiltonian, resp. the experimental setup. If a result of our calculation, say the electron density, does not respect the symmetry of the Hamiltonian it is a symmetry broken solution. At the same time, one can talk about breaking a symmetry of the system. In this case we remove a symmetry of the Hamiltonian by changing the experimental setup. In the above calculation our solution respected the symmetry of the Hamiltonian (SI-\ref{SI-sec:SymmetryConsiderations}).\\
Kohn--Sham (KS) Density Functional Theory (DFT) uses the same single determinant ansatz for its non-interacting wavefunction to generate the density, but has a different effective electron-electron interaction kernel $(h_{pqrs})$.
In both SCF theories (HF and DFT) the spin structure of the ansatz may be constrained in three different ways:
\begin{enumerate}
\item\textbf{R(estricted)HF or R(estricted)KS}: the MOs of the $\alpha$ and $\beta$ spin electrons are the same resulting in an identical spatial distribution of $\alpha$ and $\beta$ electrons and so there is no spin texture to the density.
\item\textbf{U(nrestricted)HF or UKS}: the MOs may be different for $\alpha/\beta$ electrons but each MO is either $\alpha$ or $\beta$.
\item\textbf{G(eneralised)HF or non-co-linear DFT}: each MO is now free to have $\alpha$ and $\beta$ components and a varying spin axis. This is important for relativistic theory where $m_{\t{s}}$ is not a \textit{good} quantum number.
\end{enumerate}
The ansatz is usually chosen such that the solution `best' approximates the true solution of the Schr\"{o}dinger equation. Usually `best' is decided based on the symmetry of the problem. Similar spin textures to that found in figure \ref{fig:NonEqDensityPIAndPZ} are also found in single determinant SCF calculations where the ansatz is not constrained to preserve the spin symmetries\cite{pillai_effect_2025,burton_parity-time_2019}. This is not a coincidence as at zero temperature in equilibrium the single determinant HF method and the Green's function effective one particle Hamiltonian method are exactly equivalent. The result here best matches that of a UHF determinant. \\
The spin texture found when performing a calculation with a less constrained ansatz is a symmetry-broken solution. It does not respect the symmetry of the Hamiltonian. These less constrained ansatz are used as they can give a better approximation of the ground state energy. This conundrum has been termed L\"{o}wdin's \textit{symmetry dilemma}\cite{lykos_discussion_1963}. Considerable work has gone into understanding why it comes about. It is due to the strength of exchange in HF, that which elsewhere, in a spin symmetric Hamiltonian, creates a symmetry broken solution, creates here an amplified spin texture. Therefore electron-electron exchange is the interaction responsible for the magnitude of the CISS effect.\\
\section{Conclusion}
Until now most theories of CISS have relied predominantly on single non-interacting particle pictures. As is well known to chemists, electrons in molecules cannot be modelled as non-interacting particles and due to interactions complex behaviours can emerge. It is demonstrated here that CISS is such an emergent behaviour. While the origin of CISS likely involves relativistic effects that couple spin to electronic and nuclear motion (spin orbit coupling), the magnitude of the effect can be largely explained by exchange interactions. These exchange interactions can be understood via spin symmetry breaking in Hartree--Fock solutions. These symmetry broken solutions are only realistic if the experiment has already broken the relevant (point group and time-reversal) symmetries. The calculations here show that this can qualitatively reproduce experimental results with the correct magnitude and shape. This is a promising theory as it does not rely on unphysical parameters, nor on new physics. All instances of symmetry `breaking' are physically motivated, and the computational approach does not require expensive calculations. The mechanism can be understood using existing literature on spin symmetry breaking in Hartree--Fock theory, allowing a high-level conceptual understanding without complicated mathematics. \\
The current model is simplistic relying on a mean-field approximation and therefore neglecting electron correlations which could influence spin textures. It also relies on the wide-band limit and therefore neglects electronic effects in the bulk copper as well. It also neglects the explicit source of the small initial spin polarisation. Further work is to improve the model via explicit coupling\cite{ozaki_efficient_2010} to bulk copper instead of using the wide-band limit, as well as explicitly coupling to the free photoelectron plane wave states. These would lead to more accurate transport calculations. Another direction is to extend this to larger molecules such as Helicenes and DNA where CISS has been observed experimentally in the tens of percent range. SOC could also be included to remove the artificial spin bias. While explicit coupling to the bulk copper has been developed for DFT, spin symmetry breaking in DFT is less well understood. This is because spin itself is less well quantified in DFT due to Kohn-Sham orbitals not having the same interpretation as HF orbitals.
\section{Acknowledgements}
We would like to express our gratitude to Yuthika Pillai and Andreea Filip for helpful discussions during this work. This work was funded by the Leverhulme Centre for Life in the Universe.
\section{Author Contributions}
B.C. performed the simulations and developed the code. B.C. and A.J.W.T. analysed and discussed the results. A.J.W.T. supervised the project. All authors contributed to the writing of the manuscript and approved the final version.
\section{Code Availability}
The code used to run these calculations and generate figures is available at \url{https://github.com/apu727/CISSViaExchange}.
\bibliographystyle{unsrt}
\bibliography{CISS}
\end{document}