Review—The Development of Wearable Polymer-Based Sensors: Perspectives

Tattoo-like sensors have sub-micron thickness and usually serve for continuous monitoring over a certain period. Learning from the conventional thick patch-type sensors that can be detached due to bending and stretching of the skin (up to 30%),⁸¹ the tattoo-like sensor should be stretchable and conform to the human skin, acting as an artificial, "second" skin. When the thickness of the tattoo is smaller than the natural wrinkles of the skin (15–100 μm) the user may not be able to discern the artificial skin.⁸² The sensor material should be breathable, allowing perspiration from the skin surface to cool the body temperature.

Like the aesthetic tattoo, tattoo-like sensors can be applied using an invasive method⁸³ or a non-invasive method via an adherent tattoo sticker.⁸¹ The invasive method utilises the colour change or an optical variable ink achieved by the use of chromogenic dyes, fluorescence, diffraction grating, and plasmons injected within the dermis.^83,84 This approach has been evaluated in ex-vivo porcine skin tissue for sensing pH, glucose (in mixtures with glucose oxidase and hydrogen peroxide) and albumin. However, the reversibility of glucose and albumin detection needs to be improved in order to be used as permanent tattoo-like sensor. Non-enzymatic reversible glucose sensor can be realised by phenylboronic acid with β-cyclodextrin,⁸⁵ which able to reversibly bind the cis-diol of glucose molecules.⁸⁶

Recently, most research has been carried out non-invasive tattoo-like sensors. Earlier, costly microfabrication processes were used for patterning the surface of the thin electronic tattoo, which usually involves high-cost equipment and cleanroom laboratories.⁸² Nowadays, the cost-effective methods such as screen printing,^81,87 "cut and paste" method⁸⁸ and direct writing^11,89 and have been developed for fabricating the tattoo in Fig. 6. Screen printing uses inking through a patterned stencil to create sensor electrodes, cut and paste utilises cutter plotter to draw the electrode followed by removing extraneous part, while direct writing designs the electrode using inkjet or direct ink printing. The electrode sticks to adhesive layer which also functions as cover to protect the sensor from the environment. The signal from electrode is processed in signal processing unit which can be directly interpreted qualitatively or digitised in quantitative analysis. Colour changing or optical variable ink such as chromogenic dyes, fluorescence, diffraction grating, and plasmons can be used for qualitative analysis. For quantitative measurement, the signal is translated by transducer, amplified, and filtered to reduce noise. The data can be digitised and wirelessly transferred by transceiver. A flexible printed circuit board (PCB) made of screen-printed silver ink on polymer, such as polyimide, polyethylene terephthalate (PET), and polyethylene naphthalate (PEN), is usually used for signal processing circuit. The integration of electrode, signal processing unit, transceiver, and power system in a thin tattoo-like sensor becomes a challenge for researcher in the recent years.

Zoom In Zoom Out Reset image size

Primarily, the tattoo-like sensor can be categorised in three main functions that are chemical/electrochemical,⁸¹ thermal,⁹⁰ and strain sensor.^91,92 Contrary to the invasive tattoo which detect skin interstitial fluid, tattoo stickers for electrochemical sensor can obtained information from sweat, tears, and saliva. Sweat is biofluid which can be accessed on epidermis containing a wealth of biomarker such as electrolytes (e.g., pH, sodium, and potassium ion) and metabolites (e.g., glucose, lactate/lactic acid). The sweat pH contains information such as hydration level as the Na⁺ is depleted during dehydration as well as an indicator for body odour and skin diseases.⁷³ Based on a Nernstian response, the pH sensitivity is limited to about 59 mV pH⁻¹ at 25 °C, based on the maximum change in anodic and cathodic potentials.⁹³ Electrochemical sensors based on conducting polymer usually follow the Nernst equation with the sensitivity value about 40–50 mV · pH⁻¹ at 25 °C with Ag/AgCl reference electrode stabilised by a KCl-saturated insulator.⁸¹ Ion recognition sites are synthesised during oxidation of conducting polymers exhibiting a pH response. Among the conducting polymers used, polypyrrole (PPy) and polyaniline (Pani) have the highest pH sensitivity, which are 43.2 mV pH⁻¹ and 50.1 mV pH⁻¹ for PPy⁶⁸ and Pani,⁶⁷ respectively.

The sensitivity of field-effect transistors (FETs)⁹³ and charge-coupled devices (CCDs)⁹⁴ may exceed the Nernstian limit (59 mV pH⁻¹). This could be achieved by accumulation of counter ions above critical potential causing crowding or steric effect of ions near the surface rather than following a classical Boltzmann model of ion distribution.⁹⁵ Such a crowding effect could be obtained using extended or dual gate ion sensitive field effect transistor due to capacitive coupling of top and bottom gate which the sensitivity increases as the ratio of top and bottom gate capacitance increases.^93,96–98 The same strategy can be used in a charge coupled device used in a wearable pH sensor.⁹⁴ Electrons are transferred from input gate voltage to pH sensing well when a positive bias voltage is switch on. When the positive bias applied to transfer gate, the input gate voltage return to normal stage while the electrons from pH sensing well migrate to capacitor and accumulate.⁹⁹ After 100 cycles of accumulation, the sensitivity reaches to 240 mV pH^–1 in the range of pH 2.8 to 7.1 using silicon oxide as pH sensing layer, PET film as a substrate and polyimide as a cover.⁹⁴

The glucose and lactate detections commonly exploit enzyme such as glucose oxidase and lactate oxidase. The glucose and lactate react with glucose oxidase and lactate oxidase respectively producing hydrogen peroxide which can be detected by electrochemical sensor.^100,101 Alcohol in sweat also can be detected using alcohol oxidase enzyme which shows high correlation of current response with blood alcohol concentration (BAC) at 102 μA BAC%⁻¹.²⁰ The measurement was faster (less than 60 s) than alcohol detection by transdermal ethanol vapour which has 0.5–2 h delay.¹⁸ In order to provide long term stability, the enzyme should be kept in restricted temperature and humidity ranges, which is inconvenient for highly active people such as athletes. This drawback inspires the development of non-enzymatic biosensor which is based on the reversible reaction of glucose binding. The boronic acid groups are able to reversibly bind 1,2- and 1,3-diol functionalities which are an element of glucose and lactate.¹⁰² Boronate-functional group can be imparted in polymer such as polyaniline¹⁰³ and poly(3-aminophenyl)boronic acid¹⁰² as well as in metal oxides such as titanium dioxide.¹⁰⁴ Several redox active metal oxides and hydroxides such as Fe, Cu, and Ni based also has the ability to reversibly bind glucose¹⁰⁵ and lactate¹⁰⁶ through electrochemically and chemically induced reversible redox reactions.

In a hot environment, heat stroke may occur when the body cooling system is not fully functional. Staying hydrated is one of the keys to prevent heat stroke. Wearable temperature sensor can be used as early warning of heat stroke and heat related illness which reminds us to stay hydrated. Having a thin morphology with high conductivity, graphene becomes a popular material for tattoo-like temperature sensor. Lu et al. grew graphene on copper foil and attached to Kapton^TM polyimide to be transferred via copper etching.⁶⁹ It was stuck on thermal realising tape followed by cutting, heating, and peeling to remove the excess leaving patterned graphene electrode on Tegaderm^TM (transparent film dressing) as support. The sub-micron thick graphene sensor was able to withstand 15% stretching up to 1300 cycles and had the temperature coefficient of resistance of 0.0042 °C⁻¹, which is comparable to commercial thermocouple.⁶⁹ The durability of graphene sensor can be improved significantly using an ionic crosslinked silk as support resulting in high durability over 10000 cycles of 50% strain.⁹² The toughness caused by hydrophobic - hydrophobic and ionic interactions between graphene and silk via calcium ions improving shear strength of layered graphene.¹⁰⁷ The non-covalent interaction also gave self-healing ability with 100% healing under 0.3 s.⁹² To support the temperature sensor, another functionalities such as UV detection can be added to skin-like sensor by using azobenzene, which its sheet resistance is highly influenced by the intensity of UV irradiation.¹⁰⁸

The high sensitivity of the tattoo like strain sensor enables it used for respiration monitoring. Recently, the Japanese art of paper crafting such as "Kirigami" and "Origami" give a huge influence in the design of two-dimensional strain sensor. The term "kirigami" comes from the Japanese words "kiri" cut and "gami" paper, which can be translated as the art of paper cutting. Meanwhile, "ori" in "origami" means folding, i.e., it involves paper folding rather than cutting. These paper crafting techniques have been used in space engineer to transport a solar panel in a small packaging.¹⁰⁹ The ability to transform a rigid planar material into an agile design which can be fold and unfold repeatedly without over-stretching the material is beneficial in the strain sensor. There are three methods (e.g., island-bridge, accordion, and kirigami) of paper crafting techniques which is commonly used in two-dimensional strain sensor in Fig. 7.¹¹⁰ The kirigami design enables the strain sensor to maintain its sensitivity (>80%) at 60% strain as well as more than 60000 cycles of respiration without degradation in performances.⁷⁰

Zoom In Zoom Out Reset image size

While some researchers focus on fabricating thinner patch-type sensor, the others pursue better interfacial contact between electrode and human body.¹¹¹ Polymer is commonly used for electrolyte gel in wet electrode. However, the gel may cause skin irritation and discomfort. Dry contact or capacitive electrode become a potential candidate to replace it. Dry capacitive electrode is a non-contact electrode which has a layer of insulator between electrode and skin. Although it completely eliminates the skin irritation, the signal to noise ratio is low due to high impedance between skin and sensor as well as high sensitivity to mechanical/movement noise.¹¹² In order to provide better contact without the gel, researchers fabricate microneedle pattern thus the electrode may reach the epidermis in Fig. 8. Stratum corneum act as an information barrier creating noise in the measurement. The outer surface of skin can be contaminated by environment or dead skin cells, the skin motion may loosen the electrode contact. Also, the high impedance at the interface of electrode and skin hampers the signal source. The needles need to be strong enough to puncture the stratum corneum (10–20 μm in thickness) and penetrate to the epidermis (50–100 μm) to collect biofluid. The impedance of skin can be interpreted using circuit model of parallel resistors (R) and capacitors (C) where

Zoom In Zoom Out Reset image size

E_el is the half-cell potential of electrode

C_el and R_el represent electrode-electrolyte impedance

R_gel is the resistance of electrolyte gel

E_sc is the half-cell potential of electrolyte-skin interface

C_sc and R_sc represent the impedance of skin

R_sg is the resistance of subcutaneous tissue and dermis.

In wet electrode, the electric current is hampered at the interface of electrode to gel and gel to skin which is weaken the signal to noise ratio, while direct connection of dry electrode and skin removes the resistance at the interface. In terms of the material, metal, although it is conductive, it is difficult to stretch. Conductive elastomers are the most promising candidates for this purpose.

To fabricate a microneedle array on a conductive elastomer, photo-lithography,¹¹⁴ injection moulding,⁷¹ soft lithography (e.g., replica moulding¹¹⁵ and embossing^116,117), and 3D printing¹¹⁸ are commonly used. In photo-lithography, the material surface is covered with photo-resist materials and is irradiated with masked ultraviolet followed by etching of the photoresist material developing desirable pattern on the surface. This process requires sophisticated equipment and materials. Contrary to photo-lithography, the first step in injection moulding is to create patterned negative mould which usually deploys electric discharge milling (EDM) or laser ablation followed by injection moulding of melted polymer and cooling down. In soft-lithography, a micro/nanopatterned elastomer is deployed as replica mould of patterned master surface to minimise the high cost of photolithography.¹¹⁹ The replica mould could also serve as stamp to create recessed relief pattern on the subject surface. The recent development in 3D printing techniques especially fused deposition modelling (FDM)¹²⁰ and stereolithography^121,122 enabling high resolution polymer printing up to micron size. Metal sputtering, conducting polymer or carbon deposition can be used afterwards to increase the electrical conductivity of polymer microneedles. In case of using soft elastomer such as polydimethylsiloxane (PDMS), the geometry of PDMS pillars is important to prevent pairing or lateral collapse in Fig. 9.¹²³ As a rule of thumb, the length (H) to width (L) ratio of pillar should be more than 0.5 and lower than 5 (0.5 < H/L < 5) also the ratio of length (H) to distance between pillar (D) should be more than 0.05 (H/D > 0.05).

Zoom In Zoom Out Reset image size

Instead of sensor applications, microneedle arrays are also utilised for transdermal delivery of drug or vaccine which is a minimally invasive drug or vaccine injection.^124–126 The microneedles may be encapsulated with drug or to be used for drug infusion into the skin. This technique can be ultimately combined with microneedles sensor creating a sensor that can administer drug at the same time or to monitor the drug delivery process. Wang's group at University of California has used microneedles sensor for continuous monitoring of levodopa (L-dopa) which is a medication for Parkinson's disease (PD).⁷² The symptom of Parkinson's diseases occurs when the dopamine level is decrease. Levodopa replenishes dopamine while it crosses the blood brain barrier.¹²⁷ Hence, continuous monitoring of levodopa is crucial to prevent dopamine wearing off in the patient with PD. The electrochemical microneedle sensor utilised enzymatic and non-enzymatic electrodes.⁷² The non-enzymatic needle electrode was made of graphite and mineral oil and the enzymatic electrode used tyrosinase mushroom enzyme which immobilised on the surface of graphite and mineral oil as an active material to detect transition from levodopa to dopaquinone during electrochemical measurement. The electrode was covered by Nafion^TM (e.g., ionically conductive polymer) to cover the electrode from contamination. The reference needle electrode was modified with Ag. The anodic detection of L-dopa was achieved using non-enzymatic needle through square-wave voltammograms (SWV) while the chronoamperometric detection was performed using enzymatic needle. The sensitivity could reach up to 0.038 nA μM⁻¹ and 0.082 μA μM⁻¹ for the chronoamperometric and SWV, respectively. The limit of detection (LoD) for chronoamperometric and SWV was 0.25 μM and 0.5 μM, respectively.

The microneedles sensor has shown to effective in continuous monitoring of glucose, lactate, and tyophylline as well.⁷¹ Tyophyilline is a known drug for respiratory diseases such as asthma and chronic obstructive pulmonary disease.¹²⁸ Cass et al. was utilised aluminium mould patterned with electrical discharge milling (EDM) to be used for injection moulding of polycarbonate microneedles.⁷¹ To improve the electrical conductivity, the chromium (15 nm) and platinum (50 nm) were sputtered on polycarbonate microneedles as working electrode. For reference electrode, the microneedles were sputtered with Ag (150 nm) and treated with FeCl₃. The working electrode was decorated with specific enzyme such as glucose oxidase, lactate oxidase, and xanthine oxidase for chronoamperometric measurement of glucose, lactate, and tyophylline, respectively. In which the maximum limiting current (I_max) and Michaelis-Menten constant (K_M) were ≈23 μA, 0.95 μA, 0.31 μA and ≈13 mM, 0.7 mM, 13 mM, respectively for glucose, lactate, and tyophylline sensor.

Based on the working principles, textile-based sensor can be categorised into electrochemical, transistor-based, and passive stimulus-response sensor in Fig. 10.¹²⁹ Three electrodes configuration such as reference, working, and counter electrode are sewn on the textile and are used for electrochemical sensor with sweat as electrolyte. Carbon or Ag/AgCl conductive ink is usually deposited on textile as electrodes by screen printing⁷⁵ or dip coating.¹³⁰ Carbon based electrode such as graphene can be utilised as substrate for biomarkers such as 1-Pyrenebutyric acid-N-hydrosuccinimide ester (PANHS), which is an active material for binding influenza A-specific antibody.⁷⁵ The bound and unbound reaction occurs during electrochemical impedance measurement. The resulting sensor was able to detect the influenza protein as low as 10 ng mL⁻¹ up to 10 μg · mL⁻¹. In a transistor sensor, the electrons are passed from source to drain through gate and channel in which the channel conductivity is controlled by gate. The voltage can be applied to gate electrode inducing reaction in the channel with molecules in sweat. The conducting polymers such as poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), polypyrrole (PPy), and polyaniline (Pani) are usually used as channel and modified by enzyme (e.g., glucose oxidase and dehydrogenase)¹³¹ or redox active analytes (e.g., adrenaline, dopamine and ascorbic acid),¹³² in which the redox reactions occurs during applied voltage.

Zoom In Zoom Out Reset image size

In passive stimulus-response polymer sensors, the sensing method is deployed using the observation of direct changes in materials such as resistance, optical, capacitance, which is commonly used to detect strain and chemical sensing.^74,133,134 As a smart textile is often used for highly active user such as athlete, minimal modification is sought to retain its comfort. Hence, power-free sensor such as stimulus-response sensor is preferable. A colorimetric dye is one of the most common methods for passive detection of chemical in sweat.⁷⁴ The pH of human sweat is typically between 4.5 and 7, so pH sensitive dyes (e.g., bromocresol green, methyl orange, bromothymol blue, and bromocresol purple) should exhibit a colour change in this pH range.⁶⁷ To contain the dyes within the textiles, hydrogel can be used with surfactant to prevent swelling of hydrogel due to humidity. Alternatively, ionic gel is more stable to humidity but one must note the anions or cations contain in the gel may interfere the pH indicator.⁷³ For lactate detection, the purple lactate colour intensity increases as the lactate concentration increases due to enzyme reaction of lactate dehydrogenase, where changes can be observed up to 12.5 mM as the threshold concentration.⁷⁴ As an alternative to a colorimetric dye, a plastisol based microfluidic channel can be used to detect chemical such as sodium ions.¹³⁵ The plastisol resistivity is sensitive to the concentration of sodium ions. The microfluidic channel offers focused detection of the chemical solution considering the fabric may absorb the solution as well. Instead of resistivity-based sensor, passive detection can be pursued by capacitive sensor.¹³ The humidity can be measured using polyimide which has imide group hence it is able to create hydrogen bonding with water. The capacitance change when the hydrogen ions in water fill the microvoid within the polyimide.

The stimulus-response strain sensor works based on the measurable changes in resistivity, capacitance, piezoelectric, triboelectric, optical during stretching or bending.¹³⁴ The resistive strain sensor is the most common type in textile-based sensor which utilises the fabric as resistor and measures the resistivity changes. The electrical conductivity of fibres can be manipulated through coating, spinning, and knitting of conductive materials such as conducting polymer, carbon-based structures (e.g., graphene, carbon nanotubes, carbon black nanoparticles, carbon fibres), and silver. The conducting polymer can be deposited on non-conducting fabric by chemical polymerisation.¹³⁶ Typically, the process begins with dipping the fabric into monomer solution containing dopant followed by soaking it into the oxidant at controlled temperature. The opposite can be achieved using vapor phase polymerisation where the fabric is soaked into oxidant and dopant followed by exposing the fabric to the monomer vapour.¹³⁷ Alternatively, a one-pot method can be deployed using repetitive dipping of fabric into a dispersed solution of conducting polymer.¹³⁸ However, cracking can occur in conducting polymer coated fabrics due to the brittle nature of some conducting polymers.¹³⁹ Elastomers, such as PDMS can be used either as an interface material between fabric and conducting material or as a mixture with the conducting composite. The conducting filler in the composite may create a percolative network or conducting path during stretching which increases the conductivity. Carbon/PDMS core/sheath composite strain sensors can retain their performance over more than 10000 cycles at 100% strain with a gauge factor ≈0.68.⁷⁶ Carbon nanotube coated cotton/polyurethane core/sheath composites strain sensors have also shown excellent durability with a gauge factor ≈0.65 at 40% strain over more than 300000 cycles.⁷⁷ However, the comfort of the textile can be compromised due to the hydrophobic nature of most elastomers. Instead of carbon and conducting polymer, the recent development of MXenes (two-dimensional transition metal carbides) has created high sensitivity capacitance strain sensors with a maximum strain gauge ≈6.02 and it is able to withstand 2000 cycles at ≈14% strain although the maximum strain is only 20%.⁷⁸ Such a sensor was made by two step dip coating of cotton in small diameter MXene solution followed by a large diameter MXene solution. For further work, the maximum strain cycles of MXene-coated textiles could be improved by using an elastomeric substrate.

While tattoo, patch, and textile collect data from sweat or saliva, contact lenses obtain information from tears in Fig. 11.⁷⁹ The glucose level in tears is averaged 3.6 mg · 100 ml⁻¹ (0.2 mM) and 16.6 mg · 100 ml⁻¹ (0.92 mM) for normal and diabetic patients respectively¹⁴⁰ which can be a reliable biomarker for diabetes. The challenge is to integrate the power system, rectifier, transmitter, and sensor electrodes within the lenses while maintaining biocompatibilty with delicate human eye tissue. Another issue is the glucose level from tears has lag time around 10–20 min compared to blood glucose level.²¹ Since the aim of glucose sensing contact lenses is continuous glucose monitoring, the lag time problem could be minimised. Optically transparent and biocompatible polymers such as soft silicone elastomer, Ecoflex^{TM 22} and fluorinated polymer, EFiRON^{® 79} are often used as substrate and are deposited by silver nanowires (AgNW) followed by patterning using photolithography to fabricate electrodes, rectifier, and wireless circuit. The field effect transistor-based glucose sensor can be utilised in contact lenses type sensor using graphene as active material immobilising glucose oxide enzyme with a mixed of Cu/Au as source and drain electrodes. The circuit can be connected to small light emitting diode (LED) as indicator although some people argue that the LED may contain toxic arsenic and lead.¹⁴¹ Using only a few layers of graphene, a transparency of >91% can be retained.^22,79 Chen et al. utilised phenylboronic acid (PBA) based hydroxyethyl methacrylate (HEMA) contact lens to fabricate a non-enzymatic glucose sensor.¹⁴² The sensing mechanism was done by measuring the changes in width and thickness due to swelling during glucose interaction with boronic acid. Although cytotoxicity studies showed no significant difference in cell viability, the comfort of the lenses is questionable due to swelling. Besides transistor-based and swelling based contact lenses sensors, an electrochemical tears sensor has been developed, particularly the NovioSense Glucose Sensor, which has been through phase II clinical trials with 6 patients with type 1 Diabetes Mellitus.¹⁴³ The tears sensor contained a flexible coil (diameter of 60 μm) containing working, counter, and reference electrodes as well as transmitter. The outer layer of coil was made of hydrophilic polysaccharide immobilised with glucose oxide enzyme. The coil was then put on the lower eye lid for chronoamperometric measurement. The performance of NovioSense Glucose Sensor was similar to Abbott FreeStyle Libre, blood glucose sensor.

Zoom In Zoom Out Reset image size

The pressure or strain sensor in contact lenses can be utilised to detect the intraocular pressure (IOP) of bovine eyeballs which is useful for glaucoma patients (see Fig. 12²²). Tonometry is commonly used to measure intraocular pressure of glaucoma patients although it has many weaknesses such as single point and one-time (rather than continuous) measurement not to mention the discomfort when a tonometer probe applies pressure to the eyes.¹⁴⁴ As an alternative, a wireless non-invasive pressure sensitive contact lenses sensor can be deployed either using a resistive-based⁸⁰ or capacitive-based^22,145 sensor. In resistive-based sensors, two thin Pt/Ti strain gauges have been microfabricated and sandwiched between two polyimide layers. One strain gauge was used to measure changes in the corneal curvature due to IOP and the other placed radially to minimise strain while acted as baseline measuring strain caused by thermal expansion. The gold antenna was then electrodeposited on the polyimide and connected to gauge by conductive epoxy. The gauge, antenna, and wireless microprocessor were embedded on the silicone contact lenses by cast moulding technique. The resulting sensor was tested on pig eyes and showed high linearity and a sensitivity of 109 μV · mmHg⁻¹ between 20 and 30 mmHg, which is the pressure range of glaucoma patient.⁸⁰ For capacitive-based sensors, a silicone elastomer, Ecoflex^TM, was used as dielectric and sandwiched in between a spiral coil of graphene and silver nanowire electrodes.²² A bovine eyeball has been used for in-vitro measurement and the resonance frequency was calculated as a function of inductance and capacitance with the sensitivity of 2.64 MHz · mmHg⁻¹ between 5 to 50 mmHg.²² These pressure sensitive contact lenses sensors could be combined with drug release of anti-glaucoma drugs, such as timolol, to evaluate the effectiveness of glaucoma treatment. However, calibration with a commercial tonometer may be needed due to the variation of corneal curvature between individuals.

Zoom In Zoom Out Reset image size