---
title: Lewis–Ermakov Invariant in Oscillator Dynamics
url: https://www.emergentmind.com/topics/lewis-ermakov-invariant
type: topic
---

# Lewis–Ermakov Invariant in Oscillator Dynamics

The Lewis–Ermakov invariant, also called the Ermakov–Lewis invariant or Lewis invariant, is a conserved quantity associated with nonautonomous oscillator dynamics, most prominently the time-dependent harmonic oscillator (TDHO). In its standard unit-mass form, for
\[
H(t)=\frac{p^2}{2}+\frac{1}{2}\omega^2(t)x^2,
\]
one introduces a positive auxiliary function \(\rho(t)\) satisfying the Ermakov equation
\[
\ddot{\rho}(t)+\omega^2(t)\rho(t)=\frac{1}{\rho^3(t)},
\]
and obtains the invariant
\[
I_{\mathrm{EL}}=\frac{1}{2}\left[\left(\rho p-\dot{\rho}x\right)^2+\left(\frac{x}{\rho}\right)^2\right].
\]
Its defining feature is exact conservation despite explicit time dependence of the Hamiltonian, and this property underlies a broad body of work in classical mechanics, quantum mechanics, optics, cosmology, Koopman–von Neumann mechanics, and modern invariant-based control [2006.06489].

## 1. Historical setting and standard definition

The modern theory combines two strands. Ermakov’s nineteenth-century nonlinear auxiliary equation provided the amplitude variable \(\rho\), while Lewis and Lewis–Riesenfeld established the invariant method for the quantum TDHO. In the standard one-dimensional case, \(x(t)\) solves
\[
\ddot{x}(t)+\omega^2(t)x(t)=0,
\]
and \(\rho(t)\) solves the associated Ermakov equation; the Lewis–Ermakov invariant is then the quadratic form written above. In the quantum theory the same structure becomes a Hermitian operator \(\hat I_{\mathrm{EL}}\) satisfying
\[
\frac{d\hat I_{\mathrm{EL}}}{dt}-i[\hat I_{\mathrm{EL}},\hat H(t)]=0,
\]
so its eigenvalues are time-independent even though \(\hat H(t)\) is not [2006.06489].

A commonly used generalization includes a time-dependent mass. For
\[
\hat H(t)=\frac{1}{2m(t)}\hat p^2+\frac{1}{2}m(t)\omega^2(t)\hat q^2,
\]
the invariant is written as
\[
\hat I(t)=\frac{1}{2}\left\{\left[\frac{\hat q}{\rho(t)}\right]^2+\left[\rho(t)\hat p-m(t)\dot\rho(t)\hat q\right]^2\right\},
\]
with \(\rho(t)\) now determined by
\[
\ddot{\rho}(t)+\frac{\dot m(t)}{m(t)}\dot\rho(t)+\omega^2(t)\rho(t)=\frac{1}{m^2(t)\rho^3(t)}.
\]
This form is the natural extension for variable-mass quadratic systems and appears directly in optical analog models of time-dependent oscillators [1403.1498].

The standard invariant is exact rather than adiabatic. Several sources in the literature emphasize that \(I\) is closely related to the adiabatic quantity \(E/\omega(t)\) for slowly varying frequency, but unlike the adiabatic invariant it remains conserved for arbitrary time dependence of \(\omega(t)\) once \(\rho\) satisfies the appropriate Ermakov equation [1805.08194].

## 2. Classical constructions and equivalent formulations

A canonical construction uses two real linearly independent solutions \(u_1(t),u_2(t)\) of the linear TDHO equation and their Wronskian
\[
G=u_1\dot u_2-u_2\dot u_1.
\]
Writing
\[
u_1=-\rho\sin s_\rho,\qquad u_2=\rho\cos s_\rho,
\]
one obtains
\[
G=\rho^2\dot s_\rho.
\]
This orthogonal-functions representation identifies \(\rho\) as an amplitude and \(s_\rho\) as a phase, with \(\omega(t)\equiv \dot s_\rho\). In this language the classical invariant can be expressed not only in the usual \((x,p,\rho,\dot\rho)\) form but also as an amplitude-phase quantity,
\[
I=\frac{1}{2}\left(\rho^2\dot s_\rho^2+\dot\rho^2\right),
\]
which makes explicit the interplay between envelope and phase rotation [1309.1498].

Another influential formulation interprets the invariant as ordinary conserved energy after a time-dependent change of variables. With
\[
Q(\tau)=\frac{q(t)}{f(t)},\qquad dt=m(t)f^2(t)\,d\tau,
\]
and \(f(t)\) chosen to satisfy the appropriate Ermakov-type equation, the original TDHO is mapped to a constant-frequency oscillator in the new variables. The conserved energy of that transformed system becomes, when re-expressed in the original variables, the Lewis–Ermakov invariant. This viewpoint was used to “demystify” the constancy of the invariant and to connect it to a broader family of Ray–Reid-type invariants generated by more general transformed potentials [1712.07328].

The same classical framework also extends beyond the standard Pinney nonlinearity. For Reid’s \(m\)-th order generalized Ermakov systems, the nonlinear equation is equivalent to an integrable Emden–Fowler equation, and a closed formula for the higher-order invariant can be written. In the standard \(m=2\) case the invariant is
\[
I=\frac{1}{2}\left[\left(\tilde q\,\dot q-q\,\dot{\tilde q}\right)^2+\alpha\left(\frac{q}{\tilde q}\right)^2\right],
\]
while for \(m\ge 3\) the invariant acquires the characteristic higher-power term
\[
I(t)=\frac{1}{2}\left(\tilde q\dot q-q\dot{\tilde q}\right)^2+\frac{\alpha W^{m-2}}{m-1}\left(\frac{q}{\tilde q}\right)^{2m-2}.
\]
This preserves the basic “Wronskian-like term plus nonlinear potential term” architecture of the Lewis–Ermakov construction [1402.4402].

## 3. Quantum operator theory, ladder structures, and exact solution methods

In the quantum TDHO, the invariant is not merely conserved in expectation value; it organizes the exact solution space. A standard operator realization is
\[
\hat I=\frac{1}{2}\left[\left(\frac{\hat q}{\rho}\right)^2+(\rho\hat p-\dot\rho\hat q)^2\right],
\]
and one may define invariant and time-dependent ladder operators by
\[
\hat A=\frac{1}{\sqrt{2G}}(\hat G_1-i\hat G_2),\qquad
\hat a(t)=\frac{1}{\sqrt2}\left(\frac{\hat q}{\rho}+i(\rho\hat p-\dot\rho\hat q)\right),
\]
with
\[
\hat I=\hat a^\dagger(t)\hat a(t)+\frac{1}{2}=\hat A^\dagger\hat A+\frac{1}{2}.
\]
The two ladder sets are related by a unitary phase transformation
\[
\hat a(t)=e^{is_\rho(t)\hat I}\hat A e^{-is_\rho(t)\hat I},
\]
so the explicit time dependence is a phase rotation generated by the invariant itself [1309.1498].

This operator structure supports a number-phase representation. A generalized Turski phase operator \(\hat\Phi\) satisfies
\[
[\hat\Phi,\hat I]=-i,\qquad \dot{\hat\Phi}=-\omega(t),
\]
and the coordinate operator can be written in amplitude-phase form using Dirac-type relations
\[
\hat a=\sqrt{\hat I}\,e^{-i\hat\Phi},\qquad
\hat a^\dagger=e^{i\hat\Phi}\sqrt{\hat I}.
\]
Within this formulation the invariant constrains the co-evolution of excitation number and instantaneous frequency. In particular, for each mode the paper on Manley–Rowe relations states
\[
\omega_k(t_i)\,\hat n_k(t_i)=\omega_k(t_f)\,\hat n_k(t_f),
\]
which becomes the basis for a quantum derivation of power-balance relations in nonlinear optics [1309.1498].

A distinct quantum reinterpretation identifies the invariant with second-order quantum moments. For Gaussian wave packets in a time-dependent harmonic well, the width \(\alpha(t)\) obeys an Ermakov equation,
\[
\ddot\alpha+\omega(t)^2\alpha=\frac{\hbar^2}{4m^2\alpha^3},
\]
and the quadratic uncertainty Casimir
\[
C=\langle \hat x^2\rangle\langle \hat p^2\rangle-\langle \hat D\rangle^2,\qquad
\hat D=\frac{1}{2}(\hat x\hat p+\hat p\hat x),
\]
is conserved. For Gaussian states this Casimir reproduces the Lewis–Ermakov invariant up to fixed factors and a constant term. The same work further interprets the wave-packet width as an intrinsic clock through
\[
\tau=\int^t\frac{dt'}{\alpha^2(t')},
\]
which maps the mean trajectory to a constant-frequency oscillator in the \(\tau\) variable [2212.09442].

## 4. Interpretations: angular momentum, symmetry, and hidden geometry

One classical interpretation, due to Eliezer and Gray and reused in later work, embeds the problem into an auxiliary two-dimensional isotropic oscillator. Writing \(x=\rho\cos\theta\), \(y=\rho\sin\theta\), one has conserved angular momentum
\[
J=\rho^2\dot\theta,
\]
and the Ermakov–Milne–Pinney equation becomes the radial equation
\[
\ddot\rho+\kappa^2(t)\rho=\frac{J^2}{\rho^3}.
\]
The Lewis–Ermakov invariant then reduces to
\[
I=\frac{1}{2}\left[(\rho\dot x-x\dot\rho)^2+\frac{J^2x^2}{\rho^2}\right]=\frac{J^2}{2},
\]
so, in this picture, the invariant is simply angular momentum squared over two [2102.09613].

That same angular-momentum interpretation survives in a relativistic extension. For the relativistic Ermakov–Milne–Pinney system, the invariant becomes
\[
I_R=\frac{1}{2}\left[\gamma^2(\rho\dot x-x\dot\rho)^2+\frac{J^2x^2}{\rho^2}\right],
\]
and again evaluates to \(J^2/2\). The novelty lies not in the conserved quantity’s physical meaning but in the modified radial equation, where relativistic factors introduce additional nonlinearities in both the restoring term and the inverse-cubic term [2102.09613].

A complementary symmetry-based interpretation uses time reparameterizations and the Virasoro action on the TDHO. Under the synchronizing change of variables \(Q=q/\eta\) and \(d\tau=dt/\eta^2\), with \(\eta\) solving the Ermakov equation, the dynamics becomes that of a constant-frequency oscillator in \(\tau\). The invariant is then the oscillator energy in the synchronized frame, while the underlying quadratic moments generate an \(sl(2,\mathbb{R})\) or \(su(1,1)\)-type algebra. This places the Lewis–Ermakov invariant among the conserved charges associated with the hidden conformal structure of the problem [2212.09442].

A further geometric interpretation appears in stationary Bohm–Madelung quantum mechanics. After Liouville normalization of separated Sturm–Liouville sectors, each sector obeys
\[
\psi_i''(q_i)+\Omega_i^2(q_i)\psi_i(q_i)=0,
\]
while the Bohm amplitude satisfies
\[
\rho_i''(q_i)+\Omega_i^2(q_i)\rho_i(q_i)=\frac{k_i}{\rho_i^3(q_i)}.
\]
The sectorwise invariant
\[
I_i=\frac{1}{2}\left[\big(\rho_i y_i'-\rho_i' y_i\big)^2+\frac{k_i\,y_i^2}{\rho_i^2}\right]
\]
is then independent of the spatial coordinate \(q_i\). In this formulation the quantum potential is encoded as a curvature contribution of the self-adjoint Sturm–Liouville operator rather than as an additional dynamical term [2602.00507].

## 5. Reformulations and generalizations beyond the standard oscillator

A notable nonstandard setting is Koopman–von Neumann (KvN) mechanics, the Hilbert-space formulation of classical mechanics. For the TDHO the KvN Hamiltonian can be written
\[
\hat H=p\,\hat P+k(t)\,q\,\hat Q,
\]
and, after a canonical transformation in the extended KvN phase space, it splits as
\[
\hat H=\hat H_1-\hat H_2,
\]
with each \(\hat H_j\) an ordinary quantum TDHO Hamiltonian. This permits a direct import of two standard Ermakov–Lewis invariants and yields the KvN invariant
\[
\hat I=\frac{1}{2}\left[\frac{q^2+Q^2}{\rho^2}+(\rho p-\dot\rho Q)^2+(\rho P-\dot\rho q)^2\right].
\]
An earlier KvN construction derived the invariant from a quadratic ansatz and a coupled ODE system; the later simplification avoided those coupled equations but was explicitly described as less general because it relied on the Hamiltonian splitting into two standard TDHOs [1805.08194; 2006.06489].

The invariant also survives nontrivially under deformations of the underlying oscillator. For singular parametric oscillators generated by one-parameter Darboux deformations of the classical harmonic oscillator, the associated Lewis–Ermakov invariants were shown to be independent of the Darboux deformation parameter \(\gamma\) and singularity-free, even though the deformed frequencies and linear modes possess finite-time singularities. In that setting the invariant retains the undeformed harmonic-oscillator form
\[
\mathcal I=\frac{1}{2}\left(\omega_0^2\tilde N^2-k\tilde M^2\right)
\]
for both the odd and even Darboux families [2403.12084].

An algebraic extension reaches time-dependent two-level systems. Exploiting the shared complexification \(sl(2,\mathbb C)\) of the \(su(2)\) algebra of two-level systems and the \(su(1,1)\) algebra of quadratic oscillators, a Lewis–Ermakov-type invariant was constructed for
\[
\hat H_{\mathrm{TLS}}(t)=a(t)(\hat\sigma_++\hat\sigma_-)+b(t)\hat\sigma_z.
\]
The auxiliary scaling function \(\rho(t)\) satisfies a complex Ermakov–Pinney equation, and the invariant \(\hat I_{\mathrm{TLS}}(t)\) generates a closed-form evolution operator for arbitrary time-dependent coupling \(a(t)\) and detuning \(b(t)\). The same framework was used for Landau–Zener transitions, non-Hermitian dissipative processes, adiabatic rapid passage, and inverse-engineered shortcuts to adiabaticity [2606.22709].

## 6. Applications, methodological significance, and limitations

In optics the invariant appears in several distinct guises. In nonlinear optics, writing the quantum invariant in terms of number and phase operators yields modewise relations of the form \(\omega_k n_k=\text{const}\), which were used to derive Manley–Rowe-type power-balance relations from a quantum invariant rather than from semiclassical transport equations [1309.1498]. In photonic lattices, a class of waveguide arrays was mapped to a quantum harmonic oscillator with propagation-dependent mass and frequency, so the lattice inherits Lewis–Ermakov symmetry through the invariant
\[
\hat I(t)=\frac{1}{2}\left\{\left[\frac{\hat q}{\rho(t)}\right]^2+\left[\rho(t)\hat p-m(t)\dot\rho(t)\hat q\right]^2\right\},
\]
with \(\rho\) satisfying the generalized Ermakov equation [1403.1498].

A spatial version of the invariant was developed for quantum light propagation in separable longitudinally inhomogeneous media. After generalized canonical transformations to a comoving frame, the generator of propagation becomes a physical Lewis–Ermakov invariant in space,
\[
\mathcal M_\sigma(q,p,z)=\frac{1}{2}\left[(\rho_\sigma p_\sigma-q_\sigma\rho_\sigma')^2+\beta_{0\sigma}^2\left(\frac{q_\sigma}{\rho_\sigma}\right)^2\right],
\]
and the net propagation effect is a quantum Gouy phase rather than the unphysical squeezing produced by naive quantization in the original variables [1506.08523]. A related paraxial-wave treatment showed that finite-energy free-space beams behave as if subject to an effective quadratic GRIN-like confinement, with a zero-frequency Ermakov equation
\[
\ddot\rho=\frac{b_0^2}{\rho^3}
\]
and invariant
\[
\hat I(\tau)=\frac{1}{2}\sum_{\mu=x,y}\left[(\rho\hat p_\mu-\dot\rho\,\mu)^2+\frac{b_0^2}{\rho^2}\mu^2\right].
\]
That work also stressed a limitation: for any \(b_0\neq 0\), \(\hat I(0)\) and the free-space Hamiltonian do not commute, and exact commutativity would require an infinitely wide, nonsquare-integrable field [2510.00277].

Cosmological and gravitational applications arise because many mode equations and scale-factor equations reduce to EMP form. Within an Eisenhart–Duval lift, the auxiliary function \(f(t)\) obeys
\[
\frac{d^2f}{dt^2}+\omega^2(t)f=\frac{\lambda}{f^3},
\]
and the usual invariant
\[
I=\frac12\left[\frac{q^2}{f^2}+\bigl(f\dot q-\dot f\,q\bigr)^2\right]
\]
emerges as the conserved energy of a conformally related static oscillator metric. In the cosmological extension of that construction, the scale factor itself satisfies an EMP equation,
\[
\ddot a(t)+\Omega^2(t)a(t)=\frac{\gamma^2}{a^3(t)},
\]
linking the Ermakov structure to Friedmann-type dynamics and Dmitriev–Zel'dovich equations [1802.03370].

Several methodological caveats recur across the literature. In the stochastic setting, numerical experiments with multiplicative noise suggested that the classical invariant is not affected when the same Wiener process is added to both members of the Ermakov system, but the same study explicitly noted the absence of a rigorous stochastic proof and found strong deviations when the noise was applied asymmetrically [1410.5474]. In virialised stellar dynamics, the Ermakov–Lewis–Leach invariant was expanded asymptotically in small parameters to generate an infinite series of conservation laws; the resulting constants were stated to depend on the virialised radius and virialised mass independently of the potential used, but the construction is explicitly asymptotic and tied to short-time regimes [2306.04435]. These examples show that the invariant is both robust and structurally adaptable, but also that the form of the auxiliary equation, the choice of variables, and the admissible class of transformations determine how far any given construction extends.

Source: https://www.emergentmind.com/topics/lewis-ermakov-invariant