Canalized Rotation in Polaritonic Crystals

• Canalized rotation is a mode of rotational control that uses twist-induced anisotropy and lattice orientation to channel energy into a narrow directional pathway.
• By precisely tuning the bilayer twist angle and square-lattice orientation, the approach reshapes flat iso-frequency contours to selectively excite specific Bloch modes.
• This method enables controlled polariton propagation in twisted bilayer α-MoO₃, offering practical insights for designing advanced anisotropic photonic systems.

Canalized rotation denotes a mode of rotational control in which rotation does not merely reorient a structure, but selectively constrains propagation, coupling, or admissible states into a narrow directional channel. In the most explicit arXiv treatment of the term, the concept is developed for square-lattice polaritonic crystals on twisted bilayer $\alpha$-MoO$_3$, where rotational control operates at two distinct levels: first, the intrinsic biaxial anisotropy is rotated internally by twisting the two layers, thereby reshaping the phonon-polariton iso-frequency contours (IFCs) into a nearly flat form; second, the external square lattice is rotated relative to that flat IFC so that chosen reciprocal-lattice points become co-linear with it and selectively excite Bloch or Bragg modes (Yin et al., 2024). In this sense, canalized rotation is not simple beam steering by rotation alone, but the interplay of twist-induced canalization and rotation-controlled reciprocal-space access.

1. Core definition and rotational architecture

In the polaritonic-crystal formulation, canalized rotation is a two-stage rotational design problem. The first stage rotates the material anisotropy by twisting two biaxial $\alpha$-MoO$_3$ layers relative to one another. The second stage rotates the patterned square lattice relative to the canalized IFC generated by that twist. The central physical claim is that directional control comes from this interplay, not from rotating the material or the lattice in isolation (Yin et al., 2024).

Canalization in this system arises because twisted bilayer $\alpha$-MoO$_3$ supports quasi-linear, asymmetric, highly localized IFCs at the operating frequency. The paper explicitly states that suitable bilayer thicknesses and twist angle produce flat IFCs at $931~\mathrm{cm}^{-1}$. When the contour becomes nearly vertical in momentum space, described as “quasi-linear dispersions almost parallel to the $k_y$ axis,” the group velocity points predominantly along the orthogonal real-space direction, defined in the paper as the $x$ axis. The result is highly collimated, diffractionless propagation mainly along $x$ (Yin et al., 2024).

The two rotational degrees of freedom and their demonstrated values are summarized below.

Rotational degree of freedom	Function	Demonstrated values
Bilayer twist angle	Reshapes bilayer IFC and sets canalization direction	$_3$0 for $_3$1; $_3$2 for $_3$3
Square-lattice orientation $_3$4	Aligns reciprocal-lattice points with flat IFC	$_3$5, $_3$6, $_3$7, $_3$8

This architecture also fixes the coordinate system in a nontrivial way. The $_3$9 axis is defined as parallel to the canalized propagation direction and the $\alpha$0 axis as perpendicular to it. A plausible implication is that the propagation axis is sample-defined by the twisted bilayer dispersion rather than imposed by an external laboratory frame.

2. Twist-induced canalization in bilayer $\alpha$1-MoO$\alpha$2

The prerequisite for canalized rotation is twist-angle engineering of the bilayer IFC. Twisting rotates the crystallographic principal axes of one biaxial $\alpha$3-MoO$\alpha$4 layer with respect to the other, so the hybridized bilayer dispersion is neither that of a single crystal nor a simple sum of two aligned crystals. The paper attributes flat-IFC formation to this twist-induced reshaping and confirms the chosen structures through transfer-matrix-method momentum-space maps (Yin et al., 2024).

For the periodicity-tuning series, the top and bottom layers are $\alpha$5 and $\alpha$6 thick and twisted by $\alpha$7. For the orientation-tuning series, the top and bottom layers are $\alpha$8 and $\alpha$9 thick and twisted by $_3$0. These parameter pairs are selected because they generate the desired flat IFC near $_3$1. The paper emphasizes that the flat IFC “provides the direction of polariton canalization,” which is why the subsequent lattice design is expressed relative to the canalization axis rather than to a fixed crystallographic axis (Yin et al., 2024).

The broader twistoptics literature provides direct context for this mechanism. In twisted $_3$2-MoO$_3$3 at THz frequencies, the bilayer IFC undergoes a twist- and frequency-dependent topological transition from hyperbolic to elliptic, and canalization appears at the transition where the effective IFCs become “straight and parallel.” In that work, the critical-angle condition is written as

$_3$4

and canalization is experimentally reported at $_3$5 for a twist angle of $_3$6, with a spread angle $_3$7 and $_3$8 (Obst et al., 2023). This establishes that twist-controlled IFC flattening is not specific to the $_3$9 mid-infrared design, but is a broader mechanism for creating canalized polariton transport.

3. Lattice rotation, reciprocal-space matching, and selective Bloch excitation

Once twist has fixed a canalized IFC, the second rotational degree of freedom is the square-lattice orientation. The paper defines an angle $\alpha$0 as either the angle between the reciprocal lattice and the flat IFCs or the angle between the lattice orientation and the canalized direction. Because the square lattice and its reciprocal lattice rotate together, these descriptions are geometrically equivalent (Yin et al., 2024).

The momentum-space framework is standard. For a square lattice of period $\alpha$1,

$\alpha$2

and resonances are labeled by $\alpha$3, with $\alpha$4 along $\alpha$5 and $\alpha$6 along $\alpha$7. The paper’s novelty is the use of “co-linear reciprocal points.” Because the IFC is nearly a straight line rather than a broad curved contour, only reciprocal-lattice points lying on or very near that line efficiently couple. A flat IFC therefore acts as a momentum-space selector, reducing the number of accessible intersections and improving mode selectivity (Yin et al., 2024).

Periodicity and rotation play complementary roles. Increasing $\alpha$8 shrinks reciprocal-space spacing, moving reciprocal-lattice points relative to the fixed flat IFC. The demonstrated periods are $\alpha$9, $_3$0, $_3$1, and $_3$2. In the near-field design figures, the authors double the calculated IFC values because the s-SNOM interference measurement records a fringe periodicity corresponding to half the actual polariton wavelength. Under this rule, changing $_3$3 allows the same flat IFC to intersect the $_3$4, $_3$5, $_3$6, or $_3$7 families (Yin et al., 2024).

Rotation at fixed $_3$8 provides a second selection mechanism. As the lattice rotates, each order $_3$9 traces a circle in momentum space with radius

$931~\mathrm{cm}^{-1}$0

At certain $931~\mathrm{cm}^{-1}$1, a given order intersects the flat IFC; at others it misses. The paper describes this as orientation-selective Bragg excitation. Orders with the same $931~\mathrm{cm}^{-1}$2 have similar orientation-dependent reflection curves, shifted in angle. This suggests that rotation primarily selects reciprocal-space shells, while the exact $931~\mathrm{cm}^{-1}$3 label sets angular offset (Yin et al., 2024).

A central clarification in the paper is that this is not free-space beam steering in the usual sense. Rotating the lattice does not rotate the canalized propagation direction itself. Rather, it changes which reciprocal vectors can couple into the pre-existing canalized channel. The steering therefore occurs in reciprocal-space accessibility, not in the real-space propagation axis (Yin et al., 2024).

4. Real-space, Fourier-space, and spectral manifestations

The evidence for canalized rotation is organized across transfer-matrix-method momentum maps, near-field s-SNOM images, FFT maps, and RCWA reflection calculations. In the unpatterned twisted bilayers, the TMM maps show the intended flat, quasi-linear IFCs “almost parallel to $931~\mathrm{cm}^{-1}$4” and strongly asymmetric in intensity. In patterned samples, the near-field images show highly collimated interference fringes mainly along $931~\mathrm{cm}^{-1}$5, with almost no fringe development along $931~\mathrm{cm}^{-1}$6, which the paper treats as the real-space signature of canalization (Yin et al., 2024).

Periodicity tuning appears directly in the fringe count. When $931~\mathrm{cm}^{-1}$7 is varied from $931~\mathrm{cm}^{-1}$8 to $931~\mathrm{cm}^{-1}$9, the number of bright fringes between adjacent holes along the canalization direction $k_y$0 increases from one to four, confirming selective excitation of different Bloch orders. The FFT maps show discrete asymmetric spots rather than broad symmetric rings, demonstrating that reciprocal-space selection is constrained by the asymmetric flat IFC rather than by the square lattice alone (Yin et al., 2024).

Orientation tuning produces a different effect. When $k_y$1 is varied across $k_y$2, $k_y$3, $k_y$4, and $k_y$5, the fringe positions relative to the hole geometry change, but propagation remains canalized along $k_y$6. The FFT spots shift to different resonance orders that coincide with stronger portions of the asymmetric IFC. The paper explicitly emphasizes this intensity asymmetry: the excited orders are concentrated at reciprocal-space points with stronger intensity in the unperforated bilayer momentum map (Yin et al., 2024).

The RCWA calculations reinforce the same logic. For $k_y$7, the flat IFC approximately matches the $k_y$8, $k_y$9, and $x$0 orders. For the orientation study, the paper uses $x$1 and $x$2 as an illustrative geometry and computes orientation-dependent reflection coefficients for the $x$3, $x$4, $x$5, and $x$6 orders over $x$7. The resulting curves differ in both peak count and peak position, providing a quantitative map of where each order can be selectively excited by lattice rotation (Yin et al., 2024).

5. Design rules, misconceptions, and limitations

The practical design rule extracted by the paper is sequential. One first chooses a twisted bilayer $x$8-MoO$x$9 geometry that gives a flat IFC at the operating frequency. In the demonstrated system, this means $x$0 with twist angles near $x$1–$x$2 and thickness pairs $x$3 or $x$4. One then defines coordinates so that $x$5 is along the canalization direction inferred from the IFC normal. Next, one chooses the square-lattice period $x$6 to target the desired reciprocal-lattice order, taking into account the factor-of-two correction used in the near-field design stage. Finally, one rotates the lattice by $x$7 so that the target reciprocal point becomes co-linear with the flat IFC segment (Yin et al., 2024).

Several misconceptions are explicitly precluded by this formulation. Canalized rotation is not the same as rotating the material alone. It is also not the same as rotating the photonic-crystal lattice alone. Nor is it synonymous with rotating the real-space propagation axis. The twisted bilayer determines the canalization direction, while lattice rotation selects which Bloch resonance orders are excited along that fixed direction (Yin et al., 2024).

The demonstrated limitations are also specific. For larger periods $x$8 and $x$9, experiment deviates from theory because polaritons experience more loss while traveling between more distant holes. As $_3$00 becomes large, the FFT amplitude evolves from discrete spots toward a more continuum-like distribution resembling the unpatterned IFC, because the periodic modulation weakens when the holes are farther apart. More generally, the design assumes a fixed excitation frequency and relies on prior thickness and twist optimization to create the flat IFC. The main text does not derive the bilayer anisotropic dispersion analytically; it validates the chosen structures numerically via TMM and RCWA. The paper therefore leaves open a more explicit predictive design framework linking bilayer dielectric tensors, twist angle, and canalization angle (Yin et al., 2024).

A plausible implication is that the main bottleneck is not the reciprocal-space selection principle itself, but the absence of a closed-form forward model for the twisted bilayer dispersion in the design loop.

6. Related literature and divergent uses of the term

The most closely related literature remains within polaritonics. The THz twistoptics work on $_3$01-MoO$_3$02 demonstrates that rotational misalignment alone can convert propagation from hyperbolic to canalized to elliptic, with canalization appearing at the hyperbolic-to-elliptic transition where the bilayer IFC flattens (Obst et al., 2023). A later theory of hyperbolic magnetoexciton polaritons in van der Waals semiconductors makes the rotational aspect still more explicit in a twisted-bilayer setting, stating that “The canalization direction along the optical axis of IFC features a clockwise rotation with increasing the twisted angle” (Jia et al., 13 Oct 2025). These works support a broader interpretation in which rotation of anisotropy axes or IFC optical axes can rotate the preferred direction of canalized transport.

A different but related development appears in hyperbolic plasmonic metasurfaces for all-angle retroreflection. That work does not formulate a general theory of canalized rotation, but it does show that canalized spoof surface plasmons can be redirected on a deformed anisotropic surface while preserving the phase coherence required for retroreflection. The authors explicitly characterize this as directed routing of canalized transport rather than arbitrary rotation of a canalized beam (Yin et al., 2022). This distinction parallels the one in the $_3$03-MoO$_3$04 polaritonic crystal: geometric reorientation of coupling pathways need not imply rotation of the underlying canalized propagation axis.

The phrase also appears outside wave physics with a different meaning. In OScaR, “Canalized Rotation” denotes a fixed Hadamard transform applied before Omni-Token Scaling in extreme KV-cache quantization. There, the purpose is to redistribute the energy of outlier channels across dimensions so that token-wise scaling does not create new per-channel outliers, rather than to control real-space energy flow (Su et al., 19 May 2026). This indicates that the term is not yet standardized across disciplines.

Taken together, the literature supports a narrow and a broad reading. In the narrow reading, canalized rotation refers to the two-level control of twisted anisotropy and lattice orientation that enables selective excitation of Bloch modes in twisted-bilayer $_3$05-MoO$_3$06 polaritonic crystals (Yin et al., 2024). In the broader reading, it refers to rotational degrees of freedom that create, preserve, or selectively access canalized states by reshaping IFCs, routing energy flow, or stabilizing subsequent transformations (Obst et al., 2023).