Lead-free piezoelectric film hits record performance — on plain silicon wafers

Research conducted at Osaka Metropolitan University, Graduate School of Engineering. Lead author: Takeshi Yoshimura, Associate Professor.

A lead-free piezoelectric thin film deposited on an ordinary silicon wafer has just set the performance record for its material class, delivering five times the energy-conversion efficiency of previous bismuth ferrite devices. The film, engineered by a team at Osaka Metropolitan University, does something the field has chased for years: it matches the output that once required toxic lead-based ceramics, and it does so on the same substrate the semiconductor industry already uses by the millions.

That last detail matters more than the record itself. A material that works only on exotic substrates is a laboratory curiosity. A material that works on silicon is a product waiting to happen.

The lead problem no one talks about

Piezoelectric materials convert mechanical stress into electric charge and vice versa. They are everywhere: microphones, inkjet printer heads, ultrasound probes, parking sensors. The best-performing piezoelectrics almost all contain lead zirconate titanate (PZT), a ceramic whose dominance has persisted for decades despite rising concern over lead's environmental and health toll. The European Union's Restriction of Hazardous Substances directive grants PZT a rolling exemption precisely because no lead-free alternative has matched its output in real devices.

Bismuth ferrite has long been the most promising candidate to break that deadlock. Its theoretical piezoelectric coefficients rival PZT's. In practice, though, bismuth ferrite films leak current like a sieve, and their actual energy-harvesting numbers have lagged far behind the theory. Growing them on silicon — the only substrate that matters for mass manufacturing — made things worse, because the thermal mismatch between the two materials pulls the film into tensile strain during cooling. Tensile strain, in conventional wisdom, degrades piezoelectric response.

Turning the wrong strain into the right phase

Takeshi Yoshimura's group at Osaka Metropolitan University decided to stop fighting the tensile strain and start exploiting it. Their hypothesis: if they could push bismuth ferrite past a critical strain threshold on silicon, they might trigger a structural phase transition from the material's native rhombohedral crystal arrangement into a monoclinic phase — a geometry whose atomic symmetry is far more favorable for piezoelectric output.

The challenge was control. Bismuth has a low melting point, so even small temperature swings during deposition change the film's composition. To map the landscape efficiently, the team developed what they call biaxial combinatorial sputtering. Instead of depositing one film per wafer under fixed conditions, they designed a setup in which growth temperature and chemical composition vary continuously across a single wafer's surface. One deposition run yields dozens of distinct growth conditions, collapsing weeks of trial-and-error into a single experiment.

Sputtering itself is no exotic technique — it is a workhorse of chip fabrication, used to lay down metal and dielectric layers in everything from smartphone processors to MEMS accelerometers. The fact that Yoshimura's films require no departure from that toolset is part of the point.

What the numbers actually show

By screening the combinatorial wafers, the team pinpointed the narrow window of temperature and manganese doping at which tensile strain on silicon triggers the rhombohedral-to-monoclinic transition. Films grown in that window recorded the highest piezoelectric response ever measured for bismuth ferrite thin films — a metric that quantifies how much electric charge the material generates per unit of applied mechanical stress.

Manganese doping played a dual role. It suppressed the electrical leakage that has historically plagued bismuth ferrite, and it shifted the strain threshold for the phase transition into a range achievable on silicon. Without the dopant, the transition either doesn't occur or occurs at strains that crack the film.

To translate that laboratory measurement into an engineering figure of merit, the team fabricated microelectromechanical systems (MEMS) vibration energy harvesters — tiny cantilever structures that flex in response to ambient vibration and convert that motion into voltage. The harvesters showed a fivefold improvement in energy-conversion efficiency over earlier bismuth ferrite MEMS devices. They operated reliably under both continuous sinusoidal vibrations and sharp, impulsive shocks, the kind produced by motors, compressors, and industrial equipment.

From lab cantilever to factory floor

Vibration energy harvesting sits at a bottleneck in the Internet of Things. Billions of wireless sensors are projected for industrial monitoring, structural health assessment, and environmental sensing over the next decade. Most will need power, and most will be in places where running a wire or swapping a battery is expensive, dangerous, or impossible — think bridge girders, pipeline joints, turbine housings. A sensor that scavenges enough energy from the vibrations it is already measuring to run its own radio eliminates the battery entirely.

Lead-based harvesters can do this today, but scaling them raises regulatory and disposal headaches. A lead-free alternative fabricated on silicon with standard sputtering equipment sidesteps both problems. It slots into existing MEMS foundry workflows without retooling.

Still, there are gaps between a record-setting film and a commercial product. The team's published data demonstrate high piezoelectric coefficients and improved harvester output, but long-term fatigue behavior — how the film performs after millions or billions of strain cycles — has not yet been reported. Nor has the work been validated at wafer-scale uniformity, the kind of run-to-run consistency a foundry demands. These are engineering problems, not physics problems, but they are real.

The strain-engineering playbook

What may prove most consequential here is not the specific material but the method. Strain engineering — using the mechanical interaction between a film and its substrate to force a desired crystal phase — is not new in semiconductors. Strained silicon channels have boosted transistor speeds in commercial chips since the early 2000s. Applying the same logic to piezoelectrics on silicon is a natural extension, yet it has been underexplored because the conventional assumption was that tensile strain is always detrimental to piezoelectric response.

Yoshimura's results flip that assumption. They suggest a broader design principle: for certain compositions near a phase boundary, tensile strain can be a tool rather than a defect. Other lead-free piezoelectric families — potassium sodium niobate, barium titanate, sodium bismuth titanate — may harbor similar strain-induced transitions that have gone unnoticed because nobody was looking for them on silicon.

The combinatorial sputtering technique accelerates that search. Mapping an entire composition-temperature phase space on a single wafer is the kind of high-throughput approach that materials science has embraced in other domains but has been slow to adopt for piezoelectric thin films.

What comes next

Yoshimura has stated that the group's next targets include smart sensors and self-powered IoT devices — applications where microwatts of harvested power are enough to wake a microcontroller, take a reading, and transmit a packet before returning to sleep. If the fatigue and uniformity data hold up, licensing to a MEMS foundry is a plausible near-term path.

The broader stakes are environmental. Lead-based piezoelectrics have enjoyed their regulatory exemption partly because the alternatives were not competitive. A bismuth ferrite film that matches or approaches PZT performance on silicon weakens the case for that exemption — and strengthens the argument that the electronics industry can cut lead from one more product category without sacrificing function.