Rotation-Free Shell Formulation

• Rotation-Free Shell Formulation is a class of shell theories where the rotation is derived from the deformed midsurface, eliminating independent rotational degrees of freedom.
• It distinguishes between Kirchhoff–Love displacement-only models and 6-parameter no-drilling shells that ensure energy invariance under local drilling rotations.
• Various discretization strategies, including mixed-hybrid, isogeometric, subdivision, and meshfree methods, are developed to address curvature approximation and membrane locking challenges.

Rotation-free shell formulation denotes a class of shell theories in which rotations are not independent unknowns, so that the shell orientation is recovered from the geometry of the deformed midsurface rather than carried by separate rotational degrees of freedom. In the Kirchhoff–Love sense, the director remains orthogonal to the middle surface and transverse shear is eliminated; in a distinct but related sense used in geometrically nonlinear 6-parameter resultant shell theory, shells “without drilling rotations” are characterized by invariance of the strain energy under superposed local rotations about the third director $\boldsymbol d_3$, which removes energetic sensitivity to the drilling mode but does not eliminate the independent rotation field itself (Neumeyer et al., 20 May 2025, Duong et al., 8 Jan 2025, Birsan et al., 2013).

1. Conceptual scope and terminological distinctions

In Kirchhoff–Love shell theory, the defining rotation-free property is that the director at each point remains orthogonal to the middle surface, so the rotation field is not an independent unknown. Rotations are determined by gradients of the displacement of the middle surface, and the formulation has only displacement degrees of freedom. This is the sense adopted by mixed-hybrid coordinate-free Tangential Differential Calculus formulations, nonlinear isogeometric thin-shell formulations in curvilinear coordinates, rotation-free subdivision shells, peridynamic Kirchhoff–Love shells, and rotation-free isogeometric shells for embedded fibers (Neumeyer et al., 20 May 2025, Duong et al., 8 Jan 2025, Munglani et al., 2015, Behzadinasab et al., 2021, Duong et al., 2021).

A different usage arises in the 6-parameter resultant shell theory. There, the shell kinematics include an independent rotation field $\boldsymbol Q\in SO(3)$, but the shell is said to be “without drilling rotations” when the strain energy density is invariant under the superposed drilling rotation

$$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$

for all angle fields $\theta(x_1,x_2)$. In that setting, the model is free of the rotational mode about $\boldsymbol d_3$, but it is not fully rotation-free in the Kirchhoff–Love sense (Birsan et al., 2013, Birsan et al., 2013).

This distinction is central to the literature. Kirchhoff–Love formulations are displacement-only and remove all independent rotational variables. No-drilling Cosserat-type shells retain an independent rotation field but enforce constitutive indifference to one rotational mode. By contrast, not every shell method for thin structures with directors is rotation-free: an SPH formulation based on Uflyand–Mindlin theory is explicitly “not rotation-free in the classical sense,” because each surface particle carries translational and rotational degrees of freedom and the pseudo-normal is updated from independent rotation variables (Wu et al., 2023).

2. Kirchhoff–Love kinematics and rotation-free strain measures

The geometric core of a rotation-free Kirchhoff–Love shell is the midsurface map together with the induced unit normal. In the coordinate-free TDC formulation, the shell thickness domain is described by

$x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$

and the linearized displacement field is

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$

where $u$ is the surface displacement and $w\in T_\Gamma$ is the difference vector describing the rotation of the surface normal induced by bending. The Kirchhoff–Love constraint is realized by

$$w = H\cdot u - \nabla_\Gamma(u\cdot n)=-(\nabla_\Gamma^{dir}u)^T\cdot n,$$

which enforces no transverse shear and leaves no drilling rotations and no independent rotational degrees of freedom (Neumeyer et al., 20 May 2025).

Within that framework, the membrane strain and bending strain are expressed as

$\boldsymbol Q\in SO(3)$0

and

$\boldsymbol Q\in SO(3)$1

Using the Kirchhoff–Love definition of $\boldsymbol Q\in SO(3)$2, the bending strain takes the compact in-plane symmetric form

$\boldsymbol Q\in SO(3)$3

which makes explicit that bending is governed by surface second derivatives of the displacement (Neumeyer et al., 20 May 2025).

In natural curvilinear coordinates, the same rotation-free structure appears through metric and curvature changes. If $\boldsymbol Q\in SO(3)$4 and $\boldsymbol Q\in SO(3)$5 denote the reference and current midsurfaces, with covariant tangents $\boldsymbol Q\in SO(3)$6 and $\boldsymbol Q\in SO(3)$7, then

$\boldsymbol Q\in SO(3)$8

where $\boldsymbol Q\in SO(3)$9, $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$0, $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$1, and $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$2. The tensorial form used in nonlinear isogeometric formulations is

$$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$3

and

$$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$4

with only the displacement control variables $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$5 as unknowns and no rotational degrees of freedom (Duong et al., 8 Jan 2025).

The same geometric pattern underlies rotation-free triangular subdivision shells. Under Kirchhoff–Love assumptions, lines initially normal to the midsurface remain straight, unstretched, and normal after deformation; the shell director equals the unit normal to the midsurface; there are no rotational degrees of freedom; and only translational displacement degrees of freedom at mesh nodes are used (Munglani et al., 2015).

3. No-drilling shells within the 6-parameter resultant theory

The geometrically nonlinear 6-parameter resultant shell theory augments the midsurface deformation $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$6 with an independent orthonormal director triad $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$7, assembled into $$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$8. The elastic rotation is

$$\boldsymbol Q\ \longrightarrow\ \boldsymbol R_\theta\,\boldsymbol Q$$9

the surface deformation gradient is

$\theta(x_1,x_2)$0

and the exact strain measures are

$\theta(x_1,x_2)$1

The drilling rotation at a point is the rotation of the director triad about $\theta(x_1,x_2)$2, represented by $\theta(x_1,x_2)$3, and “without drilling rotations” means that the shell energy is invariant under $\theta(x_1,x_2)$4 for arbitrary $\theta(x_1,x_2)$5 (Birsan et al., 2013).

The representation theorem gives a complete characterization of such constitutive laws: $\theta(x_1,x_2)$6 and conversely any energy of this form is invariant under superposed drilling rotations. An equivalent invariant description uses the triple

$\theta(x_1,x_2)$7

so that

$\theta(x_1,x_2)$8

The additional invariance condition

$\theta(x_1,x_2)$9

shows that the energy is independent of $\boldsymbol d_3$0, i.e. of the drilling curvature component (Birsan et al., 2013, Birsan et al., 2013).

In the linearized limit, these invariant measures reduce to Reissner-type five-parameter kinematics: $\boldsymbol d_3$1

$\boldsymbol d_3$2

and they are independent of the drilling component $\boldsymbol d_3$3. This is the precise sense in which the 6-parameter theory reduces from six to five effective parameters when drilling rotations are excluded (Birsan et al., 2013).

The constitutive consequences are stringent. For isotropic quadratic energies, the no-drilling constraints enforce

$\boldsymbol d_3$4

which explicitly eliminate drilling stiffness and remove in-plane antisymmetric dependence coupled to drilling. At the same time, the general coercivity conditions valid in the drilling-inclusive case fail; the quadratic energy becomes only positive semi-definite in some directions, and the general existence theorem does not apply directly (Birsan et al., 2013, Birsan et al., 2013). This is a persistent analytical distinction between no-drilling 6-parameter shells and genuinely rotation-free Kirchhoff–Love shells.

4. Discretization strategies for rotation-free shells

The principal computational obstacle in rotation-free Kirchhoff–Love shells is the appearance of curvature, and therefore second derivatives, without independent rotational unknowns. Several distinct discretization strategies address this obstacle.

A mixed-hybrid finite element method for Kirchhoff–Love shells elevates the bending moment tensor $\boldsymbol d_3$5 to a primary unknown and hybridizes its tangential continuity by a scalar Lagrange multiplier $\boldsymbol d_3$6 on the edge skeleton. The unknowns are the displacement field $\boldsymbol d_3$7, the in-plane symmetric bending moment tensor $\boldsymbol d_3$8, and $\boldsymbol d_3$9, which behaves “rotational-type” only as a constraint variable and is not a physical rotation. The method uses standard $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$0 Lagrange elements with equal-order interpolation, allows $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$1 to be discontinuous across elements, and eliminates the element-local moment unknowns by static condensation, yielding a condensed global system in $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$2 that is symmetric positive definite (Neumeyer et al., 20 May 2025).

A nonlinear isogeometric formulation in natural curvilinear coordinates adopts NURBS or B-splines with $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$3 to provide $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$4 continuity within patches and uses Bezier extraction for efficient FE data structures. The practical difficulty is patch-boundary transmission of bending moments. That formulation introduces a unified angle-based $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$5 constraint between adjacent patch normals, enforceable by either a penalty potential

$x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$6

or a Lagrange multiplier potential

$x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$7

The same construction is used for fixed fold angles, symmetry conditions, and rotational Dirichlet boundary conditions, all without introducing rotational degrees of freedom (Duong et al., 8 Jan 2025).

Subdivision-based triangular shells provide another route to the required smoothness. Loop subdivision surfaces furnish $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$8-continuous triangular elements suitable for Kirchhoff–Love bending with no rotational DOFs. For orthotropic materials, a one-time Total Lagrangian transformation of the first and second shape-function derivatives in the undeformed configuration aligns the element basis with a preferred material direction. Because the transformation is applied directly to derivatives rather than to stress or strain components, it remains consistent with curvature evaluation and keeps the additional computational effort minimal relative to isotropic rotation-free shells (Munglani et al., 2015).

Ultraweak and meshfree formulations remove the classical $x = x_\Gamma + \zeta\, n,\qquad |\zeta|\le t/2,$9 barrier by reformulating the problem rather than by enforcing smooth trial spaces. A discontinuous Petrov–Galerkin method for Koiter-type shallow shells is rotation-free because the primary kinematic variables are the midsurface tangential displacements $$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$0 and the transverse deflection $$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$1, with no independent rotational degrees of freedom. Its ultraweak form places $$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$2 in $$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$3, introduces trace variables on the mesh skeleton, and achieves uniform stability through problem-scaled test norms and optimal test functions (Führer et al., 2021). A peridynamic Kirchhoff–Love shell formulation is also rotation-free: it employs only midsurface velocity DOFs, constructs local surface parameterizations through Principal Component Analysis, evaluates first- and second-order parametric derivatives by kernel-based operators, and uses bond stabilization to suppress zero-energy modes on unstructured point clouds (Behzadinasab et al., 2021).

5. Constitutive generality, anisotropy, and multiphysics extensions

Rotation-free shell formulations are not restricted to linear isotropic Koiter laws. A nonlinear isogeometric Kirchhoff–Love framework in curvilinear coordinates admits general hyperelastic surface energies

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$4

with stress resultants and bending moments

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$5

Within that setting, the paper treats the Koiter solid shell model, the mixed Koiter model, the Canham model, Helfrich-type fluid membrane bending, and shells derived from 3D continua by thickness projection and integration (Duong et al., 8 Jan 2025).

A further generalization introduces embedded fiber families with in-plane bending resistance. The corresponding rotation-free isogeometric shell extends Kirchhoff–Love theory by the in-plane curvature tensor

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$6

and its relative form

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$7

This allows constitutive models that capture anisotropy in stretching, shearing, twisting, out-of-plane bending, and in-plane bending, with only displacement degrees of freedom. The formulation is applied to textiles, composites, pantographic structures, and dry woven fabrics (Duong et al., 2021).

Rotation-free shells have also been specialized to materially narrow but mechanically demanding systems. For graphene, an ab-initio-based membrane model calibrated from Kumar and Parks is combined with a Canham bending model reflecting the ab-initio data of Kudin et al., yielding a rotation-free isogeometric shell in which curvature is obtained directly from second derivatives of the surface parametrization. The membrane energy depends on anisotropic logarithmic-strain invariants, while the bending energy is

$$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$8

with $$u_\Omega(x_\Gamma,\zeta)=u(x_\Gamma)+\zeta\,w(x_\Gamma),$$9 for the QM calibration (Ghaffari et al., 2016).

For soft biological materials, three constitutive reduction strategies have been developed for rotation-free Kirchhoff–Love shells: Numerically-Projected, Analytically-Projected, and Directly-Decoupled. The first performs numerical thickness integration of a 3D hyperelastic law, the second derives closed-form first-order thickness projections, and the third decouples a fully nonlinear membrane model from a linear elastic bending model based on the reference membrane stiffness. The paper examines six isotropic and anisotropic material models, including Neo–Hooke, Mooney–Rivlin, Fung, AMR, GOH, and GOH with a compression/tension switch (Roohbakhshan et al., 2016).

The multiphysics extension of rotation-free shells proceeds by preserving the distinction between the shell midsurface and the fields in the surrounding medium. An embedding-aware continuum thin shell formulation derived from 3D hyperelasticity retains the top and bottom shell surfaces during dimension reduction, distinguishes pressure loads applied at the top and bottom surfaces, and expresses the reduced midsurface equations in terms of $u$0, curvature, and the pressure combinations

$u$1

This yields explicit face-aware terms such as $u$2 and $u$3, and provides a platform to include multi-physics coupling (Ghosh et al., 2024). A fully-coupled nonlinear magnetoelastic thin shell formulation generalizes the Kirchhoff–Love assumptions to magnetic variables, uses the general deformation map

$u$4

and explicitly discards the plane stress assumption because Maxwell stress in the surrounding free space requires careful treatment on the upper and lower shell surfaces (Ghosh et al., 2023).

6. Numerical behavior, benchmarks, and persistent limitations

The numerical record of rotation-free shell formulations is broad but heterogeneous. In the mixed-hybrid Kirchhoff–Love method, smooth solutions yield optimal higher-order rates: $u$5-error in $u$6 and $u$7: $u$8, $u$9-error in derived resultants: $w\in T_\Gamma$0, residual errors involving second derivatives: $w\in T_\Gamma$1, and stored elastic energy error: $w\in T_\Gamma$2. Benchmarks include the Scordelis–Lo roof, for which the reference vertical displacement $w\in T_\Gamma$3 is matched, as well as the partly clamped hyperbolic paraboloid, extruded arc, hemispherical shell, flower-shaped shell, and ring-shaped shell (Neumeyer et al., 20 May 2025).

The nonlinear isogeometric thin-shell formulation in curvilinear coordinates reports classical linear and nonlinear benchmarks, including the hemisphere pinching problem, simply supported plates under sinusoidal pressure, the pinched cylinder with rigid diaphragms, pure bending of a flat strip, folded strips with kinks, cantilevers under end shear, hemispheres with holes, and spreading cylinders. The mixed Koiter model avoids thickness integration and is reported as approximately $w\in T_\Gamma$4 faster than projected 3D models while maintaining essentially the same accuracy in the pinched-cylinder benchmark (Duong et al., 8 Jan 2025).

Rotation-free orthotropic subdivision shells reproduce the pinched hemispherical shell with an $w\in T_\Gamma$5 hole and the wrinkling of orthotropic sheets under shear. In the wrinkling study, coarse meshes already deliver near-asymptotic critical shear displacements, and the orthotropic case yields $w\in T_\Gamma$6, consistent with literature, without an additional wrinkling model (Munglani et al., 2015). The peridynamic rotation-free Kirchhoff–Love shell extends the scope from elastostatics to large elasto-plastic deformations, fracture, and fragmentation, and its asymptotically compatible meshfree discretization converges to the classical KL shell model while naturally handling discontinuities (Behzadinasab et al., 2021).

Two numerical limitations recur across the literature. First, membrane locking remains a central issue in thin-shell discretization. The DPG shallow-shell formulation addresses it by raising the polynomial degree only of the tangential displacement trace variable, and its built-in error estimator drives adaptive refinement capable of resolving boundary and interior layers (Führer et al., 2021). Second, the classical $w\in T_\Gamma$7 requirement of Kirchhoff–Love shells continues to motivate mixed, ultraweak, subdivision, and meshfree alternatives; this suggests that “rotation-free” does not imply a unique discretization technology, but rather a kinematic constraint that can be realized by several numerical architectures.

Analytically, the main limitation is different. In the 6-parameter no-drilling setting, the constitutive elimination of drilling stiffness leads to loss of uniform coercivity, so the general existence theorem valid for drilling-inclusive isotropic shells does not apply directly (Birsan et al., 2013). In mechanical modeling, the Kirchhoff–Love hypothesis excludes transverse shear and is therefore appropriate for thin shells but not for shear-dominated regimes. In terminology, the literature is explicit that a method with independent rotational DOFs, even if geometrically robust under large rotation, is not rotation-free in the classical sense (Wu et al., 2023). This suggests that the phrase “rotation-free shell formulation” properly denotes a kinematic class—most often Kirchhoff–Love displacement-only shells—rather than a generic claim of rotational robustness.