Theory-Guided Materials Design: Two-Dimensional MXenes in Electro- and Photocatalysis

Two-dimensional transition metal carbides and nitrides (MXenes) have made significant impact on sustainable energy research in the fields of energy storage and conversion. Unlike short-term energy storage strategies (e.g. batteries and supercapacitors), the catalytic conversion of simple molecules to value-added chemicals using renewable energy represents a more long-term solution to the world’s energy crisis. Significant advances in density functional theory and low-cost computing in the past decade have enabled the generation of reliable materials data from fundamental physics equations. The paradigm shift towards theory-guided materials design is expected to enhance the catalyst discovery and development process by providing rational guidance to screen viable MXene catalysts more rapidly than an experimental-only approach. In this review, we aim to provide a critical appraisal of the latest theoretical and experimental work on MXenes in the fields of electroand photocatalytic energy conversion, including relevant reactions involving hydrogen, oxygen, carbon dioxide and nitrogen molecules. In the process, we will also be pointing out current limitations in theoretical models, existing scientific gaps and future research directions for this field.


Introduction
The burgeoning energy demand linked to rapid global population growth and expanding industrialization calls for new sources of sustainable and renewable energy. The U.S. Energy Information Administration estimates that close to 80% of the world's primary energy consumption is derived from fossil fuels, including coal, natural gas, petroleum, and related products. 1 Fossil fuel is not a renewable source of energy, and its depletion within the next century could pose grave consequences if suitable alternatives are not found in time. 2 In addition, fossil fuel combustion is also a major source of CO2, along with other noxious pollutants like soot, SOx, NOx that further exacerbate environmental issues.
Electro-and photocatalytic energy conversion presents an attractive pathway to convert Earth abundant molecules (e.g. H2O, CO2, or N2) into high-value products and fuels (e.g. H2, hydrocarbons, oxygenates, or NH3). While sustainable conversion can be achieved by tapping into renewable energy sources, the viability of this process is contingent on the development of selective, robust and affordable catalysts that allow faster reaction rate to occur at lower overpotential. The search for high performance catalysts, especially those without precious metals content, has attracted intensive research interest. 3 MXenes, a contemporary family of 2D materials based on transition metal carbides or nitrides, have shown interesting properties and promise in catalytic energy conversion. 4 So far, of the 54 MXenes proposed to exist in silico, only 22 of them have been experimentally realised (Fig 1). One of the challenges in MXene synthesis is the instability of some of its constituents in strong HF etchant, particularly nitrogen and transition metals reactive to fluorine. 5 More recently, milder synthesis routes using in-situ HF generated from fluoride salts in acid have been demonstrated. 6,7 In this review, we will discuss the latest promising research on MXenes in the field of electro-and photocatalytic energy conversion, based on the strategy of theory-guided materials design. First, the fundamentals of catalytic energy conversion will be elaborated, including H2 evolution reaction (HER), O2 evolution reaction (OER), O2 reduction reaction (ORR), CO2 reduction reaction (CO2RR) and N2 reduction reaction (N2RR). The key properties of MXenes relevant to electro-and photocatalysis will also be discussed. A critical appraisal of the state-ofthe-art research using MXenes as catalysts in the above reactions will follow, including both theoretical and experimental aspects. Using computational chemistry to guide materials design has the potential to accelerate catalyst development by uncovering broader governing principles and providing rational design guidelines. In the process, we will also be pointing out current limitations in theoretical models, existing scientific gaps and future research directions for this field.

Basic Concepts
Electrocatalysis exploits the thermodynamic and kinetic tuning of electrochemical reactions occurring on electrode surfaces as a function of the applied external voltage. An electrocatalyst modifies the catalytic reaction rate compared to solution electrochemistry or an inert electrode, by providing favourable reaction sites, altering adsorption energies and/or introducing alternative reaction pathways. Therefore, faster product formation can be achieved at lower overpotentials. Modifications in exposed facets, 9 surface functionalisation, 10 size [11][12][13] and doping/co-catalyst addition 14, 15 are common strategies to enhance electrochemical activity and selectivity. Depending on the target reaction, MXenes can be suitable electrocatalysts due to high electronic conductivity and optimal adsorption strength for key reaction intermediates.
In photocatalysis, the driving force arises from electron/hole pairs generated in semiconducting materials upon light irradiation. The additional step of photocarrier generation makes photocatalysis generally less efficient than electrocatalysis due to the associated energy losses. Important strategies to improve photocatalyst performance usually involve bandgap alteration, 16 addition of co-catalysts, [17][18][19] or the formation of cascading junction (Z-scheme) structure [20][21][22] aimed to enhance light absorption and promote photogenerated carrier separation. 23 MXenes can be a prime choice for co-catalysts as the Fermi level is compatible with popular widebandgap photocatalysts like TiO2 and ZnO such that it encourages electron migration to MXene surfaces and simultaneously form a Schottky barrier at the interface to prevent back-flow. 18 Before going into the details of electro-and photocatalytic process, there are several fundamental terms that need to be defined.
Overpotential describes how much additional voltage above the thermodynamic requirement is needed to drive an electrochemical reaction. Expressed as a potential difference, overpotential is considered as one of the primary metric in evaluating the catalytic activity as it is strongly related to the redox reaction kinetics and the overall energy efficiency of the process.
Experimentally, it is determined by the onset potential at which a redox event is observed, usually marked by a significant rise of the current density compared to the baseline (non-redox) level.
Electro-and photocatalytic reactions such as CO2RR can be complicated, with multiple electron transfer steps involving different intermediates and overpotentials. 24 It is common to take the largest overpotential (i.e. the most difficult) of those individual steps to be the overall redox overpotential. An excellent catalyst should display overall overpotential value close to zero.
Faradaic efficiency (FE) describes how efficient a catalytic system is in using electron input to facilitate a particular chemical reaction. In catalysis, FE is usually calculated by comparing the eoutput, estimated based on the amount of product generated (or reactant consumed), with the actual einput as measured by a coulombmeter or a potentiostat. A real catalytic reaction usually shows FE less than unity, as Faradaic losses can be incurred when electrons or ions participate in unmonitored side reactions. Comparison of FE among different products can serve as insightful indicator for both energy efficiency and product selectivity, especially important for CO2RR where product selectivity is a critical challenge.
Turnover number (TON) describes the intrinsic activity of a catalyst. It is usually defined as the total number of reactant molecules converted (or the total number of product molecules generated) per active site. TON is also often used to describe the "recyclability" of a catalyst, which is calculated by the number of molecules of reagent that a molecule of catalyst can convert before it becomes inactive. In both cases, an ideal catalyst should have a very large TON, approaching infinity. Sometimes, the term turnover frequency (TOF) is used, which is simply TON per unit time. TOF describes the number of catalytic conversions that occur at the active centre per unit time, which also gives information on the overall kinetics. Thus, the higher the intrinsic reaction rate, the higher the TOF value.
Tafel plot/equation relates the rate of electrochemical reaction on a particular catalyst to the applied overpotential: where ŋ is overpotential, b is Tafel slope, i is current density, and i0 is exchange current density.
The constants and coefficients used in the equation are R, T, F and α, representing gas constant, temperature, Faraday's constant and reaction transfer coefficient respectively. Two of the most important measures in the Tafel equation are b and i0, as they can provide insights regarding interaction between reactant/electrolyte and the catalyst surface, as well as the rate-determining reaction step. Experimentally, the values of b and i0 can be obtained by plotting the applied overpotential against the logarithm of observed current density. The Tafel slope b is the overpotential difference required to increase the observed current density by a decade (thus the unit V dec -1 ), and i0 is obtained by extrapolating the Tafel plot to intersect the "log i" axis at zero ŋ. Smaller b and larger i0 values are indicators of a good catalyst, implying a larger reaction transfer coefficient and better binding between reactants/electrolyte and catalyst. Tafel slope values, subject to limitations, can also be used as indicators of reaction pathway and rate determining step in electrochemical reactions. For example, a multi-step reaction with rate determining chemical step preceded by two 1-electron electrochemical steps should display Tafel slope of around 30 mV dec -1 . 25 The Sabatier's principle states that the adsorption between the catalysts and reactants should be neither too weak nor too strong. 26 When the adsorption is too weak, it becomes difficult to bind the reactants onto the catalyst surface, and no reaction would take place. On the other hand, very strong adsorption causes the reactant to be too strongly bound, slowing down the product desorption step, and sometimes leading to rapid catalyst deactivation. 27 Hence, an optimal catalyst should bind the reactants just strong enough to allow surface adsorption of reactant, while still allowing facile desorption of the intended products. Taking a relatively simple HER as an example, the catalytic activity (expressed as i0) of the elementary steps (Volmer, Tafel or Heyrovský) is exponentially proportional to the partial pressure of hydrogen (pH2) and the standard Gibbs free energy of adsorption of atomic hydrogen (ΔGH) by: 28 where, k is the Boltzmann constant and T is temperature in Kelvin.
Plotting log i0 against ΔGH results in a triangle that yields a maximum value of i0 when ΔGH is close to 0, and rising and falling log i0 when

Computational Methods
Density functional theory (DFT) is frequently exploited to calculate the adsorption energy based on the differences between the energy of a "clean" atomic surface model plus half the energy of a H2 molecule, representing the proton/electron pair. This model is known as the computational hydrogen electrode (CHE). 29 In combination with DFT computations for the intermediates (e.g. the energy of hydrogen bound onto the same surface model), thermodynamic overpotentials can be estimated at low computational cost. The thermodynamic overpotential is defined as the potential at which each electrochemical step (coupled proton/electron transfer) is thermo-neutral.
These overpotentials can then be plotted as a function of a descriptor (e.g., the d-band centre or the adsorption energy of a particular intermediate) to obtain a volcano plot. Though simple, volcano plots have proven to be a powerful tool that can predict the activity of different materials in catalysing various reactions. 3,30,31 In its most commonly used CHE formulation, neither electrolyte presence nor electrode polarization due to applied electrochemical potential is taken into account. It also assumes that no additional kinetic barriers exist, i.e. if a reaction is thermodynamically favourable at a given potential, it will also occur at a high rate. These assumptions, which simplify the computational work significantly, often provide correct trends, 32-34 although they cannot capture the details of the electrochemical processes. [35][36][37] Furthermore, for complex reactions, where chemical and electrochemical steps have to be considered, microkinetic simulations are necessary to estimate the rate of the electrochemical reactions, rather than taking the thermodynamic overpotential as provided by CHE as a proxy for the activity of a given catalyst. 38,39 In order to lift the approximations of CHE, transition states for all processes need to be determined, which is particularly challenging for electrochemical reactions. 40,41 Second, the electrochemical potential needs to be included explicitly, which is equivalent to simulating electrode surfaces in the presence of a surface charge. Third, the solvent needs to be modelled as well. The computationally most efficient approach is to rely on an implicit solvent/electrolyte, [42][43][44] which comes at the cost of neglecting "direct" (i.e. chemical) effects of the solvent, such as the competition of the adsorption between reactants and the solvent or the assistance of solvent molecules in breaking bonds. 45 Alternatively, aqueous electrolytes are often approximated as icelike layers, 46,47 which is very cumbersome for studying various reaction intermediates, as the (arbitrary) placement of adsorbates into or under these ice-like layers is associated with uncontrolled effects on their properties. Furthermore, for non-aqueous solvents, no obvious organization for placing explicit solvent molecules is currently available. All these challenges explain why the simplistic CHE model, with its success in predicting activity trends, continues to be highly popular.

Properties of MXenes
In order to elaborate about MXenes, one has to understand a special family of the layered MAX phases. The acronym MAX comes from the materials' composition, Mn+1AXn (n = 1, 2 and 3), where 'M' stands for early transition metals such as Sc, Ti, Cr, Nb, V, Mo and so on, 'A' is group IIIA or IVA element, and 'X' represents C and/or N. MAX is composed of Mn+1Xn face-sharing octahedral layers and interlayer cation A, resulting in a hexagonal 6 3 ⁄ structure, the same symmetry as graphite (Fig 2a). By etching out the interlayer cations A, a graphene-like 2D transition metal carbide/nitride Mn+1XnTx (here T represents the surface termination groups such as fluorine, oxygen or hydroxyl) is obtained, and the name MXene was given due to their remarkable similarity to graphene and the relation to MAX phase. 5 The relatively strong metallic M-A bonds in MAX phases suggests that the mechanical method is not suitable to exfoliate the Mn+1Xn layers. However, the distinct chemical activities between M-A and M-X bonds make selective etching plausible. Thus the first reported carbide MXene, Ti3C2Tx, was obtained via aqueous hydrofluoric acid (HF) etching of Ti3AlC2 at room temperature (Fig 2b). 5 The majority of other MXene phases were synthesized with similar recipe, using either direct addition of HF or via in-situ formation of HF (e.g. by adding LiF to HCl), sometimes at elevated temperature and/or prolonged heating. 4,6 Most pristine (bare) MXenes show metallic characteristics. 48 The high conductivity can be linked to the redistribution of the outer M(1) atoms' 3d states from the broken M(1)-A bonds into delocalized M(1)-M(1) metallic-like bonding states placed in the window around the Fermi level (Fig 2c). 49 The M(1) atoms are distinct from the inner M(2) atoms that interact more closely with the X atoms.
Further delamination of the MXene layers can be achieved by intercalation strategies, using sonication in moieties such as dimethyl sulfoxide, tetrabutylammonium hydroxide or amines. 4 flowrate, which results in in-situ formation of graphene underlayer that improves the crystallinity of Mo2C (Fig 2d).
Precise characterization on the surface coverage and ratios of termination groups using Xray techniques remains challenging due to the intrinsic low-scattering strength of the light elements.
Solid state nuclear magnetic resonance (NMR) spectroscopy is one potential solution, as it has been successfully used to quantify the surface termination composition on Ti3C2Tx MXene. 60 More importantly, the study indicates that the ratio of Tx surface termination species can be tuned by adjusting synthesis method. DFT calculations also indicate that the tuneability of the Tx population is also related to the different metal species in MXenes, which governs the stability of the M-Tx interaction. 61,62 Being the outer-most layer, the composition of Tx significantly influences MXene's physical and electronic properties. For example, -O terminated MXenes have been calculated to have smaller lattice parameter and thinner interlayer thickness than -F and -OH terminated variants, leading to higher mechanical strength. 63 The presence of Tx termination also makes MXenes more semiconducting. The source of this electronic properties alteration is linked to the distinct electron orbital hybridization that yields bandgap variations. For example, the bandgap Sc2CTx MXene was found to be dependent on the Tx species in the order of Sc2CO2 > Sc2CF2 > Sc2C(OH)2. 63 Generally -F and -OH have similar effects on the electronic structures because both can only receive one electron, but -O differs significantly because it readily demands two electrons. 63 It is also shown that -O functionalisation are most favourable for high Li-ion capacity, superior to those with -F and -OH terminations, as Li prefers to adsorb on O. 64, 65 In some other applications areas such as heavy metal ion adsorption, -OH terminated MXenes are preferred. 66 The Tx composition was also found to affect surface adsorption and magnetisation behaviour. For example, -O terminated Ti3C2Tx exhibits a near zero ∆ of 0.003 eV while bare and -F terminated Ti3C2Tx give high values of -0.927 and 1.995 eV, respectively. 18 Moreover, -OH terminated Ti3C2Tx is calculated to show stronger CO2 adsorption than -F terminated ones. 19 In general, the experimentally hardly avoidable surface termination by Tx suppresses the magnetism and electronic conductivity in MXenes. For example, the semi-metallic ferromagnetic behaviour of bare Cr2C has been shown to become semiconducting antiferromagnetic upon -F, -OH and -H functionalization. 67 Similarly, other properties such as optical, thermal and charge transport are effectively modified by the presence of surface functionalizations. 68

Carbide vs. Nitride MXenes
The relatively more electronegative nitrogen compared to carbon at the MXene X sites gives stronger bonding to the electropositive M, thus variation in physical and electronic properties related to the shorter M-M and M-X interatomic distances are expected. For example, bare Tin+1NnTx has been calculated to have smaller lattice constant and monolayer thickness compared to Tin+1CnTx, accompanied by larger in-plane Young's modulus. 69 It is interesting that Tin+1NnTx MXenes, including -O terminated ones, have higher density of states (DOS) at the Fermi level, as opposed to carbide MXenes. 70 These findings lead to predictions of higher electronic conductivity and more catalytically active surface on nitride MXenes.
Despite their immense promise, nitride MXenes are generally more difficult to obtain due to less stable M-N layer in the etching solution and higher formation energy. Nevertheless, an etching strategy to produce Ti2NTx by exploiting the K + ions that can intercalate and weaken the M-X layers has recently been proposed. 7 An alternative strategy is to obtain nitride MXene via ammoniation of the carbide MXene, for example Mo2NTx synthesized from Mo2CTx, is reported to exhibit 3 orders of magnitude higher electrical conductivity than the corresponding carbide. 8 Although limited by difficulties in synthesis, we expect that nitride MXenes will continue to be the subject of intense experimentation to exploit their superior electronic and physical properties potential.

Effect of Metal Species
Many Ti3C2Tx, whereas -F and -OH are more common for V2CTx. 61,62 Consequently, alteration of M species composition will bring about significant change in the electronic properties of MXenes. 48 More subtle change in the electronic structure has been predicted on MXenes with varying M but same Tx preference, as seen in the increased bandgap in the order of Ti2CO2 < Zr2CO2 < Hf2CO2. 68 Addition of transition metal ad-atoms over the MXene layer has also been reported to yield interesting HER catalytic properties. 71 The discovery of double-transition metal MXenes, 72  to Mo3C2Tx, rather than the isostructural Ti3C2Tx, thus the former shows larger capacity when applied as Li-battery electrode. 74 The current investigation on double transition metal MXenes is very limited, but it is expected to expand rapidly. Khazaei et al. proposed that many compositions are topological insulators due to their large spin orbital coupling between metal elements, that can be widely applied in many energy applications such as thermoelectrics. 77

Effect of Layer Thickness
Due to the higher volume to surface ratio, the stability of MXenes generally increases with increasing layer thickness, such that M2XTx < M3X2Tx < M4X3Tx, in agreement with experimental observation on the better stability of Ti3C2Tx compared to Ti2CTx. 48 Effect of layer thickness on mechanical properties are investigated by first principle calculations, highest Young's moduli for the thinnest MXene are obtained from comparison of Ti2C, Ti3C2 and Ti4C3. 78 The equilibrium bond strength calculations indicate Ti2C has the shortest Ti-Ti distance but longest Ti-C bond among the three. 49 As surface properties mainly depends on the outer layer atoms, different layer thickness does not induce significant effect on their preferred surface functionalization, e.g.
Tin+1CnTx show similar Tx preference of -O > -F > -OH, regardless the number of n. 78 However, distinct metal d band degeneracy was observed on MXenes with the same atomic constituent but different n, thus displaying significantly different electronic structures. 79 DFT calculations found that the DOS near the Fermi level generally reduces with increasing n for pristine Tin+1Xn. 49 Interestingly, bulk Ti3C2Tx (n = 2) displays much smaller electrical resistivity than Ti2CTx (n = 1) possessing similar surface functionalization, probably due to the increasing difficulty for electron hopping in between the layers when n is smaller. 59 Due to the varying accessibility and tuneability of the active sites, different numbers of n have also been shown to cause variations in the HER activity. 11,80 With proper delamination, MXenes can be obtained in monolayer or multilayers; the interlayer spacing can be tuned by intercalation using multivalent cations, water or organic molecules. While it has been established that most MXenes show metallic-like behaviours, the actual bulk conductivity depends greatly on preparation method. Milder etching and delamination conditions generally result in higher bulk conductivity, 81 so is more thorough removal of intercalated species. 82 In summary, the flexibility and tuneability of atomic composition and surface functionalization are key to the excellent electronic, physical and surface properties of MXenes, making them unique among the 2D materials reported so far.
In the next few sections, we will discuss how the excellent properties of MXenes can be exploited to enhance catalytic conversion to various value-added chemicals and fuels.

Electrocatalytic Hydrogen Evolution Reaction (HER)
As one of the two half-reactions of electrochemical water splitting, HER provides a cornerstone in the exploration of other, more complex electrocatalytic processes that involve multiple electron transfer. In brief, the mechanism for HER can be separated into three distinct elementary reactions in acidic condition: Volmer (adsorption/discharge reaction) Tafel (combination reaction) Heyrovský (ion + adsorbed *H reaction) In alkaline condition, a different Volmer and Heyrovský mechanism is proposed due to low H + ion concentration in the electrolyte: Heyrovský (alkaline)  (2), is expected to favourably interact with the hydrogen 1s orbital. This is advantageous compared to other, rather inert, 2D materials such as MoS2, 86 as the entire basal plane of MXene's highly anisotropic monolayered structure has the potential to act as active sites for hydrogen adsorption. 10, 87 DFT is an ideal tool to assess MXene's potential for HER electrocatalysis, as it provides a framework for rapid, precise, systematic and yet qualitatively correct predictions of the catalytic performance. The primary metric is ∆ , and volcano plots based on calculated ∆ values are then constructed to aid in material screening (Fig 3a). 80,87 In the case of MXenes, the composition and role of Tx surface functional groups for *H adsorption site need to be clarified and taken into account for a more realistic approximation. In M2XTx type MXenes, -O or -OH was found to be the preferred Tx species, with five MXenes with theoretical HER overpotential under 100 mV identified: Sc2C(OH)2, Mo2CO1.75(OH)0.25, Hf2NO2, Nb2NO2 and V2NO2 (Fig 3b). 87  can achieve HER current density of 10 mA cm -2 at 190 mV overpotential (Fig 3c). It was found that the HER overpotential and activity of MXenes vary significantly with changing Tx species composition. 10 In particular, -F species was found to be particularly detrimental to MXene's HER Mo2Ti2C3Tx is more closely correlated to the extent of -F coverage on their surfaces than the outer metallic species (Fig 3c), as Mo is the outer M' species in all three cases above. 10 Several other promising theoretical and experimental approaches have been explored to enhance the HER activity of MXenes, such as: (1) Controlling the surface functionalization. In particular, replacing -F termination with other moieties such as -O or -OH with suitable *H adsorption strength as described in the case of Ti3C2Tx. 10 (2) Interlayer delamination. Delamination of MXene interlayers increases the basal plane accessible for HER. This effect has been demonstrated on Mo2CTx, where the overpotential to achieve 10 mA cm -2 current density can be reduced by 30 mV after delamination. 10 The improved overpotential indicates that the basal planes of Mo2CTx are catalytically active towards HER, unlike in the case of 2H-phase MoS2, where the edge sites have been shown to be more active. 86 Alternatively, the interlayer delamination may have also improved the electron hopping to the basal plane. 89 Similar improvement on HER overpotential has also been demonstrated on Ti2CTx after delamination treatment in N-methyl-2-pyrrolidone. 87 (Fig 3d).
Other approaches such as introducing structural strain on monolayer MXenes can also be promising. Strained surface has been shown to increase metal d-states energy and increase surface reactivity. 97 Strasser et al. has demonstrated that the catalytic activity of Pt-Cu alloy can be tuned by introducing surface strain through de-alloying. 98 Similar concepts may be applicable to MXenes, as strain has also been predicted to change the covalent character of the M-X bond and release delectrons. 99 These newly formed states may also show favourable interaction with adsorbed *H and increase HER activity.

MXenes in Photocatalytic HER
Although most bare and terminated MXenes have no or very small bandgaps, they have found their way as co-catalysts for photocatalytic HER due to their electrocatalytic HER performance.
Additionally, MXenes are relatively water stable with hydrophilic surface and metallic conductivity, making them a good substitute to noble metal co-catalysts. For example, a composite of 5 wt% MXene and a classic photocatalyst, TiO2 (Fig 4a), was reported to exhibit up to 940% higher H2 production rate under visible light compared to bare TiO2. 17 Similarly, 2.5 wt% loading of -O/-OH terminated Ti3C2Tx on CdS nanoparticles yields a 136.6 times higher H2 production rate, superior to Pt/CdS at the same condition (Fig 4b), presumably due to the better interfacial contact between MXenes and CdS. 18 Understandably, higher MXene loading does not necessarily result in better HER, as too much MXenes will block the light from the CdS semiconductor (Fig 4c). In another example, Ti3C2Tx and Ti2CTx were also shown to exhibit activity enhancement on g-C3N4. 100,101 As metallic-like MXene could not have generated photo-carriers, the enhanced photocatalytic HER was attributed to the inherent catalytic activity of MXene in facilitating HER.
Additionally, the formation of Schottky barrier between MXene and semiconductor encourages electron pooling and ameliorates charge recombination within the semiconductor band structure (Fig 4d). The height of Schottky barrier is proportional to the work function of the outer M' transition metals, therefore higher photocatalytic activity can be further optimised by selecting MXenes-semiconductor composite pair with suitable Fermi level.
We note that some MXenes, especially Sc and Y based ones like Sc2C(OH)2, are predicted to be good HER photocatalyst, as it has indirect bandgap of around 2 eV and conduction band minimum above HER theoretical potential, thus opening up further opportunities. 102, 103

Oxygen evolution reaction (OER)
OER is the anodic half-reaction for overall water splitting. However, the substantial overpotentials required to overcome the kinetic barrier, even with the most efficient precious metal catalyst, make this process particularly challenging. 3 Additionally, OER (along with ORR) is also the major bottleneck step for rechargeable metal-air batteries. 104 Electrochemical OER is proposed to be a four-electron transfer process with the overall reaction as follows: In acidic electrolyte: In alkaline electrolyte: Due to its relatively positive potential (with respect to most metal oxidation potential), it is proposed that oxide layer always forms at the catalyst surface during OER, 105 sometimes even involving an increase in the formal oxidation state of catalyst. [106][107][108] One of the more likely mechanism is therefore the electrochemical oxide path (assuming acidic electrolyte): 109 As can be seen, the mechanism for OER is more complicated than HER as it involves multiple steps and intermediates. Furthermore, the formation of oxide on non-oxide catalyst surface needs to be taken into account as it affects the intermediate stability and OER reaction mechanism significantly. Various other reaction mechanisms based on catalyst behaviour in electrolyte have thus been proposed. 105,110 For example, an 'associative' (or metal peroxide route 109 Regardless the route taken, a common feature is that the initial catalytic step involves the discharge of a hydroxide ion at a catalytically active surface site (*) to form a surface adsorbed hydroxide (*OH). 111 Subsequent steps in the reaction reflect the formation of surface intermediates that depends on reaction conditions and catalyst surface. This is further complicated by the possibility of involvement from lattice oxygen atoms, as demonstrated by 18 O labelling experiment on "nickelate" (MNiO3) and cobalt phosphate (CoPi) catalysts. 112,113 Admittedly, OER is a more complex process than HER, and a universal "activity descriptor" for its activity is elusive, as different rate determining steps exist for different catalysts and reaction conditions. Seminal works by Trasatti suggested that the OER reactivity of simple oxides can be correlated to the standard enthalpy of transition from lower to higher oxide, 114 (Fig 5a). 31,116 By itself MXenes do not show efficient OER electrocatalytic activity to date. However, when coupled with other catalysts, a synergistic effect has been shown to significantly accelerate oxygen evolution process. 117 FeNi-LDH and Ni(OH)2 respectively, and also to the benchmark RuO2 (397 mV). Similarly, Ni-Co mixed metal sulphide (NiCoS) incorporated with Ti3C2Tx also yields a more competitive OER rate than RuO2 in terms of smaller overpotential and Tafel slope, and larger TOF (Fig 5b-d). XPS measurement suggested that the NiCoOOH originated from NiCoS during OER acted as the active sites. 117

Oxygen Reduction Reaction (ORR)
ORR is a fundamental reaction for many processes in biology, energy conversion or materials dissolution, including fuel cells and metal-air batteries. 104 In acidic condition, ORR can be described as simple dissociation of adsorbed oxygen on catalyst surface, followed by formation of hydroxyl species via electron transfer and hydrogenation: 29 Alternatively, it is also possible for the protonation for the oxygen to occur first before the dioxygen is being dissociated. This mechanism is termed the "associative" mechanism: 2 + * → * 2 (21) * 2 + + + − → * (22) * + + + − → * + 2 (23) * + + + − → * (24) * + + + − → 2 + * Nørskov et al. proposed that the ORR activity can be closely correlated with the binding strength of the two reaction intermediates, i.e. *O and *OH (Fig 5e), 29,121 which remains valid even when the associative mechanism involving *OOH is included. This observation is rationalised later on by the discovery of the scaling relations of adsorption energies of similarly adsorbed intermediates. 122 In alkaline electrolyte, another indirect mechanism involves electron transfer (tunnelling) to solvated molecular oxygen cluster in the outer Helmholtz plane. 123 This mechanism explains the high ORR activity of non-noble metal catalysts in alkaline electrolyte, typically yielding peroxide ( 2 − ) that is detectable in the rotating ring disc electrode experiment. * + [ 2 • ( 2 ) ] + − → * + ( 2 • ) + − + ( 2 ) −1 (26) While having similar reaction intermediates and electron transfer requirement with OER, ORR (especially in alkaline solution) is more complex with multiple, direct and indirect mechanisms running in parallel. Thus, modelling ORR requires the incorporation of a realistic environment model that includes (explicit) surface water to estimate the charge capacitance of the system. 124 Combination of DFT and kinetic Monte Carlo approaches has uncovered that the ORR activity is significantly affected by the stabilisation of intermediates provided by hydrogen bonds with water molecules from the environment. 125 Some research has been dedicated to search for a "bifunctional" oxygen catalyst (i.e. both OER and ORR active), as it is of great interest to metal-air batteries. However, finding such a material turns out to be challenging, simply because the requirement of variable oxidation states and stable structures at very different working potentials is difficult to fulfill. 104 Perovskites, 126  can result in excellent ORR activity surpassing Pt/C in alkaline electrolyte (Fig 5f). 120 While FePc is known to be ORR active, 129 combining it with Ti3C2Tx doubles its activity. The source of activity enhancement is proposed to be Fe 3d electron delocalisation and spin-state transition of Fe(II) ions upon coupling with Ti3C2Tx.
In another example, Xue et al. prepared Mn3O4/Ti3C2Tx nanocomposites using hydrothermal method, and found a dominant four electron transfer ORR mechanism and an onset overpotential as low as Pt/C. 130 The enhancement mechanism is fundamentally the same as that for HER and OER as discussed earlier. The Mn3O4 nanoparticles were detected to be well dispersed on MXene nanosheets, and increased the surface area by inhibiting aggregation. 130 Interestingly, the Ti in Ti3C2Tx MXene can form strong hybridization with Ag when the alkalinetreated MXene was stirred with AgNO3 solution in the presence of poly(vinylpyrrolidone) and deionized water for 60 mins. 131 The existence of reductive low valence Ti 2+ and Ti 3+ , confirmed by XPS, results in the formation of metallic Ag, therefore a hybrid MXene with bimetallic Ag/Ti was obtained. The MXene/Ag0.9Ti0.1 nanowire exhibited an onset potential and a half-wave potential of 0.921 and 0.782 V vs. RHE at 1600 rpm, respectively, which were significantly more positive than the 20 wt% Ag/C catalyst or even pure Ag nanowire. 131 The MXene not only served as structural support, but also an efficient charge carrier conductor. The Ti doping induced vacancies or defects also provided more active sites for O2 adsorption. 131 While by itself MXene has not been considered as active ORR catalyst, the synergistic coupling between MXenes and other catalytic materials has opened up a new approach for the development of efficient electrocatalysts. Of particular interest is the fact that MXene coupling can induce significant change and delocalisation in the catalyst's d-band electronic structure. We note that there is still much to be explored in this area as only one MXene, Ti3C2Tx, has been investigated. We speculate that, with realisation of new types of MXenes, many interesting properties can be discovered by coupling existing catalyst materials with MXenes.

CO2 Capture and Electrocatalytic CO2 Reduction Reaction (CO2RR)
Electrocatalytic reduction of CO2 to value-added products has attracted a lot of attention due to its value proposition of creating a closed-loop anthropogenic carbon cycle. 132 Depending on the nature of the catalyst surface, CO2RR has been proposed to begin with the formation of either *COOH, *CO2radical or *OCHO. [133][134][135] From there, a maze of multi-stepped process continues towards one or more of the 16 distinct molecular products that have been reported on Cu surface. 136 With so many viable reaction pathway permutations, finding an excellent catalyst for CO2RR goes beyond lowering the required overpotential, but also tailoring the catalyst to enhance reaction selectivity towards a certain product. The complexity of CO2RR has called for advanced DFT approaches beyond computational hydrogen electrode (CHE) models that incorporate explicit solvent layers 44,137 and surface charge. 35 The "universal" linear dependence of Ea and ΔE on the surface structure of the reaction site proposed by Nørskov and co-workers 83,84 has an overreaching consequence for CO2RR, especially upon the revelation that the adsorption energies of similarly bound intermediates (e.g. *C and *CH; *O and *OH; etc.) are linearly correlated. 122,138 These unfavourable scaling relations suggest that the CO2RR limiting potential is not expected to surpass that of Cu (211), 139 Fig 6a- 6c) and best selectivity against competing HER (Fig 6d). This limiting potential corresponds to theoretical CO2RR to CH4 overpotential of 0.52 and 0.69 V, significantly lower than that of Cu (0.91 to 1.10 V). 145

Photocatalytic CO2RR
Research into photocatalytic CO2RR is largely inspired by the natural photosynthesis process, where complex carbohydrate molecules and oxygen can be obtained from CO2, water and sunlight. 22 Arguably, photocatalytic CO2RR is even more challenging than electrocatalytic CO2RR due to the additional processes of photocarrier generation and charge transfer. 147 These processes are generally inefficient and the quantum yield of photocatalytic CO2RR is typically less than 1%. 148 As elaborated previously, most MXenes are not able to generate photocarriers due to its metallic characteristics. Only some -O and -OH terminated MXenes like Ti2CO2 and Sc2C(OH)2 with narrow bandgap (around 0.9 to 2.0 eV) may be photocatalytically active. DFT calculations suggest that Ti2CO2 could be active for photocatalytic CO2RR to HCOOH. 149 The oxygen vacancy (Ov) on -O terminated Ti2CO2 was found to play an essential role to enhance CO2 adsorption since the adsorption on a perfect -O terminated surface was too weak. It is interesting that the energy barrier for the rate limiting step was much lower than that of anatase TiO2. 150 Instead of being the main photocatalyst, MXenes can also take the role of co-catalyst or conductive support when combined with an efficient semiconductor photoabsorber during CO2RR.
One of the sources of inefficiency is the relatively poor CO2 adsorption and activation. Ye et al.
proposed that the -OH functional groups on the MXene surface may serve as basic sites that facilitate the adsorption of relatively acidic CO2. 19 The group employed KOH treatment to replace majority -F termination with -OH on Ti3C2Tx and combined it as co-catalyst with the commercial P25 TiO2 (Fig 7a). The resulting photocatalytic CO2RR to CO rate was increased by 3 times and CH4 production rate was boosted by 277 times. 19 In another example, a hybrid Ti3C2/Bi2WO6 composite shows a higher yield of CH4 and CH3OH by 4.6 times compared to bare Bi2WO6. 151 Recently Persson et al. devised a sequential heating treatment in H2 environment during in-situ TEM experiment to realise O-deficient Ti3C2 MXene. 152 In-situ CO2 exposure to this fresh Ti3C2 results in immediate CO2 saturation on the basal plane while maintaining the 2D structure (Fig 7b-c).

N2 Reduction Reaction (N2RR)
Ammonia is an important molecule, widely used for fertilizers, and is an important precursor to most nitrogen containing products, including amines for polymers and basic chemical compounds like nitric acid. 154 Currently, 3-5% of natural gas worldwide is used to make ammonia through successive processes of reverse gas shift, Haber-Bosch and purification, releasing large amounts of CO2 to the atmosphere. 155 Although the Haber-Bosch process (N2 + 3H2 → 2NH3) is exothermic, the kinetics are very slow under ambient conditions. These kinetic limitations are overcome by the combination of an iron-based catalyst with high pressure, high temperature conditions. 153 Electrochemical N2 reduction (N2 + 6H + + 6e -→ 2NH3) aims at forming ammonia under ambient conditions, replacing H2 as hydrogen source by proton/electron pairs, therefore reducing the number of process steps, costs and the CO2 emission related to natural gas reforming. 156 Pickett and Talarmin demonstrated the possibility of electrochemically synthesising ammonia from N2 via controlled hydrogenation of dinitrogen complex (trans-[W(N2)2(Ph2PCH2CH2PPh2)2]) over Hgpool electrode. 157 While promising, the reaction rate is uncompetitive compared to the high temperature process using H2. Two approaches are common in electrochemical N2 reduction: moderate temperature (400-600 °C) using solid oxide proton conductor electrolyte 158 and aqueous alkaline electrolyte using Nafion membrane. 159 Generally the aqueous electrolyte approach is attracting more interest due to higher ammonia TOF, but the overall FE is generally below 1% due to competition from HER. Biological N2RR using extracted or synthesised Fe-nitrogenases has also been proposed, 160 Fig 7d). 153 The cell voltage required to drive 10 mA cm -2 current density was 1.2 V. However, the proposed contraption was not stable as electrostatic would eventually coagulate the nano-Fe2O3 catalyst and halt the electrochemical process.
Herein, MXene has the potential to step into the foray as the state-of-the-art N2RR catalyst is still based on the same Fe2O3 that is used in 1910. Current investigation for MXenes to be applied as N2RR electrocatalyst is led by theoretical DFT calculations. 161,162 Although N2 is relatively inert, chemisorption of N2 on bare M3C2 MXenes was found to be spontaneous. 162 Importantly, the N≡N bond appears to be stretched >20% upon adsorption compared to the gas phase, indicating weakening/activation. The first hydrogenation step *N2 + *H → N-NH • on V3C2 was found to proceed with activation barriers of 0.64 eV, 162 which compares favourably with the 1.98 eV estimated for isolated N2 to N2 • . 163 Therefore, MXenes are predicted theoretically to be promising N2 capture and reduction electrocatalysts. In order to further reduce the energy input, the effect of metal doping in MXenes has been explored, and a low overpotential of 0.47 V was obtained for Fe-doped Mo2N where Fedopants were found to weaken the metal-nitrogen interaction and promote ammonia formation. 161

Computations
To date, computational studies of MXenes are still largely dominated by (semi-local) DFT. Since these functionals are notoriously ill adapted for band-gap predictions, 164,165 applications that aim at exploiting the (small) band-gap of Tx terminated MXenes can be especially problematic. In analogy to other 2D materials such as MoS2 166 or graphitic carbon-nitrides, 167 we encourage the community to rely on the more robust many-body perturbation treatments, also known as the GW approximation. We note that first steps in this direction have already been taken. 168 Furthermore, systematic studies of MXenes' topological electronic structure 169  To obtain an atomistic understanding of the working catalyst and promising performance of MXenes in electrocatalysis, efforts towards the inclusion of the electrochemical potential and the influence of the solvent need to be made. 36,37,46,173,174 These efforts are not specific to MXenes, but generally necessary for metallic catalysts. Both aspects are highly challenging and not always fully separable; to account for the electrochemical potential, the electrolyte also needs to be modelled. On the other hand, in the context of bio-mass valorization, 175 aqueous solution-based catalytic processes over metal catalysts is on the rise. 176 In these processes, the chemisorption of water molecules at the solid/liquid interface has been shown to significantly modify the reactivity of the catalyst. 177 Although highly challenging to model even for mildly reactive surfaces such as Pt (111), 178 the structuration of the metal/liquid interface is expected to depend on the nature of the MXenes and their surface termination, which may open the door to more active and selective catalysts compared to known transition metal based nanostructures.

Applications
As Overall, MXenes as catalysts or co-catalysts are still a relatively under-explored field, and theory-guided materials design is expected to bring about significant breakthroughs in clean energy research in the near future.

Conflicts of Interest
There are no conflicts of interest to declare.

TOC Graphic
A critical review on how theory-guided materials design can enhance MXene based catalyst discovery and development process

Author Biographies
Albertus Denny Handoko is a Scientist at Institute of Materials Research and Engineering, A*STAR. He attained his core competency in solvothermal/hydrothermal synthesis method and X-ray characterisation during his PhD studies at Nanyang Technological University (Singapore). He went for postdoctoral research fellowships at University College London (UK) and National University of Singapore where he gained research experience and developed keen interest in photo-and electrochemistry. His current research effort is dedicated to understanding advanced electrocatalytic system for renewable energy conversion reactions such as CO2 reduction, water splitting and purification.