Aluminum scandium nitride, commonly written as AlScN or ScAlN, is emerging as an important semiconductor thin-film material for next-generation radio-frequency filters, piezoelectric MEMS and nonvolatile memory devices.

By adding scandium to aluminum nitride, engineers can significantly increase piezoelectric response and electromechanical coupling. At suitable scandium concentrations, the same wurtzite material system can also exhibit switchable ferroelectric polarization.

These properties create an unusual opportunity: AlScN can potentially support both high-frequency acoustic devices and CMOS-compatible ferroelectric memory. However, the two applications do not require identical material specifications.

RF filters prioritize electromechanical coupling, acoustic quality factor, thickness uniformity and crystallographic orientation. Ferroelectric memory places more emphasis on remanent polarization, coercive field, leakage current, endurance and retention.

For both applications, scandium content, residual film stress, crystal polarity, surface roughness and wafer-scale uniformity must be controlled carefully.

What Is an AlScN-on-Silicon Wafer?

An AlScN-on-silicon wafer normally consists of a thin aluminum scandium nitride film deposited or grown on a silicon substrate. It is not generally a bulk AlScN crystal wafer.

The basic structure may include:

• Silicon handle wafer
• Thermal oxide or dielectric isolation layer
• Seed or buffer layer
• Bottom electrode
• AlScN functional layer
• Optional top electrode or passivation layer

The exact stack depends on the intended device.

For an acoustic resonator, the AlScN layer may be combined with metal electrodes, acoustic reflectors, cavities or released membranes. For ferroelectric memory, it is usually integrated between conductive electrodes as a metal–ferroelectric–metal or metal–ferroelectric–semiconductor structure.

Reactive magnetron sputtering is currently one of the most practical deposition methods because it supports relatively low processing temperatures, large-area coating and compatibility with semiconductor manufacturing. Molecular beam epitaxy and metal-organic chemical vapor deposition are also being explored where higher crystallographic quality or epitaxial growth is required.

Recent research has demonstrated molecular-beam epitaxy of AlScN directly on Si(111), supporting its potential for RF filters, memory, photonics and MEMS applications.

Why Add Scandium to Aluminum Nitride?

Pure aluminum nitride has a stable wurtzite crystal structure, high thermal stability, relatively high acoustic velocity and established use in bulk acoustic wave filters.

However, its piezoelectric response and electromechanical coupling can limit the bandwidth achievable in advanced RF filters.

Substituting a controlled fraction of aluminum atoms with scandium changes the crystal lattice. The structure becomes more responsive to an applied electric field, increasing piezoelectric coefficients and electromechanical coupling.

This can enable:

• Wider RF filter bandwidth
• Stronger electromechanical conversion
• Improved actuator displacement
• Higher sensor sensitivity
• Ferroelectric polarization switching
• New tunable and reconfigurable device concepts

The material is generally expressed as Al₁₋ₓScₓN, where “x” represents the scandium atomic fraction on the metal sublattice.

Increasing scandium content does not produce unlimited improvement. Higher scandium concentration can reduce elastic stiffness, increase film stress sensitivity and make it more difficult to maintain a highly oriented wurtzite structure.

If the composition or deposition process is not controlled properly, the film may develop:

• Abnormally oriented grains
• Increased surface roughness
• Secondary or rock-salt-like phases
• Higher leakage current
• Local composition variations
• Cracks or delamination
• Reduced acoustic quality factor

The appropriate scandium concentration must therefore be selected according to the application rather than specified as “as high as possible.”

Scandium Content for RF Filters

RF acoustic filters rely on the conversion between electrical and mechanical energy. A higher electromechanical coupling coefficient can support wider operating bandwidth and more flexible filter design.

For RF applications, AlScN films often use moderate scandium concentrations that improve piezoelectric performance while preserving crystal quality, acoustic velocity and acceptable loss.

The optimum composition depends on:

• Required center frequency
• Filter bandwidth
• Resonator mode
• AlScN film thickness
• Electrode material and thickness
• Acoustic stack design
• Target quality factor
• Deposition technology

Higher scandium content may increase piezoelectric coupling, but it can also lower acoustic velocity and increase structural disorder. This means a film with the highest piezoelectric coefficient may not produce the best complete RF resonator.

The final performance must be evaluated through device-level measurements such as:

• Resonant and anti-resonant frequencies
• Effective electromechanical coupling
• Quality factor
• Insertion loss
• Return loss
• Temperature coefficient of frequency
• Power handling
• Frequency drift

AlScN is especially attractive for wideband bulk acoustic wave filters, film bulk acoustic resonators and high-frequency resonators being developed for advanced 5G, Wi-Fi and future 6G systems.

Research published in 2025 and 2026 continues to investigate polarization-inverted AlScN layers and advanced acoustic modes for filters operating above 5 GHz.

Scandium Content for Ferroelectric Memory

Ferroelectric AlScN uses a switchable polarization state to represent stored information. Unlike conventional piezoelectric operation, the polarization direction is intentionally reversed by an applied electric field.

Ferroelectric behavior generally becomes more accessible as scandium content increases and the energy barrier between polarization states decreases.

Important memory parameters include:

• Remanent polarization
• Coercive electric field
• Leakage current
• Breakdown field
• Switching speed
• Write endurance
• Data retention
• Wake-up behavior
• Imprint
• Device-to-device variation

Higher scandium content can reduce the field required for switching, but it may also make phase stability, leakage and film quality more difficult to control.

Recent AlScN memory research has demonstrated write endurance exceeding (10^{10}) cycles in sub-50-nanometer films through controlled partial polarization switching. This represents significant progress toward addressing one of the major reliability limitations of wurtzite ferroelectrics.

The best scandium concentration for a memory device depends on film thickness, electrode selection, deposition method and maximum allowable operating voltage. A composition optimized for an RF resonator should not automatically be assumed to be suitable for ferroelectric memory.

RF Filters and Memory Require Different Specifications

A supplier should know the intended device application before recommending an AlScN-on-silicon wafer structure.

Film Stress and Wafer Bow

Residual stress is one of the most difficult parameters to control in AlScN thin films.

Stress can originate from:

• Lattice mismatch
• Thermal expansion mismatch
• Ion bombardment during sputtering
• Deposition temperature
• Gas pressure and gas ratio
• Target power
• Substrate bias
• Scandium concentration
• Film thickness
• Electrode and seed-layer properties

Excessive tensile stress can cause cracking, especially in thicker films or released MEMS structures. Excessive compressive stress can produce wafer bow, buckling or delamination.

Stress also affects the internal crystal structure. It can change lattice parameters, polarization behavior, coercive field and acoustic velocity.

For RF filters, stress variation across the wafer can shift resonant frequency and reduce device matching. It may also deform released resonator membranes.

For ferroelectric memory, stress can change the energy landscape for polarization switching. Local stress variation may lead to differences in coercive voltage, remanent polarization and endurance between devices.

A 2026 study of 200 mm wafer-scale wurtzite ferroelectric films demonstrated the use of RF substrate bias to tune AlScN film stress, showing how deposition conditions can be adjusted to improve large-area integration.

Important stress-related measurements include:

• Average film stress
• Stress uniformity
• Wafer bow before deposition
• Wafer bow after deposition
• Temperature-dependent curvature
• Local stress near the wafer edge
• Stress after annealing
• Stress after electrode deposition

Stress should always be reported together with film thickness, substrate thickness and measurement temperature.

Crystal Orientation vs Crystal Polarity

Crystal orientation and crystal polarity are related but different specifications.

Orientation describes how the crystallographic axes are aligned relative to the substrate. For most AlScN piezoelectric devices, a strong c-axis orientation perpendicular to the wafer surface is desirable.

Polarity describes the direction of the crystal’s polar axis. Wurtzite nitride films may be metal-polar or nitrogen-polar, depending on which atomic termination and polarization direction are present.

Two films can have similar c-axis orientation while having different polarity or mixed-polarity regions.

For RF resonators, uniform polarity helps ensure consistent piezoelectric response. If regions of opposite polarity are mixed unintentionally, their electromechanical responses may partially cancel.

Controlled polarity inversion can also be useful. Alternating-polarity AlScN multilayers are being investigated for high-order acoustic modes and filters operating at higher frequencies.

For ferroelectric memory, polarization switching is the fundamental operating mechanism. Important polarity-related issues include:

• Initial polarization direction
• Polarization uniformity
• Domain nucleation
• Complete vs partial switching
• Built-in electric field
• Imprint toward one polarization state
• Mixed-polarity grains
• Stability after repeated switching

Film polarity can be influenced by seed-layer material, bottom-electrode texture, deposition sequence, plasma conditions and substrate bias.

Importance of the Seed Layer and Bottom Electrode

AlScN does not operate independently of the layers beneath it. The seed layer and bottom electrode strongly influence nucleation, grain orientation, surface roughness and residual stress.

Frequently considered electrode and seed materials include:

• Molybdenum
• Titanium nitride
• Tungsten
• Platinum
• Ruthenium
• Aluminum nitride
• Other conductive nitride or refractory metal layers

The preferred stack depends on the application.

RF resonators require electrodes with suitable conductivity, acoustic impedance, surface texture and low loss. Ferroelectric memory requires electrodes that can withstand switching fields while controlling interface reactions, leakage and imprint.

The bottom layer should provide:

• Smooth surface morphology
• Uniform nucleation
• Strong c-axis texture
• Stable chemical interface
• Appropriate thermal-expansion behavior
• Compatible processing temperature
• Low particle density

A rough or poorly oriented electrode can produce abnormal AlScN grains even when the deposition parameters are otherwise well controlled.

Surface Roughness and Abnormally Oriented Grains

AlScN film roughness typically becomes more difficult to control as scandium content increases.

Abnormally oriented grains are especially important because they can protrude from the film surface and disturb the local electric or acoustic field.

In an RF resonator, these grains may increase acoustic scattering, reduce quality factor and create thickness nonuniformity beneath the top electrode.

In a memory capacitor, surface protrusions can create electric-field concentration, increasing leakage and the risk of premature breakdown.

Useful inspection methods include:

• Atomic force microscopy
• Scanning electron microscopy
• Optical defect inspection
• X-ray diffraction
• Transmission electron microscopy
• Surface profilometry
• Film-thickness mapping

An average roughness value alone may not reveal isolated protrusions. Defect density and maximum feature height should be considered together with RMS or Ra roughness.

Selecting the Silicon Substrate

Silicon is attractive because it provides mature wafer manufacturing, established supply chains and compatibility with CMOS and MEMS processing.

However, silicon wafers are not interchangeable.

Important substrate parameters include:

• Diameter
• Crystal orientation
• Resistivity
• Conductivity type
• Thickness
• TTV
• Bow and warp
• Front-side roughness
• Backside finish
• Thermal oxide thickness
• Particle and metal contamination
• Edge profile

Si(100) is widely used in CMOS manufacturing, while Si(111) may be selected for certain epitaxial AlScN structures because of its surface symmetry and growth relationship.

High-resistivity silicon may be preferred for RF devices to reduce substrate-related RF losses. Standard low-resistivity silicon may be acceptable where the resonator is acoustically and electrically isolated through cavities, dielectric layers or reflector structures.

For ferroelectric memory, substrate selection depends more heavily on the target integration route. The AlScN stack may be placed above completed CMOS circuits, integrated near the transistor level or built on a dedicated test wafer.

Wafer-Scale Uniformity Requirements

A film that performs well at the wafer center may still be unsuitable for production if its properties change significantly toward the edge.

The following parameters should be mapped across the usable wafer area:

• AlScN thickness
• Scandium concentration
• Film stress
• Crystal orientation
• Rocking-curve width
• Surface roughness
• Sheet resistance of electrodes
• Dielectric constant
• Leakage current
• Piezoelectric coefficient
• Polarization
• Coercive field
• Particle and defect density

For RF devices, small thickness variations can cause resonant-frequency differences across the wafer. If a wafer contains many filters operating within a tightly defined frequency band, even modest nonuniformity may reduce the number of usable dies.

For memory, variations in thickness, composition or electrode quality can produce different switching voltages and leakage levels across the wafer.

A production-oriented specification should therefore include both average values and uniformity limits. Full-wafer maps are more informative than measurements from only the center and four edge points.

Deposition Methods for AlScN-on-Silicon

Reactive Sputtering

Reactive magnetron sputtering is widely used for AlN and AlScN thin films. It offers relatively high throughput, large-wafer compatibility and flexible control of alloy composition.

Key process variables include:

• Aluminum and scandium target configuration
• Target power
• Nitrogen-to-argon ratio
• Chamber pressure
• Substrate temperature
• Substrate bias
• Target-to-substrate distance
• Deposition rate
• Wafer rotation

Co-sputtering from separate aluminum and scandium targets provides composition flexibility, while alloy targets can simplify process control once the required composition has been established.

Molecular Beam Epitaxy

Molecular beam epitaxy offers precise control of composition and interfaces and is valuable for fundamental research and high-quality epitaxial films.

However, deposition rate, cost and large-volume manufacturing requirements must be considered.

Metal-Organic Chemical Vapor Deposition

MOCVD may support high-quality epitaxial nitride layers and established III-nitride manufacturing infrastructure. Challenges include scandium precursor chemistry, composition control and process scalability.

The correct method depends on whether the target product is a research-grade epitaxial wafer, a MEMS development substrate or a production-scale functional film.

Thermal Budget and CMOS Compatibility

One of AlScN’s advantages is the possibility of deposition at temperatures compatible with many semiconductor integration flows.

However, “CMOS compatible” is not a complete specification. The allowable temperature depends on whether deposition occurs before or after transistors, interconnects and low-k dielectric layers are formed.

The process must also control:

• Metal contamination
• Sodium and mobile-ion contamination
• Hydrogen exposure
• Plasma damage
• Electrode diffusion
• Thermal stress
• Interface reactions

For back-end-of-line integration, both peak temperature and total thermal exposure are important.

AlScN Wafer Procurement Checklist

Before ordering AlScN-on-silicon wafers, buyers should define:

1. Target application: RF, MEMS, memory or research
2. Silicon wafer diameter and orientation
3. Silicon resistivity and conductivity type
4. AlScN composition and scandium tolerance
5. AlScN film thickness and uniformity
6. Deposition method
7. Seed-layer material and thickness
8. Bottom-electrode material and texture
9. Crystal orientation and rocking-curve specification
10. Crystal polarity requirements
11. Residual stress range and uniformity
12. Wafer bow after deposition
13. Surface roughness and abnormal-grain limits
14. Particle and metallic-contamination limits
15. Electrical, piezoelectric or ferroelectric test data
16. Edge exclusion and usable wafer area
17. Thermal budget and annealing history
18. Wafer mapping and certificate-of-analysis requirements

For development programs, it is advisable to begin with a small qualification batch and correlate wafer data with finished-device performance.

Frequently Asked Questions

Is AlScN the same as ScAlN?

Both abbreviations usually refer to scandium-alloyed aluminum nitride. AlScN emphasizes aluminum nitride as the base material, while ScAlN emphasizes scandium addition. The actual alloy composition should always be defined as Al₁₋ₓScₓN.

Is higher scandium content always better?

No. Higher scandium content can increase piezoelectric response and promote ferroelectric switching, but it may also increase roughness, structural instability, leakage and process sensitivity.

Can the same AlScN wafer be used for RF filters and memory?

It may be useful for early material research, but optimized device wafers normally require different compositions, thicknesses, electrodes and electrical specifications.

Why is film stress important?

Stress affects wafer bow, cracking, crystal orientation, acoustic properties and ferroelectric switching. Stress variation can also reduce wafer-level device uniformity.

Which silicon orientation is preferred?

Si(100) is common for CMOS-compatible processing, while Si(111) may be selected for certain epitaxial structures. The choice depends on the deposition method and device architecture.

What limits the commercial use of AlScN?

Major challenges include composition control, high coercive field, leakage, film stress, abnormal grains, thickness uniformity and reliable large-wafer production.

Conclusion

AlScN-on-silicon wafers provide a versatile material platform for wideband RF filters, high-frequency acoustic resonators, piezoelectric MEMS and ferroelectric memory.

Their performance depends on more than scandium concentration. Film stress, crystal orientation, polarity, electrode texture, surface roughness and wafer-scale uniformity must be engineered as a complete system.

RF applications require a balance between electromechanical coupling, acoustic velocity and low loss. Ferroelectric memory requires controlled switching, low leakage, high endurance and stable retention. These different priorities must be reflected in the wafer specification.

As AlScN technology progresses from laboratory devices toward 200 mm and potentially larger-scale manufacturing, uniform deposition and reliable wafer-level characterization will become increasingly important. Buyers should therefore select AlScN-on-silicon wafers according to the final device architecture and request mapped material data rather than relying on a single nominal composition or center-point measurement.