Aluminum Nitride EO Transducers

• Aluminum nitride electro-optic transducers are devices that convert electrical signals into optical responses via inherent Pockels and piezoelectric effects.
• They utilize c-axis oriented AlN and related alloys or heterostructures to achieve efficient modulation across UV, visible, and infrared wavelengths with minimal free-carrier interference.
• Recent device architectures, including microring resonators and Mach–Zehnder interferometers, optimize modulation bandwidth, energy efficiency, and integration for telecommunications and quantum applications.

Aluminum nitride electro-optic transducers are devices in which an applied electrical signal is converted into an optical phase, resonance, intensity, or sideband response using aluminum nitride (AlN) or closely related AlN-family materials. In the narrow photonics-device sense, the canonical implementations are carrier-free Pockels modulators based on c-axis-oriented AlN or scandium-doped AlN waveguides and resonators. In the broader transducer sense used across the literature, the category also includes piezo-acousto-optic and electro-optomechanical devices in which AlN mediates electrical-to-mechanical-to-optical conversion on an integrated chip (Xiong et al., 2014, Xu et al., 2024, Tadesse et al., 2015, Zhou et al., 4 Aug 2025).

1. Material system and electro-optic tensor physics

AlN is treated throughout the literature as a non-centrosymmetric, wide-bandgap photonic material with intrinsic second-order nonlinearity, a usable Pockels effect, and strong piezoelectricity. The bandgap is quoted as $6.2\,\mathrm{eV}$, which underlies the repeated emphasis on suppression of two-photon absorption, suppression of free-carrier effects, and operation from ultraviolet to infrared (Xiong et al., 2014, Xiong et al., 2012). In sputtered integrated films, the decisive structural feature is strong c-axis orientation normal to the film plane. This orientation determines which tensor components are practically accessible and explains the near-universal preference for an out-of-plane electric field $E_z$.

For c-axis-oriented AlN, the nonzero electro-optic coefficients emphasized in the device literature are $r_{13}$, $r_{33}$, and $r_{51}$, with $r_{13}$ and $r_{33}$ typically quoted as $\sim 1\ \mathrm{pm/V}$. In the ring-modulator formulation used for AlN-on-insulator devices, the index perturbations under vertical drive are written as

$n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$

and

$n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$

which is why top-electrode geometries are engineered to maximize $E_z$0 rather than an in-plane field (Xiong et al., 2014). The same directional requirement reappears in AlScN microring modulators, where the index ellipsoid is explicitly specialized to $E_z$1, $E_z$2, giving

$E_z$3

In those devices, $E_z$4 corresponds to TE-polarized modes and $E_z$5 to TM modes (Xu et al., 2024).

Scandium alloying is treated as the most direct AlN-family route to stronger electro-optic behavior. One integrated AlScN study uses prior second-harmonic data to estimate that, for $E_z$6, $E_z$7 would imply $E_z$8, roughly $E_z$9 intrinsic AlN if the same enhancement translated directly to the electro-optic effect. The same study notes that increasing Sc concentration enhances non-centrosymmetric properties until about $r_{13}$0 Sc, above which the crystal begins transitioning away from wurtzite toward cubic symmetry and loses the desired nonlinear and piezoelectric response (Yoshioka et al., 2024). A separate AlScN microring work at $r_{13}$1 Sc likewise frames the material as a wurtzite-derived AlN extension with enhanced second-order optical nonlinearity and enhanced Pockels response while retaining a relatively large bandgap and CMOS-compatible sputter processing (Xu et al., 2024).

The AlN family is also being extended laterally rather than only by alloying with Sc. AlGaN/AlN heterostructures are presented as an emerging III-nitride platform with large nonlinear coefficients, high electro-optic modulation capabilities, and refractive-index engineering through Al composition. That paper does not demonstrate direct electro-optic modulation, but it explicitly cites prior AlGaN/AlN multiple quantum wells with a $r_{13}$2-fold increase in second-order susceptibility compared to bare AlN, which suggests a route to stronger nitride-family electro-optic transducers through heterostructure engineering rather than only bulk-film substitution (Gündogdu et al., 2023).

2. Integrated material platforms and optical infrastructure

The early integrated AlN electro-optic literature established two closely related thin-film platforms: AlN-on-insulator on silicon for telecom and visible integrated photonics, and suspended AlN structures for optomechanics. In the AlN-on-insulator route, bare $r_{13}$3 silicon wafers with $r_{13}$4 thermally grown $r_{13}$5 support sputter-deposited AlN thin films of $r_{13}$6 thickness for telecom and $r_{13}$7 for visible devices. These films are polycrystalline but strongly c-axis oriented, with AlN (0002) rocking-curve full width at half maximum reported as less than $r_{13}$8 and specifically $r_{13}$9 in one figure inset. On this platform, propagation loss of $r_{33}$0 and undercoupled ring quality factor $r_{33}$1 were reported at telecom wavelengths, while a closely related AlN-on-silicon platform reported waveguide loss as low as $r_{33}$2 and microring $r_{33}$3 under weak coupling (Xiong et al., 2014, Xiong et al., 2012).

These optical baselines matter because most AlN electro-optic transducers are resonance-enhanced. The same AlN-on-insulator work showed that the resonance-enhanced modulator bandwidth is often set by photon lifetime rather than electrode RC, using

$r_{33}$4

For a telecom ring with $r_{33}$5, $r_{33}$6 and $r_{33}$7, in good agreement with the measured $r_{33}$8 electrical bandwidth (Xiong et al., 2014). This cavity-lifetime limit recurs in later AlN microring modulation work (Xiong et al., 2012).

Single-crystalline epitaxial AlN on sapphire provides a different optical infrastructure. A $r_{33}$9-thick MOCVD AlN film on c-plane sapphire supported fully etched ring resonators with extracted $r_{51}$0, loaded $r_{51}$1 for the infrared mode and $r_{51}$2 for the visible mode in a dually resonant second-harmonic platform (Bruch et al., 2018). That work is not a microwave or direct electro-optic transducer demonstration, but it is directly relevant to AlN electro-optic transducers because it establishes that epitaxial AlN can simultaneously provide high optical $r_{51}$3, low visible/IR loss, and strong usable $r_{51}$4, which are the optical-side prerequisites for cavity electro-optic conversion.

AlScN extends the thin-film infrastructure in two experimentally distinct directions. One study used a $r_{51}$5 $r_{51}$6 film on $r_{51}$7 sapphire with an $r_{51}$8-Si strip-loaded waveguide and top coplanar electrodes, reporting intrinsic AlScN film loss $r_{51}$9 and device propagation loss $r_{13}$0 (Yoshioka et al., 2024). Another used a $r_{13}$1 sputtered $r_{13}$2 film on thermally grown $r_{13}$3 over silicon, with fully etched waveguides, $r_{13}$4 PECVD $r_{13}$5 cladding, and waveguide loss $r_{13}$6 (Xu et al., 2024). The two demonstrations make clear that the AlScN question is not only whether Sc increases intrinsic nonlinearity, but also how growth method, substrate, optical confinement, roughness, and electrode geometry reshape the realized device response.

3. Canonical device architectures

The direct Pockels branch of AlN electro-optic transducers is dominated by resonators and interferometers. In telecom AlN-on-insulator, the foundational active device is the microring resonator with top ground-signal-ground electrodes above an oxide cladding. The applied field shifts the ring resonance, and a laser biased on the resonance slope converts the phase perturbation into transmitted intensity modulation. This architecture was demonstrated in the telecom and visible bands, including visible modulation near $r_{13}$7 and digital modulation up to $r_{13}$8 at telecom (Xiong et al., 2014). A closely related AlN microring platform was also used to control a Kerr microcomb: the applied voltage linearly tuned the cavity resonance and reversibly switched comb generation on and off, thereby realizing a resonantly enhanced AlN electro-optically controlled photonic element rather than a stand-alone phase shifter (Jung et al., 2013).

Interferometric AlN-family transducers occupy a complementary space. The AlScN Mach–Zehnder interferometer demonstrated on sapphire is a strip-loaded, $r_{13}$9-Si-on-AlScN phase modulator designed for the fundamental TM$r_{33}$0 mode so that the optical field addresses the extraordinary axis and the expected $r_{33}$1-dominated response. Its layer stack comprises sapphire, co-sputtered AlScN, etched intrinsic amorphous silicon, PECVD $r_{33}$2, and Ti/Au electrodes, with a measured AlScN thickness of $r_{33}$3, waveguide width $r_{33}$4, waveguide height $r_{33}$5, oxide thickness $r_{33}$6, and electrode gap $r_{33}$7 (Yoshioka et al., 2024). The device is explicitly framed as an integrated electro-optic modulator and, in the broader sense, an electrical-to-optical transducer.

At shorter wavelengths, AlN is being pushed into ultraviolet photonic integrated circuits. A design-and-simulation study at $r_{33}$8 for $r_{33}$9 systems analyzes a straight-waveguide phase modulator and a $\sim 1\ \mathrm{pm/V}$0 Mach–Zehnder switch in c-axis AlN, both using coplanar electrodes to generate a vertical field component $\sim 1\ \mathrm{pm/V}$1. The phase modulator uses a ground-signal-ground line over a $\sim 1\ \mathrm{pm/V}$2 interaction length, while the switch uses a push-pull ground-signal-ground-signal-ground configuration and an unbalanced MZI biased around the quadrature point (Icli et al., 24 Mar 2025). Although that work is not experimental, it is important because it translates the AlN design rules established at telecom into the ultraviolet.

A brief comparison of representative direct electro-optic architectures is useful because the field contains both experimentally successful and deliberately cautionary demonstrations.

Architecture	Representative reported metric	Source
AlN microring resonator	$\sim 1\ \mathrm{pm/V}$3, $\sim 1\ \mathrm{pm/V}$4, down to $\sim 1\ \mathrm{pm/V}$5	(Xiong et al., 2014)
AlN comb-control microring	$\sim 1\ \mathrm{pm/V}$6, $\sim 1\ \mathrm{pm/V}$7, $\sim 1\ \mathrm{pm/V}$8 switching	(Jung et al., 2013)
AlScN MZI	$\sim 1\ \mathrm{pm/V}$9	(Yoshioka et al., 2024)
AlScN microring	$n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$0 at $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$1, $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$2, bandwidth $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$3	(Xu et al., 2024)
AlN $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$4 switch design	$n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$5 for TE	(Icli et al., 24 Mar 2025)

This range shows that “AlN electro-optic transducer” is not a single device class. It spans passive-compatible low-energy microrings, resonance-controlled nonlinear photonic elements, interferometric phase modulators, and wavelength-specific designs for ultraviolet control hardware.

4. Performance landscape and governing tradeoffs

The mature baseline for direct AlN modulation is the microring platform on silicon dioxide on silicon. In one telecom implementation, a ring resonance near $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$6 with extinction ratio $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$7 and $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$8 shifted by $n_{x,y} = n_o - \frac{1}{2} r_{13} n_o^3 \cdot E_z$9 over $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$0 to $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$1, corresponding to a tuning efficiency of roughly $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$2. The same platform demonstrated $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$3 NRZ, $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$4 PRBS, and $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$5 PRBS operation, with measured $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$6 extinction ratio at $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$7 and up to $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$8, and an estimated capacitive energy down to $n_z = n_e - \frac{1}{2} r_{33} n_e^3 \cdot E_z,$9 using $E_z$00 with $E_z$01 (Xiong et al., 2014). A related AlN ring-modulator study reported a $E_z$02 resonance shift from $E_z$03 to $E_z$04, measured $E_z$05 EO bandwidths of $E_z$06 for $E_z$07, $E_z$08 for $E_z$09, and $E_z$10 for $E_z$11, and explicitly identified cavity photon lifetime rather than RC as the dominant speed limit (Xiong et al., 2012).

The AlN microring can also be used as a resonance-controlled nonlinear switch rather than only as a linear intensity modulator. In the optical-frequency-comb work, the resonance tuning is linear with applied voltage, with measured efficiency $E_z$12 or $E_z$13, and $E_z$14, $E_z$15 pulses reversibly switch the comb off in about $E_z$16. The same work gives a photon lifetime $E_z$17 and a slower thermal component with $E_z$18 time constant, thereby making a general point: in resonant AlN transducers, the intrinsic electro-optic response can be faster than the observed thermal settling (Jung et al., 2013).

Recent AlScN results split sharply into two experimental regimes. The negative result is the integrated $E_z$19 Mach–Zehnder modulator. That device reported $E_z$20 and $E_z$21 from DC measurements, $E_z$22 and $E_z$23 from lock-in measurements at $E_z$24, and $E_z$25 on a second device at $E_z$26. The measured response was about $E_z$27 smaller than the AlScN-enhanced expectation $E_z$28, about $E_z$29 smaller than an intrinsic-AlN-like simulation $E_z$30, and smaller than a cited prior AlN TM-mode value $E_z$31 (Yoshioka et al., 2024).

The positive result is the silicon-integrated $E_z$32 microring platform. There, a DC bias from $E_z$33 to $E_z$34 shifted the resonance by $E_z$35, with fitted slope $E_z$36, corresponding to an in-device effective EO coefficient of $E_z$37 under DC conditions. High-frequency optical-sideband measurements then extracted $E_z$38 for one $E_z$39-radius all-pass ring and $E_z$40 for a $E_z$41-radius add-drop ring, with the maximum $E_z$42 reached at $E_z$43. The same study reported minimum RF $E_z$44 at $E_z$45 and a $E_z$46 EO bandwidth of approximately $E_z$47 (Xu et al., 2024).

These two 2024 AlScN results define an important controversy. One might expect that enhanced second-harmonic coefficients automatically imply enhanced Pockels modulation. The low-frequency AlScN MZI result explicitly challenges that expectation, while the high-frequency AlScN ring result supports a more optimistic view for a different film composition, substrate, and resonant architecture. The literature therefore does not support a single generic statement such as “AlScN is simply better AlN for electro-optic modulators.” A more precise reading is that the realized response depends strongly on tensor component access, optical/electrical overlap, material loss, and possibly intrinsic dispersion or phonon-related contributions (Yoshioka et al., 2024, Xu et al., 2024).

The ultraviolet design study reinforces this architecture dependence. At $E_z$48, the simulated TE phase modulator gives DC $E_z$49, whereas the push-pull Mach–Zehnder switch gives $E_z$50. The paper explicitly attributes the improvement to push-pull operation, quadrature-point operation, and stronger interferometric conversion of phase shift into power transfer, rather than to a change in the underlying AlN material coefficient (Icli et al., 24 Mar 2025).

5. Beyond direct Pockels modulation: piezo-acousto-optic and microwave-optical transduction

A second major branch of AlN electro-optic transducers uses the material’s piezoelectricity and mechanics rather than only its direct Pockels response. In a suspended $E_z$51 AlN membrane, an interdigital transducer launches Lamb waves that modulate a photonic crystal nanobeam cavity mainly through the elasto-optic effect. This piezo-acousto-optic device reached $E_z$52 in the microwave K band using a $E_z$53-period IDT, with stronger modulation generally obtained for the symmetric $E_z$54 mode than for antisymmetric modes because the strain symmetry improves overlap with the optical field (Tadesse et al., 2015). Although this is not direct Pockels modulation, it belongs to the broader transducer category because the chain is electrical microwave drive $E_z$55 piezoelectric actuation $E_z$56 Lamb wave $E_z$57 optical resonance modulation.

The suspended AlN optomechanics literature established the optical-mechanical half of this broader transducer picture before direct electrical conversion was realized. A suspended AlN ring resonator with $E_z$58 film thickness, $E_z$59 ring width, and spokes engineered for reduced anchor loss exhibited loaded optical $E_z$60 up to $E_z$61 and optomechanical transduction of a $E_z$62 contour mode with displacement sensitivity $E_z$63 in ambient air (Xiong et al., 2012). That work did not demonstrate electrical drive, but it explicitly framed AlN resonators as a basis for tunable, electrically driven, optically sensed oscillator systems.

The piezo-optomechanical route was then formalized theoretically as an AlN microwave-to-optical transducer. In that proposal, a microwave cavity mode couples directly to an AlN thickness mode via the piezoelectric effect, and the same mechanical mode couples to an optical cavity mode through optomechanics. The linearized tripartite Hamiltonian is written as

$E_z$64

and the optimized conversion analysis projects efficiencies near $E_z$65 once the piezomechanical coupling exceeds a few MHz, with performance then limited mainly by intrinsic optical loss rather than piezoelectric coupling (Zou et al., 2016). This proposal is not a direct electro-optic Pockels transducer, but it is central to the AlN transducer literature because it identifies a physically distinct route to high-efficiency microwave-optical conversion on an AlN chip.

The most system-level realization in the dataset is the coherent link between two dilution refrigerators. That work uses a pair of frequency-matched AlN cavity electro-optic transducers connected by $E_z$66 of telecom fiber. Each transducer achieves $E_z$67 on-chip efficiency, the optical link uses $E_z$68 fiber, and the pairwise transduction loss represents an overall $E_z$69 improvement over using two commercial electro-optic modulators for the same fridge-to-fridge link. The devices employ a double-ring AlN photonic molecule and a Nb microwave resonator, satisfy the matching condition $E_z$70, and demonstrate coherent microwave-signal transfer through a $E_z$71 fiber channel (Zhou et al., 4 Aug 2025). Here the term “AlN electro-optic transducer” has fully expanded from a photonic modulator element into a network component for superconducting-circuit interconnects.

6. Limitations, misconceptions, and current research directions

One persistent misconception is that AlN electro-optic transducers are equivalent to direct Pockels modulators. The literature is broader. Direct ring and interferometric Pockels devices are the canonical integrated implementations, but AlN also supports piezo-acousto-optic modulation in Lamb-wave membranes, optomechanical readout of GHz contour modes, and cavity electro-optic microwave-optical conversion in cryogenic systems (Tadesse et al., 2015, Xiong et al., 2012, Zhou et al., 4 Aug 2025). A second misconception is that bandwidth is automatically RC-limited. In resonance-enhanced AlN modulators, the dominant limit is often the optical photon lifetime, and the measured $E_z$72 bandwidth tracks cavity $E_z$73 rather than electrode capacitance (Xiong et al., 2014).

The most important materials controversy concerns Sc doping. One AlScN paper is explicitly cautionary: despite prior evidence of enhanced $E_z$74, its integrated MZI measured $E_z$75, and the authors discuss fabrication imperfections, higher dielectric constant, possible oxidation, tensor-component cancellation, dispersion, phonon resonances, ionic contributions, piezoelectric strain, and charge screening as possible reasons for the unexpectedly weak response (Yoshioka et al., 2024). Another AlScN paper, however, reports $E_z$76 at $E_z$77, $E_z$78, and $E_z$79 bandwidth on a silicon-integrated microring platform (Xu et al., 2024). The two results do not invalidate one another; they show that the AlScN problem is highly geometry- and process-dependent.

Current research directions therefore focus as much on architecture and interfaces as on the material coefficient itself. The AlScN MZI study points to directly etched AlScN waveguides, lower-loss films, better substrate and electrode geometries, and systematic tensor characterization versus Sc concentration as necessary next steps (Yoshioka et al., 2024). The ultraviolet AlN design study emphasizes RF/optical co-optimization, especially $E_z$80 matching, velocity matching, and the tradeoff between stronger $E_z$81 and higher optical loss from nearby metal; it explicitly suggests capacitively loaded electrodes as a route toward better RF-optical index matching (Icli et al., 24 Mar 2025). The AlGaN/AlN platform paper suggests a complementary direction: reducing interface-related loss by moving the optical mode into an AlGaN guiding layer above AlN, with TE propagation loss $E_z$82 and intrinsic ring $E_z$83 at $E_z$84, but without yet demonstrating a direct EO device (Gündogdu et al., 2023).

At the system scale, the cryogenic telecom-link demonstration identifies the remaining bottlenecks with unusual clarity. Frequency matching across multiple cavity electro-optic transducers is achievable through asymmetric photonic molecules and DC tuning, but higher efficiency, lower optical-induced microwave noise, better microwave $E_z$85 under illumination, and lower fiber-chip coupling loss remain essential before a fully quantum-enabled link is possible (Zhou et al., 4 Aug 2025). This suggests that the field has moved beyond the question of whether AlN can modulate light. The active questions are now how to reconcile optical $E_z$86, electrical overlap, microwave loss, spectral alignment, and fabrication compatibility across direct modulators, electro-acousto-optic devices, and networked microwave-optical transducers.