Complex-Band-Engineered Mechanical Links

• Complex-Band-Engineered Mechanical Links are nanomechanical interconnects defined by lithography, with transport set by the complex continuation of a periodic flexural-wave band diagram.
• They employ serpentine geometries to achieve exponentially decaying couplings, enabling precise control over inter-resonator connectivity for applications in optomechanics and sensing.
• A predictive design workflow—from unit-cell Bloch analysis to full-device simulation and optical readout—transforms empirical coupling challenges into a parametric engineering problem.

Complex-band-engineered mechanical links are lithographically defined nanomechanical interconnects whose transport properties are set by the complex continuation of a periodic structure’s flexural-wave band diagram. In the silicon nanobeam platform of Alonso-Tomás et al., the relevant implementation is a serpentine link that functions as a compact mechanical mirror or evanescent coupler for megahertz in-plane flexural motion, while transverse asymmetry in the photonic-crystal cavity renders the same flexural modes optically addressable without ancillary structures (Alonso-Tomás et al., 15 Jun 2026). The resulting coupled-cavity architecture combines optical transduction, mechanical confinement, and inter-resonator connectivity within a single integrated platform, with the central observable being an exponentially decaying normal-mode splitting governed by the link attenuation constant.

1. Conceptual basis and device architecture

The platform addresses a specific problem in integrated nanomechanical circuits: megahertz flexural modes are attractive for optomechanics, sensing, and signal processing because of their large mechanical response and nonlinear dynamics, yet their extended spatial character makes local confinement and controlled coupling difficult in dense devices (Alonso-Tomás et al., 15 Jun 2026). The reported solution is a nanobeam architecture in which both the optical readout and the mechanical interconnection are engineered at the lithographic level.

The optical element is a photonic-crystal nanobeam cavity. The mechanical element is an in-plane flexural resonator supported and connected by a serpentine interconnect. In coupled-cavity devices, two nominally identical flexural resonators are linked through a finite chain of serpentine cells. The interconnect is not treated merely as a compliant tether; rather, it is designed through its complex band structure so that, at the cavity-mode frequency, it supports evanescent rather than propagating flexural motion. This makes the link analogous to a tunnel barrier in wave mechanics: motion leaks across the link with an amplitude that decays exponentially with the number of cells.

Within this architecture, “complex-band-engineered” refers specifically to the use of a unit-cell Bloch analysis extended into stop bands, where the Bloch wavevector acquires an imaginary part. The imaginary component determines a per-cell attenuation constant, and that attenuation constant in turn predicts the inter-cavity coupling strength. This suggests a calibrated route from unit-cell geometry to device-level coupling.

2. Complex Bloch bands and evanescent flexural transport

The serpentine link is defined by a periodic unit cell whose inner contour is an ellipse of semi-axes $r_t$ (horizontal) and $r_s$ (vertical), plus an offset by beam width $w$. Its projected period is

$a' = 4r_t + 2w.$

For in-plane flexural modes, the unit-cell problem is posed as a Bloch-periodic eigenproblem: one seeks $(\omega, k_x)$ such that

$D(\omega, k_x)\cdot u(x,y)=0,$

with

$u(x+a')=u(x)e^{ik_x a'}.$

The real band structure over $0 \le \mathrm{Re}\,k_x \le \pi/a'$ contains propagating bands separated by stop bands. The crucial step is the analytic continuation into the stop-band regime. Near the $\Gamma$ point, the continuation is obtained from

$\det D(\omega, k_x=\kappa i)=0,$

and near the $r_s$0 point from

$r_s$1

For frequencies inside a stop band, the Bloch wavevector therefore takes the form

$r_s$2

with attenuation per cell

$r_s$3

In the example geometry $r_s$4, the one-antinode mode at $r_s$5 crosses the complex branch with $r_s$6 in the unit-cell analysis, while the three-antinode mode at $r_s$7 gives $r_s$8 (Alonso-Tomás et al., 15 Jun 2026). These values quantify how strongly the serpentine chain suppresses transmission of flexural motion at the relevant cavity frequencies.

A common misconception is to treat the link as if only its static compliance mattered. The reported treatment shows that its wave-attenuation properties are frequency-selective and band-structure-dependent. In that sense, the interconnect is a mechanical spectral element, not only a structural connector.

3. Exponential splitting in coupled cavities

For two identical flexural resonators of frequency $r_s$9 coupled through $w$0 evanescent cells, the normal modes are symmetric and antisymmetric combinations at

$w$1

where the coupling rate decays as

$w$2

The measurable normal-mode splitting is therefore

$w$3

This relation is the central analytic link between the complex-band calculation and the experimentally observed spectrum. Because $w$4 is extracted from the imaginary part of the Bloch wavevector, a unit-cell analysis predicts how rapidly the cavity-cavity splitting should fall with interconnect length. The paper reports that, in full-system FEM simulations, $w$5 and $w$6, in good agreement with the corresponding unit-cell values $w$7 and $w$8; moreover, $w$9 versus $a' = 4r_t + 2w.$0 is linear, as expected from the exponential law (Alonso-Tomás et al., 15 Jun 2026).

For the one-antinode mode with $a' = 4r_t + 2w.$1, measured coupled-cavity data for $a' = 4r_t + 2w.$2 fit

$a' = 4r_t + 2w.$3

with

$a' = 4r_t + 2w.$4

The same summary reports $a' = 4r_t + 2w.$5 for $a' = 4r_t + 2w.$6, along with corresponding values $a' = 4r_t + 2w.$7 for $a' = 4r_t + 2w.$8 and $a' = 4r_t + 2w.$9 for $(\omega, k_x)$0. Taken together, the unit-cell, full-device, and measured attenuation constants establish that the coupling decay can be predicted quantitatively from the complex band structure.

Geometry	Attenuation data	Note
$(\omega, k_x)$1	$(\omega, k_x)$2, $(\omega, k_x)$3, $(\omega, k_x)$4	Splitting decay measured for $(\omega, k_x)$5–$(\omega, k_x)$6
$(\omega, k_x)$7	$(\omega, k_x)$8, $(\omega, k_x)$9	Deeper in gap, stronger attenuation
$D(\omega, k_x)\cdot u(x,y)=0,$0	$D(\omega, k_x)\cdot u(x,y)=0,$1, $D(\omega, k_x)\cdot u(x,y)=0,$2	Near band edge, weaker attenuation

This agreement is significant because it turns an otherwise empirical coupling problem into a parametric design problem. A plausible implication is that large arrays of flexural resonators could be designed by specifying attenuation targets at the unit-cell level and then validating finite-device corrections by full FEM.

4. Geometry dependence of stop bands and attenuation

The reported geometry study holds $D(\omega, k_x)\cdot u(x,y)=0,$3 and $D(\omega, k_x)\cdot u(x,y)=0,$4 fixed while varying $D(\omega, k_x)\cdot u(x,y)=0,$5, the vertical-ellipse parameter of the serpentine cell. Under this variation, the lower-gap center $D(\omega, k_x)\cdot u(x,y)=0,$6 shifts approximately linearly with $D(\omega, k_x)\cdot u(x,y)=0,$7, and the per-cell attenuation at the one-antinode frequency, $D(\omega, k_x)\cdot u(x,y)=0,$8, decreases as $D(\omega, k_x)\cdot u(x,y)=0,$9 approaches the band-edge value $u(x+a')=u(x)e^{ik_x a'}.$0.

For $u(x+a')=u(x)e^{ik_x a'}.$1, an approximate fit is given by

$u(x+a')=u(x)e^{ik_x a'}.$2

with

$u(x+a')=u(x)e^{ik_x a'}.$3

An effective-mass expansion near the $u(x+a')=u(x)e^{ik_x a'}.$4 point provides a related local picture,

$u(x+a')=u(x)e^{ik_x a'}.$5

implying

$u(x+a')=u(x)e^{ik_x a'}.$6

These relations formalize the intuitive statement that attenuation is strongest when the operating frequency lies deeper within the stop band and weakens as the mode approaches the band edge (Alonso-Tomás et al., 15 Jun 2026). They also explain why small geometric adjustments in the serpentine cell can produce large changes in coupling.

The design rules stated in the summary are correspondingly direct. To reduce coupling, one increases $u(x+a')=u(x)e^{ik_x a'}.$7 by increasing $u(x+a')=u(x)e^{ik_x a'}.$8 toward $u(x+a')=u(x)e^{ik_x a'}.$9, which moves the gap down so that the working frequency lies deeper in the gap, or by increasing $0 \le \mathrm{Re}\,k_x \le \pi/a'$0 or $0 \le \mathrm{Re}\,k_x \le \pi/a'$1 to lengthen the pitch $0 \le \mathrm{Re}\,k_x \le \pi/a'$2 and thereby widen or relocate gaps. To increase coupling, one decreases $0 \le \mathrm{Re}\,k_x \le \pi/a'$3 toward $0 \le \mathrm{Re}\,k_x \le \pi/a'$4, or places the operating frequency closer to the band edge, or uses fewer cells $0 \le \mathrm{Re}\,k_x \le \pi/a'$5. For a target coupling $0 \le \mathrm{Re}\,k_x \le \pi/a'$6 at $0 \le \mathrm{Re}\,k_x \le \pi/a'$7 cells, the summary gives

$0 \le \mathrm{Re}\,k_x \le \pi/a'$8

The need to infer $0 \le \mathrm{Re}\,k_x \le \pi/a'$9 from unit-cell analysis and then confirm with full-device simulation is central to the workflow.

An apparent inconsistency in the summary table—where the note for $\Gamma$0 labels the geometry “near band-edge, weak $\Gamma$1,” whereas the prose design rule states that increasing $\Gamma$2 toward $\Gamma$3 increases $\Gamma$4—should be treated cautiously. The numerical entries in the table indicate lower attenuation at $\Gamma$5 than at $\Gamma$6 or $\Gamma$7, so the table and fitted values support the “near band-edge, weak $\Gamma$8” description more directly than the contrary wording. This suggests that interpretation should follow the explicit attenuation data when applying the design rules.

5. Optical brightening of in-plane flexural modes

In a laterally symmetric photonic-crystal nanobeam, dispersive optomechanical coupling to in-plane flexural motion is strongly suppressed because the moving-boundary and photo-elastic contributions to $\Gamma$9 nearly cancel. The total vacuum coupling is written as

$\det D(\omega, k_x=\kappa i)=0,$0

with the moving-boundary term given schematically by

$\det D(\omega, k_x=\kappa i)=0,$1

The device overcomes this cancellation by introducing a transverse asymmetry in the cavity defect wings such that $\det D(\omega, k_x=\kappa i)=0,$2. This shifts the optical mode laterally and increases the net moving-boundary overlap, restoring strong dispersive coupling to the in-plane flexural modes. Both FEM and experiment show that the symmetric case gives $\det D(\omega, k_x=\kappa i)=0,$3, whereas the asymmetric case yields $\det D(\omega, k_x=\kappa i)=0,$4–$\det D(\omega, k_x=\kappa i)=0,$5 depending on mode order $\det D(\omega, k_x=\kappa i)=0,$6, with the coupling eventually plateauing beyond a saturation asymmetry ratio (Alonso-Tomás et al., 15 Jun 2026).

This optical brightening is important because the complex-band-engineered link would be of limited utility without a direct, integrated readout channel. The platform’s distinctiveness lies in combining optical accessibility and mechanical spectral engineering in the same lithographically defined structure. A plausible implication is that the optical cavity serves not merely as a transducer but as an integral part of the calibration loop that links FEM, Bloch analysis, and measured spectral splitting.

6. Experimental operating regime, model limits, and design workflow

The measured single-beam mechanical quality factor depends on pressure and temperature. At $\det D(\omega, k_x=\kappa i)=0,$7, the one-antinode mode has $\det D(\omega, k_x=\kappa i)=0,$8–$\det D(\omega, k_x=\kappa i)=0,$9, consistent with an internal-loss-limited regime, and at $r_s$00 the quality factor increases by approximately a factor of two (Alonso-Tomás et al., 15 Jun 2026). These numbers delimit the damping background against which the coupling-induced splittings are resolved.

The simple two-resonator model is valid only while the serpentine interconnect contributes no additional nearby modes. The summary identifies a breakdown regime: for $r_s$01 or $r_s$02, extra peaks appear as finite-link modes cross the gap edge. In that case, the interconnect modes hybridize with the cavity modes and the two-resonator description no longer suffices; full-system FEM is required to predict the resulting additional splittings. This is not a minor correction but a qualitative change in modal structure. It also serves as an objective rebuttal to the idea that evanescent coupling in these devices can always be summarized by a single scalar $r_s$03.

The stated design workflow proceeds in three stages. First, unit-cell Bloch analysis is used to extract the stop-band structure and the attenuation constant $r_s$04. Second, full-device FEM checks finite-size effects, including link-mode proximity and hybridization. Third, optical readout experiments validate the predicted splitting law. The summary explicitly characterizes this sequence as a “predictive, lithography-only approach” for building MHz-band mechanical circuits with controllable coupling and integrated optical access.

Within the scope of the reported work, complex-band-engineered mechanical links therefore denote a methodology as much as a component: the interconnect is designed by complex Bloch theory, calibrated by full-system simulation, and verified spectroscopically through optically bright flexural resonances. The resulting picture is one of mechanically coupled nanocavities in which confinement, readout, and coupling are co-designed rather than appended as separate subsystems.