Liquid Crystal Elastomer (LCE) Muscles

• Liquid Crystal Elastomer muscles are smart, soft polymer networks that couple liquid-crystalline order with rubber elasticity to enable large, reversible deformations.
• They convert stimuli such as heat, light, and electric fields into controlled motions like contraction, bending, twisting, and shape morphing.
• Advanced studies integrate continuum models, director patterning, and programmable metrics to tailor mechanical functions for applications in artificial muscles and soft robotics.

Searching arXiv for papers on liquid crystal elastomer muscles and related actuation. Liquid crystal elastomer (LCE) muscles are soft, anisotropically contractile polymer networks in which liquid-crystalline order is coupled directly to rubber elasticity, so that changes in orientational order generate macroscopic deformation along a preferred axis. In the most common picture, a monodomain LCE contracts along its nematic director and expands transversely when heated, illuminated, or otherwise driven across an order-changing transition; in patterned, bilayer, rod, fiber, balloon, annular, or metamaterial geometries, the same coupling is converted into bending, twisting, inflation, whirling, snap-through, locomotion, or distributed shape morphing. Across the literature, LCEs are therefore treated both as artificial muscles in the narrow sense of reversible uniaxial contraction and as a broader class of active solids whose internal metric, spontaneous curvature, or effective constitutive state can be programmed, reprogrammed, or trained to yield task-specific mechanical function (Knezevic et al., 2014).

1. Molecular origin and contractile mechanism

The defining physical feature of an LCE is a cross-linked polymer network with embedded mesogenic order. In the nematic state, the network is anisotropic because the polymer chains are statistically elongated along the director; when the nematic order decreases, the network becomes less anisotropic, and the body contracts along the director while expanding in transverse directions because of incompressibility. This coupling is encoded in standard step-length-tensor descriptions such as

$$l_{ij}=l_{\perp}\delta_{ij}+\left(l_{\parallel}-l_{\perp}\right)n_i n_j,$$

or, in neo-classical form,

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$

with the anisotropy parameter $r$ or the order parameter $Q$ controlling the magnitude of spontaneous deformation (Babakhanova et al., 2018, Lee et al., 2022).

Several sources in the literature emphasize that the most pronounced contraction occurs near the nematic–isotropic transition, where the order parameter changes rapidly. In the optomechanical engine model of Corbett and Warner, the spontaneous contraction of a free elastomer between a nematic state and an illuminated pseudo-temperature state is written as

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$

and the corresponding isotropic-to-nematic recovery stretch can be large; for $Q=0.5$, the recovery stretch is $1/\lambda_m=1.58$, while elongations up to $400\%$ are noted (Knezevic et al., 2014). The same general mechanism underlies thermally driven strip contraction, balloon actuation, and photomechanical fiber motion.

A recurrent misconception is that light-driven LCE muscles are actuated by direct optical forcing. In azo-LCEs, illumination instead produces trans $\to$ cis photoisomerization of azobenzene dopants, lowering nematic order and thereby creating an internal spontaneous strain. In the fiber model of Armon et al., the spontaneous axial strain is

$\varepsilon_c=-\lambda n_c,$

with cis density $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$0 governed by

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$1

so the optical field acts by evolving an internal stress-free state rather than by prescribing shape directly (Maghsoodi et al., 2022). This distinction is central to LCE muscle theory: the actuator is active because stimulus changes the material’s preferred metric or spontaneous curvature, not because an external field directly imposes displacement.

The same order–deformation coupling also admits electrical actuation. In an electroclinic LCE based on pendent smectic mesogens tethered to a polymer backbone, an applied electric field produces a mesogen tilt angle $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$2, inferred optically from

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$3

and the mechanical response is interpreted as a competition between field-driven mesogen tilt and polymer-network resistance (Spillmann et al., 2010). This suggests that the “muscle” concept in LCEs is stimulus-agnostic at the continuum level: heat, light, and electric field differ primarily in how they modify orientational order and its kinetics.

2. Constitutive descriptions, soft elasticity, and reduced theories

The constitutive description of LCE muscles is unusually rich because the active response is inseparable from anisotropic elasticity, director evolution, and, in many geometries, microstructure relaxation. At the simplest level, the optomechanical engine paper models the reduced tension as

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$4

with free energy per unit unstretched length

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$5

where $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$6 is the shear modulus, $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$7 the unstretched cross-sectional area, and $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$8 the stretch ratio (Knezevic et al., 2014). This captures the axial work cycle of an LCE muscle under changing natural length.

More general theories describe LCEs as incompressible hyperelastic solids whose effective energy depends on deformation relative to a director-dependent metric. In ideal isotropic-genesis LCEs, Biggins, Warner, and Bhattacharya study relaxed energies in terms of principal stretches $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$9, parameterized by $r$0 and $r$1, and show that the relaxed response contains a fully relaxed zero-energy region $r$2, a soft but not fully stress-free region $r$3, and a strictly stressed region $r$4. In region $r$5, the constitutive stress vanishes: $r$6 This is the mathematical basis of soft elasticity: large deformations may occur with almost no constitutive stress because director rotation and microstructure accommodate the imposed strain (Lee et al., 2022).

That soft behavior is not universal across all operating regimes. The macroscopic constitutive model for isotropic-genesis, polydomain LCEs (I-PLCEs) introduces coarse-grained internal variables $r$7 and $r$8 to represent the hidden evolution of domain microstructure, with admissible set

$r$9

The reduced incompressible energy takes the form

$Q$0

where the residual hardening term is chosen as

$Q$1

A key consequence is that minimization with respect to $Q$2 enforces equality of the first two principal stresses, $Q$3, yielding the experimentally observed in-plane liquid-like response under biaxial loading (Lee et al., 2023). This places LCE muscles conceptually between classical contractile rubbers and internally reconfiguring soft matter.

Rigorous dimension reduction has extended this framework to muscle-relevant structural elements. For nematic bilayer plates, Agostiniani and coauthors derive a two-dimensional bending theory as a $Q$4-limit from three-dimensional nonlinear elasticity, with elastic and Oseen–Frank contributions coupled by a spontaneous-curvature term. For homogeneous isotropic material, the reduced energy contains

$Q$5

as a preferred-curvature tensor, making explicit how anisotropic eigenstrain in the active layer is converted into plate bending (Bartels et al., 2022). Analogous $Q$6-limit rod theories exist for nematic LCE rods and strips, where spontaneous curvature and twist emerge from director coupling. In the one-dimensional rod model of Alouges, DeSimone, and coauthors, the energy includes curvature and torsion penalization, a Frank–Oseen term, and a spontaneous bending–twisting shift induced by the director field (Bartels et al., 2022). In the transversely curved strip model of Tarantino and coauthors, activation enters through a multiplicative decomposition with a stimulus-induced distortion tensor, while pre-existing transverse curvature enables tape-spring-like localization and snap-through (LoGrande et al., 2023).

At a more fundamental continuum level, Bladon, Terentjev, and Warner–type anisotropic elasticity is generalized in models of rod networks and compressible cross-linked polymers by energies of the form

$Q$7

or, in the compressible case, with $Q$8. In these models, the step-length tensor is linked directly to nematic order through

$Q$9

formalizing the principle that order changes induce preferred elastic anisotropy and thereby actuation (Calderer et al., 2013). For LCE muscles, this is the core constitutive idea from which more specialized geometries inherit their behavior.

3. Programming deformation: director fields, metrics, curvature, and topology

Simple monodomain LCE muscles contract uniaxially, but much of the field is concerned with programming that local contractility into distributed shape change. A central mechanism is director patterning. In coatings with predesigned in-plane director textures, temperature-driven contraction along the local director and transverse expansion convert the two-dimensional orientation field into deterministic surface topography. The activation-force density proposed by Ware et al.,

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$0

explains why pure splay produces depressions, pure bend produces elevations, and mixed or half-integer defects generate paired topographic features and defect-core motion (Babakhanova et al., 2018). This formulation places patterned LCE muscles close to active matter: local anisotropic contraction plus director gradients generate a body-force density that redistributes material.

A complementary programming language is metric design. For monodomain sheets with crosslink-density patterning, the target metric is written as

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$1

with $\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$2 for elastomers. In the discontinuous metric programming strategy of White and coauthors, post-functionalization raises crosslink density in selected regions, thereby changing the local contraction factor from $\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$3 to $\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$4. The resulting metric mismatch across an interface generates Gaussian curvature localized near the boundary (Hebner et al., 2022). For a straight interface, the zero-curvature condition for director angle $\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$5 is

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$6

and the experimentally predicted special angle is about $\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$7 (Hebner et al., 2022). This shows that complex LCE muscle behavior need not require spatially varying director fields; spatial variation of actuation magnitude alone can suffice.

Annular and multiply connected geometries add a topological dimension to programming. In annular sheets with axisymmetric spiral director patterns,

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$8

the activated metric

$\lambda_m=\left[\frac{1-Q}{1+2Q}\right]^{1/3},$9

supports a catalog of developable shapes including flat irises, cylinders, truncated cones, and everted annuli (Duffy et al., 2021). The distinguishing quantity is the integrated Gaussian curvature threading the hole,

$Q=0.5$0

Here the hole permits globally distinct Gauss-flat embeddings even when the local Gaussian curvature vanishes. This suggests that LCE muscle design is governed not only by local anisotropy and geometry but also by topology.

Thin active composites and rods translate the same programming logic into curvature and twist rather than full surface metrics. In nematic LCE bilayers, the active top layer has an orientation-dependent eigenstrain, while the passive layer is stress-free; the mismatch is relieved by bending (Bartels et al., 2022). In rod and strip models, through-thickness director variation generates spontaneous bend–twist coupling (Bartels et al., 2022, LoGrande et al., 2023). A plausible implication is that LCE muscle programming spans a hierarchy of internal descriptors—director field, spontaneous metric, spontaneous curvature tensor, and effective topological class—rather than a single design variable.

4. Stimulus pathways, dynamics, and non-equilibrium regimes

LCE muscles are often discussed as high-stroke actuators, but their practical behavior is strongly conditioned by stimulus pathway and rate dependence. Thermal actuation remains the canonical mechanism: heating above a transition temperature reduces order and drives contraction along the director, while cooling restores elongation. A review of LCE actuation for the built environment reports a maximum strain of about $Q=0.5$1 in its comparison table, notes bidirectional reversible actuation, and describes actuation speed as slow compared with shape-memory alloys or dielectric elastomers; specific examples include LCE fibers with $Q=0.5$2 actuation strain and $Q=0.5$3–$Q=0.5$4 length reduction within $Q=0.5$5 second (Schwartz et al., 2021). In dynamically crosslinked xLCEs based on siloxane exchange, thermal contraction begins around $Q=0.5$6, saturation strain exceeds $Q=0.5$7, and the response is repeatable over $Q=0.5$8 cycles without noticeable creep in the actuation range (Saed et al., 2019).

Photoactuation introduces both photochemical and transport constraints. In the photomechanical fiber theory, shallow optical penetration is assumed and modeled by Beer’s law, with local intensity

$Q=0.5$9

which evolves a spontaneous curvature according to

$1/\lambda_m=1.58$0

Under steady illumination, a pre-stressed helix-like fiber can enter a self-sustained periodic whirling state without time-varying forcing, gravity, or snap-through. The paper emphasizes that this whirling motion is not instability-driven: the smallest eigenvalue of the stiffness matrix remains positive throughout the cycle (Maghsoodi et al., 2022). This is an important corrective to the frequent assumption that all periodic LCE motions arise from bistability or buckling.

Electrically driven LCE dynamics also depart from naive monotonic expectations. In the electroclinic elastomer studied by Hiraoka et al., the switching time is defined as the time for $1/\lambda_m=1.58$1 to go from $1/\lambda_m=1.58$2 to $1/\lambda_m=1.58$3 of its full value after field reversal, and it exhibits a maximum at intermediate field rather than decreasing monotonically. The maximum occurs at approximately $1/\lambda_m=1.58$4–$1/\lambda_m=1.58$5 for the elastomer samples, and the critical field does not change much with temperature in the studied range (Spillmann et al., 2010). In Region II of the tilt dynamics, the creep exponent saturates at $1/\lambda_m=1.58$6, interpreted as approximately $1/\lambda_m=1.58$7 entanglement-limited polymer motion (Spillmann et al., 2010). This indicates that field-responsive LCE muscles are jointly limited by mesogen torque and polymer viscoelasticity.

High strain-rate loading exposes a related limitation in isotropic-genesis polydomain systems. In quasi-static tension at $1/\lambda_m=1.58$8, I-PLCEs show a soft plateau of about $1/\lambda_m=1.58$9, but at $400\%$0 peak stress rises to about $400\%$1, and at $400\%$2–$400\%$3 to about $400\%$4, with the classical soft plateau becoming masked or absent (Wihardja et al., 7 Oct 2025). The constitutive extension proposed there decomposes the deformation gradient as

$400\%$5

and attributes the rate dependence to finite kinetics of internal variables $400\%$6, $400\%$7, and viscous deformation. This shows that the same microstructural mechanism that enables soft, muscle-like deformation at low rate can suppress softness at high rate.

By contrast, induction-heated liquid-metal/LCE composites emphasize the opposite regime: rapid, remote actuation. In the LCE–LM architecture, an LM layer is sandwiched between two DIW-printed LCE layers, and eddy current induction heating can raise the temperature to over $400\%$8 in milliseconds. The printed LCE had a measured shear modulus of $400\%$9, actuation strain was about $\to$0 at $\to$1 and about $\to$2 at $\to$3, and a circular disk actuated in $\to$4 (Maurin et al., 2023). The contrast between these results and slower thermally cycled systems does not imply contradiction; it reflects the difference between heat delivery times, specimen geometry, and recovery conditions.

5. Actuator architectures and modes of mechanical function

The term “LCE muscle” encompasses a wide range of actuator architectures. The optomechanical turbine engine is an instructive early example in which LCEs serve as the active working material of a work-producing cycle. A closed loop of photo-LCE belt passes through a cycle $\to$5: stretching in the dark on a tapered spindle, illumination at fixed stretch, contraction while moving down the taper to deliver torque, and relaxation back to the nematic state. The net work per unit initial length is

$\to$6

and the geometry is designed so that contraction occurs while the belt is in contact with a changing spindle radius, converting free shortening into useful work (Knezevic et al., 2014). In this architecture the LCE muscle is not merely an actuator but a thermodynamic working body.

Balloon and shell geometries use geometric instability as an amplifier. In cylindrical LCE balloons, the deformation is described by

$\to$7

while the spontaneous LCE strain for an azimuthal director is encoded through

$\to$8

giving a spontaneous volume strain

$\to$9

This spontaneous strain can drive classical ballooning, producing phase separation at constant volume or a volume jump at constant pressure (Giudici et al., 2020). In the universal-deformation analysis of ideal LCEs, spherical and cylindrical balloon inflation, cavitation, and bending are all treated within the same relaxed-energy framework; for cylindrical balloons, the system may exhibit a pressure plateau over a range of large deformations and very large radial inflation at moderate pressure (Lee et al., 2022). These results frame LCE balloon muscles as actuators whose apparent stroke is magnified by geometry and soft elasticity.

Fiber and microswimmer architectures exploit slender-body dynamics rather than static shape change. In the pre-stressed helix-like azo-LCE fiber, constant illumination generates a periodic whirling motion whose frequency and propulsive force increase with light intensity, and the bead attached to such a fiber may follow a helical trajectory analogous to flagellar swimming (Maghsoodi et al., 2022). The total propulsive force is written as

$\varepsilon_c=-\lambda n_c,$0

Here the LCE muscle acts as a remote-controlled propulsor and micromixer, extending the muscle analogy from force generation to fluid-mediated locomotion.

Metamaterial architectures further broaden the concept. In trainable disordered LCE lattices cut from polydomain sheets, directed aging under compression for $\varepsilon_c=-\lambda n_c,$1 hours at $\varepsilon_c=-\lambda n_c,$2 can tune the Poisson’s ratio into the auxetic regime, while heating above the nematic–isotropic transition erases the trained state so that the same array can be retrained for allostery (Gowen et al., 24 Feb 2025). The isotropic two-dimensional relation

$\varepsilon_c=-\lambda n_c,$3

is invoked to interpret how directed aging lowers the effective bulk modulus relative to the shear modulus. A plausible implication is that future LCE muscles may be specified not only by actuation strain and blocking force but by history-dependent functional modes such as auxetic response, long-range mechanical communication, or rewritable strain routing.

6. Processing, reconfiguration, and system-level applications

Fabrication chemistry and network topology strongly condition how LCE muscles can be aligned, welded, reshaped, and integrated into devices. Dynamic siloxane-exchange LCEs provide a representative route to reprocessable muscle materials. In the xLCE system of Saed et al., the network is formed by a two-stage, one-pot thiol-acrylate/thiol-ene “double-click” reaction using RM82, EDDT, and the vinyl siloxane crosslinker TMTVCTS. The resulting networks have $\varepsilon_c=-\lambda n_c,$4 about $\varepsilon_c=-\lambda n_c,$5, tunable $\varepsilon_c=-\lambda n_c,$6 from about $\varepsilon_c=-\lambda n_c,$7 to $\varepsilon_c=-\lambda n_c,$8, vitrification or creep-onset temperature about $\varepsilon_c=-\lambda n_c,$9–$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$00, and strong plastic flow above about $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$01 (Saed et al., 2019). Director alignment can be reprogrammed after curing by heating to $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$02, applying constant tensile stress, allowing stress-induced plastic flow to about $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$03 extension, and cooling under load (Saed et al., 2019). This resolves a longstanding processing constraint of conventional monodomain LCE muscles, where alignment is normally fixed during synthesis.

Additive and hybrid fabrication methods enable further control. In the LCE–LM composite, direct ink writing aligns mesogens along the printing path, while masked airbrush deposition of liquid metal sets the heating geometry and thickness; one LM layer corresponds to about $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$04 thickness (Maurin et al., 2023). In annular metric mechanics, photoalignment and UV crosslinking are used to realize axisymmetric director patterns that actuate into irises and cylinders (Duffy et al., 2021). These methods indicate that “programming” in LCE muscles operates simultaneously at the levels of chemistry, alignment, thickness, and embedded energy-delivery architecture.

The application spectrum is correspondingly broad. The built-environment review organizes LCE functionality into sensing, actuation on demand, and actuation by sensing, and cites active hinges, self-regulating irises, sun-tracking louvers, self-cleaning solar-panel covers, humidity-responsive ventilation panels, and strain-warning coatings (Schwartz et al., 2021). Patterned surfaces with crosslink-density discontinuities can dynamically flip curvature sign and dislodge debris, supporting self-cleaning functions (Hebner et al., 2022). Annular actuators are proposed for optical apertures, flow regulation in pipes, sphincter-like devices, and filters (Duffy et al., 2021). Fiber whirling suggests remote-controlled microswimmers and micromixers (Maghsoodi et al., 2022), while induction-heated composites demonstrate pop-up structures, omnidirectional robotic motion, underwater manipulation, and crawling (Maurin et al., 2023).

The term “muscle” therefore has several technically distinct meanings in this field. In some cases it denotes axial contraction along a director; in others it denotes a shape-programmed active solid whose contractile anisotropy is converted into curvature or topology change; in still others it denotes a reconfigurable or trainable mechanical substrate. This broader usage is consistent across the cited literature, though it requires care when comparing devices with very different actuation metrics and operating principles.

7. Performance limits, misconceptions, and unresolved directions

Several recurring misconceptions are addressed explicitly in the literature. First, high stroke does not imply monotonic softness or high speed. The built-environment review lists LCE actuation speed as slow relative to several other actuator classes (Schwartz et al., 2021), whereas induction-heated LCE–LM composites achieve millisecond heating and sub-second deformation (Maurin et al., 2023). This suggests that response rate is not an intrinsic scalar property of “LCE muscles” but a systems property determined by heat or light transport, sample size, geometry, and cooling pathway.

Second, soft elasticity is not equivalent to universal zero-stress behavior. Ideal theories indeed admit fully relaxed zero-energy regions (Lee et al., 2022), but macroscopic polydomain models require residual hardening and kinetic variables (Lee et al., 2023), and high-rate tests show pronounced stiffening with suppression of the soft plateau (Wihardja et al., 7 Oct 2025). In practical muscle design, softness is therefore contingent on rate, loading path, and network microstructure.

Third, periodic photomechanical motion need not result from instability. The whirling fiber under steady illumination remains stable in the sense that the smallest stiffness eigenvalue stays positive throughout the cycle (Maghsoodi et al., 2022). By contrast, planar flapping in related settings is associated with snap-through and negative stiffness eigenvalues. The distinction matters for controllability, robustness, and model selection.

Fourth, energetic figures of merit are highly architecture-dependent. The optomechanical turbine paper estimates an efficiency

$$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$05

with $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$06, and reports $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$07 for $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$08 and $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$09 (Knezevic et al., 2014). The same paper lists the assumptions behind this estimate: quasi-equilibrium operation, comparable moduli in the two states, complete photo-conversion, finite but manageable sliding losses, the stated material-parameter ratio, and an idealized work-cycle geometry (Knezevic et al., 2014). It would therefore be inaccurate to treat $$\ell_n=r^{-1/3}\left(I+(r-1)\mathbf n\otimes\mathbf n\right),$$10 as a generic efficiency of LCE muscles.

A final unresolved direction concerns the balance between programmability and adaptability. Conventional LCE muscles are often programmed during synthesis by director alignment or architecture, whereas recent work shows post-fabrication reprogramming through dynamic covalent exchange (Saed et al., 2019) and even training, erasure, and retraining of macroscopic mechanical function in disordered lattices (Gowen et al., 24 Feb 2025). This suggests a shift from fixed-function actuators toward pluripotent active materials. A plausible implication is that future LCE muscle systems will be evaluated not only by strain, stress, or speed but by reconfigurability, recoverability, and the degree to which function can be reassigned after deployment.

Taken together, the modern literature presents LCE muscles as a family of active solids in which molecular order, elastic anisotropy, and geometry are co-designed. Their distinctive features are large reversible strain, strong coupling between internal order and macroscopic shape, access to soft and metric-driven deformation modes, and an unusual capacity for integrating actuation, sensing, programmability, and even trainability within the material itself.