---
title: Nonlinear Geometric Phase
url: https://www.emergentmind.com/topics/nonlinear-geometric-phase
type: topic
---

# Nonlinear Geometric Phase

Nonlinear geometric phase is a phase acquired by a nonlinear state, nonlinear polarization, frequency-conversion amplitude, or nonlinear oscillator during evolution through a parameterized state space or through an orientation-dependent nonlinear interaction. It differs from a conventional dynamical phase, which is associated with energy, propagation, dispersion, or optical path length, because it depends on geometric transport, spin–rotation coupling, nonlinear mode mixing, parameter-space holonomy, or the orientation of a nonlinear susceptibility. The term encompasses several related but nonidentical constructions: non-unitary and non-cyclic phases of nonlinear quantum states, multiphoton Pancharatnam phases, nonlinear Berry phases in harmonic generation, geometric phases in depleted three-wave mixing, geometric-phase-induced localization, and orientation-controlled phases of nonlinear metasurfaces and thin-film crystals.

## 1. Conceptual foundations and phase decomposition

For a state vector that may be unnormalized, the normalized state is $|\psi(t)\rangle/\|\psi(t)\|$. The Pancharatnam total phase between initial and final normalized states is

$$
\chi=\arg\left\langle
\frac{\psi(0)}{\|\psi(0)\|}
\middle|
\frac{\psi(t)}{\|\psi(t)\|}
\right\rangle .
$$

The dynamical phase is

$$
\delta=-i\int_0^t
\left\langle
\frac{\psi(\tau)}{\|\psi(\tau)\|}
\middle|
\frac{d}{d\tau}
\left|
\frac{\psi(\tau)}{\|\psi(\tau)\|}
\right\rangle
d\tau ,
$$

and the geometric phase is $\gamma=\chi-\delta$. In Pati’s reference-section formulation,

$$
|\chi_0(t)\rangle=
\frac{\langle\psi(t)|\psi(0)\rangle}
{|\langle\psi(t)|\psi(0)\rangle|}
\frac{|\psi(t)\rangle}{\|\psi(t)\|},
$$

so that

$$
\gamma=i\int_0^t
\left\langle\chi_0(\tau)\middle|
\frac{d}{d\tau}
\middle|\chi_0(\tau)\right\rangle d\tau
\quad \mathrm{mod}\ 2\pi .
$$

The reference-section prefactor imposes the Pancharatnam condition that the transported state remains in phase with the initial state. The resulting phase is gauge invariant and reparametrization invariant. For an open path closed by a geodesic, the geodesic contribution vanishes, and Stokes’ theorem converts the remaining line integral into a surface integral of curvature. The phase consequently depends on the path through projective Hilbert space rather than only on its endpoints [1011.0501].

In nonlinear systems, the separation between total, dynamical, and geometric phases is usually more involved than in a linear two-level system. Nonlinear coupling changes the state trajectory itself, while depletion, mode weighting, dissipation, or nonlinear susceptibility can alter both the endpoint phase and the connection term. A geometric phase may therefore be nonlinear in time, intensity, parameter displacement, harmonic order, or polarization angle.

The term is not uniform across all applications. In nonlinear metasurfaces, it often denotes an orientation-controlled phase of a generated harmonic field. In fully nonlinear three-wave mixing, it denotes a phase obtained after subtracting a dynamical contribution from the phase of a nonlinear eigenstate. In dissipative oscillators, it is a holonomy of the slowly varying limit-cycle phase. In multiband systems, “nonlinear geometric phase” is used for gauge-covariant geometric factors entering higher-order response tensors rather than for a single universally defined scalar phase [2502.14128].

## 2. Nonlinear quantum states and oscillator dynamics

### Nonlinear coherent and squeezed states

A standard deformation introduces a real function $f(N)$ into the oscillator operators:

$$
A=af(N),\qquad A^\dagger=f(N)a^\dagger ,
$$

with

$$
[A,A^\dagger]=(N+1)f^2(N+1)-Nf^2(N).
$$

The inverse-deformed operators are

$$
B=a\frac{1}{f(N)},\qquad B^\dagger=\frac{1}{f(N)}a^\dagger,
$$

and satisfy $[A,B^\dagger]=[B,A^\dagger]=1$. Two nonlinear coherent states arise from different non-unitary displacement-like operators,

$$
D_1(\beta)=\exp(\beta B^\dagger-\beta^*A),
\qquad
D(\beta)=\exp(\beta A^\dagger-\beta^*B).
$$

Their number-state expansions differ by the placement of the nonlinear factor:

$$
|\beta_1\rangle =
e^{-|\beta|^2/2}
\sum_{n=0}^{\infty}
\beta^n\frac{1}{f(n)!\sqrt{n!}}|n\rangle ,
$$

and

$$
|\beta\rangle =
e^{-|\beta|^2/2}
\sum_{n=0}^{\infty}
\beta^n\frac{f(n)!}{\sqrt{n!}}|n\rangle .
$$

Under free harmonic evolution, each number-state component acquires $e^{-iE_nt}$, with $E_n=\omega(n+1/2)$. The nonlinear weighting changes both the endpoint interference phase and the effective energy contribution to the connection. For the first coherent state, the cyclic phase at $T=2\pi/\omega$ contains inverse factors such as $1/[f(n+1)]^2$, whereas the second contains factors $[f(n+1)]^2$. Their geometric phases therefore provide a proposed criterion for distinguishing the two constructions. Both reduce to the ordinary coherent state when $f(n)=1$ [1011.0501].

The physically relevant trapped-ion example uses

$$
f(n)=\frac{L_n^1(\eta^2)}
{(n+1)L_n^0(\eta^2)},
$$

where $L_n^m$ is a generalized Laguerre polynomial and $\eta$ is the Lamb–Dicke parameter. For nonzero $\eta$, the phase evolution becomes nonlinear in time; the nonlinear function can accelerate or slow phase accumulation. The phase is consequently tunable through the trapped-ion coupling parameters.

The nonlinear squeezed-state constructions generated by

$$
S_1(\zeta)=
\exp\left[\frac12(\zeta B^{\dagger 2}-\zeta^*A^2)\right],
$$

and

$$
S(\zeta)=
\exp\left[\frac12(\zeta A^{\dagger 2}-\zeta^*B^2)\right]
$$

coincide. Their expansion contains only even number states. For $f(n)=1$, the standard squeezed-vacuum result is recovered,

$$
\gamma_{\rm sq}^{\rm standard}(t)
=
\chi+\omega t\left(\frac12+\sinh^2r\right)
\quad \mathrm{mod}\ 2\pi .
$$

At $T=2\pi/\omega$, the standard cyclic phase is $2\pi\sinh^2r$ modulo $2\pi$. In the nonlinear case, the large-squeezing limit $r\rightarrow\infty$ can approach a plateau whose value is modified by the nonlinear function. This plateau is therefore not universal.

### Nonstatic coherent states

A distinct nonlinear mechanism occurs when the medium is static but the quantum wave itself periodically contracts and expands. A time-dependent function

$$
f(t)=c_1\sin^2\widetilde{\varphi}(t)
+c_2\cos^2\widetilde{\varphi}(t)
+c_3\sin[2\widetilde{\varphi}(t)]
$$

controls the width, phase curvature, and quadrature-space shape of a generalized coherent state, subject to

$$
c_1c_2-c_3^2=1,\qquad c_1c_2\geq1.
$$

The static limit is $c_1=c_2=1$, $c_3=0$, for which $f(t)=1$. The time-dependent phase variable is

$$
T(t)=\int_{t_0}^t\frac{dt'}{f(t')}.
$$

The geometric phase can be written as

$$
\gamma_G(t)
=
-\frac{\omega}{2}T(t)
-\Gamma_D(\bar t)(t-t_0)
+\gamma_G(t_0),
$$

while the total phase is

$$
\gamma(t)=-\frac{\omega}{2}T(t)+\gamma(t_0).
$$

The nonlinear component is controlled by

$$
T(t)-(t-t_0)
=
\int_{t_0}^{t}
\left[\frac{1}{f(t')}-1\right]dt'.
$$

It vanishes in the static limit. Periodic wave collapse and expansion produce nonlinear phase oscillations superposed on a linearly shifting center; under extreme nonstaticity, the total phase can exhibit periodic downward changes at wave nodes [2401.12560].

### Nonlinear limit-cycle oscillators

For a weakly nonlinear oscillator,

$$
\ddot y+\omega^2 y=\epsilon(t)f(y,\dot y),
$$

perturbation theory generates secular terms. A renormalization-group construction absorbs these terms into a slowly varying complex amplitude $A$. Writing

$$
A=\frac{r}{2}e^{i\theta},
$$

gives amplitude and phase equations,

$$
\dot r=f(r,\boldsymbol\epsilon,\dot{\boldsymbol\epsilon}),
\qquad
\dot\theta=\Omega(r,\boldsymbol\epsilon,\dot{\boldsymbol\epsilon}).
$$

For a stable limit cycle $R(\boldsymbol\epsilon)$, slow parameter variation displaces the oscillator from the instantaneous cycle by an amount linear in parameter velocity. The phase then decomposes into a dynamical term and a geometric connection,

$$
\theta_{\rm geom}=\oint_\gamma\vartheta,
\qquad
\vartheta=a_1(\boldsymbol\epsilon)d\epsilon_1+
a_2(\boldsymbol\epsilon)d\epsilon_2.
$$

For a closed cycle, the phase equals the curvature flux,

$$
\theta_{\rm geom}
=
\int_G\chi\,d\epsilon_1\wedge d\epsilon_2,
\qquad
\chi=
\frac{\partial a_2}{\partial\epsilon_1}
-\frac{\partial a_1}{\partial\epsilon_2}.
$$

For the Van der Pol oscillator, the stable limit cycle is $R=2$, and the curvature in the $(\mu,\omega)$ parameter plane is

$$
\chi_{\rm VdP}=\frac{1}{8\omega^2}.
$$

For the Van der Pol–Duffing oscillator, the dominant curvature in the regime $\mu^2\ll\beta\ll\omega^2$ is

$$
\chi_{\rm VdPD}
=
\frac{3}{2}\frac{\beta}{\mu^2\omega^3}.
$$

The phase is path dependent but, in the adiabatic limit, independent of the detailed traversal rate of a fixed path. The construction presumes a smoothly varying attracting limit cycle and becomes unreliable near bifurcations or when radial relaxation becomes arbitrarily slow [2309.16587].

### Dissipative quantum synchronization

A spin-1 quantum van der Pol oscillator provides a nonunitary setting. Its density matrix evolves under a Lindblad equation with nonlinear gain and damping. Slow rotation of the quantization axis causes the density matrix to follow an instantaneous rotated steady state. The geometric phase is evaluated with Tong’s kinematic mixed-state expression,

$$
\gamma[\mathcal P]
=
\arg\left[
\sum_k
\sqrt{p_k(0)p_k(\tau)}
\langle\phi_k(0)|\phi_k(\tau)\rangle
\exp\left(
-\int_0^\tau
\langle\phi_k(t)|\dot\phi_k(t)\rangle dt
\right)
\right].
$$

An external signal generates coherences $c_{+1,0}$ and $c_{0,-1}$. Synchronization depends on their sum,

$$
\mathcal S(\hat\rho_{\rm ss})
=
\frac{3}{8\sqrt2}\,
T|c_{+1,0}+c_{0,-1}|.
$$

The geometric phase depends separately on the imaginary parts of the coherences weighted by population differences. It therefore develops an Arnold-tongue-like structure in signal strength and detuning, but the geometric-phase tongue is not identical to the synchronization tongue. In synchronization blockade, destructive interference can eliminate the synchronization measure even when individual coherences remain nonzero; numerical results nevertheless show suppression of the geometric-phase structure because the coherences become globally small [2302.08866].

## 3. Multiphoton and higher-dimensional state-space phases

For three normalized states, the three-vertex geometric phase is the argument of the Bargmann invariant,

$$
\gamma(\psi_1,\psi_2,\psi_3)
=
\arg[
\langle\psi_1|\psi_3\rangle
\langle\psi_3|\psi_2\rangle
\langle\psi_2|\psi_1\rangle].
$$

Because each state occurs once as a bra and once as a ket, arbitrary state rephasings cancel. For polarization qubits, the phase equals half the oriented solid angle of the geodesic triangle on the Poincaré sphere.

For $N$ identically polarized photons, the phase is multiplied by $N$:

$$
\gamma_N=N\gamma_1.
$$

In a two-photon Mach–Zehnder interferometer with post-selection, the gauge-invariant phase is

$$
\gamma(\psi_A,\psi_B,\psi_2)
=
\arg[
\langle\psi_A|\psi_B\rangle
\langle\psi_B|\psi_2\rangle
\langle\psi_2|\psi_A\rangle].
$$

Post-selection can restore interference visibility even when the two arm states are orthogonal by erasing which-path information. For the polarization arrangement studied in [1105.1855], the phase varies as

$$
\phi_f=
2\tan^{-1}
\left(\frac{\tan\theta_1}{\tan\theta_2}\right)
+\frac{\pi}{2}.
$$

Its slope near $\theta_1=0$ is

$$
\left.
\frac{\partial\phi_f}{\partial\theta_1}
\right|_{\theta_1=0}
=
\frac{2}{\tan\theta_2}.
$$

For $N$ photons the slope becomes $2N/\tan\theta_2$. The increased phase response is accompanied by a reduced post-selection probability, so the enhancement is conditional. In the shot-noise regime, the lower successful-event rate offsets the larger phase slope; in a technical-noise-dominated regime, the phase response can improve the signal-to-noise ratio [1105.1855].

A genuine two-photon polarization qutrit occupies the symmetric subspace of two qubits, which is three dimensional. A standard triplet can be written as

$$
|\varPsi_1\rangle=|\psi_1\rangle|\psi_1\rangle,
\qquad
|\varPsi_2\rangle=|\psi_2\rangle|\psi_2\rangle,
$$

and

$$
|\varPsi_3\rangle=
|\psi_3\rangle|\psi'_3\rangle+
|\psi'_3\rangle|\psi_3\rangle.
$$

Using the Majorana representation, the qutrit phase separates into two qubit-like phases,

$$
\gamma(\varPsi_1,\varPsi_2,\varPsi_3)
=
\gamma(\psi_1,\psi_2,\psi_3)
+
\gamma(\psi_1,\psi_2,\psi'_3).
$$

The two spherical-triangle contributions can undergo separate or coincident rapid $2\pi$ increases as the third state is varied. For $\chi=0^\circ$, the increases coincide and produce a total unwrapped increase of $4\pi$; for $\chi=180^\circ$, they occur separately at $90^\circ$ and $270^\circ$. The phenomenon is characteristic of the higher-dimensional qutrit structure and cannot be represented by a single qubit Bloch-sphere triangle [1504.01838].

## 4. Nonlinear frequency conversion and geometric evolution

### Two-mode frequency conversion

In sum-frequency generation with an undepleted pump, the signal and generated-frequency amplitudes obey

$$
i\frac{d}{dz}a_2(z)
=
\kappa^*e^{-i\Delta\beta z}a_3(z),
$$

$$
i\frac{d}{dz}a_3(z)
=
\kappa e^{i\Delta\beta z}a_2(z).
$$

After a rotating-frame transformation, the amplitudes form an effective two-level state with Hamiltonian

$$
\hat H=
\begin{pmatrix}
-\Delta\beta/2 & \kappa^*\\
\kappa & \Delta\beta/2
\end{pmatrix}.
$$

The state traces a circle on a Poincaré sphere, and the geometric phase is half the enclosed solid angle,

$$
\gamma=-\frac{\Omega}{2},
$$

up to orientation convention. When $\Delta\beta=0$, the trajectory can reach a singular point where the geodesic closure changes branch and the phase jumps by $\pi$. For large mismatch, $|\Delta\beta|\gg|\kappa|$, the trajectory contracts toward its initial point and the geometric phase approaches zero.

The pump controls $|\kappa|$, and therefore the conversion rate, trajectory, and accumulated phase. Momentum-dependent phase matching can make the phase non-reciprocal: a forward wave can be phase matched and acquire approximately a $\pi$ geometric phase, while backward propagation is strongly mismatched and acquires essentially no geometric phase. This mechanism uses frequency conversion rather than a Kerr-biased transmission contrast and is therefore distinct from dynamic reciprocity limitations in Kerr isolators [1704.03701].

### Fully nonlinear three-wave mixing

When all three waves are depleted, the normalized amplitudes satisfy

$$
\frac{dq_1}{d\tau}
=
i\Delta\Gamma q_1-igq_2^*q_3,
$$

$$
\frac{dq_2}{d\tau}
=
i\Delta\Gamma q_2-igq_1^*q_3,
$$

$$
\frac{dq_3}{d\tau}
=
i\Delta\Gamma q_3-ig^*q_1q_2,
$$

where

$$
g(\tau)=\Xi(\tau)e^{i\phi_{\mathrm d}(\tau)}.
$$

The Manley–Rowe invariants are

$$
K_1=|q_1|^2+|q_3|^2,
\qquad
K_2=|q_1|^2-|q_2|^2,
\qquad
K_3=|q_2|^2+|q_3|^2.
$$

The nonlinear dynamics reduces to a closed surface in $(X,Y,Z)$ coordinates,

$$
X+iY=q_1q_2q_3^*,
\qquad
Z=|q_3|^2,
$$

with

$$
X^2+Y^2-Z(Z-K_1)(Z-K_3)=0.
$$

The quasi-phase-matching parameters form an effective control vector,

$$
\mathbf B=
\left(
\Xi\cos\phi_{\mathrm d},
\Xi\sin\phi_{\mathrm d},
\frac{\Delta\Gamma}{2}
\right).
$$

When this vector is slowly rotated around a closed path, an instantaneous nonlinear eigenstate can follow adiabatically and return to its initial physical state with an accumulated phase. The adiabatic geometric phase is

$$
\beta_j=\Phi_j-\mathfrak D_j,
$$

where $\Phi_j$ is the total phase and $\mathfrak D_j$ is the dynamical phase. It can also be written as the phase along a circular QPM trajectory minus the phase along its vertical projection. This circular-minus-projection relation applies in both the undepleted and depleted regimes, although the nonlinear state surface is not generally a sphere [2003.03075].

Full-wedge and half-wedge rotations provide two forms of QPM control. Full-wedge rotation introduces a discontinuous jump in $\phi_{\mathrm d}$ at a pole where $\Xi=0$ and produces a geometric phase that is essentially independent of depletion except at a fully depleted cusp. In the undepleted limit,

$$
\beta_1=-\Delta\phi_{\mathrm d}
\quad\text{for SFG},
\qquad
\beta_3=+\Delta\phi_{\mathrm d}
\quad\text{for DFG}.
$$

Half-wedge rotation varies all QPM parameters continuously. Its phase is approximately linear in the wedge angle but depends on initial intensity imbalance,

$$
\beta_1=\kappa_1\Delta\phi_{\mathrm d}
\quad\text{for SFG},
\qquad
\beta_3=\kappa_3\Delta\phi_{\mathrm d}
\quad\text{for DFG}.
$$

Thus full-wedge rotation suppresses depletion uncertainty, whereas half-wedge rotation converts depletion dependence into a calibratable coefficient [2104.13022].

### Adiabatic geometric phase matching

In multilayer nonlinear thin-film crystals, each layer contributes a nonlinear polarization with an orientation-dependent phase. For z-cut lithium niobate, the relevant cross-circular nonlinear polarization contains

$$
P^{\mathrm{NL}}_{-\sigma}
=
2\sqrt2\,d_{22}\,i\sigma
e^{i\sigma3\theta}
\left(E^{\mathrm{FF}}_\sigma\right)^2.
$$

The geometric phase is therefore $\varphi_{\mathrm{GP}}=\sigma3\theta$. For successive layers, the effective structural wave vector is

$$
k_{\mathrm{GP}}=\frac{\Delta\varphi_{\mathrm{GP}}}{l},
$$

and geometric phase matching occurs when

$$
k_{\mathrm{GP}}=\Delta k.
$$

For lithium niobate,

$$
\Delta\theta=\frac{\Delta k\,l}{3}.
$$

A bilayer provides continuous constructive-to-destructive modulation. An eight-layer structure can produce nearly quadratic coherent buildup, with the same physical layer stack reconfigured for different wavelengths by changing layer orientations. Opposite circular input spins produce opposite geometric-phase progressions, enabling spin-selective phase matching and polarization conversion [2602.18418].

## 5. Nonlinear geometric phases in optical materials and metasurfaces

### Spin–rotation coupling and harmonic generation

For a nanostructure rotated by $\theta$, a circularly polarized input field acquires a local-frame phase $e^{i\sigma\theta}$. In an $n$-th-order nonlinear process, this becomes $e^{in\sigma\theta}$. Transforming the emitted harmonic back to the laboratory frame yields

$$
\Phi_{\mathrm{geo}}^{\mathrm{same}}
=
(n-1)\sigma\theta,
\qquad
\Phi_{\mathrm{geo}}^{\mathrm{opposite}}
=
(n+1)\sigma\theta.
$$

For third-harmonic generation, the two phase laws are $2\sigma\theta$ and $4\sigma\theta$. A $C_4$ structure suppresses the same-spin channel and leaves a clean opposite-spin channel with phase $4\sigma\theta$. Spatially varying orientations can consequently produce phase gradients for beam steering, focusing, vortex generation, nonlinear holography, and continuous quasi-phase matching [1407.4012].

The phase is embedded in the nonlinear source term rather than added solely to a linearly scattered field. This distinguishes it from conventional linear geometric phase and from periodically poled quasi-phase matching, which primarily provides discrete phase shifts of $0$ or $\pi$.

### Dielectric nonlinear metasurfaces

All-dielectric silicon nanofins generate third-harmonic radiation through a volume nonlinear polarization,

$$
\mathbf P^{(3)}(3\omega,\mathbf r)
=
\varepsilon_0
\boldsymbol{\chi}^{(3)}_{\mathrm{eff}}
:
\mathbf E(\omega,\mathbf r)^3.
$$

The confined fundamental mode can contain multiple circular, longitudinal, forward, and backward components. Different tensor pathways therefore carry different rotation phases. For LCP excitation, representative pathways include

$$
P_L^{(3)}\sim
\varepsilon_0\chi_{LLLL}^{(3),\mathrm{eff}}
E_L^3e^{2i\theta},
$$

and

$$
P_R^{(3)}\sim
\varepsilon_0\chi_{RLLL}^{(3),\mathrm{eff}}
E_L^3e^{-4i\theta}.
$$

The observed phase is determined by the coherent sum of pathways, their amplitudes, the tensor coefficients, and coupling between neighboring nanofins.

For $C_1$, $C_2$, and $C_4$ structures, different channels are available. $C_2$ nanofins support principal $2\theta$ and $4\theta$ channels; $C_4$ symmetry suppresses the co-polarized channel and predominantly supports a cross-polarized $4\theta$ response. These phases generate nonlinear beam steering and holography. Reversing the fundamental helicity reverses the phase gradient and produces a conjugate holographic image [2003.08623].

Dense-array coupling can modify the effective principal-axis rotation,

$$
\theta_{\rm eff}=l\theta,
$$

where $l$ is determined by the linear polarization response of the entire array rather than simply by isolated-particle symmetry. In a C3 meta-atom array on a square lattice, $\theta_{\rm eff}=3\theta$ produces an observed co-polarized third-harmonic phase proportional to $6\theta$. C4 meta-atoms on a hexagonal lattice produce generalized phase factors such as $e^{\pm4i\theta}$ and $e^{-8i\theta}$. Thus structural symmetry and lattice symmetry jointly determine accessible nonlinear phase channels [2311.07186].

AlGaAs provides another mechanism. Its linear response can be nearly isotropic in the plane while its zincblende second-order susceptibility is anisotropic. The tensor rotation gives

$$
\chi^{(2)\prime}
=
\cos(2\beta)\chi_1^{(2)}
+
\sin(2\beta)\chi_2^{(2)}.
$$

For a selected SH channel,

$$
E_{\rm ch}^{2\omega}(\beta)
\propto
a\cos(2\beta)+b\sin(2\beta),
$$

so the tensor-induced phase is

$$
\varphi_\chi(\beta)
=
\arg[
a\cos(2\beta)+b\sin(2\beta)
].
$$

The complete phase is

$$
\varphi(\beta)
=
\varphi_\chi(\beta)
-m^{2\omega}\beta
+2m^\omega\beta.
$$

This phase need not be linear in the rotation angle because the tensor contribution is the argument of a complex superposition. The approach has been used for nonlinear beam steering and unit-charge vortex generation [2601.18246].

### Nonlinear localization

In reorientational liquid crystals, light changes the director angle $\theta(x,y,z)$, which changes the local Pancharatnam–Berry phase. In the circular basis, the anisotropic coupling contains $e^{\pm2i\theta}$, and the resulting effective propagation equation has a geometric-phase potential. For small director rotations,

$$
V(x,y)\approx-\frac{m\pi}{\Lambda}\Gamma(x,y),
$$

where $m=\pm1$ is beam helicity and $\Gamma(x,y)$ is the transverse director-rotation amplitude.

The feedback sequence is

$$
I
\longrightarrow
\theta(x,y,z)
\longrightarrow
\text{polarization evolution}
\longrightarrow
\Phi_{\mathrm{PB}}
\longrightarrow
\nabla_\perp\Phi_{\mathrm{PB}}
\longrightarrow
\text{wavefront self-modulation}
\longrightarrow
\text{localization}.
$$

The beam can therefore self-confine without writing a scalar refractive-index well. Opposite helicities experience opposite effective potentials and can repel when overlapping. The mechanism differs from Kerr self-focusing and from conventional nematic solitons based on an intensity-written dynamic-phase index channel [1811.01819].

### Pump shaping and harmonic structured light

A spatially patterned liquid-crystal half-wave plate applies opposite geometric phases to the two circular components,

$$
\hat{\mathbf e}_L\longrightarrow
e^{+i2\alpha(\mathbf r)}\hat{\mathbf e}_R,
\qquad
\hat{\mathbf e}_R\longrightarrow
e^{-i2\alpha(\mathbf r)}\hat{\mathbf e}_L.
$$

A Gaussian pump can thereby be converted into conjugate spatial modes before second-harmonic generation. If the two pump components have relative phase $\delta$, SHG doubles it:

$$
\delta_{2\omega}=2\delta_\omega.
$$

The nonlinear mapping of the polarization weights is

$$
b=\frac{a^2}{a^2+(1-a)^2}.
$$

The output can consist of conjugate Laguerre–Gaussian modes with opposite orbital angular momenta, producing cylindrically vectorial harmonic fields. The method has been demonstrated for $\mathrm{LG}_1^{\pm2}$ and $\mathrm{LG}_1^{\pm4}$ mode pairs and is designed to preserve propagation-invariant vectorial structure [2506.17167].

## 6. Applications, interpretation, and limitations

Nonlinear geometric phase supports several classes of functionality:

- **Phase-sensitive measurement**: multiphoton phases can amplify the response to small polarization changes, although post-selection losses prevent a universal improvement over shot-noise limits [1105.1855].
- **Frequency-conversion control**: pump amplitude, mismatch, QPM modulation, layer orientation, or metasurface rotation can control the geometric phase of converted fields [1704.03701, 2003.03075].
- **Nonlinear phase matching**: orientation-dependent source phases can replace or supplement domain inversion and birefringent phase matching, enabling reconfigurable wavelength and spin-selective conversion [2602.18418, 2605.11914].
- **Wavefront shaping**: local orientation profiles can generate beam steering, focusing, vortex beams, nonlinear holograms, and far-field diffraction patterns [1407.4012, 2003.08623].
- **Nonlinear localization and routing**: intensity-dependent Pancharatnam–Berry modulation can create self-written geometric-phase potentials and helicity-dependent beam interactions [1811.01819].
- **Structured harmonic generation**: geometric-phase pump shaping can generate higher-order cylindrically vectorial modes directly at harmonic frequencies [2506.17167].
- **Nonlinear multiband response**: Berry connections, Berry curvature, and quantum-geometric tensors can organize gauge-covariant nonlinear polarization, rectification, and Hall responses [2502.14128].
- **Mechanical and elastic holonomy**: in nonlinear elastic rotor systems, cyclic shape deformation produces net rigid rotation through the holonomy of a mechanical connection. The shape manifold for a double rotor can have hyperbolic geometry for positive angular momentum or a two-dimensional Robertson–Walker geometry for negative angular momentum, depending on the rotation-sign convention [2302.07441].

Several qualifications are essential. A nonlinear geometric phase is not always a canonical Berry holonomy. In nonlinear metasurface literature, the term can denote an orientation-controlled phase of a nonlinear polarization without a demonstrated fiber-bundle construction. In fully nonlinear frequency conversion, the state-space surface is generally not a sphere, so the phase cannot always be represented by a simple solid angle. In open quantum systems, the phase depends on the chosen mixed-state kinematic definition and may differ from a no-jump interferometric phase. In multiband nonlinear optics, “nonlinear geometric phase” may refer collectively to gauge-invariant products of geometric tensors rather than to a separately defined scalar phase.

Practical limitations include pump depletion, nonadiabaticity, cusp singularities, absorption, finite conversion efficiency, fabrication disorder, imperfect polarization purity, orientation-dependent amplitude modulation, resonant phase shifts, near-field coupling, finite angular resolution, mode mixing, post-selection loss, and unstable interferometric alignment. In nonlinear metasurfaces, the ideal phase law can be distorted when multiple tensor pathways coexist or when rotating a resonator shifts its resonance. In multilayer geometric phase matching, Fresnel reflections, substrate phases, layer misalignment, thickness variations, and Fabry–Pérot effects reduce external efficiency. In nonlinear oscillators, the adiabatic construction fails near bifurcations or when the limit-cycle relaxation rate becomes too small.

The common structural principle is that a nonlinear process transforms a geometric control variable into a phase of a nonlinear source or nonlinear state. The control may be a path in projective Hilbert space, a polarization angle, a crystal orientation, a QPM vector, a material tensor, an internal oscillator parameter, or a spatially varying optical axis. The resulting phase is distinct from a purely dynamical phase because it is determined by the geometry of transport or orientation-dependent mode mixing, while its nonlinear character arises from field multiplication, state-dependent coupling, depletion, higher-order susceptibility, or nonlinear feedback.

Source: https://www.emergentmind.com/topics/nonlinear-geometric-phase