3D-printed epifluidic electronic skin for machine learning–powered multimodal health surveillance

ENGINEERING

3D-printed

epifluidic

electr

onic

skin

for machine

learning

–

wered multimodal

health

surv

eillance

Yu Song

†

Roland

Yingjie

Tay

†

, Jiahong

Li, Changhao

Xu, Jihong

Min,

Ehsan

Shirzaei

Sani,

Gwangmook

Kim,

Wenzheng

Heng,

Inho

Kim,

Wei Gao*

The

amalgama

tion

of wearable

technologies

with

physiochemical

sensing

capabilities

promises

to create pow-

erful

interpr

etive and

predictiv

e platforms

for real-time

health

surv

eillance.

However, the cons

truction

of such

multimodal

devices

is difficult

to be implemented

wholly

by traditional

manufa

cturing

techniques

for at-home

personalized

applica

tions.

Here, we present

a univ

ersal

semisolid

extrusion

–

based

three-dimensional

printing

technology

to fabrica

te an epifluidic

elastic electr

onic

skin

-skin)

with

high-performance

multimodal

physi-

ochemical

sensing

capabilities.

We demons

trate that the e

-skin

can serve as a sustainable

surv

eillance

platform

to captur

e the real-time

physiological

state of individuals

during

regular

daily

activities.

We also

show that by

coupling

the informa

tion

collected

from the e

-skin

with

machine

learning,

we were able

to predict

an individ-

ual

’

degr

ee of beha

vior

impairments

(i.e.,

reaction

time

and

inhibitory

contr

ol) after alcohol

consumption.

The

-skin

paves the path for futur

e autonomous

manufa

cturing

of customizable

wearable

systems

that will enable

widespr

ead

utility

for regular

health

monitoring

and

clinical

applica

tions.

Authors,

some

rights

reserved;

exclusive licensee

American

Associa

tion

for the Advancement

of Science.

No claim to

original

U.S. Government

Works. Distributed

under

a Creative

Commons

Attribution

License

4.0 (CC BY).

INTRODUCTION

Maintaining

a well-balanced

lifestyle and effectiv

e recognition

premedical

symptoms

to obtain

early intervention

is paramount

to sustain one

’

s physical well-being

and attain longevity

. With the

advent of wearable technology

, traditional

healthcar

e practices

are

rapidly

changing

their course

through the implementa

tion of per-

sonalized

medicine

and digital

health

(

–

Skin-interfa

ced wear-

able devices

that deliver intima

te details

relating to the users

’

health

status in real-time

are deemed

integral enablers

to this endeavor (

–

). Real-time

tracking of vital signs such as heart rate and body tem-

perature from the skin provides insightful

informa

tion on the phys-

iological

condition

of the human

body. On the other side, in situ

microfluidic

sampling

and analysis of sweat, a key noninvasiv

ely ac-

cessible

body fluid,

could

offer rich biomolecular

informa

tion

closely

associa

ted with our health

state (

–

To this end, there

is an unprecedented

need

to develop multimodal

wearable

systems with both molecular

sensing

and vital sign tracking capabil-

ities for more compr

ehensiv

e informa

tion of our bodily

responses

(

). Such multimodal

data, when

coupled

with modern

data

analysis approaches (such as machine learning),

will enable

numer-

ous practical health

surveillance

and clinical

applica

tions (

Despite

the high demand,

fabrica

tion and integration of trans-

disciplinary

modules

for such wearable device

involve processes

that use highly

customizable

materials

and designs.

For example,

patterned

nanoma

terials

and composites

are often used to increase

the active surface area of electrochemical

sweat sensors

for en-

hanced

sensing

capabilities

(

); use of various

biorecogni-

tion molecules

(e.g., enzymes

and ionophor

es) in a polymer

matrix

is often necessary

for selectiv

e detection

of specific

biomark

ers (e.g.,

metabolites

and ions) (

). Polymeric

hydrogels are commonly

patterned

on the electrodes for transdermal

deliveryof the nicotinic

agents

(e.g.,

pilocarpine

and carbachol)

via iontophor

esis for

autonomous

sweat induction,

while microfluidic

channels

that reg-

ulate and sample

the sweat flow are typically

fabrica

ted through

polymer

molding

or laser cutting

of plastic films (

). Con-

versely, three-dimensional

(3D) micro/nanos

tructur

es are often re-

quired for highly

sensitiv

e pressure and strain sensing

with

compr

essible

and stretchable

functionalities

(

). Hence,

the in-

corpor

ation of such comple

x fabrica

tion, which

encompasses

diverse range of materials

and processes,

traditionally

requires the

combina

tion of a series

of conventional

cleanr

oom facilities

and

manufa

cturing

technologies.

Moreover, complementary

laborious

interventions

such as manual

deposition

and assembly

are usually

performed

at a laboratory scale. The development

of a scalable

and

customizable

prototyping

and fabrica

tion method

that caters to the

aforementioned

fabrica

tion needs

would

be vital for the future

widespr

ead implementa

tion of multimodal

wearable sensors

in per-

sonalized

healthcar

e but has not been realized

yet.

To addressthese challenges,

here, we present an epifluidic

elastic

electronic skin (e

-skin)

with multimodal

physiochemical

sensing

capabilities,

which

is constructed

exclusiv

ely using a highly

adapt-

able and versatile semisolid

extrusion

(SSE)

–

based

3D-printing

technology

involving

direct ink writing

and selectiv

e phase

elimina-

tion (Fig. 1A). This 3D-printed

-skin,

coupled

with machine

learning,

enables

remote

multimodal

personalized

health

assess-

ment.

To prepare the e

-skin with optimal

performance

for on-

body biosensing,

epifluidic

modula

tion, and energy

efficiency

functional

inks comprising

various

multidimensional

nanoma

teri-

als, polymers,

and hydrogels are custom-tailor

ed to pattern all mul-

tidimensional

architectur

es in the wearable system with high

precision

(figs. S1 and S2). All inks were formula

ted to fulfill

the

desirable rheological

properties

for SSE, which

necessita

te suitable

viscoelas

ticity and shear-thinning

behaviors as instructed

by the

choice

of materials

combina

tion (Fig. 1B, fig. S1, and table S1)

(

). A phase

elimina

tion strategy, which

involves the selectiv

removal of the sacrificial

component

in the ink (

–

was used to

transform

as-printed

3D filaments

into porous architectur

es to

enhance

the performance.

This technology

enables

low-cost

Andrew and Peggy Cherng

Department

of Medical

Engineering,

California

Insti-

tute of Technology

, Pasadena,

CA 91125,

USA.

*Corresponding

author.

Email:

weigao@caltech.edu

†

These

authors

contributed

equally

to this work.

SCIENCE

ADVANCES

RESEARCH

ARTICLE

Song

et al.

Sci.

Adv.

, eadi6492

(2023)

13 September

2023

1 of 13

Downloaded from https://www.science.org at California Institute of Technology on September 29, 2023

customizable

prototyping

of sustainable

and wearable multifunc-

tional

physiochemical

sensing

systems

via a simple

process,

ideally

suitable

for remote

healthcar

e surveillance

(fig. S2).

The SSE-based

3D-printed

-skin is composed

of an array of

electrochemical

sweat biosensors

(e.g., glucose,

alcohol,

and pH

sensors)

and biophy

sical sensors

(e.g.,

temper

ature and pulse

sensors),

a pair of hydrogel-coa

ted iontophor

esis electrodes for lo-

calized

sweat induction,

a microfluidics

forefficient

sweatsampling,

and a micro-super

capacitor (MSC)

as energy

storage module

inter-

faced with an energy-harv

esting device

(e.g., solar cell) for sustain-

able wearable operation (Fig. 1C). Further

integrated with wireless

electronic module,

the e

-skin could

perform

prolonged

physio-

chemical

data collection

from the daily activities

(Fig. 1, D and

E). Such multimodal

data collection,

coupled

with machine

learn-

ing

–

based

data analytics,

opens

the door to a wide range of person-

alized

healthcar

e applica

tions in the era of digital

health.

As an

exemplar,

we demons

trate that simultaneous

monitoring

of the

pulse waveform,

temper

ature, and alcohol

levels using

a machine

learning

–

coupled

-skin is able to accurately predict an individu-

’

s behavior response

(Fig. 1F).

RESUL

3D-printed

biophy

sical

sensors

The interconnects

and biophy

sical sensors

in the e

-skin were pre-

pared primarily

based

on high-pr

ecision

SSE with an aqueous

(MXene)

ink (Fig. 2A, figs. S3 and S4, and movie S1).

Owing

to the bidimensionality

with high aspect

ratio, negatively

charged

surfaces, and intrinsic

hydrophilicity

, mono-

to few-layer

MXene

nanosheets,

with an average lateral size of 2.63 μm (fig.

S5), feature strong electrostatic repulsion

properties,

making

them

highly

dispersible

and stable in water (

–

The printed

line-

width

of the MXene

filaments

can be modula

ted by tuning

the pres-

sure and speed

of the extrusion

printer.

Uniform

arrays of intrica

lines could reach a minimum

average linewidth

of 160 μm and a line

gap down to 10 μm (Fig. 2, B and C, and fig. S4). The MXene

fila-

ments

can be readily

printed

onto a variety

of flexible substrates, as

evidenced

by the identical

MXene

fingerprints

in Raman

spectra (Fig. 2D).

In addition

to its applica

tions as interconnects,

MXene

was also

used as an active material

for wearable temper

ature sensing

(Fig. 2E). In the e

-skin,

an MXene-based

temper

ature sensor

was

patterned

by adopting

a strain-insensitiv

e serpentine

design

withstand the stress during

daily wear (Fig. 2F). With a linewidth

of 350 μm, it exhibits

a negative temper

ature coefficient

behavior

Fig. 1. SSE-based

3D-printed

-skin.

(

) Schema

tic illustration of the SSE-based

3D printing

that features highly

customizable

inks based

on versatile materials

construct all main building

blocks

of the wearable e

-skin with multimodal

sensing

and power management

capabilities.

(

) Schema

tic illustration of SSE printing

pro-

cedures to prepare 2D and 3D architecture

s. Top right inset, typical

rheological

properties

of printable

inks; bottom,

optical

images

of as-printed

2D and 3D MXene

architectur

es. G

, storage modulus;

, loss modulus.

Scale bars, 2 mm. (

) Schema

tic illustration of the 3D-printed

-skin that comprises

multiplex

ed biophy

sical and

biochemical

sensors

forpulsewaveform,

temper

ature,andsweatbiomark

ermonitoring,

amicrofluidic

iontophor

eticmodule

forlocalized

automa

ticsweatinduction

and

sampling,

and MSCs forenergy

storage. (

and

) Optical

images

of ane

-skin (D)and afullyassembled

wirelesse

-skinsystem (E)worn ona human

subject.

Scale bars,1

cm. (

) Machine learning

–

wered multimodal

-skin for personalized

health

surveillance.

AI, artificial

intelligence.

SCIENCE

ADVANCES

RESEARCH

ARTICLE

Song

et al.

Sci.

Adv.

, eadi6492

(2023)

13 September

2023

2 of 13

Downloaded from https://www.science.org at California Institute of Technology on September 29, 2023

with a sensitivity

−

1.07%

°C

−

across a physiologically

relevant

range of 25° to 50°C (Fig. 2G). Other

printed

temper

ature sensors

withdifferentlinewidths

showedsimilar

sensitivities

(fig.S6).Stable

temper

ature sensing

performance

with fast response

time was ob-

served during

mechanical

bending

tests and upon

placing onto

human

skin (Fig. 2H and fig. S7).

The e

-skin

’

pulse monitoring

capability

was based

on a pres-

sure sensor

composed

of an interdigital

MXene

electrode and a

porous carbon

nanotube

(CNT)

–

polydimethylsiloxane

(PDMS)

foam as the active sensing

component

(Fig. 2I). The latter was pre-

pared via 3D printing

with a customizable

ink containing

homoge-

neously

mixed

PDMS,

CNT, and finely

grounded

salt

microparticles,

followed by selectiv

e salt removal to form

the

porous structur

e with an average pore size of 30 μm (Fig. 2J and

fig. S8A).Suchhigh porosityplays acrucial

roleto realize high-pr

es-

sure sensitivity

(fig. S8, B and C) and can be optimized

by control-

ling the size of salt microparticles

and varying

the compositional

ratio of CNT-PDMS

and salt to fulfill

the rheological

criteria

for

SSE with robust mechanical

stability

(figs. S9 to S11). Although

multiple

3D-printed

surface architectur

es (e.g., cone, semi-cylinder,

and cross-line

architectur

es) were able to improve the sensitivity

(Fig. 2K and fig. S12), pressure sensor

based

on a cross-line

archi-

tecture yielded

the highes

t sensitivity

due to the increased

contact

area and enabled

reliable

radial pulse monitoring

on human

sub-

jects (Fig. 2L). The printed

CNT-PDMS

foam is mechanically

resil-

ient and superelastic, demons

trating repetitiv

e and reproducible

resistance changes

under

20,000

pressing-r

eleasing

cycles

(fig. S13).

3D-printed

biochemical

sensors

The proposed

SSE-based

3D-printing

technology

was successfully

implemented

to prepare a variety

of electrochemical

biosensors

on the e

-skin (Fig. 3A). For example,

enzyma

tic biosensors

were

fabrica

ted through sequential

printing

of porous CNT

–

tyrene-bu-

tadiene-s

tyrene (CNT-SBS)

as the working

electrode (WE),

MXe-

–

Prussian

blue (MX-PB)

as the redox

media

tor, and the

bioactive polymer

[e.g.,

chitosan

and bovine serum

albumin

(BSA)]

loaded

with enzymes

[e.g., glucose

oxidase

(GOx)

and

alcohol

oxidase

(AOx)]

as the target

recognition

element.

Fig. 2. Design

and characteriza

tion of 3D-printed

interconnects

and biophy

sical sensors.

(

) Schema

ticillustrationofhigh-pr

ecision

SSE-based

3Dprinting

usingan

aqueous

MXene

ink. Inset, microscopic

image

displaying an array of 3D-printed

MXene

filaments

with narrow gaps down to 30 μm. Scale bar, 100 μm. (

) Scanning

electron microscopy

(SEM) image

of a 3D-printed

MXene

filament.

Scale bar, 100 μm. (

) Dependence

of the linewidth

of the 3D-printed

MXene

on printing

speed

and

pressure. Error bars represent the SD from five measur

ements.

(

) Raman

spectra of 3D-printed

MXene

on various

substrates. SBS, styrene-butadiene-s

tyrene; PDMS,

polydimethylsiloxane;

PET, polyethylene

terephthala

te; PI, polyimide.

(

) Optical

image

of MXene-based

temper

ature and pulse sensors.

Scale bar, 2 mm. (

) Resistive

change

of temper

ature sensors

based

on different designs

under

varying

bending

curvatures. (

) Dynamic

response

of an MXene-based

temper

ature sensor

under

varying

temper

atures (

). Inset, calibration plot within

the physiological

temper

ature range.

Error bars represent the SD from five measur

ements.

(

) Responses

the temper

ature sensor

under

mechanical

deforma

tions and periodically

changing

temper

ature. (

) Schema

tic illustration of the 3D-printed

pressure sensor

consisting

of an interdigital

MXene

electrode and a porous CNT-PDMS

active layer. (

) SEM image

of the porous 3D-printed

CNT-PDMS.

Scale bar, 50 μm. (

) Resistive responses

thepressuresensors

based

ondifferentsurfacearchitectur

esunder

applied

pressure.(

) Real-time

monitoring

oftheradialpulseofahuman

subject

usingthe3D-printed

pulse sensor

at rest and after exercise. a.u., arbitrary units.

SCIENCE

ADVANCES

RESEARCH

ARTICLE

Song

et al.

Sci.

Adv.

, eadi6492

(2023)

13 September

2023

3 of 13

Downloaded from https://www.science.org at California Institute of Technology on September 29, 2023

Considering

that high electrochemical

activity

and active surface

area are highly

desired for electrochemical

WEs,

CNT-SBS

was

chosen

over multiple

other

carbon-based

composite

inks [e.g.,

graphite-SBS

and carbon

black (CB)

–

SBS]

that fits the rheological

criteria

for SSE (fig. S14). Uniformly

distributed

porous structur

were introduced

via phase

elimina

tion of polyethylene

glycol

(PEG)

within

a printed

CNT-SBS-PEG

composite

(with optimized

CNT

and PEG ratio) to maximize

the active surface area of the WE

(Fig. 3, B and C, and fig. S15). The optimized

3D-printed

porous

CNT-SBS

WE displayed a superior

electrochemical

performance

compar

ed to othercommer

cial WEs such as glassy carbon

electrode

(GCE),

screen-printed

carbon

electrode (SPE),

and Au electrode

(AuE)

(Fig. 3D), and was able to detect

ultralow-level uric acid

(UA) in sweat through direct oxidiza

tion (fig. S16).

To prepare the enzyma

tic electrochemical

sensors,

PB was

chosen

as the redox media

tor as it enables

low-voltage

operation

(~0 V versus reference electrode) and minimizes

the interfer

ences

from other electroactive molecules

(

). Here, we introduced

an in

situ reduction

strategy to formula

te high-performance

printable

MX-PB

as a mediator layer on top of the CNT-SBS

WE (Fig. 3E,

note S1, and fig. S17). Transmission

electron microscopy

(TEM)

images

depict

uniformly

deposited

PB nanoparticles

over the

MXene

film surface with an average size of ~8 nm (Fig. 3F and

fig. S18). The concentr

ation of MX-PB

was tailored with desired

sensitivity

and linear

operating range according

to the levels of

target

biomark

ers in sweat (Fig. 3, G and H). In particular,

because

of the charge

transfer

properties

including

electron

hopping

and counter-ion

movement

within

the PB layer, MX-PB

(1 mg ml

−

) resulted

in a high sensitivity

, while increased concen-

tration (10 mg ml

−

) of MX-PB

led to a decreased

sensitivity

and

wide linear

operating windo

w due to the slower charge

transfer

ki-

netics

(

). The incorpor

ation of MX-PB

in the mediator layer sub-

stantially

enhanced

both sensitivity

and detection

limit as compar

to neat PB due to the synergis

tic hybridized

network

that enhances

electronic coupling

for interfa

cial electron transfer

(fig. S19).

Cocktails

comprising

enzymes

(i.e., GOx and AOx) and bioac-

tive polymers

(i.e., chitosan

and BSA) were then directly printed

the MX-PB

mediator layer to prepare glucose

and alcohol

sensors

suitable

for wearable sweat analysis (fig. S20). Figure 3 (I and J)

shows the representative amper

ometric

responses

of the optimized

Fig. 3. Design

and characteriza

tions

of the 3D-printed

biochemical

sensors.

(

) Schema

tic illustration of the design

and SSE-based

3D-printing

process of the en-

zymatic sensors.

PEG, polyethylene

glycol;

MX-PB,

MXene

–

Prussian

blue. (

) SEM image

of the porous CNT-SBS

after PEG dissolution.

Scale bar, 10 μm. (

) Magnified

SEM

image

of the porous CNT-SBS

with exposed

CNTs. Scale bar, 1 μm. (

) Cyclic

voltammetry

(CV) scans of a gold electrode (AuE),

an SPE, a GCE, and a 3D-printed

CNT-SBS

electrode in a solution

containing

5 mM [Fe(CN)

]

−

and 0.1 M KCl. (

) Schema

tic illustration of in situ reduction

of PB onto MXene.

(

) TEM image

of an MX-PB

film. Scale

bar, 500 nm. Inset, magnified

TEM image

revealing

the PB nanoparticles

as circled in yellow. Scale bar, 5 nm. (

and

) Amper

ometric

calibration plots of CNT-SBS

electrodeswithvarying

MX-PB

loadings

inhydrogenperoxide(H

)solutions

withlow(0to500μM)(G)andhigh(0to5mM)(H)concentr

ations.(

and

) Amper

ometric

responses

of the glucose

sensor

(I) and alcohol

sensor

(J). Insets,

the corresponding

calibration plots. PBS, phospha

te-buffer

ed saline;

Glu, glucose;

Alc, alcohol.

Error bars

represent the SD from 10 sensors.

(

) Schema

tic illustration of the working

mechanism

of the 3D-printed

CNT-SBS-P

ANI

–

based

pH sensor

and the corresponding

SEM

image.

Scale bar, 1 μm. (

) Open-cir

cuit potential

response

of a pH sensor

in Mcllvaine

’

buffer.

Inset, the corresponding

calibration plot. Error bars represent the SD from

10 sensors.

SCIENCE

ADVANCES

RESEARCH

ARTICLE

Song

et al.

Sci.

Adv.

, eadi6492

(2023)

13 September

2023

4 of 13

Downloaded from https://www.science.org at California Institute of Technology on September 29, 2023

glucose

and alcohol

sensors,

measur

ed at physiologically

relevant

concentr

ations between 0 to 200 μM and 0 to 20 mM, respectiv

ely.

A linear

relationship

between the current and analyte

’

concentr

tionwithsensitivities

of0.036μAcm

−

μM

−

forglucose

sensor

and

0.204 μA cm

−

for alcohol

sensor

was observ

ed. Considering

that sweat alcohol

concentr

ation after alcohol

intake can be as high

as tens of mM, a polyur

ethane

(PU) diffusion

–

limiting

membr

ane

was introduced

to improve the current stability

and widen

the linear

range (fig. S21), and the developed

alcohol

sensor

demons

trated

high stability

over prolonged

periods

and multiple

cycles

of mea-

surements

(fig. S22). Both enzyma

tic sensors

showed high repro-

ducibility

(fig. S23) and selectivity

over other metabolites

typically

found

in sweat (fig. S24) and were able to reliably

detect

the analyte-

level changes

when sampled

at physiological

sweat rates (fig. S25).

The pH sensor

on the e

-skin was designed

on the basis of a

CNT-SBS-poly

aniline

(PANI) electrode printed

with an ink made

of PANI powder and CNT-SBS

(fig. S26A).

pH was measur

ed as a

function

of potential

changes

caused

by protonation/depr

otonation

on the PANI surface (Fig. 3K). A near-Nerns

tian sensitivity

of 55.6

mV pH

−

with high reproducibility

over a physiologically

relevant

pH range of 4 to 8 was obtained

(Fig. 3L and fig. S26, B and C). To

accurately quantify

sweat glucose

and alcohol

levels, real-time

sensor

calibrations were performed

on the basis of the simultane-

ously

obtained

pH and temper

ature informa

tion to compensa

the influence

of pH and temper

ature on the enzyma

tic reactions

(figs. S27 and S28).

3D-printed

microfluidics

for biofluid

extraction,

sampling,

and multiple

xed analy

sis

To enable

on-demand

and continuous

molecular

monitoring,

miniaturized

microfluidics

with a built-in

iontophor

etic sweat in-

duction

module,

composed

of hydrogels containing

carbachol (car-

bagel),

and iontophor

esis electrodes (IP cathode

and IP anode],

was

developed

to interfa

ce with the 3D-printed

biochemical

sensors

(Fig. 4A). Autonomous

and long-las

ting sweat induction

was real-

ized via transdermal

delivery of muscarinic

agent

carbachol

(Fig. 4B). The sweat-sampling

microfluidics

was 3D-printed

using

an SBS ink of 25 wt % with an appropriate viscosity

to achieve high

printing

resolution

(Fig. 4C and movie S2), while the iontophor

esis

module

was prepared by printing

a pair of CNT-SBS

electrodes fol-

lowed bygelatin-agar

osecarbagels

(Fig.4, D and E).Localized

sweat

induction

using the 3D-printed

iontophor

esis module

was realized

by delivering a very small dose of carbachol with a current ranging

from 1 to 3 μA mm

−

. The secreted sweat volume

was found

to be

linearly

correlated with the applied

current/deliv

ered drug dose

(Fig.

4F). During

the on-body

test, the induced

sweat was

sampled

through the microfluidics

to ensure that newly secreted

sweat flows through the sensing

reservoir with high-tempor

al reso-

lution

toward real-time

wearable analysis (Fig. 4G and movie S3).

The assembled

3D-printed

microfluidic

-skin could

confor-

mallyadhere to the skin (Fig. 4H) and displayed excellent

selectivity

and stable sensor

performance

under

mechanical

deforma

tions

(Fig. 4, I and J, and fig. S29). The e

-skin

’

high biocompa

tibility

and low cytotoxicity

were validated by culturing

human

dermal

fi-

broblasts (HDFs)

and normal

human

epidermal

keratinocyte

(NHEK)

cells using

a commer

cial live/dead

kit and PrestoBlue

assay, as represented

in Fig. 4 (K and L) and fig. S30. The viability

of HDF and NHEK

cells remained

>95%,

and their metabolic

activ-

ities consistently

increased during

the 7-day culture.

3D-printed

wearable

energy

system for the e

-skin

Wearable systems with miniaturized

energy-harv

esting and storage

devices

are highly

desired to promote

sustainability

and untether

battery-fr

ee operations (

–

Here, we designed

high-perfor-

mance

3D-printed

MXene

MSCs

that can interfa

ce with a solar

cell to power the e

-skin.

A highly

concentr

ated MXene

ink (MX-

H; 120 mg ml

−

) was used to print the 3D freestanding

interdigital

MSC (Fig. 5A, fig. S31, and movie S4). The resistance and thickness

can be readily

tuned

by adjusting the number

of printed

layers and

bysubjecting

them to different posttreatments

[i.e., air-drying

(AD)

and freeze-drying

(FD)]

(Fig. 5, B and C). Compar

ed to AD post-

treatment,

the FD MX-H

exhibits

superior

electrochemical

perfor-

mance

owing to the highly

porous structur

es with substantially

enhanced

active surface area and reduced

impedance

(Fig. 5D

and fig. S32). Using

this approach, a variety

of comple

x 3D archi-

tectures can also be stably printed

and well-preserved (fig. S33 and

movie S5).

The charge-s

torage performance

of the MXene

MSC was evalu-

ated after printing

a layerof poly(vinyl

alcohol)

(PVA)

–

sulfuric

acid

) gel electrolyte on top of the interdigital

MXene

electrodes

(figs. S34 and S35). Such gel electrolyte is commonly

used in wear-

able energy

devices

(

). Here, the encapsula

ted gel electrolyte used

in this work demons

trated high biocompa

tibility

(fig. S36) and was

assembled

without

direct contact with the skin to avoid any poten-

tial irritation. As illustrated in Fig. 5E, proportional

to the area en-

closed

in the cyclic

voltammogr

am (CV),

the areal capacitance

FD-H

MXene

MSC

was outstanding

compar

ed to other

MXene

MSC.

For the detailed

investment

of FD-H

MXene

MSC,

the electrochemical

performance

and ion transport

property

were

further

improved with

the increased

electrode dimensions

(number

of interdigital

pairs,

length,

and number

of printed

layers) and the reduced

electrode gaps (Fig. 5F and figs. S37 and

S38). The CV curves and galvanos

tatic charge-discharge

(GCD)

profiles of the FD MSCs

showed an electric

double-la

yer capacitive

andhigh-r

ate behavior(Fig.5, Gand H). Inparticular,

MSCwith 10

printed

layers (MSC-10L)

exhibited

an extremely

high areal capac-

itance

of 8.61 F cm

−

at a scan rate of 5 mV s

−

and was able to

discharge

at a current of up to 30 mA that is adequa

te for practical

wearable applica

tions and for initiating wireless Bluetooth

commu-

nications

(Fig. 5I). Compar

ed to previously

reported

printed

MXene

MSCs

(

–

this MSC-10L

showed a superior

energy

density

that reached as high as 12.91

μWh

−

at a power

density

of 439.35

mW cm

−

and 43.18

μWh cm

−

at the highes

power density

of 35.99 mW cm

−

(Fig. 5J). Moreover, it displayed

robust mechanical

and satisfied

electrochemical

cycling

stability

with capacitance

retention

of 95 and 87% after 2000 bending

and

scanning

cycles,

respectiv

ely (Fig. 5K and fig. S39). This can be at-

tributed

to the excellent

mechanical

properties

and strong adhesion

between the MXene

nanosheets

(

). To achieve desired

working

potential

and capacitance,

the 3D-printed

MSCs

can be se-

rially connected,

where a voltage

output

of 4.8 V was achieved by

connecting

eight MSCs

in series to potentially

power our wearable

sensor

with Bluetooth

transmission

(Fig. 5L).

Todemons

tratethefullpotential

ofSSE-based

3D-printing

tech-

nology

in wearable sensing,

the disposable

3D-printed

-skin (con-

sisting of biophy

sical temper

ature and pulse sensors,

biochemical

sensors,

iontophor

esis-integr

ated microfluidics,

and MSC)

was in-

tegrated with a reusable

flexible printed

circuit board

(FPCB)

coupled

with a commer

cial solar cell that is equipped

for energy

SCIENCE

ADVANCES

RESEARCH

ARTICLE

Song

et al.

Sci.

Adv.

, eadi6492

(2023)

13 September

2023

5 of 13

Downloaded from https://www.science.org at California Institute of Technology on September 29, 2023