Universal Pole–Plus–Branch Decomposition

• Universal Pole–Plus–Branch Decomposition is a framework that defines the analytic structure of scattering amplitudes through coordinated zeros, poles, and branch cuts.
• The method enables deterministic phase engineering by exploiting symmetry breaking, resulting in a universal 2π phase winding when branch cuts cross the real axis.
• Its application in metasurface design allows tuning of dipole resonances and navigating exceptional points for full spectral phase control.

The universal pole–plus–branch decomposition describes the analytic structure of frequency-dependent scattering amplitudes in non-Hermitian photonic systems, particularly metasurfaces and related resonant electromagnetic interfaces. In these systems, the reflection or transmission coefficients—denoted generically as $S(\omega)$—possess singularities (poles and zeros) in the complex frequency plane, and their analytic continuation leads to the emergence of branch cuts linking these singularities. The topology of these branch cuts enforces fundamental phase effects, notably a universal $2\pi$ phase winding when a zero–pole pair is arranged so its branch cut crosses the real frequency axis. This framework is pivotal for understanding and engineering full $0\to2\pi$ phase modulation in photonic devices (Colom et al., 2022).

1. Analytic Structure and Decomposition

For a one-port (scalar) non-Hermitian photonic structure, the scattering amplitude $S(\omega)$ is meromorphic except for branch cuts that emanate from its branch points (zeros and poles), enabling the decomposition: $$S(\omega) = \sum_{n} \frac{R_n}{\omega - \omega_n} + \frac{1}{2\pi i}\int_{C} \frac{\Delta S(\omega')}{\omega - \omega'}\,d\omega'$$ Here, the residue $R_n$ is defined by $$R_n = \underset{\omega=\omega_n}{\mathrm{Res}\;} S(\omega)$$ at each simple pole $\omega_n$, and the contour $C$ follows each branch cut in the complex frequency plane, with the discontinuity $\Delta S(\omega') = S(\omega'+i0) - S(\omega'-i0)$. Each cut connects a zero–pole pair, reflecting the underlying logarithmic branch point at each zero in $S(\omega)$.

This representation generalizes directly to $N\times N$ multiport scattering matrices by considering the decomposition for each eigenchannel.

2. Zeros, Poles, and Symmetry Breaking

Zeros and poles are defined by $S(\omega_z)=0$ and $S(\omega_p)\to\infty$, respectively. In time-reversal ($\mathcal{T}$)-symmetric, lossless systems, the singularities appear in complex-conjugate pairs. For instance, $\omega_z$ and $\omega_z^*$ for zeros, and analogously for poles. No singularity is positioned off the real axis in isolation under these symmetries.

When $\mathcal{PT}$ (parity–time) symmetry is broken for reflection, or $\mathcal{T}$ is broken for transmission, the zeros (but not poles in passive systems) can be displaced into the upper or lower half of the complex frequency plane. This shift is crucial for engineering phase behavior and is associated either with explicit symmetry breaking (e.g., asymmetric substrates) or spontaneous symmetry breaking at exceptional points (EPs).

3. Branch Cuts and Real-Axis Intersections

Branch cuts are constructed to join each simple zero–pole pair $(\omega_z, \omega_p)$, ensuring the cut does not intersect other singularities. When $\mathrm{Im}\,\omega_z>0>\mathrm{Im}\,\omega_p$, the branch cut must cross the real axis at a frequency $\omega_c \in \mathbb{R}$.

The table below summarizes the conditions for branch cut topology:

Symmetry Condition	Zero–Pole Locations	Branch Cut Behavior
$\mathcal{T}$-symmetric, lossless	Paired on real axis or as conjugates	No real-axis crossing
$\mathcal{PT}$/$\mathcal{T}$-broken	Zero straddles real axis	Cut crosses real axis

Branch cut crossings have direct photonic consequences, as detailed below.

4. Universal $2\pi$ Phase Winding and Topology

Defining the phase $\Arg S(\omega) = \Im \ln S(\omega)$, a core result is that as the real frequency $\omega$ traverses the real-axis intersection $\omega_c$ of such a branch cut, the phase accumulates exactly $2\pi$. This arises from the argument principle: integrating the logarithmic derivative of $S(\omega)$ around a contour enclosing a single zero (and no poles) leads to a $2\pi i$ increment, such that: $\Arg S(\omega_2) - \Arg S(\omega_1) = 2\pi,$ where $\omega_1<\omega_c<\omega_2$. This universal $2\pi$ ramp is Dirac-like and topological; each zero–pole pair with a real-axis-crossing cut contributes one such phase winding. This effect underpins the control of phase in non-Hermitian photonic devices (Colom et al., 2022).

5. Practical Criteria for Branch Cut Crossing

A sufficient condition for a branch cut to intersect the real axis is that a zero–pole pair straddles it. Two main routes are:

• Explicit $\mathcal{PT}$-breaking for reflection: Placing the metasurface on an asymmetric substrate/superstrate shifts the reflection zero into $\Im\omega>0$, with all poles in $\Im\omega<0$.
• Spontaneous $\mathcal{T}$-breaking at an exceptional point (EP): In Huygens metasurfaces, two zero–pole pairs coalesce at an EP on the real axis; beyond this, zeros bifurcate off-axis and the corresponding branch cut for one pair crosses the real line, producing the $2\pi$ phase sweep.

These scenarios enable deterministic engineering of phase behavior essential for wavefront modulation.

6. Applications in Metasurface Design

The pole–plus–branch decomposition provides a rigorous basis for metasurface phase control:

• PT-symmetry-broken reflectionless metasurfaces: A single reflection zero is rendered in $\Im\omega>0$, producing full $0\to2\pi$ phase modulation in reflection (see Fig. 2b in (Colom et al., 2022)).
• All-dielectric Huygens metasurfaces: Tuning system parameters enables the approach to, and traversal of, an EP in the transmission zeros, yielding $2\pi$ phase coverage across transmission (see Fig. 3).
• GSTC-dipole-susceptibility model: For reflection and transmission described by

$$R(\omega)=\frac{2ik(\chi_e-\chi_m)}{(2-ik\chi_e)(2-ik\chi_m)},\quad T(\omega)=\frac{4 + k^2\chi_e\chi_m}{(2-ik\chi_e)(2-ik\chi_m)},$$

arranging $\chi_e$, $\chi_m$ (Lorentz poles) so that the Kerker condition $\chi_e=\chi_m$ is met yields a cut crossing the real axis whenever the zero and its associated pole straddle it. This insight enables the tuning of dipole resonances for full $0\to2\pi$ spectral phase coverage.

7. Broader Implications and Topological Significance

The universal pole–plus–branch decomposition clarifies the topological origin of phase effects in non-Hermitian photonic systems. Each real-axis-crossing zero–pole cut contributes exactly one $2\pi$ phase winding, facilitating arbitrary wavefront control. By engineering the position and symmetry environment of zeros—either through explicit symmetry breaking or by navigating parameter space through exceptional points—one obtains deterministic access to phase spans essential for advanced photonic functionality. This framework has broader relevance for resonant photonic structures where non-Hermitian topological features dictate their electromagnetic response (Colom et al., 2022).