RESEARCH ARTICLE

www.afm-journal.de

Multifunctional Thermoelectric Temperature Sensor for

1. Introduction
Using sensing devices for noncontact information transfer enhances data
security in human-machine interaction by eliminating direct contact between The booming development of artificial
intelligence, augmented reality, and soft
the information carrier and the communication interface. However, current
robotics has driven the demands for flexi-
contactless information transfer strategies struggle with complex application ble wearable sensors,[1] which can not only
scenarios due to limitations in sensing mechanisms. Here, this work proposes perceive external environmental stimuli,[2]
an innovative noncontact information transfer approach that leverages invisible but also serve as communication interfaces
thermal radiation and the Seebeck effect, and demonstrates its feasibility for extracting, decoding, and transmitting
by designing a high-resolution temperature sensor based on a laminar information in human-machine interac-
tion (HMI).[3] As intelligent electronic
thermoelectric aerogel of carbon nanotube/PEDOT:PSS/nanocellulose. The
devices, sensors can convert applied exter-
sensor exhibits exceptional sensitivity, ultralow detection limits (0.02 K), rapid nal stimuli into detectable electrical signals
response time, and reliable cycling stability in temperature detection without through various sensing mechanisms,
interference from pressure stimuli. This work further presents a sensor such as piezoresistive,[4] piezoelectric,[5]
array-based communication interface capable of extracting, decoding, and and triboelectric.[6] By assigning spe-
cific meanings to these electrical signals,
transmitting high-capacity encrypted information contactlessly. The interface
encrypted-information reading can be
also serves as a contactless input terminal for precisely monitoring finger achieved by detecting customized stimuli
movements. Additionally, the sensor can convert pressure into resistance with a sensor. For instance, Morse code can
signals, functioning as tactile electronic skin in contact scenarios. This be sent by regularly pressing a sensor with
study presents a dependable strategy for transmitting information without a finger.[7] In the era of the information
physical contact, contributing to the development of future integrated devices. explosion, more stringent requirements are
placed on the security of information trans-
fer during human-machine interactions.
Noncontact information transmission
F.-L. Gao, P. Min, Z.-Z. Yu, X. Li techniques are gaining increasing attention because they can ef-
State Key Laboratory of Organic-Inorganic Composites fectively avoid the risk of information disclosure by eliminat-
College of Materials Science and Engineering ing the physical contact between the information carrier (the
Beijing University of Chemical Technology customized stimulus source) and the sensing device.[8] Further-
Beijing 100029, China
E-mail: xfl[email protected] more, contactless interaction can also prevent bacterial infection
Q. Ma and virus transmission.[9] The common approach of available
State Key Laboratory of Chemical Resource Engineering sensors for noncontact information transfers is to utilize invisible
Beijing University of Chemical Technology environmental stimuli (e.g., temperature,[10] magnetic field,[11]
Beijing 100029, China and humidity[12] ) to induce changes in physical parameters (e.g.,
T. Zhang, Z.-Z. Yu capacitance,[13] resistance[14] ) of the sensors. For example, Qian
Beijing Key Laboratory of Advanced Functional Polymer Composites
et al. introduced a capacitive magnetic sensor based on a ciliary-
Beijing University of Chemical Technology
Beijing 100029, China like inclined flexible micropillar array for contactless, efficient,
J. Shang, R.-W. Li convenient, and programmable information communication.[13]
CAS Key Laboratory of Magnetic Materials and Devices Liu et al. proposed a resistive humidity sensor based on hexag-
Ningbo Institute of Materials Technology and Engineering Chinese onal boron nitride nanosheets for noncontact Morse code
Academy of Sciences communication.[14] Despite these achievements, in complex ap-
Ningbo 315201, China
plication scenarios where interference from unrelated objects
The ORCID identification number(s) for the author(s) of this article and multiple environmental stimuli are present, these methods
can be found under https://doi.org/10.1002/adfm.202309553 may cause overlapping signals and reduce the accuracy of the
DOI: 10.1002/adfm.202309553 derived information.[8a] Till now, reliable contactless extraction

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (1 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

and transmission of encrypted information using wearable sens- handwritten Greek letters with the aid of machine learning. Ad-
ing devices for efficient HMI is still a major challenge. ditionally, unlike conventional rigid TE temperature sensors, the
Well-selected sensing mechanisms and stimulation sources excellent resilience and fatigue resistance of aerogel allows the
are at the core of solving the above problems.[9c] The Seebeck CPN sensor to detect pressure through the piezoresistive effect,
effect, which refers to the generation of thermoelectric (TE) volt- and the sensing mechanism based on the lamellar structure
ages arising from temperature differences,[15] is an essential phe- allows the sensor to detect much lower pressures (12 Pa). This
nomenon within the TE field. Interestingly, apart from direct dual-mode operating mechanism of temperature and pressure
contact, invisible thermal radiation can also induce temperature sensing makes the CPN sensor promising for tactile electronic
variations that change the output voltage of the TE materials.[16] skin as well.
It is important to note that the generated TE voltage is depen-
dent on the intrinsic Seebeck coefficient of the active material 2. Results and Discussion
and the temperature difference between the device ends.[17] In
contrast, other physical stimuli such as pressure have minimal 2.1. Formation of Laminar-Structured CPN Aerogel and Its
impact on the TE voltage,[18] which is advantageous for avoiding Structural Superiority
overlapping output signals and improving information reliabil-
ity. Combining thermal radiation and the Seebeck effect provides Figure 1a illustrates the fabrication of the laminar-structured
a novel platform for noncontact information transfer, but its fea- CPN aerogel with a bidirectional freezing process followed by
sibility has not been fully explored. TE temperature sensors that freeze-drying and thermal treatment. A homemade stepped
convert temperature stimuli into voltage signals for temperature PDMS cavity (Figure S1, Supporting Information), composed of
detection hold potential as electronic devices for implementing a small cavity on the left (Cavity I) and a large cavity on the right
this strategy. However, existing TE temperature sensors typically (Cavity II), is placed on a copper bridge that is immersed in liquid
have low detection limits, making them challenging to generate nitrogen and ice water at both ends, respectively. Subsequently,
reliable voltage signals for tiny temperature variations caused by an aqueous suspension consisting of CNT, PEDOT:PSS, NFC,
thermal radiation. Thus, it is necessary to develop a sensor with and 3-glycidyloxypropyltrimethoxysilane (GOPS) is injected into
highly accurate temperature resolution to achieve information the cavity, where laminar-shaped ice crystals grow along both X-
transmission based on thermal radiation and the Seebeck effect. and Z-directions due to the generated temperature gradients.[22]
Herein, we report a multifunctional TE temperature sen- This well-designed PDMS cavity facilitates well-controlled ice
sor based on a designed carbon nanotube (CNT), poly(3,4- crystal growth by providing a buffer zone (Cavity I) for the transi-
ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), tion of isotropic ice nuclei to bidirectional ice lamellae, and avoid-
and nanofibrillated cellulose (NFC), achieving contactless ing cellular or short-range lamellar structures.[23] After freeze-
extraction, decoding, and transmission of information us- drying to cross-link the epoxy groups on GOPS with the hydroxyl
ing thermal radiation and the Seebeck effect. A special groups of PSS and NFC,[19a] the resultant aerogel is heated in an
bidirectional freezing technique is applied to fabricate the oven at 140 °C for 30 min to enhance the mechanical strength of
CNT/PEDOT:PSS/NFC (CPN) aerogel with alaminar structure, the aerogel. Notably, the cross-linking reaction does not massively
where the CNT/PEDOT:PSS components provide excellent damage the internal lamellar structure of the aerogel because of
electrically conductive and TE properties,[19] while the NFC com- the low GOPS dosage.
ponent enhances strength and resilience of the aerogel.[19a,20] As shown in Figure 1b, the CPN aerogel is ultralight with an
The three-dimensional (3D) CPN aerogel is well suitable for apparent density of as low as ≈7.5 mg cm−3 , allowing it to stand
fabricating TE temperature sensors. Firstly, unlike the randomly on the Setaria viridis beard. The internal structure of the aerogel
distributed porous structure, the long-range ordered lamellar determines its performance and application potential. For com-
structure provides the CPN aerogel with a lower thermal con- parison, isotropic freezing is used to fabricate an isotropic CPN
ductivity (0.041 W m−1 K−1 ) in the direction perpendicular to (i-CPN) aerogel. As shown in Figure S2a (Supporting Informa-
the layers, which enables the CPN sensor to detect tempera- tion), the PDMS mold filled with CNT/PEDOT:PSS/NFC/GOPS
ture variations stably and accurately. Secondly, the interlayer suspension is immersed in liquid nitrogen for uniform nucle-
connections within the aerogel provide efficient pathways for ation and growth of ice particles. Like porous foams, the i-CPN
carrier transport, thereby reducing electrical noise and ensuring aerogel exhibits a typical porous structure with random and ir-
an ultralow detectable temperature of the sensor.[21] The CPN regularly distributed pores (Figure S2b–d, Supporting Informa-
aerogel-based temperature sensor exhibits high sensitivity (30.5 tion). In contrast, due to the generation of lamellar ice crystals in
μV K−1 ), ultralow detection limits (0.02 K), excellent stability, the initial suspension during the bidirectional freezing, the CPN
and fast response time in temperature detection without in- aerogel displays a highly aligned lamellar porous structure with a
terference from external pressure. Especially, by utilizing the scale-like morphology observed along the Z-direction (Figure 1c–
excellent TE characteristics of the sensor and its independence e), and junctions between the lamellar layers (Figure 1f). Differ-
of temperature sensing, a contactless communication interface ent from the irregularly distributed porous structure, the regular
is designed on the basis of the CPN sensor array for efficient and long-range lamellar structure can provide the aerogel with lower
invisible transfer of large-capacity encrypted information with thermal conductivity and superior resilience in the direction per-
thermal radiation and the Seebeck effect. Furthermore, by using pendicular to the laminar.[22b,23,24]
a finger as a stable thermal radiation source, the interface can The elasticity and durability of the aerogel are investigated,
serve as an intelligent contactless input terminal that can track as demonstrated in Figure 1g, the CPN aerogel has excellent re-
finger movements and achieve 100% accuracy in recognizing silience, recovering its initial strength and shape at strains of up

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (2 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 1. a) Schematic illustrating the fabrication of the CPN sensor. b) Photograph of the CPN aerogel standing on Setaria viridis beard. Scanning elec-
tron microscope (SEM) images of the CPN aerogel in c) z–y plane, d) x–y plane, e) z–x plane, and f) x–y plane. g) Photographs of the CPN aerogel during
the compression-recovery process. h) Compressive stress–strain curves of the CPN aerogel under different compressive strains. i) Cyclic compressive
stress–strain curves of the CPN aerogel under the strain of 50%.

to 70% (Figure 1h). After 200 compression cycles at a strain of based TE temperature sensor. The CPN sensor is fabricated by
50%, the stress–strain curve shows negligible plastic deformation attaching conductive copper foils to two opposing surfaces of the
(Figure 1i), indicating the remarkable fatigue resistance and sta- CPN aerogel perpendicular to the laminar orientation (Figure S3,
bility of the CPN aerogel.[25] These superior mechanical charac- Supporting Information).
teristics are primarily attributed to the enhancement of the aero-
gel backbone by the NFC and the cross-linking reaction between
GOPS and both PSS and NFC.[20] In addition, the lamellar struc- 2.2. Excellent Temperature Sensing Performances of the CPN
ture induces the inherent low thermal conductivity of the aerogel Sensor
while the junctions between the layers maintain the high electri-
cal conductivity, both of which endow the CPN aerogel with su- The Seebeck effect-based temperature sensing mechanism of
perior TE properties, making it promising for the Seebeck effect- the CPN sensor is elucidated with the finite element analysis.

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (3 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 2. a) Structural simplification of the CPN sensor based on its scanning electron microscope (SEM) image. b) Simulations of temperature and
potential distribution of the sensor under various temperature differences. c) Plot of voltage output versus temperature gradient for the CPN sensor.
d) Responses of the CPN sensor to the increase in temperature gradients. e) Voltage outputs under a subtle temperature difference of 0.02 K. f) Voltage
curve with the temperature at one end of the CPN sensor fixed (15 °C) and that of the other end changed, and corresponding infrared images. g) Voltage
outputs of the CPN sensor induced by finger temperatures of two different testers and corresponding infrared images. h) Voltage outputs of the CPN
sensor during the temperature stimulus (25 °C) cycled for 100 times. Voltage output curves of the CPN sensor under different scenarios: i) touching a
hot plate (>80 °C), and j) touching the wall of a container filled with liquid nitrogen (<−40 °C).

The 3D CPN sensor is simplified to a 2D structure based on its ture range (Figure 2d). Notably, superior to the TE temperature
scanning electron microscope (SEM) image (Figure 2a).[17] As sensors obtained from conventional processing, the minimum
shown in Figure 2b, when a temperature difference exists be- detectable temperature of the CPN sensor fabricated from bidi-
tween the two ends of the sensor, a TE voltage (Vtherm ) is gen- rectional freezing is as low as 0.02 K (Table S1, Supporting Infor-
erated, defined as ST × ΔT, where ST and ΔT are the Seebeck co- mation, and Figure 2e). Such high temperature resolution could
efficient, and temperature difference, respectively.[26] Especially, be attributed to both the inherent low thermal conductivity of
when the temperature at one end of the sensor is constant (T0 ), the layered aerogel that facilitates the sensing of tiny tempera-
the temperature stimulus (Ts ) at the other can be calculated from ture differences and the low electrical noise of the output signal
T0 + Vtherm /ST .[18a] The surface isotherm distribution of the CPN due to the high electrical conductivity sustained by the interlayer
sensor at ΔT = 60 K shows that the temperature varies rapidly junctions.[28]
at both ends while remaining close to the setting ambient tem- As illustrated in Figure 2f, where the temperature at one end of
perature (298.15 K) in the middle (Figure S4a, Supporting In- the sensor remains constant (15 °C), the temperature variations
formation), which is because most of the heat is dissipated in at the other end can be calculated from TE voltages, and the cal-
the air between the internal layers of the sensor during thermal culations are consistent with the corresponding infrared images.
conduction (Figure S4b, Supporting Information). Theoretically, Furthermore, the CPN sensor can accurately distinguish the fin-
the interlayer air hindrance to heat transfer is beneficial for the ger temperature of different individuals through the voltage out-
rapid establishment of temperature differences and the capture put (Figure 2g), suggesting its great potential for high-precision
of tiny temperature variations, allowing the fast response of the temperature detection. Additionally, when the applied tempera-
sensor to temperature variation and a lower temperature detec- ture stimulus (25 °C) is cycled for 100 times, the output voltages
tion limit.[27] remain constant within the same magnitude (Figure 2h), verify-
The Seebeck coefficient of the CPN sensor is 30.5 μV K−1 ing the excellent cyclic stability of the CPN sensor in temperature
(Figure 2c) with good reproducibility (Figure S5, Supporting In- sensing.
formation), and the output voltage exhibits a linear increase with Consistent with the simulations, the CPN sensor exhibits
the temperature gradient ranging from 11 to 46 K, indicating an excellent immediate stimulus response to external tempera-
that the ST of the sensor remains stable over a wide tempera- ture changes due to its low thermal conductivity. As shown in

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (4 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure S6 (Supporting Information), the response time of the sensors are integrated into an array, the ternary coding system
sensor to the temperature difference of 45 K is less than 1 s (0.97 is more efficient for large-capacity information transfer than a
s). Furthermore, as illustrated in Figure 2i,j, where the sensor is simple binary coding system[13,29] (Figure 3e). Therefore, a 4 × 4
attached to the robot finger and used to touch the hot table and Peltier element array (Figure S8, Supporting Information) is de-
the wall of a container filled with liquid nitrogen, the obtained signed as a customized stimulus source (information carrier),
TE voltages exhibit significant variations within very short times where each element in the array can be considered as an indi-
of 0.34 and 0.45 s, respectively. These results demonstrate that vidual heat or cold source providing thermal radiation. By con-
the CPN sensor can be applied for high-cold temperature warn- trolling the temperatures of each element with DC supplies, the
ings, which is beneficial for avoiding damage of electronics and large-capacity information to be transmitted can be encoded in
prolonging the service life of devices. the thermal field generated from the array. Correspondingly, a
4 × 4 CPN sensor array-based communication interface (Figure
S9, Supporting Information) is also developed to extract and de-
2.3. Noncontact Information Transmission Based on Thermal code the encrypted information within the customized stimulus
Radiation and the Seebeck Effect source.
To demonstrate the capability of the communication interface
Depending on the Seebeck effect, a TE voltage can be generated to transmit large-capacity encrypted information, the Chinese
whenever a temperature difference exists between both ends of Zodiac Chronology, seasons, months, and 24 solar terms are
the sensor. In fact, besides the heat/cold source directly contact- selected as the information contents (Figure 3f), and unique
ing the sensor, thermal radiation can also cause temperature vari- codes are allocated to each piece of information using the ternary
ation at both sensor ends, inducing TE voltage generation even coding system (Figure S10, Supporting Information). The trans-
if there is no contact (Figure 3a). Our proposed CPN sensor has mission and decoding process of the encrypted messages is
outstanding temperature-sensitive properties that enable it to de- illustrated in Figure 3g. Firstly, the information is pre-encrypted
tect not only direct contact with thermal stimuli, but also has into the thermal field generated by the Peltier element array, by
the ability to perceive invisible thermal radiation. Hence, by as- applying the necessary voltages on each pixel according to the
signing specific meanings to the different signals of voltage vari- codes. After placing the customized stimulus source over the
ations, it is theoretically possible to transfer information con- sensor array (≈2.5 mm), a series of voltage signals from each
tactless through thermal radiation with the CPN sensor as the pixel of the sensor array are obtained (Figure 3h,i). Then, the re-
medium. ceived signals are decoded by calculating the difference between
To demonstrate this concept, a Peltier element positioned ap- the output voltages before and after the voltage stabilization.
proximately 2.5 mm above the sensor is employed as a heat/cold As shown in Figure 3j,k, the resultant 2D voltage heat maps
source to generate thermal radiation at room temperature. The accurately depict the pre-encrypted information of “Year of the
voltage applied to the terminals of the Peltier element is varied Rat, Spring, February, Spring begins” and “Year of the Ox, Sum-
in polarization. Specifically, when the voltage polarity is positive, mer, May, Summer begins”. Furthermore, more information
the surface temperature of the Peltier element increases above is encoded into the thermal field and transmitted through the
room temperature due to heating. Conversely, when the volt- CPN sensing array (Figure S11a,c,e,g, Supporting Information).
age polarity is negative, the surface temperature decreases below The results show that the sensing array can extract and decode
room temperature due to cooling. The resulting voltage varia- the information accurately, regardless of how the information
tions caused by thermal radiation from the heat or cold source changes (Figure S11b,d,f,h, Supporting Information). Notably,
are represented as “dashes” or “dots”, respectively, resembling a the invisibility of the thermal field and the noncontact character
conventional Morse code system (Figure 3b). By continuously ad- of the entire process ensure that the encrypted information
justing the voltage polarity of the Peltier element terminals, a se- is delivered stealthily and securely. These results demonstrate
ries of voltage variation signals are generated. Figure 3c,d vividly the promising application of the CPN sensor in the field of
depicts the transmission of Morse code from A to Z through the contactless encrypted information transfer.
thermal radiation stimulus. Furthermore, real-time voltage sig-
nals accurately identify the desired messages of “food & water” 2.4. Sensor Array-Based Contactless Intelligent Input Terminal
and “yes & no” (Figure S7, Supporting Information). These re-
sults validate that integrating thermal radiation with the Seebeck It has been demonstrated that the finger, an essential organ of
effect, which converts thermal signals into electrical signals, is a the human body, plays an important role in the next generation
reliable approach for achieving contactless information transfer. of human-machine interactions.[30] Therefore, to extend the ap-
Because it takes time to heat/cool the sensor surface by radia- plication of the 4 × 4 noncontact sensor array, we propose to use
tion and to return the surface temperature to the ambient temper- the finger as a stable thermal radiation emitter to track and recog-
ature after the removal of the heating/cooling source, it is diffi- nize finger motion trajectories. When the finger is placed above
cult to apply a single CPN sensor for efficient large-capacity infor- the sensor, one end of the sensor is heated by the emitted thermal
mation transfer. To solve such an inefficiency, we take advantage radiation, resulting in an increase in TE voltage. After the finger
of the high sensitivity of the CPN sensor to temperature stim- leaves, the TE voltage drops due to the disappearing temperature
uli and define the differences in voltage changes (ΔV) caused by difference (Figure 4a). As the finger moves over the array, the TE
heat source radiation, no radiation, and cold source radiation as voltage signals within the corresponding channels are recorded
1 (ΔV > 0), 0 (ΔV = 0), −1 (ΔV < 0), respectively, to obtain (Figure 4b). When the finger is above the pixel-1A, the induced
a ternary information coding system. Especially, when multiple voltage on the channel-1A increases rapidly, generating a valid

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (5 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 3. a) Schematic of generating thermoelectric (TE) voltages induced by thermal radiation and the Seebeck effect. b) Schematic of contactless
Morse code communication by varying the voltage polarity of the Peltier element. c,d) Real-time voltage signals and decoded Morse code information
of letters from A to Z. e) Schematic of the ternary coding system. f) Schematic of Chinese Zodiac Chronology, and the relationship among seasons,
months, and the 24 solar terms. g) Illustration of no-contact encrypted information transfer utilizing customized stimulus source and communication
interface. h,i) Real-time voltage signals of each channel in the CPN sensor array under two different conditions. j,k) Corresponding 2D voltage heat
maps obtained by ΔV (the voltage difference between the 30th and 5th s) to demonstrate the decoded information. To avoid experimental errors,
ΔV < −0.015, −0.015 to 0.015 mV, and >0.015 mV are considered as “−1”, “0”, “1”, respectively.

trigger signal (Figure 4c). As the finger moves to the next pixel- As a further initiative, six adjacent pixels in the array are cho-
2A, the voltage on the channel-1A gradually decreases to its orig- sen to design an intelligent contactless input terminal, aiming to
inal state, while the voltage on the channel-2A starts to increase. recognize handwritten fonts. As shown in Figure 4d, to demon-
During the finger movement, the pixels can be triggered in se- strate the recognition ability of the input terminal, we choose six
quence. Thus, tracking the position and sequence of pixels being common Greek letters of “𝛼, 𝜌, 𝜖, 𝜂, 𝛾, 𝜄”, and establish a system-
triggered within the array enables the identification of the finger atic association between the pixel-triggered sequence and the spe-
motions, which is of significance for developing contactless HMI cific Greek letter based on the handwriting orders. Figure 4e–g
systems. and Figure S12 (Supporting Information) show the 6-channel

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (6 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 4. a) Schematic of the finger as a thermal radiation emitter to induce thermoelectric (TE) voltage variations. b) Photographs of finger movements
over the CPN sensor array and c) corresponding output voltage signals for each channel. d) Schematic of intelligent contactless input terminal based
on CPN sensor array for recognition of handwritten Greek letters. The 6-channel voltage signals when the array identifies Greek letters of e) 𝛼, f) 𝜌, and
g) 𝜖. h) Flow chart of a machine learning algorithm based on convolutional neural networks for Greek letters classification. i) Confusion matrix of the
verification result.

voltage signals corresponding to the 6 Greek letters of “𝛼, 𝜌, 𝜖, the dataset used for training and testing for each letter in a ra-
𝜂, 𝛾, and 𝜄”. It is apparent that the order of the pixels being trig- tio of 4:1 (Figure 4h). After data processing, the voltage-signal
gered coincides with the expected sequence. sets for each letter are trained and validated by the CNN clas-
To facilitate human-machine interaction, a machine learning sification algorithm, resulting in a confusion matrix indicating
algorithm based on Convolutional Neural Networks (CNN) is de- 100% classification accuracy (Figure 4i). The model obtained
signed to train and classify data obtained from recognizing char- from the CNN algorithm shows a high degree of generalization,
acters. For accurate recognition, the voltage signal correspond- thereby avoiding overfitting, as shown by the learning accuracy
ing to each letter is tested 40 times to build the database, with and loss function (Figure S13, Supporting Information).[31] More

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (7 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 5. a) Output voltages of the CPN sensor with increasing the strain at different ΔT. b) Seebeck coefficients of the CPN sensor at different strains.
c) Real-time monitoring of the Morse code transmission of “J” under the wire interference and “O” under the interference of nuts. d) Plot of resistance
change versus pressure for the CPN sensor. e) ΔR/R0 variations and f) I–V curves of the CPN sensor at different pressures. g) Relative resistance changes
(ΔR/R0 ) of the CPN sensor with increasing ΔT at different strains of 50%, and 60%. h) A schematic diagram of regularly pressing the CPN sensor with
a finger, and (i) corresponding changes in both output voltage and ΔR/R0 .

importantly, this model can be easily integrated into modern elec- is only related to the temperature gradient without being affected
tronic devices, which greatly broadens the application scenario of by the strain, and the Seebeck coefficient of the sensor shows only
the contactless input terminal we proposed.[32] These results in- slight variations under different strains (Figure 5b), which indi-
dicate that an intelligent input terminal, designed on the basis cates that the temperature sensing of the sensor based on the
of the thermal radiation emitted by the human finger and the Seebeck effect is unrelated to its own deformation or externally
Seebeck effect, offers a novel and efficient path for contactless applied pressure. Benefiting from the temperature response in-
human-machine interaction. dependence, the CPN sensor can stably perceive the customized
thermal radiation stimuli to extract Morse codes, even in the pres-
ence of irrelevant object interferences, suggesting that our pro-
2.5. Independent Temperature-Pressure Sensing Characteristics posed solution for noncontact transfer can be adapted to com-
of the CPN Sensor plex scenarios (Figure 5c). Furthermore, it is interesting to find
that external pressure not only has no effect on the temperature-
It is critical for the communication interface and input terminal sensitive characteristics of the CPN sensor, but is also detectable
in HMI to minimize interference from external environmental due to the unique laminar structure and resilience of the sensor
stimuli to achieve high reliability of the resulting electrical sig- (Figure S14, Supporting Information).
nals. Thus, the independence of the CPN sensor for temperature The pressure sensitivity is defined as S = 𝛿(ΔR/R0 )/P, where
detection is explored. As shown in Figure 5a, the output voltage ΔR denotes the change in electrical resistance, R0 refers to the

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (8 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

Figure 6. a) Homemade PDMS cubes are placed in specific positions on the E-skin array to identify “BUCT,” with physical photographs (left column)
and corresponding 3D pressure distribution maps (right column). b) Cold water, e) hot water, h) cold water and hot water and PDMS cubes used as
e-skin stimulants, c,f,i) corresponding temperature distributions, and d,g,j) corresponding pressure distributions.

initial resistance, and P is the applied pressure.[33] As shown 2.6. Sensor Array-Based Tactile Electronic Skin
in Figure 5d, the CPN sensor has high pressure sensitivities
of −4.48% and −1.95% kPa−1 for 0–2.5 and 2.5–4 kPa, respec- A 5 × 5 sensor array is fabricated to explore the application of the
tively, and exhibits satisfactory repeatability for pressure detec- CPN sensor in the tactile E-skin field (Figure S18, Supporting In-
tion (Figure S15, Supporting Information). For dynamic pres- formation). Figure 6a shows the 3D pressure distribution maps
sures, the CPN sensor can output real-time, and stable resistance obtained by adjusting the positions of the homemade PDMS
signals (Figure 5e) with good linear ohmic behavior (Figure 5f). cubes (each cube is ≈0.4 g) placed on the E-skin array, which
The CPN sensor can detect pressure as low as 12 Pa, and ex- identifies the information of “BUCT” accurately. In contrast, the
hibits a fast response-recovery time and excellent stability to pres- temperature distribution maps show no obvious changes since
sure (Figure S16, Supporting Information). Moreover, the rela- the PDMS cubes do not introduce temperature stimuli (Figure
tive resistance changes of the sensor are only relevant to the ap- S19, Supporting Information).
plied compressive strain without being influenced by the tem- To achieve simultaneous temperature and pressure stimula-
perature gradients (Figure 5g), demonstrating the temperature- tions, plastic containers filled with cold and hot water are placed
irrelevance of the CPN sensor for pressure sensing based on at different locations in the array (Figure 6b,e). Corresponding
the piezoresistive effect.[18b] The signal-decoupling mechanism temperature and pressure distributions (Figure 6c,d,f,g) are ac-
of the CPN sensor, that is, the voltage signal generated by the tem- curately obtained by measuring the TE voltages and resistance
perature gradients reflects the carrier energy difference, while the changes of each pixel. Additionally, Figure 6h illustrates a more
resistance signal produced by the pressure stimulus represents complex environmental stimulus, where plastic containers with
the change in the carrier migration path, and these two signals cold and hot water, and homemade PDMS cubes are all placed
do not interfere with each other.[34] Furthermore, the influence on the electronic skin array. Thanks to the excellent separate
of humidity on the temperature sensing performance of the CPN temperature-pressure sensing characteristics of the CPN sensor,
sensor is investigated, providing further evidence of its indepen- the relevant temperature and pressure mappings are observed
dent temperature detection capability (Figure S17, Supporting In- correctly (Figure 6i,j). These results demonstrate the potential of
formation). the CPN sensor in the contact E-skin field.
Apparently, the unique noninterference temperature-pressure
sensing feature of the CPN sensor allows for multiple application 3. Conclusion
scenarios. As shown in Figure 5h,i, the sensor can precisely cap-
ture temperature and pressure variations generated by the finger We propose, for the first time, a strategy for noncontact infor-
pressing Morse code at room temperature. This accurate and in- mation transfer using the Seebeck effect and invisible thermal
dependent detection of temperature-pressure changes is the ba- radiation in human-machine interaction, and confirm its effec-
sis for electronic skin (E-skin).[1c,35] tiveness and feasibility with a well-designed multifunction TE

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (9 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

temperature sensor, which is based on a CNT/PEDOT:PSS/NFC temperature differences, and the measured data from six samples were
lamellar aerogel fabricated by bidirectional freezing. The aerogel- employed to calculate the average Seebeck coefficient. To investigate the
based CPN sensor exhibits high sensitivity (30.5 μV K−1 ) and pressure-sensing performance of the sensor, a Mark-10 ESM303 mechan-
ical tester was applied to provide forces of different frequencies and mag-
resolution (0.02 K), fast response time, and good cycling stabil-
nitudes, and a Mark-10 force gauge (range: 0–0.5 N; accuracy: 0.0001 N)
ity in temperature sensing. More importantly, benefiting from was used to record the dynamic forces during compression under exter-
the excellent ability of the CPN sensor to perceive temperature nal load. The resistance changes were recorded by the Keithley DMM7510
changes, a CPN sensor array-based communication interface is multimeter. The current–voltage (I–V) characteristics of the sensor were
designed to enable the secure and efficient contactless transmis- measured with a Keithley 4200-SCS semiconductor characterization sys-
sion of large-capacity encrypted information using thermal radia- tem. Details of the finite element analysis and machine learning were pre-
sented in the Supporting Information.
tion and the Seebeck effect. Furthermore, a contactless input ter-
minal assisted by machine learning is developed to track finger
movements and recognize handwritten Greek letters with 100% Supporting Information
accuracy, achieving contactless human-machine interaction us-
Supporting Information is available from the Wiley Online Library or from
ing the finger as the medium. In addition, the CPN sensor also the author.
has the capability of pressure sensing and does not interfere with
temperature detection, promising an application for integrated
electronic skin in contact model as well. Acknowledgements
Financial support from the National Natural Science Foundation of China
(52090034, U22A20248), National Natural Science Foundation of China
4. Experimental Section Innovative Research Group Project (52221006), and the Fundamental Re-
search Funds for the Central Universities (XK1802) was gratefully acknowl-
Materials: PEDOT:PSS was used as an aqueous solution (Heraeus edged.
Clevios, PH1000, 1.3 wt%). Single-walled carbon nanotubes (model
101364, purity >95%, 5–30 mm length, and 1–2 nm diameter) were
purchased from XF NANO (China). Nanofibrillated cellulose (NFC, Conflict of Interest
carboxymethylated, 0.5 wt%) and 3-glycidyloxypropyltrimethoxysilane The authors declare no conflict of interest.
(GOPS, A.R., 97%) were supplied by Aladdin (China).
Fabrication of CPN Aerogels and i-CPN Aerogels: The CNT, PEDOT:PSS,
NFC, and GOPS were mixed in a dispersion with a solid content weight ra- Data Availability Statement
tio of 2.4:0.6:7:1. Then, after stirring for 30 min at room temperature, the
mixture was treated with probe-sonication for 15 min to obtain homoge- The data that support the findings of this study are available from the cor-
neous dispersion, and the total solids content of the final mixed dispersion responding author upon reasonable request.
was about 7.5 mg mL−1 . To prepare aerogels with a long-range laminar
structure, the mixtures were injected into homemade molds composed of
a stair-like PDMS cavity and a copper bridge. The copper bridge was im- Keywords
mersed at one end in liquid nitrogen while the other in an ice-water mix- bidirectional freezing, contactless information transmission, lamellar
ture. Following complete freezing of the mixtures, the generated ice crystal aerogels, temperature-pressure tactile sensors, the Seebeck effect
pillars were sublimated by a freeze-dryer (−68 °C, 0.2 Pa). After freeze-
drying, the obtained aerogels were placed in an air-atmosphere oven
(140 °C) for 30 min for the thermal processing. For comparison, isotropic Received: August 12, 2023
freezing was applied to fabricate isotropic CPN (i-CPN) aerogels. The mix- Revised: August 26, 2023
tures were added into PDMS containers and completely immersed in liq- Published online: September 15, 2023
uid nitrogen. After fully freezing, the frozen suspensions were freeze-dried
and thermotreated with similar fabrication conditions to the CPN aerogels
obtained by bidirectional freezing.
Fabrication of CPN Sensor, Peltier Elements Array, Contactless Sensing Ar- [1] a) J. C. Yang, J. Mun, S. Y. Kwon, S. Park, Z. Bao, S. Park, Adv. Mater.
ray, and Tactile E-Skin Array: To obtain a CPN sensor, two customized 2019, 31, 1904765; b) J. An, P. Chen, Z. Wang, A. Berbille, H. Pang, Y.
copper foils were pasted on the top and bottom surfaces of the CPN aero- Jiang, T. Jiang, Z. L. Wang, Adv. Mater. 2021, 33, 2101891; c) B. Shih,
gel with the assistance of silver paste. To obtain the Peltier elements ar- D. Shah, J. Li, T. G. Thuruthel, Y.-L. Park, F. Iida, Z. Bao, R. Kramer-
ray, 16 Peltier elements were pasted onto the aluminum plate in a 4 × 4 Bottiglio, M. T. Tolley, Sci. Rob. 2020, 5, eaaz9239; d) L. Li, S. Zhao, W.
arrangement. For preparing the 4 × 4 contactless sensing array, 16 CPN Ran, Z. Li, Y. Yan, B. Zhong, Z. Lou, L. Wang, G. Shen, Nat. Commun.
aerogels were adhered to the patterned copper electrodes with silver paste. 2022, 13, 5975.
For the 5 × 5 tactile electronic skin array, 25 CPN aerogels were attached
[2] S. Chun, J.-S. Kim, Y. Yoo, Y. Choi, S. J. Jung, D. Jang, G. Lee, K.-I. Song,
to patterned copper electrodes using silver paste.
K. S. Nam, I. Youn, D. Son, C. Pang, Y. Jeong, H. Jung, Y.-J. Kim, B.-D.
Characterization: The morphologies and microstructures of CPN and
Choi, J. Kim, S.-P. Kim, W. Park, S. Park, Nat. Electron. 2021, 4, 429.
i-CPN aerogels were characterized with a Hitachi S4700 field-emission
SEM. Compressive stress–strain curves were obtained on a Mark-10 ap- [3] a) S. Pyo, J. Lee, K. Bae, S. Sim, J. Kim, Adv. Mater. 2021, 33, 2005902;
paratus at a compressive rate of 10 mm min−1 . b) Z. Yu, P. Wu, Adv. Mater. 2021, 33, 2008479; c) X. Di, J. Hou, M.
Sensing Performance Measurements: To gauge the temperature sens- Yang, G. Wu, P. Sun, Mater. Horiz. 2022, 9, 3057.
ing capability of the CPN sensor, two Peltier elements were connected to [4] a) Y. Wei, X. Li, Y. Wang, T. Hirtz, Z. Guo, Y. Qiao, T. Cui, H. Tian,
DC supplies to customize the temperature gradient and an infrared cam- Y. Yang, T. L. Ren, ACS Nano 2021, 15, 17738; b) Y. Pang, K. Zhang,
era (FLIR E40) was used to measure the temperature with an accuracy of Z. Yang, S. Jiang, Z. Ju, Y. Li, X. Wang, D. Wang, M. Jian, Y. Zhang,
0.1 °C. The voltage signals were recorded by using a Keithley DMM7510 R. Liang, H. Tian, Y. Yang, T. L. Ren, ACS Nano 2018, 12, 2346; c) D.
7½ digit graphical sampling multimeter. The Seebeck coefficient was de- Song, G. Ye, Y. Zhao, Y. Zhang, X. Hou, N. Liu, ACS Nano 2022, 16,
termined by linearly fitting the slopes of V/ΔT plots for seven different 17168.

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (10 of 11) © 2023 Wiley-VCH GmbH
 16163028, 2024, 1, Downloaded from https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202309553 by National Yang Ming Chiao Tung University, Wiley Online Library on [01/06/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
www.advancedsciencenews.com www.afm-journal.de

[5] a) Y. H. Jung, S. K. Hong, H. S. Wang, J. H. Han, T. X. Pham, H. Park, L. Gao, J. Liu, Y. Li, P. Wan, Z. Z. Yu, X. Li, ACS Appl. Mater. Interfaces
J. Kim, S. Kang, C. D. Yoo, K. J. Lee, Adv. Mater. 2020, 32, 1904020; b) 2022, 14, 43783.
X. Cao, Y. Xiong, J. Sun, X. Zhu, Q. Sun, Z. L. Wang, Adv. Funct. Mater. [19] a) S. Han, F. Jiao, Z. U. Khan, J. Edberg, S. Fabiano, X. Crispin, Adv.
2021, 31, 2102983. Funct. Mater. 2017, 27, 1703549; b) D. Kim, Y. Kim, K. Choi, J. C.
[6] a) F. R. Fan, Z. Q. Tian, Z. L. Wang, Nano Energy 2012, 1, 328; b) Z. Grunlan, C. Yu, ACS Nano 2010, 4, 513.
Sun, M. Zhu, X. Shan, C. Lee, Nat. Commun. 2022, 13, 5224. [20] X. Wang, L. Liang, H. Lv, Y. Zhang, G. Chen, Nano Energy 2021, 90,
[7] J. Huang, J. Li, X. Xu, L. Hua, Z. Lu, ACS Nano 2022, 16, 8161. 106577.
[8] a) Q. Zhou, B. Ji, F. Hu, Z. Dai, S. Ding, H. Yang, J. Zhong, Y. Qiao, [21] L. Su, D. Wang, S. Wang, B. Qin, Y. Wang, Y. Qin, Y. Jin, C. Chang,
J. Zhou, J. Luo, B. Zhou, Adv. Funct. Mater. 2022, 32, 2208120; b) L. L.-D. Zhao, Science 2022, 375, 1385.
Li, D. Wang, D. Zhang, W. Ran, Y. Yan, Z. Li, L. Wang, G. Shen, Adv. [22] a) H. Zhuo, Y. Hu, X. Tong, Z. Chen, L. Zhong, H. Lai, L. Liu, S. Jing,
Funct. Mater. 2021, 31, 2104782. Q. Liu, C. Liu, X. Peng, R. Sun, Adv. Mater. 2018, 30, 1706705; b) M.
[9] a) H. Zhou, W. Huang, Z. Xiao, S. C. Zhang, W. Z. Li, J. H. Hu, T. X. Yang, N. Zhao, Y. Cui, W. Gao, Q. Zhao, C. Gao, H. Bai, T. Xie, ACS
Feng, J. Wu, P. C. Zhu, Y. C. Mao, Adv. Funct. Mater. 2022, 32, 2208271; Nano 2017, 11, 6817.
b) W. Pan, C. Liang, Y. Sui, J. Wang, P. Liu, P. Zou, Z. Guo, F. Wang, X. [23] P. Min, X. Li, P. Liu, J. Liu, X. Q. Jia, X. P. Li, Z. Z. Yu, Adv. Funct. Mater.
Ren, C. Yang, Adv. Funct. Mater. 2022, 32, 2204166; c) S. Li, Y. Zhang, 2021, 31, 2103703.
X. Liang, H. Wang, H. Lu, M. Zhu, H. Wang, M. Zhang, X. Qiu, Y. [24] H. Bai, Y. Chen, B. Delattre, A. P. Tomsia, R. O. Ritchie, Sci. Adv. 2015,
Song, Y. Zhang, Nat. Commun. 2022, 13, 5416. 1, e1500849.
[10] a) Z. Cao, Y. Yang, Y. Zheng, W. Wu, F. Xu, R. Wang, J. Sun, J. Mater. [25] X. Shi, X. Fan, Y. Zhu, Y. Liu, P. Wu, R. Jiang, B. Wu, H. A. Wu, H.
Chem. 2019, 7, 25314; b) Z. Wang, T. Wang, M. Zhuang, H. Xu, ACS Zheng, J. Wang, X. Ji, Y. Chen, J. Liang, Nat. Commun. 2022, 13, 1119.
Appl. Mater. Interfaces 2019, 11, 45301; c) L. Chen, X. Wei, F. Wang, [26] F. Li, Y. Liu, X. Shi, H. Li, C. Wang, Q. Zhang, R. Ma, J. Liang, Nano
S. Jian, W. Yang, C. Ma, G. Duan, S. Jiang, Chin. Chem. Lett. 2022, 33, Lett. 2020, 20, 6176.
2635. [27] a) Y. Wang, H. Wu, L. Xu, H. Zhang, Y. Yang, Z. L. Wang, Sci. Adv.
[11] a) G. S. Cañón Bermúdez, D. Makarov, Adv. Funct. Mater. 2021, 31, 2020, 6, eabb9083; b) L. Yang, Z.-G. Chen, M. S. Dargusch, J. Zou,
2007788; b) X. Wei, Q. Wu, L. Chen, Y. Sun, L. Chen, C. Zhang, S. Li, Adv. Energy Mater. 2018, 8, 1701797.
C. Ma, S. Jiang, ACS Appl. Mater. Interfaces 2023, 15, 10030. [28] S. J. Kim, H.-J. Koh, C. E. Ren, O. Kwon, K. Maleski, S.-Y. Cho, B.
[12] a) T. Li, L. Li, H. Sun, Y. Xu, X. Wang, H. Luo, Z. Liu, T. Zhang, Adv. Anasori, C.-K. Kim, Y.-K. Choi, J. Kim, Y. Gogotsi, H.-T. Jung, ACS Nano
Sci. 2017, 4, 1600404; b) L. Lu, C. Jiang, G. Hu, J. Liu, B. Yang, Adv. 2018, 12, 986.
Mater. 2021, 33, 2100218. [29] M. Lei, K. Feng, S. Ding, M. Wang, Z. Dai, R. Liu, Y. Gao, Y. Zhou, Q.
[13] Q. Zhou, B. Ji, F. Hu, J. Luo, B. Zhou, Nano-Micro Lett. 2021, 13, 197. Xu, B. Zhou, ACS Nano 2022, 16, 12620.
[14] H. Liu, J. Qin, X. Yang, C. Lv, W. Huang, F. Li, C. Zhang, Y. Wu, L. [30] a) Y. Tang, H. Zhou, X. Sun, N. Diao, J. Wang, B. Zhang, C. Qin, E.
Dong, C. Shan, Nano Res. 2023, 16, 10279. Liang, Y. Mao, Adv. Funct. Mater. 2019, 30, 1907893; b) F. Wen, Z. Sun,
[15] a) L. Zhang, X.-L. Shi, Y.-L. Yang, Z.-G. Chen, Mater. Today 2021, 46, T. He, Q. Shi, M. Zhu, Z. Zhang, L. Li, T. Zhang, C. Lee, Adv. Sci. 2020,
62; b) S. Xu, X.-L. Shi, M. Dargusch, C. Di, J. Zou, Z.-G. Chen, Prog. 7, 2000261; c) H. Guo, Y. J. Tan, G. Chen, Z. Wang, G. J. Susanto, H.
Mater. Sci. 2021, 121, 100840; c) J. Ding, W. Zhao, W. Jin, C. a. Di, D. H. See, Z. Yang, Z. W. Lim, L. Yang, B. C. K. Tee, Nat. Commun. 2020,
Zhu, Adv. Funct. Mater. 2021, 31, 2010695. 11, 5747.
[16] a) A. L. Hsu, P. K. Herring, N. M. Gabor, S. Ha, Y. C. Shin, Y. Song, [31] Y. Luo, X. Xiao, J. Chen, Q. Li, H. Fu, ACS Nano 2022, 16, 6734.
M. Chin, M. Dubey, A. P. Chandrakasan, J. Kong, P. Jarillo-Herrero, T. [32] J. Xu, X. Li, H. Chang, B. Zhao, X. Tan, Y. Yang, H. Tian, S. Zhang, T.
Palacios, Nano Lett. 2015, 15, 7211; b) C. Chen, B. Zhao, R. Wang, Z. L. Ren, ACS Nano 2022, 16, 6687.
He, J. L. Wang, M. Hu, X. L. Li, G. Pei, J. W. Liu, S. H. Yu, Adv. Mater. [33] a) H. Liu, T. Xu, C. Cai, K. Liu, W. Liu, M. Zhang, H. Du, C. Si, K. Zhang,
2022, 34, 2206364; c) X. Guo, X. Lu, P. Jiang, X. Bao, Adv. Mater. 2022, Adv. Funct. Mater. 2022, 32, 2113082; b) L. Q. Tao, K. N. Zhang, H.
34, 2204355. Tian, Y. Liu, D. Y. Wang, Y. Q. Chen, Y. Yang, T. L. Ren, ACS Nano 2017,
[17] Y. Wang, H. Mao, Y. Wang, P. Zhu, C. Liu, Y. Deng, J. Mater. Chem. 11, 8790; c) S. R. A. Ruth, V. R. Feig, H. Tran, Z. Bao, Adv. Funct. Mater.
2020, 8, 15167. 2020, 30, 2003491.
[18] a) F. Zhang, Y. Zang, D. Huang, C. A. Di, D. Zhu, Nat. Commun. 2015, [34] a) Y. Li, R. Wang, G.-E. Wang, S. Feng, W. Shi, Y. Cheng, L. Shi, K. Fu,
6, 8356; b) F. L. Gao, P. Min, X. Z. Gao, C. J. Li, T. T. Zhang, Z. Z. Yu, J. Sun, ACS Nano 2022, 16, 473; b) F.-L. Gao, J. Liu, X.-P. Li, Q. Ma, T.
X. F. Li, J. Mater. Chem. 2022, 10, 18256; c) Z. U. Khan, J. Edberg, M. Zhang, Z.-Z. Yu, J. Shang, R.-W. Li, X. Li, ACS Nano 2023, 17, 16036.
M. Hamedi, R. Gabrielsson, H. Granberg, L. Wagberg, I. Engquist, [35] J. Wu, X. Fan, X. Liu, X. Ji, X. Shi, W. Wu, Z. Yue, J. Liang, Nano Lett.
M. Berggren, X. Crispin, Adv. Mater. 2016, 28, 4556; d) X. Z. Gao, F. 2022, 22, 4459.

Adv. Funct. Mater. 2024, 34, 2309553 2309553 (11 of 11) © 2023 Wiley-VCH GmbH