Performance analysis of solution-processed nanosheet strain sensors—a systematic review of graphene and MXene wearable devices

Nanotechnology has led to the realisation of many potential Internet of Things devices that can be transformative with regards to future healthcare development. However, there is an over saturation of wearable sensor review articles that essentially quote paper abstracts without critically assessing the works. Reported metrics in many cases cannot be taken at face value, with researchers overly fixated on large gauge factors. These facts hurt the usefulness of such articles and the very nature of the research area, unintentionally misleading those hoping to progress the field. Graphene and MXenes are arguably the most exciting organic and inorganic nanomaterials for polymer nanocomposite strain sensing applications respectively. Due to their combination of cost-efficient, scalable production and device performances, their potential commercial usage is very promising. Here, we explain the methods for colloidal nanosheets suspension creation and the mechanisms, metrics and models which govern the electromechanical properties of the polymer-based nanocomposites they form. Furthermore, the many fabrication procedures applied to make these nanosheet-based sensing devices are discussed. With the performances of 70 different nanocomposite systems from recent (post 2020) publications critically assessed. From the evaluation of these works using universal modelling, the prospects of the field are considered. Finally, we argue that the realisation of commercial nanocomposite devices may in fact have a negative effect on the global climate crisis if current research trends do not change.


Introduction
The global pandemic has highlighted the enormous stress national healthcare systems are under, pushing the societal need and the commercial gap for high precision personal health monitoring to the forefront of conversations [1][2][3][4].The realisation of affordable, discreet, and inclusive wearable devices which can measure bodily signals like heart rate, breathing and joint movement are believed to be the key to next generation healthcare [3][4][5].In the future, these sensors would alleviate stress on global healthcare systems by ushering in a new era of preventative medicine [4].Currently, medical professionals only become aware of changes in Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.patient status upon physical examination.However, these future wearable devices would facilitate continuous health monitoring, allowing doctors or nurses to make early diagnostic decisions remotely before medical complications arise [6].Additionally, such wearables would have expansive use in the private personal healthcare market from sports athletes performance analysis to facilitating increased independence for boarders in care homes or retirement facilities [7].
One research area leading the way in bringing about the realisation of these wearable devices are flexible electronic materials based on conductive two-dimensional (2D) nanosheet networks embedded in or layered onto elastic polymers [8,9].One of the key reasonings why these nanocomposite devices are of such great research interest can be attributed to currently available commercial devices applying rigid metal components and stiff polymer substrates [10,11].Two attributes that make commercial devices unsuitable for wearable electronics as they cannot comfortably conform to the body through skin-on adhesion or clothes integration.As nano-based sensors can have their mechanical properties highly tuned based on what polymer is being applied and the sensing properties optimised based on what nanomaterial is utilised; these sensor types are far more attractive for health applications.Two of the most commonly applied nanomaterials used to create these nanocomposites are graphene and MXene nanosheets [10].In comparison to other nanomaterials, their devices have shown properties that far exceed other nanomaterial types [12].What truly maximises the usage of these nanosheets in particular is the ease at which they are produced in the liquid phase to facilitate the aforementioned mixing with liquid polymer or deposition on solid polymers substrates to create nanocomposite materials [13,14].In application, these solution processed devices based on percolative networks of graphene [15,16], and MXenes [17,18], have demonstrated the capabilities of measuring a wide range of bodily signals that surpass the degree of accuracy and sensing capabilities of current commercial devices.
In this review, we consider the methodologies for production and models for performance interpretation surrounding these devices.Vitally, we discuss in detail the standard reporting method for performance metrics and dispel any merit in high strain extrapolation procedures for said metrics.Furthermore, we use universal modelling to assess the performances of 70 devices published after the year 2020, to assess the recent trends and suitability of these devices for commercialisation.Finally, we conclude with a discussion on the critical requirement for the research field to trend towards more sustainable means of device production.

Graphene: background and properties
The crystal lattice of graphite, first studied by Bernal in 1924 using x-ray crystallography, consists of a structure of 2D stacked carbon monolayers [19].Individually these layers are known as graphene, bonded together by van der Waals forces (vdW) [20].These graphene sheets consist of an atomically thin array of sp2-hybridised carbon atoms organized in a planar hexagonal arrangement.Originally, thermodynamic arguments suggested that graphene could not exist as a freestanding entity [21], and it was not until 2004 when a group in the University of Manchester isolated graphene through mechanical exfoliation [22].The isolation of graphene from the bulk graphite crystal was first achieved with the 'scotch tape' method, whereby layers of graphene were peeled off using tape.This discovery heralded a material with unprecedented mechanical, electrical, and thermal properties [20,[22][23][24], thus changing the landscape of material science forever.Graphene has received unparalleled attention due to its extensive array of superlative properties [25].Graphene is both the strongest and stiffest material known, displaying tensile strength and elastic modulus values of σ B = 130 GPa and Y = 1 Tpa respectively [26].Optically, graphene is also one of the most absorbent materials known with a single monolayer of graphene shown to absorb 2.3% of incident light [27].Electrically, graphene is a zero bandgap semiconductor or semi-metal.Unlike normal semiconductors, which present a gap between the valence and conduction bands, graphene's bands touch at two points along the edge of the Brillouin zone, known as the Dirac Points [24].At these points the charge carriers act as relativistic quasiparticles called Dirac fermions, moving at a speed that is independent of their energy and direction.The charge carriers share this property with photons, which always moves at the speed of light irrespective of their energy.However, the Dirac fermions in graphene move at a speed about 300 times less [24].With very low defect density and few scattering centres, electron mobility in graphene at room temperature has been observed to be as high as 2 × 10 5 cm 2 Vs −1 [28].Other zero bandgap semiconductor/semi-metal properties observed in graphene include a tuneable bandgap [29], the breakdown of the Born-Oppenheimer approximation [30], the ambipolar field effect [22], and the quantum Hall effect at room temperature [31].

Thermodynamic stabilisation
Though mechanical cleavage was effective in producing high quality sheets of monolayered graphene, its rate of production did not match the demand and ambition of possible future applications.This however led to the demonstration of liquidbased exfoliation of bulk graphite in a solvent using sonic energy to yield large quantities of colloidal graphene suspensions (figure 1) by Coleman et al in 2008 [32].Dubbed liquid phase exfoliation or LPE, these colloidal suspensions are known to be meta-stable for long time periods, with the method also being broadly applicable to a wide range of other 2D materials based on layered [33], and non-layered [34], bulk crystals.Namely, for the sake of this review, the method has been instrumental with regards to the discovery and development of MXene materials.We will now discuss the two mechanisms which make these suspensions possible.Firstly, we will consider thermodynamically suitable solvents which allow for exfoliated material to be suspended in a liquid without chemical modification [35].In its first instance, by applying sonic energy to a system of carbon nanotubes (CNTs) in a solvent it was found a defect free suspensions of exfoliated CNTs could be achieved [36].When mixing a solute, such as CNT powder, and solvent together a change in the enthalpy (ΔH mix ) and entropy (ΔS mix ) of the system occurs.At constant pressure and temperature, the energy change of the system is described by the following where ΔG mix is the Gibbs free energy of mixing and T the temperature.In order for the chemical reaction induced by the mixing to be chemically favourable for spontaneous dissolution of the solute, ΔG mix 0. However, the large molecular weight and high rigidity of the CNT solute leads to an extremely low ΔS mix , meaning controlling ΔH mix becomes very important.In order to satisfy this criterion, the enthalpy of the system must be minimized to allow for dissolution, i.e.H mix ≈ 0. Thus, According to Flory-Huggins solution theory [37], ΔS mix can calculated by where j is the mass fraction of solute, v 0 is the solvent molecular volume and x is the degree of polymerization.This calculation is important as it shows that ΔS mix , in accordance with thermodynamic law, is always positive and the disorder of the system will increase with the mixing of the solute with the solvent.The enthalpy term of the system is approximated to describe the energy cost per unit volume of solvent [36,37] where V is the solvent volume, D is the CNT bundle diameter, E S,CNT and E S,Sol are the surface energy of the CNT solute and solvent respectively.With the CNT surface energy being the energy required to completely overcome the vdW binding force and desorb a nanotube from a bundle.From equation (4), the energy expenditure of the system can be minimized when the the surface energy of the solvent matches that of the solute.Leading on from this work, Coleman demonstrated that graphite, which possess a surface energy similar to that of CNTs, could also be exfoliated in certain solvents to produce pristine exfoliated graphene [32].In this pioneering work, graphite was sonicated in N-methyl-pyrrolidone (NMP), to produce suspensions containing ∼0.01 mg ml −1 of exfoliated graphene.To date this methodology has been universally applied to a broad range of layered materials to produce 2D nanosheets from boron nitride (BN), transition metal dichalcogenides (TMDs) [38], black phosphorus (BP) [39], clays [40], phyllosilicates [41], metal oxides [42], double layer hydroxides (DLH) [43], and metal organic frameworks (MOFs) [44].Generally, for a nanosheet the energy expenditure per unit solvent can be approximated as, where t is the nano sheet thickness and E S,S and E S,Sol is the surface energy of the nanosheet solute and solvent respectively.The nanosheet surface energy in this case being the energy required to completely overcome the vdW binding force and peel two flakes apart.However, surface energy is not an ideal predictor for solvent choice as not all solvents with surface energy close to that of a nanosheet will facilitate the creation of a stable colloidal suspension [32].It is for this reason that a more sophisticated and comprehensive method for choosing a solvent is needed.The Hildebrand solubility parameter (δ T ) of a material is easily found by the square root of the total molar cohesive energy density where V in this case is the molar volume of the solvent.Using the Hildebrand-Scratchard expression [36,37], the ΔH mix of the mixed system is approximated as Where δ T,A and δ T,B are the Hildebrand solubility parameters of the nanosheet and solvent respectively.This expression clearly shows that dissolution is favoured in systems where the solubility parameters of solvent and solute match, therefore satisfying the criteria for dissolution (i.e.H mix ≈ 0).It is important to take into account that equation (7) takes only dispersive contributions to the cohesive energy density into account.Most systems however also show hydrogen bonding and polar interactions.This leads to the Hansen solubility parameters, which are the square roots of the dispersive (δ D ), hydrogen bonding (δ H ), and polar (δ P ) components of the cohesive energy density of a material [45,46], and are a measure of intermolecular binding strength where E C,D , E C,H , and E C,P are the cohesive energy densities associated with dispersive, hydrogen bonding, and polar components respectively.The sum of the squares of these three parameters equals the square of the Hildebrand solubility parameter, 9 Hansen [45], suggested that the ΔH mix per unit volume of solvent could be approximated to, Where the subscripts A and B denote nanosheet and solvent respectively.For graphene, a 'good' solvent for exfoliation is described by a Hildebrand solubility parameter of δ T ∼ 23 Mpa 1/2 and Hansen solubility parameters of δ D ∼ 18 Mpa 1/2 , δ P ∼ 9.3 Mpa 1/2 , and δ H ∼ 7.7 Mpa 1/2 .For two systems to mix effectively, all three Hansen parameter must be close, usually within a few Mpa 1/2 [45].Interestingly, it was noted that graphene can be exfoliated in solvents irrespective of their Hansen parameters, though the resultant dispersion would aggregate quickly over time due to 'poor' solvent interactions [47].
However, despite its broad utility, the initial LPE method has many drawbacks.For LPE, the most effective solvents for promoting dissolution are seen to have very high toxicity levels and high boiling points, specifically NMP [48][49][50], making the removal of the solvent difficult.However, a small number of solvents with relatively low boiling points (notably isopropanol, chloroform and acetone) have been shown to disperse nanosheets well [51].Additionally, sonic energy is very limited in terms of commercial upscaling.This however can be replaced in the LPE methodology with shear mixing [52,53], or high pressure homogenisers [54,55], which replicate the shear forces required for exfoliation produced by sonication.Through these methods, producing rate of >9.3 g h −1 and a 100% exfoliation yield have been demonstrated [54].
To note, when exfoliating materials like graphene through LPE, the resultant suspension will be populated with material in various stages of exfoliation.From nanosheets that are very thin and small, to very large unexfoliated aggregates of graphite.Naturally these larger aggregates, which have not been stabilized in solution, will fall out of their own accord due to gravitational force over time.However, this same force is insufficient in overcoming the molecular forces between liquid and colloid to cause the sedimentation of the smaller material.To fully utilize the potential of LPE and of the subsequent 2D material created, size selection procedures are required.With the most effective method of size selection being cascade centrifugation techniques [56].When considering nanosheet influence on suspension viscosity [57], nanosheet dimension was found to scale inversely with centrifugal force and centrifugal time [58].
Though being one of the most effective production methods, one of the largest issues with LPE is the direct coupling of nanosheet aspect ratio to nanosheet thickness [59], Essentially, if you want large aspect ratio nanosheets via LPE, thickness will in turn always be large and vice versa.Thin nanosheets will always have small aspect ratios.For electronic devices this coupling will have a large effect on the transfer of charge carriers across a percolative nanosheet network [60].Meaning charge transfer will be capped by the quality of intersheet junctions.For small LPE monolayers, though they are more flexible and will create more efficient electronic contact with their neighbours in a network, their small aspect ratios will increase the overall number of junctions and thus decrease network transport properties.For larger LPE nanosheets, their thicknesses will make them less flexible, decreasing junction quality and charge transport.Even though large nanosheets will mean less network junctions, due to nanosheet stiffness the network will still present suboptimal charge transport.

Surfactant stabilisation
Water is one of the most abundant and readily available solvents.Unfortunately, water also has a surface energy that is too high to stabilize graphene nanosheets in an aqueous solution on its own.As water is a polar solvent, the stabilization of colloids often relies on the presence of a surface charge, which a non-polar material like graphite/graphene lack [61,62].Through amphiphilic molecules consisting of polar heads and non-polar tail groups in the form of surfactants or soap, water's surface tension can be broken to allow for mixing to occur.Surfactants tend to accumulate at interfaces between phases such that the polar portion can immerse itself in the polar aqueous phase while the non-polar part can interact with the non-polar basal plane of graphene.However, permanent, induced, and fluctuating dipole moments associated with atoms produce both net attractive and repulsive forces, which draw and repel atoms to one another.The addition of surfactant to a graphene's surface not only acts as a mixing agent, but due to the head group dissociating from the chain group, as an interfacial stabiliser between other exfoliated species in water.The head groups form a cloud of counter-ions, known as the double-layer, which counteract attractive forces.Through this repulsive force, via Coulomb interactions, colloids overcome their respective attractive vdW interactions and repel other nearby charged particles [61].It is through this balancing of net forces that colloids resist reaggregation.The balance between attractive and repulsive forces that leads to this stability is described by DLVO theory [61,62], which details how charged surfaces interact through a liquid medium.The total attractive vdW potential per unit area between two sheets is approximated as where ρ is density, C is a constant dependent on the polarizability of the atoms and D is intrasheet distance [61].For graphene, the energy required to separate two sheets from their vdW minimum separation (d 0 ) to infinity is given by the surface energy, For graphene the surface energy has been estimated to be ∼70 mJ m −2 [63].It is also possible to quantify the degree of Coulomb repulsion against reaggregation via the zeta potential.For the double layer, as per the name, it is composed of two parts [64].The innermost region, known as the Stern Layer, consists of counter-ions tightly bound to the charged particle surface.The potential at the edge of this region is known as the Stern potential [64].Outside the Stern layer, ions are more diffuse, mixing readily with water molecules.The zeta potential is the potential just beyond this layer of bound surfactant ions at the hydrodynamic slipping plane edge.
Beyond this point, the ions and particles cease to exist as one entity.The potential at any distance, x, from the charged surface is V = e −κx .Where κ −1 is the thickness of the double layer, known as the Debye screening length [61,65].If all dispersed particles have a large enough zeta potential, the resultant repulsion between colloids will be large enough to counteract the vdW attraction.In the literature, the most common surfactant used for colloidal stabilisation is an industrial soap called sodium cholate [66].Though, household soaps have been shown to be effective stabilisers as well [53].Generally speaking, for effective dispersal and stabilization, all surfactants show similar properties and the choice of stabiliser is really down to the user [66,67].Effectively, when surfactant concentration is below a critical value of ∼5 mg ml −1 (10 mM), all surfactant types produced suspensions that had a similar nanosheet concentration of ∼1 mg ml −1 [66].It has been shown however that the presence of surfactant on the surface of nanosheet does affect network conductivity [55], and charge mobility [68].Though this problem can be overcome by annealing or using a solvent wash [69].

Functionalisation of graphite materials
One important nanosheet derivate of graphite that has also been widely applied in nanocomposite strain sensing research is graphene oxide (GO).First prepared by Benjamin C Brodie in 1860 [70].GO was initially created by treating graphite with a mixture of potassium chlorate and fuming nitric acid to produce graphite foils ∼0.05 mm in thickness.In 1958 Hummers and Offeman developed a more efficient process called the Hummers Method [71].This improved method used mixtures of sulphuric acid (H 2 SO 4 ), sodium nitrate (NaNO 3 ), and potassium permanganate (KMnO 4 ) to functionalize graphite into thin oxidized flakes.Using a modified version of Hummers' Method [72,73].large volumes of colloidal suspensions of chemically modified exfoliated graphene have been demonstrated [74] GO consists of graphene-like sheets that have been chemically functionalized with groups such as hydroxyls and epoxides on the basal plane.The presence of these functional groups alters the vdWs interactions between the layers of graphite whilst changing it from hydrophobic to hydrophilic.With the application of sonic energy, functionalized graphite sheets at a high monolayer percentage are readily dispersed and stabilized in water (figure 2) [75].Though GO has been shown to be very useful in many applications, the functional groups alter the properties of graphene enormously compared to the pristine material.Rather than being a semimetal, GO is in fact an insulator [76], which is problematic for electronic applications.However, the functional groups can be removed by chemical reduction or thermal annealing [77], to make reduced graphene oxide (rGO).Though large defect populations remain in the structure and continue to disrupt the electronic properties [76,77], Similar to pristine graphene, rGO can also be readily dispersed in solvents with appropriate energetics to create stable colloidal suspensions [78].Processing of bulk graphite through the Hummer's method to create graphite oxide, which can be delaminated via sonic energy to create a graphene oxide suspension.

Background and properties
MXenes are 2D nanosheet derivatives of transition metal carbides, nitrides, and carbonitrides isolated through the chemical treatment of ternary materials known as MAX phase (M n+1 AX n ) crystals [79].Where M denotes an early transition metal (Ti, Sc, Nb, V, Mo, Zr, Hf, Cr, etc), A is an A-group element (Al, Si, Ga, etc), X can be carbon and/or nitrogen and n = 1, 2 or 3. First discovered in the 1960s by Nowotny et al [80].MAX phases known as H phases were the first to be synthesised.To date, 60+ MAX pure phases compositions are known based on various stoichiometric compositions [79,81].However, precursor MAX phases are not naturally occurring and are generally fabricated through solid state, high temperature reactions [82,83].The first MXene to be isolated from a MAX phase was in 2011 by Naguib et al [81], with 2D Ti 3 C 2 nanosheets produced through the delamination of Ti 3 AlC 2 .In this pioneering study, Ti 3 AlC 2 was prepared by ball-milling Ti 2 AlC and TiC together.The subsequent mixture was then heated to 1350 °C for 2 h in an argon (Ar) atmosphere to form the starting MAX phase material.The researchers then through wet-etching with hydrofluoric acid (HF) removed the A element (i.e.Al) from the MAX phase and exfoliated the multilayered MXene structure via sonication to yield the first MXene nanosheets.Where the MXene nanosheets were described by the general formula M n+1 X n .This initial discovery led to an explosion of publications investigating the exfoliation of the various elemental combinations of MAX phases [79].Furthermore, it has led to a range of studies which have explored utilising the intrinsic nanosheet properties for a board range of applications such as EMI shielding [84], energy storage/harvesting [85][86][87], communication [88][89][90], and logic [91][92][93], devices.Most importantly, in the context of this review, MXenes have led to the demonstration of many novel nanocomposite strain sensing devices.
The attractiveness of MXenes specifically for strain sensing research lies with their physical properties, with arguably the most investigated MXene being titanium carbide (Ti 3 C 2 T x ) [94][95][96][97].Where T x represents the hydrophilic groups (i.e.-OH, -O, -F) terminated on the M surface.With regards to nanocomposite formation, the weak bonding between MXenes facilitates the exfoliation of large aspect ratio nanosheets in the liquid phase to create monolayer rich colloidal suspensions [98].Similar to graphene, with MXene nanosheets in the liquid phase, it allows for researchers to utilise a wide range of scalable ink preparation techniques [99], required to deposit or mix nanosheets onto/into polymers to make composite systems for strain sensing.Ti 3 C 2 T x is reported to have a Y = 0.33 TPa [100] and electron mobilities of 2.3 cm 2 Vs −1 [101].Additionally, owing to the high monolayer population in suspensions and the large metallic conductivity of Ti 3 C 2 T x , [102].large aspect ratio MXenes form efficient intersheet junctions due to excellent nanosheet flexibility [60].This results in large electrical conductivities for nanosheet networks, with annealed Ti 3 C 2 T x networks just ∼4 nm in thickness reporting values as high as ∼5736 S m −1 [103].In comparison, an LPE graphene network using a similar preparation method and network thickness, but with nanosheets that were thicker and less flexible, reported a network conductivity of <10 3 S m −1 [69].For application, devices based on high conductivity networks would better facilitate battery-powered operation and the integration of devices into data transmission electronics to create wireless wearables.

MXene synthesis
As stated above, the most common procedure for producing MXenes from the MAX phase is via selective etching of the element A [81].This method can be summarised by the procedure in figure 3, where powdered MAX phases are stirred into solutions with molar concentrations of HF at room temperature.After which, solids within the mixture are removed from solution via filtration or centrifugation and washed with deionised water to bring the pH to within ∼6 [104,105].The resultant material from this procedural step is multilayered MXene structures that have a loosely packed accordion-like appearance.Owing to their weak intersheet vdW and hydrogen bonding (via T x terminations), these nanosheet aggregates can be placed in a suitable liquid medium and broken apart through mild sonication or even vigorous shaking of the liquid's vessel [98].Most advantageous is the fact that due to the T x terminations in MXenes like Ti 3 C 2 T x , ecofriendly additive-free aqueous colloidal dispersions can be easily processed [86,87].
Despite their superlative properties, one issue that arises with the MXenes are their ambient stability [106].Due to oxidation effects, MXenes will degrade from the edge inwards, with smaller nanosheets degrading quicker than larger species [107].Through this degradation process, MXenes decompose into amorphous carbon and anatase TiO 2 [106,107].However, though using the same procedural steps to produce the MXene nanosheets, there are wide variations with regards to stability.With stability lasting for days ambiently in water to weeks or months at low temperature in a low O 2 , Ar rich atmospheres [106][107][108].
6.A standard method for reporting strain sensing

Mechanisms of nanosheet network electromechanical response
Before even beginning to discuss nanocomposite strain sensors from literature, we must first understand how they work.One commonality between all nanosheet-based strain sensing systems is the mode in which applied strain induces changes in a sample's electrical resistance [109].In basic terms, as strain is applied to the nanocomposite system, the nanosheet network will deform.This deformation will result in the nanosheets moving out of contact with one another, which results in the electrical resistance of the nanocomposite to increase [110].For a network of nanosheets made up of many junctions, this is obviously a phenomenon that is occurring a large number of times simultaneously in a variety of ways.With the overall nanocomposite electrical response to strain a manifestation of all scenarios combined.
We can begin to understand the mechanisms for resistive change in nanocomposites through a simple, single junction scenario shown in figure 4.Here in Stage 1, where no strain is being applied to the system, we have two nanosheets roughly parallel to one another.The nanosheets are separated by a distance d and their surfaces overlap each other by an area A. An electron hoping to travel through this network will firstly encounter intrasheet resistance (R S ) described as For R S , the value can vary not only between MXenes and graphene but also between the different preparation methods used to create nanosheets [33,111].Many methods will result in minor surface functionalisation which will increase electron scattering [112].Namely the applied use of surfactants or intercalation methods cause such occurrences.As mentioned previously, these surface contaminants can be removed via annealing or acid baths for other applications like printed electronics or electrochemical electrodes [60,113].However, for mixed phase nanocomposites, these optimisation steps cannot be used due to the inaccessibility of the nanosheets in the polymer matrices and general polymer degradation that would occur during these steps.For layer deposition or coated nanocomposites, these methods would still be problematic as they would still degrade most common elastic polymer substrates.
As current continues to flow through our simple network, the electron will also encounter intersheet junction resistance (R d ) when tunnelling from the first nanosheet to the second.For the mixed phase systems, this intersheet junction will contain an interphase of polymer which results in quite large values for R d .While for layer or coated nanocomposites, R d is dependent on network porosity and the sheet-on-sheet contact quality/distance, which is controlled by how the sheets have stacked and their intrinsic nanosheet flexibility.We can describe the total resistance (R T ) of our minimalist scenario as a summation of all resistive factors, Essentially, equation ( 14) describes the nanosheets as resistors in series.With the electron's path going through the first nanosheet (R S1 ), the nanosheet junction (R d ) and finally to the second nanosheet (R S2 ).When strain (ε) is applied to our network in Stage 2 of figure 4, here we assume this to be isostrain, the nanosheets begin to drift apart.To note, ε is below the yield strain (ε Y ) of the bulk system and will be a more relevant observation later in our description of electromechanical response.Nonetheless, increased separation is noted by an increase in R S1 and decrease in A. While R d appears to remain more or less constant.From equation (13), the change in R S1 's value is simply due to the increased path length the electron must travel.Through equation (14), this would result in R T increasing.As strain continues to increase, for Phase 3 in figure 4, nanosheets diverge to a point where they no longer overlap (i.e.A = 0) and R S1 's value is maximised.At this point, ε is greater than ε Y and R d now begins to rise as d subsequently increases.This would then lead to very large increases in the value of R T .
In figure 5, we have a sample electromechanical curve that shows how the various stages of network deformation noted in figure 4 would appear in physical data.Vitally, this figure shows how data should be presented in publications.For an atypical electromechanical curve, fractional resistance change (ΔR/R 0 ) versus absolute strain should be plotted on a log-log to decompress data points at low strain and better highlight this region where critical performance metrics are derived.Commonly in literary data, the low strain regime is often masked by large ΔR/R 0 values at high strains on linearlinear plots.For our sample curve in figure 5, it can be broken up into three distinct regimes: linear, exponential and failure.How our simple network description in figure 4 relates to an expanded network of nanosheets is through Simmons [114], where h, e m , m and λ are Plank's constant, electron charge, electron mass and potential barrier height respectively.Through this more complete description, bulk network R T changes with ε can be attributed to both increases in R S (via changes in electron path length) and R d through the lefthand side of equation (15) (A diminishing).Therefore, ΔR/R 0 will linearly increase.Recent simulations of nanosheet networks deforming in a polymer matrix have also confirmed that linear response is directly correlated to overlapping nanosheets [115].However, as A → 0, R d will begin to exponentially increase as d values rise, the righthand side of equation (15) will begin to dominate bulk electromechanical response.Hence, ΔR/R 0 will exponentially increase with strain.As the rarification of the network with strain continues (i.e.A ≈ 0), final breaks in network connections will result in R T jumping to infinitesimally high values as the network begins to fail.Changes in ΔR/R 0 will occur so quickly that limitations in data sampling will generally make this region appear to be linear and have a non-zero intercept.

Performance metrics and falsehoods
The ability and quality of a nanocomposite for strain sensing measurements is commonly assessed through the gauge factor (G) metric [116], whereby G is defined as Through equation (16), G can essentially be interpreted as the slope of an electromechanical curve beginning at low strain when applying a linear fit with an intercept of zero.To maximise values of G, generally networks approaching their percolation threshold (f 0 ) present the largest value for a system as network connections will be minimised [117].However, due to necking effects, which result in the strain across the networks being non-uniform, values for G in some systems are reported to be maximised at loadings far above percolation when increased network entanglement occurs [118,119].In figure 6, we can see an example of such a fit, with the value of G being found to be ∼0.2 when equation ( 16) was applied.To note for researchers and journal editors alike, this is the only correct way in which G can and should be reported.However, it has become commonplace among researchers to report upon a fictious metric that has been dubbed as 'regional gauge factor'.In principle, a large number of literary reports have taken to reporting high strain slopes in an effort to artificially increase publication merit to sway editors.This has then had a severe knock-on effect in Figure 4.A minimalistic mechanistic view of deformation in nanosheet networks.At zero applied strain (ε) (Stage 1), nanosheets described by intrasheet resistances R S1 and R S2 will overlap with one another by an area A and reside a distance d away from each other with a respective intersheet resistance of R d .When strain is applied to the system, whereby the strain is less than the yield strain (ε Y ) of the bulk system, areal overlap will decrease while inter-sheet distance will remain unchanged, resulting in R S1 increasing.When applied strain is larger than the yield strain, nanosheets no longer overlap and inter-sheet distance increases, causing R S1 and R d to increase in value.Examples of such 'false' gauge factors are shown in figure 6, where the method of forcibly applying linear fits over nonlinear data at high strains was applied.The falsehood of these values is proven by the simple fact that these fits do not have an intercept of zero.Thus, they do not satisfy the definition of G from equation (16).Furthermore, at high strains, simulations have shown that data will never be linear due to strain's effect on intersheet junctions [115], making the use of equation ( 16) at high strains inappropriate.For our sample curve in figure 6, each of the false gauge factor fits has a different intercept value that we called i 1 and i 2 for the exponential and failure regimes respectively.Hence, the slopes of these fits are not and cannot be called G. The misleading nature of these false gauge factors is also highlighted by the fact that the slope value at high strain in the failure regime is ∼0.7, or a factor of ∼3.5 larger than true G.This however is a modest representation of the problem, with some published reports quoting slopes in the failure regime that are 10 4 times larger than the nanocomposite's true G value.The reporting of these embellished values is particularly frustrating as many authors that report them exclusively demonstrate low strain applications that are in the true G linear regime, like pulse measurements.This makes the quoting of these false values in abstracts completely redundant and merely a reflection of the publish or perish culture in research.
Where these false gauge factor values truly fall flat, is the viscoelastic nature of the polymer matrices and substrates applied to make the devices [120].Essentially, these false sensitivities are completely inaccessible for applications.To utilise these high strain values, it would imply that the nanocomposites would need to somehow be permanently prestrained to the onset point of these false gauge factors.This however is impossible due to mechanical creep and stress relaxation [121].Additionally, at high strains, the nanocomposites will be stiffer and would be unable to deform with or conform to the body [120].Furthermore, prestrains are in excess of their ε Y , causing permanent plastic deformation [122].

Gauge factor-electrical percolation coupling
For nanocomposite sensors, maximising G makes devices more suited for high precision applications and thus more desirable from a commercial standpoint.In literature, there are a lack of simple models that can be generally applied to a wide range of nanocomposite systems to describe how system parameters effect or yield large G values.There are however minor nuances across literature such as the inverse dependence between G and loading level (f), which have allowed for rudimentary tuning, and optimising of G for individual systems [123,124].This understanding has led to the broadly disseminated mantra that low density network have fewer connections and are thus more sensitive to strain.Whereas more dense networks that are well connected are less electromechanically sensitive.Based off this common knowledge, Garcia et al [125], showed that a simple model could be derived where values for G were highly dependent on two system constants for mixed phase systems, Here, G 0,f describes how strain affects the overall physical structure of a filler network and G 1 defines how percolative properties are affected by strain.For this expression to yield large values for G, the authors stated that G 1 would need to be maximised and in doing so, a large interphase of insulating polymer between fillers would be required.This statement however can also be interpreted as, large G 1 values alternatively requiring the orientation of the fillers in the network to being more anisotropic to maximise filler distance.Previously, increases in filler alignment showed large increases in G, consistent with equation (17) [117,[126][127][128].This model has been applied broadly across many strain sensing systems, with values for G 0,f and G 1 reported to vary between −200 and 80 and 10 −2 and 10 2 respectively [123,129].Fundamentally, this created a connection between the percolative properties of the network and its electromechanical sensitivity that could be derived as Where σ 0 , and n 0 are the zero-strain conductivity and the zero-strain percolation exponent respectively.For threedimensional charge transport, the universal value for n 0 is generally ∼2 [130][131][132].However, due to the width of junction resistance distributions, many nanocomposite systems report values >2 [133,134].To note, the reported values of n 0 from equation (18) fittings were observed to be similar but not the same as those found via standard percolation-like fits through σ 0 verse f plots.For equation (18), G 0,σ and σ 1 are system constants.With G 0,σ reported to vary between −200 and 80, while σ 1 had values from 10 -4 to 10 13 S m −1 .For both equations ( 17) and (18), they have been observed to also describe non-nanosheet systems [129,135].
The importance of network percolative properties have also been highlighted for thin film layers of graphene on flexible substrates [136].It was reported that G scaled with zero strain resistance (R 0 ) according to the following relationship, G R . 1 9 Like mixed phase systems, G values for thin layers of graphene were observed to also follow an inverse scaling with percolative properties through the network thickness parameters, where, t 0 and t c,0 are the network thickness and critical network thickness for charge transfer respectively.Once again, t c,0 values extrapolated from equation (20) fittings were similar but not the same as those found via standard percolation-like fits for σ 0 verse t 0 plots [136].For G TNS and t TNS , they are system constants that are related to the combined effect of strain on intersheet charge transport and network structure/dimensionality, and the change in structure as a function of strain respectively.For equation (20), to optimise the G of thin networks, similar to the case in equation ( 17), the t TNS figure of merit would ideally be maximised.Reported values of G TNS and t TNS for graphene thin film networks where ∼22 and ∼2.3 μm respectively [136].
Expressed in terms of σ 0 , G can be described as Where σ TNS is a constant related to increased G values and had a reported value of ∼10 7 S m −1 for thin graphene networks [136] Though not examined to date, with MXene layers having similar morphological composition (i.e.porous, size dependent network) to graphene, equations (20) and (21) would similarly describe network electromechanical properties.Additionally, this highlights the lack of and applied use of modelling with regards to MXene specific systems.

The modified kraus model: a universal approach
Critically, the applied usage of true G unlocks a wealth of information that would otherwise be inaccessible to those using the incorrect interpretations of the metric.For the fitting of equation ( 16) in figure 7, Boland et al [12,117,119,129,135], demonstrated that an intrinsic metric known as the working factor (W) can be extrapolated.Strictly, W is defined as the absolute strain limit at which the fit for equation (16) begins to fail and electromechanical data deviates from the linear regime to the exponential one [12].Additionally, the value for W is reported to be directly related to the ε Y of a network at low loading levels [117].Using W, an alternative expression for G can be formed through a modified version of the Kraus model [12,116,117].
where n e is a metric for alignment.Values for n e are reported to vary between ∼0.1 and ∼5, with larger values for the metric indicative of increased alignment [12,117], equation ( 22) solidifies that G is directly proportional to network alignment, which runs many parallels with the Garcia et al work [125].With regards to the magnitude of W, it is dependent only on the areal overlap of constituents in a network, with larger A values leading to higher W values [117].The value for W can be artificially improved through loading level increases, whereby nanosheets are forced to overlap more [117,135].However, this will lead to diminishing returns in relation to G, as there is an inverse correlation with the number of network connections.The relationship between G and W can be described through a universal scaling law [117].
Where m is a universal metric known as the Kraus constant, which is related to the fractal dimensions of the network [137].The expected value for m is 0.5, though experimental values for nanocomposite systems have reported its value to vary between 0.3 and 0.6 [116,137].For a study using the electromechanical data from 200 publications [12], values for G and W universally scaled with a power law exponent of 0.5, per equation (23).Furthermore, in recent studies on fibrous [129,135], mouldcasted [119], and core-shell nanocomposites [117], all reported G to scale with W according to equation (23).What is most profound about this simple universal scaling law is that it fundamentally sets limitations for nanocomposite electromechanical performances like no other model has.Intrinsically, for large G values the linear sensing range defined by W will always be small and vice versa.Large W will always result in a small G value.This comes from W's direct relationship with ε Y , and through the Kraus model, W ≈ K 1/2 m [137].In this expression, K is the ratio between the rate at which network bonds reform and break.Meaning G, a metric controlled by the rate of network connections break, is intrinsically inversely proportional to W when m = 0.5.Though this model is based on a theory involving the deformation of a filler network in a high viscosity polymer matrix, the metrics and scalings described here can be universally applied to any nanomaterial network being strained.Specifically, for coating-or layer-based nanocomposites, the networks can be interpreted as having their voids filled with air and interphases comprised of surface functionalisations, rather than polymer for both cases.For these nanocomposite types, their stacked heterostructures [138] and elastic substrates give their networks their robust yet flexible mechanical behaviour, akin to mixed phase systems.Uniform deformation applied to these networks also leads to linear electromechanical response beginning at low strains and values for W that scale according to equation (23) [12].We will again show that this is the case in the subsequent critical review of literary performances.

Cyclic signal decay
For nanocomposite strain sensors to reach their potential commercial apex, they must demonstrate sensing performances that are sustainable over many measurement cycles.For these nanocomposites, electrical resistance change is fundamentally linked to a mechanical strain being applied.Hence, the name electro-mechanical response.Thus, it is no surprise that hysteresis effects have ramifications on the electrical response of these nanomaterial doped polymers [139][140][141][142][143][144].Generally, for repeated mechanical measurements, pristine and filler loaded elastomers will have the maximum attainable stress decrease as a function of cycle number.Similarly, for electromechanical measurements, the sensitivity of a material will also decrease with cycle number [145].This phenomenon is known as the Mullins effect [146], and is a result of the polymer chains or filler materials debonding from one another or themselves due to viscoelastic effects.Post-cycling analysis of carbon black/carbon nanotube fibre nanocomposites show that these viscoelastic effects can be fully described through the standard linear solid model [147].Essentially, nanocomposites display signal fatigue, whereby repeated measurements result in a softening of the polymer and polymer/filler interphase [148,149].This then causes the reported diminished sensitivity (i.e.G) as stress/ strain are no longer being effectively transferred to the filler network.However, there are exceptions to fatigue occurrences in nanocomposites, specifically those based on selfhealing polymers [123,150].Though the softening that does commonly occur in nanocomposites is also not permanent, with reported recovery occurring over a much longer time scale [151].Using a Wöhler's plot, where signal amplitude ((ΔR/R 0 ) R,C ) is plotted as a function of cycle number, a modification of Basquin's law of fatigue [152] Where α = G 0 σ 0 /Y 0 and B is the Basquin exponent.Here G 0 , σ 0 and Y 0 are the gauge factor, stress, and Young's modulus after one cycle respectively.To note, Basquinian fitting can only be applied to testing below the yield strain (i.e.below the W value) [153].We essentially see, intrinsically, the previously highlighted electromechanical metrics dictate performance in a multifaceted way.
A key component of the observed fatigue properties is that the decay in signal is not a constant effect with cycling.Through equation (24), a conditioning metric can be derived.The endurance limit (C E ) describes how many conditioning cycles must be performed on a sample to reach a steady state signal.With C E calculated via a system's B value through the following expression [152].
Through equation (25), extrapolated values for C E were found to match precisely experimental values observed for the number of conditioning cycles required to achieve a steady state signal [129,135,152], equation (25) also vitally allows for the C E value to be calculated without the need for extensive testing, as a good fit for B can be attained from modelling 10s of cycles.This allows for high throughput screening of potential sensing devices to find those who have low C E values.Alternatively, equations ( 24) and (25) can be used as tools to aid in the signal processing of nanocomposite sensing devices that present fatigue due to large C E values.Furthermore, the utility of this Basquinian model lies not only in its dependence on universal electromechanical metrics, but its rate independent nature [154].We will now review some of the most recently published works on nanosheets being applied in strain sensing nanocomposites.Through this investigation, trends in literature will be discussed in terms of literary data trends and their ramifications on the research area achieving commercialisable devices.For the nanocomposites we will assess, they can be broken down into two classes.The first being one of the simplest forms of nanocomposite strain sensors, mixed-phase systems [155][156][157].Through this methodology, nanosheets are incorporated into an elastic [157][158][159][160], or ionic gel [161][162][163], polymer matrix either through solution-solution or powder-solution mixing.Regardless of the type of mixing, the basis of the process involves the mixing of nanosheets and polymers in the liquid phase [164][165][166][167].To note, for efficient mixing to occur, both colloid types must have Hansen solubility parameters that match that of the liquid they are suspended in, per equation (10).The resultant mixture is then generally left to dry out in a mould, with a solid film of polymer now embedded with nanosheets formed after all the solvent has been removed [164][165][166][167].However, for hydrogel systems, mix-phase nanocomposites form a soft gelated material that is swollen with water once all materials have been added [162,[168][169][170][171][172][173][174][175].The second form of nanocomposite to be analysed are layer deposition or coated nanocomposites.For these materials, nanosheet colloidal suspensions are drop-cast [176][177][178][179][180][181][182][183][184], dip-coated [185][186][187][188][189][190][191][192][193][194][195][196], spread coated [197], spray coated [198][199][200][201], spin coated [202] or screen printed [203][204][205][206] onto a flexible substrate or scaffold.Generally, the substrates are heated during these processes to facilitate solvent evaporation and allow for thin percolative networks of nanosheets to evenly form.For the sake of completeness, we also assessed and compared recent studies on layers of nanosheet formed via laser scribing [207][208][209][210][211][212] and growth methods [213,214], to show the universality of Krausian electromechanical scalings.
With regards to the 70 recent literary nanocomposites that were assessed, in figure 8(a) we see that for both graphene and MXenes that layer-or coated-based (LC) nanocomposites were more frequently applied to create sensing devices.A breakdown of our publication pool by numbers, we see 22 graphene and 19 MXene reports (total of 41) on LC nanocomposites.Conversely, 13 graphene and 16 MXene publications (total of 29) that were surveyed reported on mixed-phase nanocomposites.S1), we find that mean gauge factor (〈G〉) for LC and mixed-phase systems were 74.85 ± 28.77 and 9.71 ± 7.23 respectively.The much lower 〈G〉 value associated with mixed-phase MXene nanocomposites is partially due to an observed preference for hydrogel-based systems for this nanosheet type.This is due to the previously discussed hydrophilic functional groups which pattern the surface of Ti 3 C 2 T x , facilitating the easy production of aqueous suspensions and effective mixing with highly hydrophilic polymers.However, due to ion conduction [215], hydrogel matrices essentially 'short' the nanosheet network.This results in hydrogel nanocomposite electromechanical response being not as reliant on network deformation effects [12], as the matrix can affect intersheet charge transport.Highlighting now the best performing MXene works, one interesting study on an LC MXene nanocomposite by Pu et al [216], presented MXene flakes that were drop-casted onto a polyurethane substrate.The nanosheets were then controllably oxidised to create a micro-crack network made up of nanosheets with TiO 2 nanoparticles edges.Most impressively, this device displayed G values as high as 530 for strains up to ∼0.2%.This methodology takes advantage of the previously reported 'cracked' sensor composition which had been widely shown to create devices with very large G for small strains [217][218][219] and is a reflection of equation (22).For a mixedphase MXene system, a polypyrrole/hydroxyethyl cellulose based nanocomposite by Yang et al [156] reported G ∼ 118 for strains up to ∼12%.Overall, the MXene nanocomposites had a 〈G〉 of 45.05 ± 16.72, which is in line with a previously reported value observed across multiple nanocomposite systems that applied a diverse range of nanomaterial types [12].
For graphene-based devices also in figure 8(b) (see table S2), 〈G〉 was found to be 46.48 ± 17.46 and 52.60 ± 16.85 for LC and mixed-phase composites respectively.In comparison to the MXene works, the overall sensitivity of graphene-based devices was similar but slightly higher at 48.76 ± 12.49.This implies that with regards to G value, the type of nanosheet is less important than the manner in which a device or network has been engineered.Where the importance of nanosheet type will lay is the efficiency of the network junctions and device conductivity.MXenes overall can be produced as much larger aspect ratio and thinner nanosheets when compared to graphene.However, issues of chemical stability for MXenes, and thus long-term device performance stability, do arise.For the top performing LC graphene nanocomposite, Caffery et al [136], demonstrated a simple spray coated PDMS substrate which presented G ∼ 350 for strains up to 1%.For the mixed-phase graphene nanocomposites, a g-putty [116], sample by Garcia et al [125], reported a G ∼ 225 for 0.75% strains.
In figure 8(c), we now compare the values for mean working factor 〈W〉 for each nanosheet and nanocomposite type.What is self-evident from the plot is that for the MXene samples, the LC nanocomposites have a much smaller 〈W〉 value than the mixed-phase systems, 0.49 ± 0.20 as opposed to 2.28 ± 0.51 respectively.This is a direct reflection of the inverse scaling of G with W noted in equation (22), where the large G values seen in the LC reports always led to smaller W values.For the graphene samples, as 〈G〉 was similar between the two nanocomposite types, LC and mixed-phase systems reported similar values of 0.17 ± 0.05 and 0.22 ± 0.08 respectively for 〈W〉.Due to the prevalence of hydrogel systems for mixed-phase MXene nanocomposites, comparing the overall W values between nanosheet types would lead to a skewed statement on performance.Though this would lead to a less conclusive vision for researchers to utilise, the analysis of these systems does highlight an interesting point that hydrogels, like conductive polymer-based nanocomposites in the past, lead to very large W values.For applications where a large linear sensing range is required, large G is not necessarily required.However, when making a more direct comparison via the LC systems for the two nanosheet types this paints a clearer picture on how W is controlled experimentally.As noted above, MXenes report a larger value of 0.49 ± 0.20 in comparison to graphene (0.17 ± 0.05), which is simply due to the aspect ratio of MXenes in devices being much larger.With the applied use of large nanosheets leading to larger values for A in a network (equation ( 15)).Looking back at the 〈G〉 values for the LC systems, we find this to also be reflective of the applied use of large nanosheets, which lead to less network connections and thus a more sensitive network [128].Most notably from the systematic review of literary data is that 86% of publications report a linear strain limit for electromechanical response.However, the vast majority do not correlate this value to the figure of merit W or note the universal inverse relationship with G (equation ( 23)).

Universal performance scalings
In figure 9, all data values for MXene and graphene devices are placed in a G verse W master plot.For the plot, we find the data to scale according to a power-law of exponent −0.50 ± 0.08 in accordance with equation (23).This finding is precisely in line with a previous study on a wider range of nanomaterial-based sensing devices [12] and further validates that G will universally scale with W according to the Kraus constant [117].The implications of this scaling has been touched upon previously in this review, however it is worth reiterating the ramifications of this universal scaling as it has large implications for applications.Essentially, the previously envisioned idea of an 'all-in-one' sensor [12], with both large G and W is not possible.Strain sensing nanocomposites based on nanosheet networks can only measure small strain with a high degree of accuracy or large strains with a lower sensitivity.

Suitability for application
Without question, the superlative properties demonstrated by the surveyed devices certainly makes their potential use as commercial health applications very enticing [220].However, the combination of properties these devices need to display to even be suitable for application must now be discussed.At a minimum, these sensing devices should either present a sensitivity or measurement range larger than the performance standards set by commercial devices (i.e.G > 7; W > 0.1).Through this very basic criterion, ∼52% of surveyed literature had a G > 7, while ∼60% had an W > 0.1.Additionally, ∼26% had optimum values for both metrics.
Most critically, for application as a wearable device the mechanical properties of a nanocomposite must match that of the body to even be considered for potential commercialisation [221].For skin on applications of devices, whereby the nanocomposite is worn on the surface of the skin, they must present Young's moduli (Y) similar to skin (i.e.Y < 300 kPa) [12,222,223].The stiffness of a device is of such critical important due to the potential for signal dampening if a nanocomposite is too stiff.Surprisingly, ∼54% of reports surveyed did not quote values for Y, and those that could have values sampled, ∼44% had ones that were above the critical stiffness threshold (tables S1 and S2).For the studies that Y was taken, only ∼47% reported a value of their own accord.With the other studies requiring the value to be extrapolated from stress-strain plots.This is again a reflection of the previously mentioned fixation on attaining large G values at the cost of other device attributes.This has then led to a lack of consideration for the importance of mechanics on health sensing.So critical is the reporting of mechanical properties that if a manuscript does not quote them, it cannot be considered a complete study that demonstrates a device for health sensing.The central reason as to why these nanocomposites are investigated in the first place are for health applications and to not consider all factors which would affect this goal calls into question the usefulness of a study.When comparing the surveyed MXene and graphene nanocomposite devices in figure 10, overall, the MXene reports more values below the application threshold.The reason being is that mixed-phase MXene nanocomposites were predominantly based on very soft hydrogel systems.

Conclusions and questions of sustainability
Looking at just the electromechanical properties of nanosheetbased nanocomposites this review arguably shows MXene devices to have superior strain sensing performances in comparison to graphene.We define this superior performance through the metrics G and W, whereby MXene LC systems overall yielded larger sensitivities and linear sensing ranges respectively.Through the modified Kraus model, MXenes' superlative performances are mainly attributed to their large achievable aspect ratios.These larger aspect ratios allow for more areal overlap of nanosheets in a network, facilitating larger W values, and less network connections, leading to large G values.Furthermore, when plotting both MXenes's and graphene's G and W values on a master plot, they followed a universal scaling dictated by the Kraus constant.However, as noted through the systematic review of nanosheet sensor literature, more consideration must be made with regards to device mechanical properties.It was found that a large majority of research does not consider the critical requirement that wearable electronics must be exceedingly soft and compliant to effectively function in health application.
Though good research process is being made overall, there is also quite a large shortcoming with regards to sustainable methodologies in the field.The climate crisis is something that affects us all and consideration for how nanocomposite research will contribute to the issue needs to be considered now [224].Essentially, we must use horizontal principles for sustainable development and consider not only what the effect a commercial nanocomposite material will have on healthcare globally, but how the mass production of these polymer-based devices will also effect society.Most research in literature uses harsh solvents and non-biodegradable polymers to make their devices.Thus, there manufacturing on a large scale would contribute to waste products and forever chemicals that negatively impact human health [225,226].Essentially their production would inadvertently harm human health, which is the opposite of their intended use.Furthermore, these devices are viewed as being cheap and disposable.However, we must also then consider what happens when thousands of these polymer-based devices need to be replaced each year, month or even day.Their wrongful disposal could then contribute greatly to the harmful microplastic particulates that we currently are only beginning to grasp the seriousness of [227][228][229][230].As no nanocomposite device has yet reached the market, there is still time for research to move forward with more sustainable means via green solvents and naturally sourced biodegradable polymers.Many forms of biodegradable polymers have been demonstrated in research which possess the qualities needed to function as wearable electronics [231][232][233].However, with no readily available commercial material existing to date, this has likely hampered widespread adaptation of such polymers for general nanocomposite research.It is thus believed that research into more ecofriendly devices hinges on cheap, readily available materials coming available.

Figure 1 .
Figure 1.Procedural process of the liquid exfoliation of graphite to produce suspensions of graphene.

Figure 2 .
Figure 2.Processing of bulk graphite through the Hummer's method to create graphite oxide, which can be delaminated via sonic energy to create a graphene oxide suspension.

Figure 3 .
Figure 3. Etching of MAX phase materials to yield multilayerd MXene that can be exfoliated to create large aspect ratio nanosheet suspensions.

Figure 7 .
Figure 7. Zoom in of low strain region of figure 6. Fitting of gauge factor (G) from equation (16) and the extrapolation of working factor (W) are shown.

10 .
Systematically assessing recent literature-from 2020 to present day 10.1.Types of nanocomposite strain sensors

10. 2 .
Sensing performance by nanosheet and composite type Looking first at MXene-based materials in figure 8(b), using the reported and extrapolated G values via equation (16) from literature (see table

Figure 8 .
Figure 8. Statistical breakdown of literary data.(a) Breakdown of MXene and graphene nanocomposites in terms of whether the sampled study was based on a layer or coating of nanosheets on a scaffold/substrate or whether the system was a nanosheet/polymer mixture.(b)-(c) Mean gauge (〈G〉) and working (〈W〉) factors of MXene and graphene of mixed phase and layer or coating nanocomposites.

Figure 9 .
Figure 9. Universal scaling across literature.A plot of MXene and graphene nanocomposite gauge factors (G) as a function of their respective working factors (W) show data to expectantly scale according to the Kraus constant (m).