receptor deformation upon a given pertur-
bation ( 28 , 29 ), whereas piezoionic current
spikes are modified by permeability, modulus,
and dimensions (as described by the poroe-
lastic time constant; eq. S15).
Energy output of the piezoionic sensors is
maximized by using classical impedance match-
ing. Peak power transfer occurs when an exter-
nal resistor of 500 ohms is employed (Fig. 2K),
which matches the internal ionic resistance
of the sensor. Perhaps unsurprisingly, these
materials are not efficient energy harvesters.
Electromechanical coupling was highest at
0.3% (Fig. 2L), as measured by applying a 10%
compression at 1 Hz on a 1-mm-thick pAAm
hydrogel swollen with 1.5 M NaCl. In com-
parison with stretchable piezoelectric and
triboelectric devices ( 3 , 9 ), piezoionic voltageoutput is much lower, but peak power per
volume is similar at 0.85mW/cm^3 (Fig. 2L).
Further exploration of dimensions, as well
as electrolyte and polymer content, creates
opportunities for investigating the upper lim-
its of power density in piezoionic materials.
These should scale well as dimensions are
reduced, because the poroelastic time con-
stant (eq. S15) is proportional to the square
of the dimensions. Unlike piezoelectric ( 30 )
and triboelectric ( 9 ) generators, piezoionics
can yield very high charges—in this case,
80 mC/cm^2 or 800mC/cm^3 .InPVDF,thecharge
per area is only 20 pC/cm^2 or 200 pC/cm^3 for
the same sample dimensions and applied pres-
sure. The highly stretchable electrets and
triboelectric generators produce 4 nC/cm^2
and 0.01% of piezoionics in terms of charge per
volume ( 9 , 30 ). Further comparisons of charge
and power output are made with piezoelectric
and triboelectric devices in the supplemen-
tary materials. The high charge density of
piezoionics is particularly attractive for use in
neural interfaces, where charge injections at
tens of microcoulombs per square centimeter
are routinely required while maintenance of
low voltages is a key safety requirement ( 31 ).
Piezoionics’similarities to mechanorecep-
tors in terms of charge, voltage, and time
response led us to explore the creation of a
piezoionic skin that can interact with the
nervous system. We began by emulating trans-
membrane potential. Sodium-potassium pumps
maintain a resting transmembrane potential
of approximately−70 mV, which is partially
depolarized and repolarized in a mechano-
receptor cell. The threshold potential change
to produce an action potential is ~15 mV. To
create a resting potential analog in the hy-
drogels, a boundary between a region of fixed
polymer backbone charge and one of uncharged
matrix was built into a hydrogel (Fig. 1B) The
resulting artificial mechanoreceptor is com-
posed of a 2-mm diameter, 2-mm-tall poly(AA-
co-AAm) hemisphere that is surrounded in
plane by pAAm swollen with 0.1 M KCl and
formed into an array of 16 individually ad-
dressable touch sensors on an 8-mm pitch
(Fig. 3D). An orange dye is used to identify the
acrylic AA regions in the transparent sensor.
In this case, protons are the only mobile ions,
released by the dissociation of the AA. The
immobility of the negative charge on the AA
yields a negative Donnan potential relative to
the AAm region, as protons diffuse into the
pAAm plane. By increasing the ratio of AA to
AAm units in the charged region, this built-in
“resting potential”becomes as large as−50 mV
(Fig. 3A). By applying a light pressure (0.5-Hz
compressive sinusoidal indentation at 20%
strain, corresponding to ~20 kPa) over the
hemispheres, displaced water entrains the pro-
tons, and the potential becomes more negative,
with larger response for a higher-AA content506 29 APRIL 2022•VOL 376 ISSUE 6592 science.orgSCIENCE
Fig. 4. Demonstration of piezoionic neuromodulation in a rodent model.(A) Setup for the piezoionic
peripheral nerve stimulation experiment, showing electrode placement. (B) Area-normalized current
responses to a square wave (peak-to-peak voltage = 5 mV) applied to the stainless-steel electrodes (red) and
PEDOT-coated Pt-Ir electrodes (black) in a saline bath. Note the significant increase in the peak current
as well as the steady-state transients using the PEDOT electrodes. (C) Sensor voltage and EMG signal reading
during piezoionic stimulation, with stainless-steel electrodes at the sciatic nerve. The EMG signal lags the
sensor input by around 5 to 8 ms as a result of propagation delay. (D) Piezoionic tibial nerve stimulation. Currents
are generated by tapping on the piezoionic sensor directly, eliciting hindlimb movement (movie S5).
RESEARCH | REPORTS