For more than a century, diamond occupied a settled place in materials science: extraordinarily hard, thermally conductive and electrically insulating, but not piezoelectric. Its crystal structure was simply too symmetrical to turn an applied force into a net electrical polarisation.
A team led by researchers at the University of Hong Kong has now shown that this rule has a carefully defined exception. Ultrathin polycrystalline diamond membranes produced repeatable voltage signals when bent, with the strongest response appearing in films about five micrometres thick.
The finding does not mean that an engagement ring can charge a phone when squeezed. The samples were synthetic diamond films grown from many microscopic crystals, released from a silicon substrate and made thin enough to flex. Their internal grain boundaries supply something that a perfect diamond crystal lacks: local asymmetry.
The peer-reviewed study in Science Advances reports a genuine electromechanical response in that engineered form of diamond. It also leaves a substantial distance between a laboratory membrane that produces millivolts and a useful commercial power source.
Piezoelectricity requires charge not to cancel out
The direct piezoelectric effect is the production of electrical charge when a material is mechanically stressed. Quartz, certain ceramics and materials such as zinc oxide can separate positive and negative charge centres slightly when compressed, stretched or bent.
That separation creates an electric polarisation and a voltage that electrodes can collect. The inverse effect also exists: an applied electric field can deform a piezoelectric material. The two-way conversion makes piezoelectrics useful in microphones, ultrasound transducers, precision actuators, vibration sensors and small energy harvesters.
Crystal symmetry determines whether the microscopic changes add up. In a material with a centre of inversion, movements on one side of the structure are mirrored on the other. Their electrical contributions cancel, leaving no net piezoelectric polarisation.
Ordinary cubic diamond has this centrosymmetric arrangement. Each carbon atom forms four strong bonds in a tetrahedral geometry, and a high-quality single crystal preserves the larger symmetry that makes its piezoelectric coefficient effectively zero.
The century-old classification was therefore not a simple experimental oversight. It correctly described ideal bulk diamond. The new result changes what counts as “diamond” in that statement by examining a very different microstructure.
The samples were flexible films grown in a laboratory
The researchers did not slice membranes from natural gemstones. They deposited diamond onto silicon using microwave plasma chemical vapour deposition, starting with diamond seeds smaller than 10 nanometres. The seeds expanded, joined and grew upward into a continuous polycrystalline film.
A previously developed edge-exposed exfoliation process allowed the film to be separated from the growth substrate. The broader fabrication work had already shown that large, ultraflat polycrystalline diamond membranes can be produced and transferred without destroying them.
Thinness changes the mechanics. Diamond remains hard, meaning it strongly resists local scratching and indentation, but a sheet only a few micrometres thick can bend because flexural stiffness falls sharply with thickness. Hardness and flexibility are not opposites when the geometry changes this much.
For the electrical tests, the team coated both surfaces with gold electrodes and mounted the membrane on a flexible polyethylene terephthalate support. Insulating tape separated the film from the support to reduce the chance that contact charge from the polymer would masquerade as a diamond signal.
Pressing and releasing produced opposite voltage pulses
When the membrane was bent, the electrodes registered a voltage pulse. Releasing it produced a pulse with the opposite sign. Repeating the mechanical cycle reproduced the electrical pattern, the signature expected when polarisation reversibly follows deformation.
The team measured the intrinsic longitudinal piezoelectric coefficient, written d33. It expresses generated charge per unit of applied force and is commonly reported in picocoulombs per newton. Commercial bulk single-crystal diamond has a d33 of zero.
In the polycrystalline membranes, d33 rose from about 2 picocoulombs per newton at a thickness of one micrometre to 3 at 2.5 micrometres and 4 at five micrometres. Above seven micrometres, it gradually fell from around 3 towards 1.
The rise and fall are as important as the non-zero measurements. A contaminant or a simple surface artefact would not necessarily be expected to peak at an intermediate thickness in the same way as a response controlled by evolving grain structure.
The largest voltage was about 70 millivolts
One-centimetre-square membranes were bent through strains between 0.35 and 1.05 per cent, with voltage amplitude increasing as strain grew. The five-micrometre sample produced the highest voltage and current across the tested thicknesses.
At the largest reported strain of 1.4 per cent, its output reached approximately 70 millivolts. The authors calculated a piezoelectric voltage coefficient, g33, of about 82.2 millivolt metres per newton.
That voltage coefficient compares favourably with a number of conventional piezoelectric materials. Diamond benefits from a low dielectric constant: a modest charge coefficient can produce a relatively large electric field and voltage instead of being screened within a highly polarizable material.
Voltage should not be confused with useful electrical power. Power depends on both voltage and current, the impedance of the source and load, device area, cycling rate and energy lost in conditioning electronics. The paper demonstrates a material response, not a diamond battery.
The researchers worked to exclude false electricity
Bending experiments are vulnerable to artefacts. Surfaces touching and separating can exchange triboelectric charge. Moving cables can pick up interference. A flexible polymer support can generate its own signal, while humidity and temperature can alter leakage currents.
The researchers used insulating layers, controlled mechanical cycling and multiple measurement approaches. They also subjected a glass sheet and bulk single-crystal diamond to comparable loading and found no obvious electrical output.
A five-micrometre membrane continued to generate consistent voltage through more than 7,000 bending cycles at 0.35 per cent strain. Measurements of one-, five- and ten-micrometre membranes also found that the piezoelectric coefficients remained stable at temperatures up to 600 kelvin, roughly 327 degrees Celsius.
Those controls make a simple noise or rubbing explanation less persuasive. They do not certify every future device geometry, and independent replication would still strengthen an unexpected result, but the study goes beyond a single voltage trace from a bent film.
Grain boundaries break the symmetry locally
A polycrystalline material is a mosaic. Within each grain, carbon atoms follow the diamond lattice, but neighbouring grains meet at different orientations. The boundary cannot preserve every symmetry operation of an ideal infinite crystal.
The authors used first-principles calculations to compare single-crystal, symmetric twin-crystal and asymmetric polycrystalline models. Only the polycrystalline configurations produced non-vanishing piezoelectric coefficients.
Under simulated strain, changes in carbon-carbon bond lengths became uneven near the grain boundaries. Calculated changes in electrical polarisation were likewise localised within a limited distance of those boundaries and did not cancel symmetrically across the model.
This offers a coherent explanation for the experiment. The diamond lattice inside a grain has not suddenly lost its inversion symmetry everywhere. Instead, the interfaces between differently oriented grains create small regions where the symmetry is broken, and bending changes their polarisation.
The mechanism is an inference supported by calculation and thickness dependence rather than a direct movie of charge accumulating atom by atom. Point defects can also break local symmetry. The team modelled that possibility, but its predicted monotonic decline with thickness did not match the observed peak around five micrometres as well as the grain-boundary model did.
Five micrometres appears to be a structural sweet spot
The peculiar thickness curve follows how a chemical-vapour-deposited film grows. Nanoscale seeds first spread sideways until they coalesce, then continue upward. As thickness increases, the grains enlarge and the network of boundaries changes.
In the team’s simplified atomic model, early growth increasingly displaced the centre of the crystal structure from the centre of its ideal lattice, strengthening asymmetry. Past an intermediate size, the two centres began moving closer again.
Real membranes contain a much messier collection of grain shapes, orientations, defects and boundaries than the model. Even so, the physical balance is plausible. An extremely thin film has small grains but not necessarily the strongest net directional asymmetry. A thick film has larger grains, leaving a smaller proportion of atoms within boundary-influenced regions.
Around five micrometres, the tested membranes apparently combined enough asymmetric boundary structure with enough active material to maximise the response. That optimum belongs to this growth process and measurement arrangement; other fabrication methods may shift it.
Potential applications remain proposals
Diamond is attractive because the electrical effect would arrive alongside unusual mechanical and thermal properties. It is chemically resistant, biocompatible, highly thermally conductive and able to tolerate large electric fields. A piezoelectric diamond component could therefore interest engineers working in hot, corrosive or biologically sensitive environments.
The HKU team points to deformation sensors, implantable devices, wearable electronics, microelectromechanical systems and harvesters that convert vibration or body motion into electrical energy.
Each application introduces requirements not settled by this paper. Engineers would need reproducible wafer-scale properties, durable electrodes, packaging, useful power density, stable performance under realistic loading and manufacturing costs that compete with established piezoelectrics.
Three authors, including senior researchers Zhiqin Chu and Yuan Lin, are named as inventors on a US provisional patent related to the device. That disclosed interest does not invalidate the measurements, but it is relevant context when assessing predictions about commercial use.
The old rule survives inside the new result
It would be tempting to summarise the study as proof that diamond was piezoelectric all along. That loses the most interesting part of the work.
Perfect single-crystal cubic diamond remains non-piezoelectric for the symmetry reasons taught for generations. The measured effect appeared after researchers changed the scale, the geometry and the internal structure, making an ultrathin membrane from many grains whose boundaries do not share the perfect lattice’s symmetry.
The classification was not discarded so much as made more precise. A material’s behaviour is not dictated only by its chemical formula. Interfaces, defects, thickness and manufacturing history can create functions that the ideal bulk crystal does not possess.
Diamond did not break its oldest rule. The researchers found the narrow places between crystals where that rule no longer applies.