A flexible bimodal sensor array for simultaneous sensing of pressure and temperature

Further-more, quantitative measurements for precise sensing of arbi-trary input stimuli, e.g., strain, pressure and temperature, from the sensing devices in single or multimodal e -skin

A Flexible Bimodal Sensor Array for Simultaneous Sensing

of Pressure and Temperature

Nguyen Thanh Tien , Sanghun Jeon , Do-Il Kim , Tran Quang Trung , Mi Jang ,

Byeong-Ung Hwang , Kyung-Eun Byun , Jihyun Bae , Eunha Lee , Jeffrey B.-H Tok ,

Zhenan Bao , Nae-Eung Lee ,* and Jong-Jin Park *

An electronic skin ( e -skin) is comprised of arrays of pixels

that function as sensing devices for various targeted external

stimuli [ 1–17 ] It also consists of signal processing circuits

embedded in large-area fl exible or stretchable substrates

Typi-cally, a defi ned number of pixels are designed to sense specifi c

types of external stimuli In many previously reported works,

single modality e -skin in which one sensor in a single pixel

measures a unique sensing parameter in either pressure [ 11–19 ]

or strain [ 19,20 ] has been investigated In practical applications

of highly functional e -skin, e.g., artifi cial fi nger, multimodality

in simultaneous sensing of multiple stimuli such as

tempera-ture, strain and pressure, is required In previous reports on

multimodal sensing, [ 21–23 ] multiple sensors that are integrated

into a single pixel are able to sense multiple stimuli

simultane-ously However, large-area integration of the multiple sensors

with different sensing principles in a pixel of the multimodal

e -skin requires sophisticated fabrication processes

Further-more, quantitative measurements for precise sensing of

arbi-trary input stimuli, e.g., strain, pressure and temperature, from

the sensing devices in single or multimodal e -skin have rarely

reported even though their electrical responses to known input

stimuli were often measured [ 1–17 ]

In fl exible e -skin, the target signals from sensing elements

under multiple stimuli are often infl uenced by the other

stimuli including strain experienced by the e -skin For example,

signals from both temperature and pressure sensors in a matrix are infl uenced by other external stimuli of pressure and temperature, respectively [ 21,22,24 ] To date, signals originating from mechanical interferences are minimized at the level of sensing device by simply calibrating the output signals, which are usually achieved via designing compensating circuits [ 18,19 ] and reference devices [ 20,21 ] However, this approach has only been met with limited success to date Another approach is to employ sensing materials that are resistant to strain such that strain induced interference can be avoided [ 16,22 ] This approach requires the usage of rigid materials, which often require spe-cial processing steps, complex fabrication designs and integra-tion processes onto fl exible substrates [ 23 ] In addition, other lim-itations such as structural complexity due to the integration of heterogeneous sensing materials and devices, large power con-sumption, cross-sensitivity to multiple stimuli, and inaccurate signal read-out, have all greatly complicated this approach In

most of fl exible e -skin investigations, issues relating to signal

interference by mechanical deformation or other stimuli have yet been addressed in detail

To address the above mentioned challenges, we hypoth-esized that another viable approach is to separate or decouple a target sensing signal from interferences stemming inherently from stress-related deformation of fl exible sensing elements in

fl exible e -skin In this work, we resolve the subjected

interfer-ences by using a fi eld-effect transistor (FET) sensor platform integrated with multi-stimuli responsive materials as sub-components of gate dielectrics and channel FET is an ideal

platform for multi-stimuli e -skin sheets, and various fl exible

or rigid stimuli-responsive FETs for chemical, biological, and physical sensors have been previously reported [ 24–31 ] As men-tioned, signal interference generated via global straining of the

fl exible FET sensors has yet been addressed Although FET’s device physics and analytical equations have extensively been developed to allow quantitative analysis of the device’s para-meters, [ 32–34 ] however a detailed analysis of the FET sensors under multiple physical stimuli has rarely been reported In this report, we describe the direct integration of both the piezo-pyroelectric gate dielectric and piezo-thermoresistive organic semiconductor channel into FET platforms, and the resulting device is able to respond to two stimuli in pressure (or strain) and temperature simultaneously and disproportionally In our previous works , we had shown the possibility of extracting temperature [ 35 ] or pressure [ 36 ] responsitivities of functional

DOI: 10.1002/adma.201302869

N T Tien,[+] D.-I Kim, T Q Trung,

B.-U Hwang, N.-E Lee

School of Advanced Materials Science & Engineering

and Sungkyunkwan University

Suwon , Kyunggi-do , 440-746 , Republic of Korea

E-mails: nelee@skku.edu

M Jang, N.-E Lee

SKKU Advanced Institute of Nano Technology (SAINT)

Sungkyunkwan University

Suwon , Kyunggi-do , 440-746 , Republic of Korea

S Jeon,[+] K.-E Byun, J Bae, E Lee, J.-J Park

Samsung Advanced Institute of Technology

Samsung Electronics Corporation

Yongin , Kyunggi-do , 446-712 , Republic of Korea

E-mails: jongjin00.park@samsung.com

S Jeon

Department of Applied Physics

Korea University

Sejong City , 339-700 , Korea

J B.-H Tok, Z Bao

Department of Chemical Engineering

Stanford University

Stanford , California , 94305 , USA

[+]These authors contributed equally to this work

gate dielectrics integrated in FET platforms by using AC gate

bias technique Herein, we extend our methods to elaborate

the precise and quantitative separation of unknown,

simul-taneously applied stimuli of pressure-temperature or

strain-temperature by decoupling the output signal from the single

FET platform with stimuli-responsive gate dielectric and

semi-conductor channel, and further demonstrate the applicability

to bimodal sensing from the FET array fabricated on fl exible

substrate

In our fi rst step to realize sensor functions for application to

e -skin, we construct a device array that mimics human fi nger

functions, which is comprised of an array of pressure and

tem-perature sensor pixels displayed on top of a fl exible platform

Our sensor has high sensitivity to external stimuli, while being

able to differentiate these exposed stimuli We have chosen to

fi rst incorporate a nanocomposite material as the gate dielectric

and an organic semiconductor as the channel to the physically

responsive FET (physi-FET) platform ( Figure 1 a) The

nano-composite material serves three primary functions, namely:

(i) enhance the electro-physical coupling effects, (ii) improve the

stability in their stimuli-responsive properties, and (iii) allow

simultaneous analysis of pressure-temperature or

strain-temperature sensing parameters Since the organic FET

struc-ture has functional piezo- and pyroelectric nanocomposite gate

dielectrics, as well as piezo- and thermoresistive organic

semi-conductor, pentacene, channel, it is able to thus simultaneously

measure changes in both strain and heat This is because the two chosen materials are able to respond to pressure (or strain) and temperature simultaneously, but in a disproportionate manner

Specifi cally, the nanocomposite material we have used as a gate dielectric is a mixture of poly(vinylidenefl uoride-trifl uoro-ethylene) (P(VDF-TrFE)) and BaTiO 3 (BT) nanoparticles (NPs) The transmission electron microscopy (TEM) image of BT NPs

in the nanocomposite are depicted as insets in Figure 1 a When introducing an AC gate bias to the FET having poled functional gate dielectric with an electrical fi eld (Figure 1 b), the amplitude

and offset values ( I D amp and I D offset , respectively) of the modu-lated drain current ( I D ) include the information on channel transconductance ( g m ) and the effective remnant polarization ( P r ) of the piezo-pyroelectric gate dielectric layer We can hence estimate the g m value and its equivalent voltage in Pr V0= P C r , via both Equations (1) and (2) below:

V0= V G0 I

offset D

gm= IDamp

V G0

(2)

where C and V G0 are capacitance of gate dielectric and

ampli-tude of applied AC gate bias, respectively

Figure 1 Illustration of our approach a) The structure of physically responsive fi eld-effect transistor (physi-FET) with the bottom-gated and top-contact

structure, where the gate dielectric is comprised of P(VDF-TrFE) or nanocomposite of P(VDF-TrFE) and BaTiO 3 nanoparticles and the channel is organic

semiconductor of pentacene b) Changes in I D signals of physi-FET with P(VDF-TrFE) upon applying pressure and temperature c) Responses of channel

trans-conductance (g m ) in a FET with highly crystalline P(VDF-TrFE) when applying pressure and temperature simultaneously Error bars were drawn

with 3X magnifi cation d) Responses of equivalent voltage V 0 upon applying pressure and temperature Error bars were drawn with 30X magnifi cation

However, in most general cases in which functional materials showed non-linear relationships, e.g., Arrhenius behavior of effective channel mobility, µ eff , with T, μeff = μeff 0e



−kB T E a

 ,

and reciprocal dependence of C = ε0εr d 1

0 ( 1−P) with high pres-sure (Figure S3), a Jacobian matrix approach is instead needed (for details, see Equation S2) Extending the temperature range toward 70–80 ° C also result in reduction of accuracy due to non-linear responses of P(VDF-TrFE) [ 35 ]

For crystalline P(VDF-TrFE), its pyroelectricity ( µ C/K) prop-erty was found to be more signifi cant than piezoelectricity (pC/N) Upon comparing their responsiveness, which is

indi-cated by µ C/K vs pC/N, we realized that the value of M 3 is sig-nifi cantly smaller than M 4 (Table S1) We attribute such a huge difference in its matrix’s components to a phenomenon known

as “ill-conditioning” in linear algebra [ 37 ] Generally, it will lead

to inaccuracy when used to extract both parameters in P and

T from the data in Figure 1 e and 1 f To resolve this issue, we

instead use the approach in tuning the electro-physical cou-pling properties of the functional gate dielectric materials For tuning of functional gate dielectrics, our approach described here employs the nanocomposite properties of the highly crys-talline materials in both P(VDF-TrFE) and BaTiO 3 NPs

To study contributions from the NPs’s crystallinity to piezoe-lectricity, we have utilized the Piezo-response Force Microscopy

(PFM) method Figure 2 a and 2 d illustrated the topographic

images of the nanocomposite comprised of 20 wt% NPs Figure 2 b and 2 c exhibited PFM images of highly crystalline P(VDF-TrFE) with the NPs right after poling and after 40 hrs, respectively; while Figure 2 e and 2 f depicted PFM images of low crystalline P(VDF-TrFE) with the NPs In this experiment, both negative (V = –30 V, dark line) and positive (V = +30 V, bright line) polarization patterns were written on the surface

of P(VDF-TrFE), with and without the NPs together with the crystalline P(VDF-TrFE) The scan speed of the tip was 0.1 Hz for 5 µ m/s, and the effective poling time was about 50 s, which allowed the writing of 5 equidistant lines PFM images were obtained by the application of an AC bias of amplitude 3 V and frequency 40 kHz during highly crystalline P(VDF-TrFE) showed pronounced piezo-response images for both initial polarization pattern even after 40 hrs (Figure 2 b and 2 c), indi-cating that the crystallinity of P(VDF-TrFE) has signifi cantly enhanced its piezo-response characteristics

The effect of initial polarization performance and polariza-tion, with respect to relaxation time, were also quantifi ed in a series of PFM measurements PFM signals of Figure 2 g were consistent with piezo-response images of Figure 2 b, 2 c, 2 e, 2 f, and Figure 2 h, 2 i, 2 j depicted the time dependence of the max-imum piezo-response in both positive and negative polarization patterns and the cross-section of the resulting piezo-response signal, in comparison with the calculated distribution of the electric fi eld created by the tip The signs and the intensities of PFM signals indicated the direction and the amount of sample polarization, respectively As shown from the PFM signal panels in Figure 2 h and 2 i, highly crystalline P(VDF-TrFE) with BT NPs exhibited enhanced piezoelectric performances, together with increasing signal with respect to relaxation time Simple exponential decay mechanism could not suffi ciently account for the relaxation behaviors of the prepared nanocom-posites in both positive and negative polarization directions

Derivations and fundamental electrical characteristics for

extracting g m and V 0 values from the modulated I D are detailed

in Equation S1 and Figure S1, Supporting Information In

our previous works, we validated our model by fi tting output

( I D –V D ) and transfer ( I D –V G ) characteristic, as well as using

AC gate bias technique for extraction of V o values under the

change of individual stimulus (i.e., pressure ( P ) and

tempera-ture ( T )) [ 35,36 ] The method can be extended to extract g m as well

as V o under simultaneous changes of both T and P

Next, we proceed to investigate temperature sensing while

applying a pressure to the FET Specifi cally, we measured the

changes in g m and V 0 when subjected to varying applied P and

T , simultaneously Methods to heat and pressurize the FET was

described in the Methods section Figure 1 c and 1 d showed

the responses of g m and V 0 with P and T , which were extracted

from a FET comprised of highly crystalline P(VDF-TrFE) gate

dielectric Sampling T was measured from 35 to 40 ° C with an

increment interval of 1 °C, and P was from 0 to 0.5 N mm –2

with an interval of 0.1 N mm –2 Each point plotted in Figure 1 e

and 1 d was derived from averaging the values derived from a

total of 10 collected data points, as measured under the same

condition Our obtained data indicated that responses to P

changes were relatively smaller in both g m and V 0 values (≈5%

in the applied pressure range) when compared to those of T

changes Even though the g m and V 0 responses by P changes

are relatively small, the observed linearity in Figure 1 c and

1 d suggested that the relationship of g m and V 0 vs P and T

can be correlated via a characteristic matrix M, as shown in

Equation (3) below:

g m

V0



=

M1 M2

M3 M4

 P

T

 (3)

where g m and V0 are changes in g m and V 0 values under

subjection to various stimuli ( P , Δ P , T , Δ T ); and M 1 , M 2 , M 3 ,

and M 4 are all sub-components of the matrix M

By using Equation (3) , both values in M 3 and M 4 were

directly correlated to piezoelectric coeffi cient ( d 33 ) and

pyro-electric coeffi cient ( p 3 ), which were again associated with the

changes in P r of the nanocomposite material under

applica-tion of normal stress and temperature, respectively We have

previously reported the results for both positive pyro- and

piezoelectricity in highly crystalline P(VDF-TrFE) [ 35,36 ]

Posi-tive pyroelectricity indicated that P r would increase at elevated

temperature, while positive piezoelectricity means P r would

decrease under applied pressure In combining highly

crys-talline P(VDF-TrFE) and BT NPs with positive piezoelectricity

and negative pyroelectricity, d 33 and p 3 of BT/P(VDF-TrFE)

NCs are observed to increase and decrease, respectively In

regard to both M 1 and M 2, corresponding responses in g m

with varying P and T are not straightforward However, our

experimental data indicated that piezo-resistivity of

penta-cene channel with a larger P-dependence of effective channel

mobility are the main contributions to M 1 , while gate

dielec-tric capacitance with higher T sensitivity contribute more

to M 2, respectively (Figure S2) The linear responses of the

materials, e.g., T- and P-dependences of measured g m and

V 0 , are directly utilized in sensing applications In such a

situation, the proposed matrix approach can then be utilized

Instead, a stretched exponential dependence equation was used to fi t the obtained experi-mental data in both polarization directions The fi tting was done with the Kohlrausch-Williams-Watts formula (Equation (4) ), which

is commonly used to describe relaxation in the system with dipole-dipole interactions,

(d33)ef f ≈ e−

t t0

b

where t is the time, [ 38 ] b is a measure of the

distribution of defects that control the decay

process, and t o is the measure of the effective relaxation time In our investigation, the b

value varies with the crystallinity of P(VDF-TrFE) and the NP content in nanocomposites For nanocomposite of highly crystalline

P(VDF-TrFE) and NPs, the parameter b was

much higher than the control sample, low crystalline (P(VDF-TrFE) High b value for high crystallinity P(VDF-TrFE) incorporated with BaTiO 3 NPs exhibit high homoge-neity and low defect distribution Also, the effective relation time of high crystalline P(VDF-TrFE) with the NPs, low crystalline P(VDF-TrFE) with the NPs, and low crys-talline P(VDF-TrFE) without the NPs were found to be 693, 277 and 35 hrs, respec-tively This indicates that the crystallinity of the piezoelectric polymer and the employ-ment of the NPs both played crucial roles in improving the effective relaxation time Our PFM study demonstrated the advantages in using the NCs to greatly improve their long term stability as well as their electro-phys-ical coupling properties by adding inorganic

BT NPs having higher stability and piezo-electricity than P(VDF-TrFE), thus enabling high reliability when extracting the output sensing parameters The PFM data were also obtained at 70 °C (Figure S4) In comparison with the PFM signal at room temperature, the measurement at an elevated tempera-ture of 70 °C presented a higher PFM signal immediately after applying polling bias on the gate dielectric; however, the reduction in the retention was also faster This condition

Figure 2 Piezo-response force microscopy (PFM) images and PFM signals of various nanocomposite materials a) Surface topography image of

BaTiO 3 NPs in highly crystalline P(VDF-TrFE) matrix PFM images of line polarization patterns obtained with V = 30 V (bright line) and –30 V (dark

line) applied to nanocomposite material comprised of highly crystalline P(VDF-TrFE) matrix and BaTiO 3 NPs, b) right after poling and c) after 40 hrs d) Surface topography image of BaTiO 3 NPs in low crystalline P(VDF-TrFE) matrix PFM images of line polarization patterns obtained with V =

30 V (bright line) and –30 V (dark line) applied to the nanocomposite, e) right after poling, f) after 40 hrs g) Comparison of cross-section of piezo-response signal across the pattern right after poling and after 40 hrs for both highly crystalline and low crystalline P(VDF-TrFE) matrix h) Comparison

of PFM signal evolution with the relaxation time for both samples i) PFM signal window, the difference between the maximum and the minimum

PFM signal which were the response of V = +30 V and –30 V, respectively, right after poling ( V = ±30 V) and after 40 hrs for various nanocomposite materials j) Relaxation-related time constant, b , of various samples Incorporating BaTiO 3 NPs to the highly crystalline P(VDF-TrFE) matrix presents superior characteristics

of g m_ref , V 0_ref at certain conditions of P ref and T ref This

ref-erence determining step, which can be simultaneously per-formed for all devices after the poling process, helps to

deter-mine Δ g m and Δ V 0 and to also estimate P and T from Δ P and

Δ T , respectively If all the devices in an array format can be

ren-dered in a uniform manner, the pre-determing step need only

be performed for only one device The visual schematic of an

algorithm that allows us to extract P and T simultaneously from measured I D of FETs is shown in Figure 3 a This procedure can

be expressed by Equation (5) :

 P T



=

Pref 

Tref

+ M1 M2

M3 M4

−1 gm− g m ref

V0− V0 ref



(5)

By using Equation (5) , we proceeded to determine the accu-racy in our collected dataset that were used to calculate the data in Supplemental Table S1 Figure 3 b shows the relative accuracy percentage of the read-out values from P(VDF-TrFE),

might reduce the lifetime of the device at an

elevated temperature However, the on-chip

polling procedure can be used to reset the

sensing accuracy and to extend the device

lifetime

Another advantage in using crystalline

P(VDF-TrFE) and BT NPs is that the

dif-ferent piezo-pyroelectricity of both materials

enable us to easily tune their electro-physical

coupling properties This in turn can affect

the resulting sensitivity and accuracy when

we extract a target sensing para meter

Sup-plemental Table S1 shows characteristic

matrices in Equation 3 of highly crystalline

P(VDF-TrFE) and its nanocomposites with BT

NPs of 20 and 40 wt% The values of matrix

elements in Table S1 were calculated by

using the least-mean-square method, while

employing a dataset of more than 5000

meas-ured points from 12 devices According to

the data in Table S1, upon adding BT NPs to

P(VDF-TrFE) with a concentration of 40 wt%

NC, the pressure sensitive coeffi cient, M 3 ,

increases; while the temperature sensitive

coeffi cient, M 4, decreases This observation

was explained by the enhancement and

coun-terbalance between positive

piezoelectricity-positive pyroelectricity of highly crystalline

P(VDF-TrFE) and positive

piezoelectricity-negative pyroelectricity of BaTiO 3 NPs [ 35,36,39 ]

In another report, [ 7 ] it was also demonstrated

that well-alinged crystalline β -phase in

nano-fi bers also contributed to a large

enhance-ment in piezoelectricity, and can enable the

tunability of piezo-pyroelectricity Similarity

in both values for M 1 and M 2 in all devices

confi rmed the reliability in our approach to

estimate the components of M matrices for

the pentacene channel in FETs for all our

experiments We also observed that deviation

in estimated values of characteristic matrix M

was approximately 1 to 5% in all the examined materials, except

for M 4 in 40 wt% NC, which is up to 267% This aberration

value may be due to the non-linearity in temperature response

of the material (Figure S5)

Since we know a priori the values of M, both unknown

values in P and T can thus be estimated by using the extracted

g m and V 0 values in arbitrary FETs via the inverse matrix, M –1 ,

based on Equation (3) It should be noted that Equation 3 does

not express any direct relationship between the absolute values

of g m , V 0 and P , T ; but rather, it simply highlights their

rela-tive changes The relarela-tive changes are expected to be similar

in all FETs as they are comprised of the same semiconductor

and gate dielectric, even though they have slightly different g m

and V 0 values However, FETs comprised of the same

semicon-ductor, but with different gate dielectrics, should have equal

values in both M 1 and M 2

To resolve the issue of different initial g m and V 0 , each FET

should have a pre-determining step that gives reference values

Figure 3 Realization of P-T decoupling and bimodal sensing in a single FET sensor a, Our

algorithm to extract two sensing parameters (i.e., P and T ), simultaneously b, Standard devia-tion in estimated T ( σ T ) value c, Standard deviation in estimated P ( σ P ) value d, Accuracy

in values as measured by pressure and temperature gauges e, Demonstration of real-time

bimodal sensing of P-T f, Demonstration of real-time bimodal sensing of both tensile strain

( ε )-temperature (T)

the limiting factor to realize high-speed tactile sensors as the P(VDF-TrFE) was reported to tolerate and respond up to a very high frequency above kHz range [ 40 ]

Moreover, we also observed a slight decrease in accuracy after a continuous measurement duration of >4 h We attrib-uted this degradation to the increasing bias stress phenomenon

in the FET, which originates from charge trapping in either the pentacene channel or at the interface of the gate dielectrics and pentacene [ 41 ] The de-trapping process by applying posi-tive gate bias at high temperature, re-calibration after a certain time interval, and further optimization of the gate dielectric

or passivation layer in the future may help address the reli-ability issue of our devices We emphasize that the procedure described herein not only decouples the target sensing

param-eter, T , in which its accuracy is greatly reduced by interference from P ; but our approach also provides capability in reading P

This capability hence enables the bimodality sensing capability within a single device

Upon substituting the sub-components of the character-istic matrices M from a FET with different physical constants (related to other electro-physical couplings), this approach can be further extended to general heterogeneous bimodal sensing applications For instance, if we were to replace the

coeffi cient d 33 by e 31 (which is often associated with applied planar strain) and P r (piezoelectric phenomenon), the

tempera-ture coupled to tensile strain (under bending) can be extracted

in a similar methodology as decoupling P from T Figure 3 f showed the read-out values of strain ( ε ) and T in simultaneous

strain-temperature stimulation in real-time Since the device was mounted on a fl exible polyimide heater, the low heat capacity of the heater resulted in a large fl uctuation in control-ling temperature The sensing accuracy of our fl exible device was reduced slightly when subjected to repeated bending pro-cesses at the bending radii (Figure S7) However, this issue can be easily resolved by performing the pre-determining

step of g m_ref and V 0_ref values intermittently (Figure 3 a) after

certain mechanical bending cycles After conducting the pre-determining step after 10 000 bending cycles, for example, the readout strain was observed to recover to the value prior

to the cyclic bending Our demonstration of bimodal sensing capability also indicated that this approach should readily be extended to simultaneously extract multiple sensing param-eters for multimodal sensing

To apply our described multimodal sensing concept to a FET array, real-time bimodal sensing of a 4 × 4 device array was investigated by measuring real-time responses of the device array in a custom-built bending system integrated with real-time, multiplexed electrical measurement units The pixel size containing an FET was approximately 0.5 × 1 mm, and our six-teen devices were uniformly distributed over a 1.3 × 1.3 cm 2 area Our array system can simultaneously measure the arbi-trary T-P values of the contained 16 devices in real-time The measurement details of the array system were described in

the Methods section Figure 4 a and 4 b showed a

two-dimen-sional mapping of measured T and P , respectively, in a device

array, when half of the devices, i.e., eight devices, were pressed with a human fi nger A picture of this experiment was shown

in Figure 4 It was clear that the read-out temperature and pressure of pressed devices were higher than the untouched

20 wt% NCs and 40 wt% FETs, that matched with the reading

values of temperature ( T ) and pressure ( P ) using commercial

gauges This quantitative comparison excluded the absolute

uncertainty of commercial temperature and pressure gauges

We observed that 20 wt% NC showed the best accuracy in

extracting both the T and P parameters Figure 3 c, and 3

showed the different distribution of both applied and estimated

values of T and P It was observed that with decreasing M 4

from –4.173 to 0.003 V/K, the standard deviation in extracted

T ( σ T ) increases from 0.00581 °C to 0.08391 °C (Figure 3 d) and

corresponds well with the decreased accuracy for T

measure-ments (from 100% to 77.78%, shown in Figure 3 d) Similarly,

an increase of M 3 from 1.094 for highly crystalline

P(VDF-TrFE) to 2.615 V/MPa for 20 wt% NC was observed, along with

a signifi cant increase in accuracy from 64.58% to 99.31% with

a tighter standard deviation in extracted P ( σ P ) from 0.02159

of highly crystalline P(VDF-TrFE) to 0.00689 N mm –2 of NC

20 wt% (Figure 3 c) However, further increase in M 3 to 40 wt%

NC led to a decrease in accuracy, as indicated by the increased

deviation in σ P This was attributed to the non-linear response

in temperature of 40 wt% NC (Figure S5), which increases the

error values of the M’s components We note that accuracy is

estimated as half of the minimum measurement resolution In

the case of 20 wt% NC, pressure resolution of the sensor can

be considered as 0.07 N mm –2 (≈7 kPa) Pressure sensitivity of

the sensor, response slope of V o vs P , can be estimated as the

M 3 value, 2.615 mV kPa –1 for 20 wt% NC (Table S1) Higher

pressure sensitivity required for e -skin applications requiring

detection of 100–500 kPa range can be achieved by enhancing

the piezoelectric responsitivity of the NCs through

microstruc-turing Collectively, these results indicate the importance in

adjusting the characteristic matrix M through tuning of the

electro-physical coupling properties in the nanocomposite

material to allow precise quantifi cation of both P and T values

Figure 3 e showed the real-time responses of both P and T , as

extracted from a FET with 20 wt% NCs, upon time-dependent

pressurizing and heating of the device A video recording of

this experiment is shown in Video 1, Supporting Information

Solid red and blue lines indicate the applied T and P , while

blank orange circles and green squares are read-out values of

T and P Figure 3 e also indicated that our obtained accuracy

in measuring both sensing parameters is very good The time

required for the readout of the pressure or strain at a

steady-state signal in our measurement method of “apply and hold”

is ≈10 s This sensor response time is suffi cient for practical

applications that mimic human fi nger such as holding a hard/

soft-hot/cold object in which target sensing signals are typically

constant force, pressure, or strain and temperature In order to

further understand the response time of physi-FET under

var-ious pressurizing condition, the dynamic responses of a

penta-cene resistor, metal-20 wt% NC-metal structure and physi-FET

to dynamic pressurizing with varying forcing time were

meas-ured in a dynamic loading system similar to the system used by

Takei et al [ 1 ] The results indicated that the FET responses at a

slow forcing condition are attributed to a slow relaxation time

for steady-state signal of the pentacene channel due to the long

equilibration time of the mechanically stimulated pentacene

channel, [ 36 ] rather than that of the response of the

nanocom-posite gate dielectric (Figure S6) As observed, the NCs are not

FET array indicate that our approach is highly amenable for

e -skin application As designed for fl exible e -skin, the pres-sure–temperature bimodal sensing was also demonstrated in

a bent state with the bending radii of 1 and 2 cm (Figure S9) Last, this approach can be generalized for other heterogeneous multi-stimuli sensing application For instance, chemical-strain differentiation can be achieved by utilizing chemical-strain (M 1 ) and chemical responsiveness (M 2 ) of semiconducting channel, and strain-responsiveness of functional gate dielectric (M 3 )

In summary, we have successfully demonstrated a general

approach to fabricate e -skin that can: (i) extract effects from the target sensing signals, such as P or T , while the fl exible sensor

is under multimode stimulus; and (ii) enable real-time bimodal sensing using a single FET device by extracting parameters asso-ciated with mechanical deformations This concept for real-time bimodal sensing of unknown multi-stimuli was realized in an array format The advantages in integrating FET arrays with

mul-timodal sensing elements in fl exible e -skin greatly reduced the

complexity in structural integration, eliminated or minimized the signal interferences coupled by strain, signifi cantly reduced the power consumption, and decreased the failure rate in pro-duction due to facile integration of FET devices into the circuits Furthermore, it has potential in reducing fabrication costs of

large-area fl exible e -skins This approach may be extended to

devices Moreover, the read-out temperatures of the devices

being pressed were also measured to be very close to that of

human body temperature

Figure 4 c and 4 d showed the measured values of T and P

when all of the devices were pressed by a human thumb As

expected, both the temperature and pressure responses in all

devices were highly identical with the pressed ones in Figure 4 a

and 4 b In addition, when only one device within the array was

individually pressed by using a narrow blunt object, only the

read-out pressure of the pressed device was higher than those

of the rest; while the read-out temperature of the unaffected

devices remained close to room temperature (Figure 4 e and 4 f)

The non-uniformity in performance of the arrayed devices

(a critical issue for practical applications) was solved upon

applying our pre-determining step of g m_ref and V 0_ref values on

the initial characteristics of the devices Moreover, such small

power consumption of an individual sensing device, which is

as low as 600 nW per pixel using the sensor pixel of the one

sensor and one switch element (Figure S8), may enable

realiza-tion of a portable and self-powered e -skin We anticipate that

upon optimizing the gate dielectric thickness, we could further

reduce operation voltage more and, thus, minimize the devices

power consumption Hence, our observed results of spatially

resolved responses to both pressure and temperature of fl exible

Figure 4 Realization of P-T decoupling and bimodal sensing in an array of FET sensors a,b) Read-out pressure and temperature measurements of

a sensor array with half of the sensor’s devices are being depressed by a human thumb Pressed devices are observed to show higher pressure and temperature responses, when compared to unaffected devices c,d) Read-out pressure and temperature measurements of a sensor array with all devices being depressed by a human thumb e,f) Read-out pressure and temperature measurements of a sensor array with only one device at (column, row) = (3, 2) being depressed by a blunt object (i.e., cotton swab) The pressure response of the depressed device is measured to be higher than all other unaffected devices, with the temperature of all other unaffected devices being close to room temperature The devices at (2, 3) and (3, 3) are defective

the cantilever provides the polarization orientation of the sample Cr/

Au coated cantilevers with the resistance of 0.01-0.02 Ω and tip apex radius of less than 50 nm were used (Nanosensors) PFM imaging was acquired under an applied AC voltage with amplitude of 3 V and

frequency f = 40 kHz The measurement frequency was chosen far away

from the resonant frequencies of the cantilever-sample holder system to avoid the ambiguity of experimental data

TEM Measurement : Cross-sectional high angle annular dark fi eld scanning transmission electron microscopy (HAADF-STEM) image was taken for detailed analysis of the gate dielectric materials Our obtained HAADF-STEM Z-contrast image clearly distinguishes the BaTiO 3 NPs from P(VDF-TrFE) The order of brightness is proportional to mean square atomic number of composed elements Energy dispersive spectroscopy analysis was used for confi rming the chemical composition

of the NPs We also performed nano-beam diffraction pattern analysis The diffraction pattern clearly confi rmed the presence of crystalline BaTiO 3 NPs within P(VDF-TrFE) (Figure S10)

Supporting Information

Supporting Information is available from the Wiley Online Library or from the author

Acknowledgements

This research was supported by the Basic Science Research Program (Grant No No 2009-0083540 and 2013R1A2A1A01015232) through the National Research Foundation of Korea (NRF), funded by the Ministry of Education, Science and Technology (MEST), Republic of Korea

Received: June 24, 2013 Revised: August 9, 2013 Published online:

[1] K Takei , T Takahashi , J C Ho , H Ko , A G Gillies , P W Leu ,

R S Fearing , A Javey , Nat Mater 2010 , 9 , 821

[2] S C B Mannsfeld , B C K Tee , R M Stoltenberg , C V H H Chen ,

S Barman , B V O Muir , A N Sokolov , C Reese , Z Bao , Nat

Mater 2010 , 9 , 859

[3] G Schwartz , B C.-K Tee , J Mei , A L Appleton , D H Kim ,

H Wang , Z Bao , Nat Comm 2013 , 10.1038/ncomms2832 [4] B C K Tee , C Wang , R Allen , Z Bao , Nat Nanotech 2012 , 7 ,

825 [5] Y J Yang , M Y Cheng , W Y Chang , L C Tsao , S A Yang ,

W P Shih , F Y Chang , S H Chang , K C Fan , Sens Actuators A

2008 , 143 , 143 [6] M Ramuz , B C K Tee , J B H Tok , Z Bao , Adv Mater 2012 , 24 , 3223

[7] L Persano , C Dagdeviren , Y Su , Y Zhang , S Girardo , D Pisignano ,

Y Huang , J A Rogers , Nat Comm 2013 , 10.1038/ncomms2639 [8] W Wu , X Wen , Z L Wang , Science 2013 , 340 , 952

[9] D J Lipomi , M Vosgueritchian , B C K Tee , S L Hellstrom ,

J A Lee , C H Fox , Z Bao , Nat Nanotech 2011 , 6 , 788

[10] C Pang , G.-Y Lee , T.-i Kim , S M Kim , H N Kim , S.-H Ahn ,

K.-Y Suh , Nat Mater 2012 , 11 , 795

[11] M Zirkl , A Sawatdee , U Helbig , M Krause , G Scheipl , E Kraker ,

P A Ersman , D Nilsson , D Platt , P Bodö , S Bauer , G Domann ,

B Stadlober , Adv Mater 2011 , 23 , 2069

[12] T Someya , Y Kato , T Sekitani , S Iba , Y Noguchi , Y Murase ,

H Kawaguchi , T Sakurai , Proc Natl Acad Sci 2005 , 102 , 1 2321 [13] J Engel , J Chen , Z Fan , C Liu , Sens Actuators A 2005 , 117 , 50 [14] T Sekitani , T Someya , Adv Mater 2010 , 22 , 2228

realize multimodality in large area fl exible e -skin with

heteroge-neous input stimuli of physical, chemical or biological natures,

and also to solve problems associated with strains induced

during operative services of fl exible electronic systems

Experimental Section

Preparation of Materials and Devices : The P(VDF-TrFE) (65 mol% of

VDF) was purchased from Piezotech S.A The BT NPs with an average

diameter of 50 nm and the coupling agent 3-aminopropyltriethoxy silane

(APTES) were purchased from Sigma Aldrich The BT NPs were fi rst

ball-milled for 2 hrs to disperse the aggregated NPs After milling, the BT

NPs were dispersed in an APTES–ethanol solution with a pH of 4–4.5

(adjusted by HCl) for 1 hr The BT NPs and APTES–ethanol mixture

was fi ltered and washed in ethanol to remove the residual APTES The

treated BT NPs were then cured at 110 °C for 5 min on a hot plate

and mixed with N,N-dimethylformamide (DMF) Centrifugation was

performed to produce a solvent–particle mixture with small BT NPs The

TrFE) and DMF were added to produce solutions of 1 g

P(VDF-TrFE) in 10 mL DMF with a predefi ned weight percentage of the BT NPs

with respect to P(VDF-TrFE) An inverted-staggered bottom-contact

bottom-gate OFET structure (with Ni as the gate electrode, pentacene as

the organic semiconductor, and Au as the source/drain electrodes) was

used An organic material of tetratetracontane (TTC) was deposited by

thermal evaporation at 50 °C as the passivation layer The thickness of

the NC gate dielectric layers ranged from 600–700 nm Highly crystalline

BT/P(VDF-TrFE) NC gate dielectric layers were obtained by annealing at

140 °C for 2 hrs The channel dimensions of the characterized devices

were a length (L) of 40 µ m and a width of 800 µ m

Temperature-Pressure Bimodal Sensing from Single Device : In order to

induce piezoelectricity and pyroelectricity of functional gate dielectric

that responds to pressure and temperature, an on-chip poling process

was conducted by grounded source (S) and drain (D) electrodes while

gate electrode is applied –80 V [ 36 ] Applying the pressure to the device

positioned on a heating block having a feed-back temperature control

was achieved by supplying pressurized N 2 gas in a sealed stainless

measurement chamber, while the pressure inside the chamber was

monitored by a commercial pressure gauge The temperature of the device

on the heating block was measured with a thermocouple embedded

near the device Sinusoidal AC (alternative current) V G was created by

the Tektronix AFG 3102 two-channel Arbitrary/Function Generator The

applied V G has the frequency of 0.3125 Hz, 5 V of amplitude and without

offset I D was measured with an HP 1415B semiconductor parameter

analyzer with time interval of 0.1 s and –5 V of drain bias V D with

grounded source Amplitudes and offset values of I D were calculated by

fast-Fourier-transform (FTT) method with 64 measured I D values

Temperature-Strain Bimodal Sensing from Single Device : Bending

of the heated devices was carried out in a custom-built bending

system with the polyimide heater attached to the backside the fl exible

polyimide substrate having the devices by using a heat-conductive epoxy

Methodology for estimation of strain in strain-temperature bimodal

sensing was shown in Equation S3 For electrical measurements, the

same method as the temperature-pressure bimodal sensing was used

Piezoelectric-Force-Microscopy Measurement : In order to measure

and characterize piezoelectric materials in the nanoscale regime,

piezoelectric-force-microscopy (PFM) equipped with the function

generator and lock-in amplifi er (E-Sweep, SII) was employed Surface

topography was obtained in atomic force microscope mode while

piezo-response image was obtained in piezo-responsive mode to detect

piezoelectric vibrations PFM measures the mechanical response when

an electrical voltage is applied to the sample surface with a conductive

tip Under the electrical stimulus, the sample locally expands or

contracts If the polarization is parallel with the applied electric fi eld, the

piezoelectric effect becomes positive, and the sample will locally expand

If the polarization is anti-parallel with the electric fi eld, the sample

will shrink This sign-dependent behavior indicates that the phase of

www.MaterialsViews.com

wileyonlinelibrary.com

© 2013 WILEY-VCH Verlag GmbH & Co KGaA, Weinheim

[26] M D Angione , R Pilolli , S Cotrone , M Magliulo , A Mallardi ,

G Palazzo , L Sabbatini , D Fine , A Dodabalapur , N Cioffi , L Torsi ,

Mater Today 2011 , 14 , 424 [27] S Liu , X Guo , NPG Asia Mater 2012 , 4 , e23 [28] P Lin , F Yan , Adv Mater 2012 , 24 , 34

[29] W Yang , K R Ratinac , S P Ringer , P Thordarson , J J Gooding ,

F Braet , Angew Chem Int Ed 2010 , 49 , 2114 [30] C Bartic , G Borghs , Anal Bioanal Chem 2005 , 384 , 354 [31] M J Schöning , A Poghossian , Analyst 2002 , 127 , 1137

[32] S M Sze , Physics of Semiconductor Devices , John Wiley , New York ,

1981

[33] B Streetman , S Banerjee , Solid State Electronic Devices , Prentice

Hall , 2005

[34] Bioelectrochemistry Fundamentals, Experimentals Techniques and

Applications , Wiley , 2008

[35] N T Tien , Y G Seol , L H A Dao , H Y Noh , N.-E Lee , Adv Mater

2009 , 21 , 910

[36] N T Tien , T Q Trung , Y G Seoul , D I Kim , N.-E Lee , ACS Nano

2011 , 5 , 7069

[37] R Larson , D C Falvo , Elementary Linear Algebra , Brooks Cole ,

2009

[38] F Alvarez , A Alegra , J Colmenero , Phys Rev B 1991 , 44 , 7306 [39] C L Choi , Ferroelectrics 1999 , 224 , 113

[40] A V Shirinov , W K Schomburg , Sens Actuators A 2008 , 142 ,

48 [41] S G J Mathijssen , M Kemerink , A Sharma , M Cölle , P A Bobbert ,

R A J Janssen , D M de Leeuw , Adv Mater 2008 , 20 , 975

[15] T Someya , Proc Natl Acad Sci 2004 , 101 , 9966

[16] I Graz , M Krause , S Bauer-Gogonea , S Bauer , S P Lacour ,

B Ploss , M Zirkl , B Stadlober , S Wagner , J Appl Phys 2009 , 106 ,

034503

[17] D H Kim , N Lu , R Ma , Y S Kim , R H Kim , S Wang , J Wu ,

S M Won , H Tao , A Islam , K J Yu , T i Kim , R Chowdhury ,

M Ying , L Xu , M Li , H J Chung , H Keum , M McCormick , P Liu ,

Y W Zhang , F G Omenetto , Y Huang , T Coleman , J A Rogers ,

Science 2011 , 333 , 838

[18] F V Chantz , U Schoneberg , W Brockherde , P Kopytsynski ,

T Mehlhorn , F Obermeter , H Benzel , Sens Actuators A 1992 , 34 , 77

[19] D Hammerschmidt , F V Schnatz , W Brockherde , B J Hosticka ,

E Obermeier , IEEE International Solid-State Circuits Conference 1993 ,

93 , 128

[20] S C Kang , S Y Kim , S B Lee , S W Kwon , S S Choi , B Lee , IEEE

Photon Technol Lett 1998 , 10 , 1461

[21] Y S Hsu , L Wang , W F Liu , Y J Chiang , IEEE Photo n Technol

Lett 2006 , 18 , 874

[22] S M Nalawade , S S Harnol , H V Thakur , IEEE Sensors Journal

2012 , 12

[23] S.-I Park , J.-H Ahn , X Feng , S Wang , Y Huang , J A Rogers , Adv

Funct Mater 2008 , 18 , 2673

[24] M Schreiter , R Gabl , J Lerchner , C Hohlfeld , A Delan , G Wolf ,

A Blüher , B Katzschner , M Mertig , W Pompe , Sens Actuators B

2006 , 119 , 255

[25] A N Sokolov , B C.-K Tee , C J Bettinger , J B.-H Tok , Z Bao , Acc

Chem Research 2011 , 45 , 361

Adv Mater 2013,