Figure 4. Characterization of the artificial haptic perception
neurons. a) The circuit diagram of the haptic perception neuron. b)I –V curves of the piezoresistive sensor with different
applied pressures/weights. c) The output resistance of sensor as a
function of weight. The solid line is fitting result by Equation (3). d)
Neuronal response of the haptic perception neuron under a constant bias
voltage (5 V, 20 μs) and the influence of varying pressure. The output
frequencies are 0.45, 0.55, 0.75, and 0.95 MHz at the pressures/weights
of 100, 150, 200, and 250 g, respectively e) Recognition of braille
characters (“p”, “k”, “u”) using tactile sensing. Without convex
patterns (that is, two sensors have no pressure), the output frequency
is zero. With convex patterns both
on the left and right (that is, two sensors both have pressure by
applying 100 g), the output frequency is ~1.1 MHz. With
convex patterns only on one of the
left and right (that is, the only one of two sensors has pressure), the
output frequency is ~0.4 MHz.
2.3. Artificialtemperature perception
neuron based on VO2volatile
memristor
In human sensory system, temperature sensation is also an indispensable
sensory ability, besides tactile sensation, which can help the human
body respond to the temperature of the outside world so that the central
nervous system can initiate a motor response, for example, avoiding
injury of the human body. The VO2 device is inherently
sensitive to temperature due to its thermally-driven metal-insulator
transition,[41, 42] and hence this feature can be
exploited to construct an artificial temperature perception neuron, as
illustrated in Figure 5 a. Figure 5b firstly shows theI-V characteristics of VO2 volatile memristor at
different temperatures ranging from 284 to 306 K, showing apparent
impact on the TS characteristics.
Only the VO2 device was heated in the probe station
during the temperature-elevated test, while the other components were
all kept at room temperature. In order to further investigate the impact
of temperature on the characteristics of VO2 memristor,
critical paramters including V th andV hold are extracted. Figure 5c further showsV th and V hold values at
different temperatures. One can see that both V thand V hold decrease as the temperature increases,
which agrees with the switching behavior of VO2 observed
in the previous study and variation in switching voltages originates
from thermal processes induced by Joule heating and its
dissipation.[43-46] In order to exclude the
possibility of stochastic fluctuations, we have systematically measured
the cycle-to-cycle variation of the VO2 memristor under
different temperatures (Figure S11). The experimental results in Figure
S11 demonstrate that the temperature dependence of the threshold
voltages in the VO2 memristor is much more significant
than the parameter fluctuations and hence contribute to the temperature
sensing capability of the sensory neuron. We have characterized the
operation temperature range of our VO2 device. The
results demonstrate that the TS characteristics remain stable at least
below 0 °C but gradually disappear above 35 °C (Figure S12, Supporting
Information), which could be related to the low phase transition
temperature of VO2 (~340
K)[47]. This issue can potentially be addressed by
using Mott system with higher phase transition temperature, such as
NbO2 (~1080
K)[48].
Lee et al.[44] proposed a Joule heating model to
point out that the relationship betweenV th,V hold and temperature (T ). In this model,
the mott transition occurs when the voltage-induced Joule heating is
sufficient to raise the crystal temperature to the mott transition
temperature. Assuming the heat obtained in the VO2crystal is mainly due to the balance between resistive Joule heating and
heat loss from the crystal to the environment via heat conduction, this
relation can be expressed by the following simple heat
equation:
(4)
where C , T , R th, and R are
the heat capacitance, temperature, effective thermal resistance, and
resistance of the VO2 film. V VO2is the applied voltage to the
VO2 film andT e is the environmental temperature. Equation (4)
is a first differential equation, where the relationship betweenV th, V hold andT e can be obtained by integrating (see Supporting
Information), which shows the T e-dependence ofV th and V hold. BothV th and V hold decreased
with increasing T e.
Furthermore, Figure 5d illustrates the oscillation frequency of the
structure shown in Figure 5a as a function of R Lat different temperatures from 284 to 306 K, when a constant voltage
pulse (5 V, 20 μs) is applied. The oscillation frequency generally
increases linearly with decreased resistance, and the oscillation
frequency is increased at evelated temperature whenR L is fixed (Figure 5d). Figure 5e further shows
the oscillation waveforms at varied temperature from 284 to 306 K, with
constant R L of 3.6 kΩ and the same input voltage
pulse (5 V, 20 μs). Once again the output oscillation frequency
increases (from 0.3 to 0.8 MHz) as the temperature increases (from 284
to 306 K). Since the oscillation is activated betweenV hold and V th, the
amplitude of oscillation decreases as the temperature increases, which
is consistent with the results in Figure 5b,c. Therefore, the
VO2 neuron in Figure 5a can realize temperature
perception based on intrinsic thermal sensitivity of
VO2, which is able to perceive temperature changes and
serves as a temperature sensory neuron.