Regulating local coordination environment of Mg-Co single atom catalyst for improved direct methanol fuel cell cathode

Kiwen Wng ,Hnjun Zou ,Jizhi Meng ,Chogng Bn ,Xue Liu ,Jingping M ,Cong Wng,Liyong Gn,c,∗,Xiodong Hn,∗,Xioyun Zhou,c,d,∗

a College of Physics and Center of Quantum Materials and Devices,Chongqing University,Chongqing 401331,China

b Beijing Key Laboratory of Microstructure and Property of Advanced Materials,Beijing University of Technology,Beijing 100024,China

c Institute of Emerging Energy Storage Materials and Equipment,Chongqing 401135,China

d Analytical and Testing Center,Chongqing University,Chongqing 401331,China

Abstract Fuel cells operated with a reformate fuel such as methanol are promising power systems for portable electronic devices due to their high safety,high energy density and low pollutant emissions.However,several critical issues including methanol crossover effect,CO-tolerance electrode and efficient oxygen reduction electrocatalyst with low or non-platinum usage have to be addressed before the direct methanol fuel cells (DMFCs) become commercially available for industrial application.Here,we report a highly active and selective Mg-Co dualsite oxygen reduction reaction (ORR) single atom catalyst (SAC) with porous N-doped carbon as the substrate.The catalyst exhibits a commercial Pt/C-comparable half-wave potential of 0.806 V (versus the reversible hydrogen electrode) in acid media with good stability.Furthermore,practical DMFCs test achieves a peak power density of over 200mWcm-2 that far exceeds that of commercial Pt/C counterpart(82mWcm-2).Particularly,the Mg-Co DMFC system runs over 10 h with negligible current loss under 10 M concentration methanol work condition.Experimental results and theoretical calculations reveal that the N atom coordinated by Mg and Co atom exhibits an unconventional d-band-ditto localized p-band and can promote the dissociation of the key intermediate ∗OOH into ∗O and ∗OH,which accounts for the near unity selective 4e- ORR reaction pathway and enhanced ORR activity.In contrast,the N atom in SAC–Co remains inert in the absorption and desorption of ∗OOH and ∗OH.This local coordination environment regulation strategy around active sites may promote rational design of high-performance and durable fuel cell cathode electrocatalysts.

Keywords: Single atom catalyst;Mg-N-Co;DMFC;Cathode.

1.Introduction

Global concerns over energy crisis and environment pollution motivate clean and sustainable energy conversion technologies.Proton exchange membrane fuel cells (PEMFCs) that directly convert fuels into electricity can supply energy with efficiencies exceeding 80% [1,2].Among them,direct methanol fuel cell (DMFC) technology with the key advantages of high energy density and quick refueling makes it become a potentially dominant player in the rapidly expanding market for portable electronic devices,e.g.,laptops,smartphones,and others [3–5].The commercialization of DMFCs is currently hindered by two main challenges: the high cost of electrode catalysts and the methanol crossover issue [6–10].Specifically,Pt/Pd-based nanomaterials are well-accepted leading catalysts for cathodic oxygen reduction reaction (ORR) [11–16].Nevertheless,simultaneous achievements of desirable methanol tolerance and high ORR activity within them remain a grand challenge.Besides,the permeation of methanol from the anode to cathode results in oxidation and mixed potential,causing severe CO (methanol oxidation byproduct) poisoning of noble metals and reduced cathodic ORR performance and fuel efficiency.In search of high-performance and anti-poisoning ORR catalysts,several strategies such as Pt-M (M=Co,Ni,Sn,Ru,Mo and etc.) alloying,Pt facet engineering,size control,and strain engineering have been reported [17–23].Nevertheless,the pursuit of cost-effective catalysts remains a long-term topic.Thus,developing a platinum-group-metals-free cathode catalyst with both high ORR activity and excellent methanol/CO tolerance is a challenging yet critical goal [24–27].

Single atom catalysts (SACs) are an emerging type of catalysts,in which metal sites are atomically dispersed on welldefined supports [28–34].These isolated metal atoms have unique properties such as unconventional electronic structures,100% atom utilization efficiency and unsaturated coordination environments,contributing to boosted performances in various catalysis reactions.Among SACs,N-doped conductive carbon embedded with metal atom catalysts(M-N-C)have been reported to play crucial roles in accelerating the sluggish kinetics of ORR[35–37].Especially,Fe–N–C and Co–N–C are the most investigated ORR catalysts hitherto due to their decent activity [36,38-40].However,electrodes using such catalysts in fuel cells have shown a rapid decline in performance over time possibly due to carbon corrosion from radicals and the loss of metal sites through demetallation [41–43].Additionally,Fe–N–C is notorious as it favors the Fenton reaction,further exacerbating the degradation of the catalyst and the membrane electrode assembly (MEA).The demetallized Fe ions promote the formation of even more radicals from 2e-ORR reaction product H2O2.Despite inert in Fenton reaction,weaker ORR activity and even higher yields of H2O2are found in the Co–N–C case [44–46].Therefore,it is crucial in the development of ORR catalysts to suppress the 2e-ORR reaction pathway.

The conventional single-site mechanism of proton-coupled electron transfer in the ORR process involves multiple reaction intermediates,including∗OOH,∗O and∗OH [47].It is recognized that the improper binding strength of∗OOH results in sluggish kinetics and increases 2e-pathway possibility [48].To resolve this predicament,one potential solution is the implementation of the dual-site mechanism (M–O–O–M),which avoids the production of oxygen intermediates (∗OOH)and directly promotes the breaking of O–O radicals [49].For the dual-site mechanism to be effective in the oxygen reduction reaction(ORR),it is crucial to ensure that the interatomic spacing between the catalytic sites is optimized to promote the dissociation of molecular oxygen (∗O2) and facilitate the cleavage of the O–O bond.This is necessary to prevent the production of undesirable M–OOH species,which can hinder the ORR process and reduce its efficiency.By promoting O–O bond breakage,the dual-site mechanism can limit the two-electron reaction path selectivity,which is a key step in achieving high-performance ORR catalysts.Therefore,careful consideration of the interatomic spacing between the dual catalytic sites is necessary for successful implementation of this mechanism.However,the successful implementation of the dual-site mechanism requires strict adherence to the required geometric and electronic configurations of the active sites,i.e.,adequate interaction strength with oxygen atom and proper inter-site distance,which is currently a great challenge.

Here,we report a dual-site Mg–Co SAC synthesized by a pyrolysis-impregnation-pyrolysis procedure,which possesses densely accessible active sites to enhance the acidic ORR activity and DMFC performance.The plenty amount of Co salt addition in the first pyrolysis step ensures dense Co sites,while the subsequent impregnation ensures abundant Mg sites.Meanwhile,the SiO2hard template usage leads to a porous conductive carbon skeleton,which can promote mass transfer and induce much more accessible active sites.The diverse alloy clusters and single atom sites convert into adjacent single atoms under acidic work condition due to easy corrosion property of metallic alloys,which provide possible single atom supply reservoir source and maintain relatively high-density Mg-N-Co sites.The catalyst exhibits a high ORR activity and DMFC performance,with a half-wave potential (E1/2) of 0.806 V versus the reversible hydrogen electrode (RHE) at a catalyst loading of 0.6 mg cm-2and a peak power density of 0.204 W cm-2in DMFC (Mg-Co as cathode) tests.The ORR activity of Mg–Co is greatly enhanced compared with single atom Co catalyst (SAC-Co) and approaches that of commercial Pt/C,and the peak power density surpasses that of commercial Pt/C.Further,the Mg–Co catalyst was found to be much more durable than the commercial Pt/C catalyst due to the anti-methanol/CO-poison property.Theoretical calculations reveal that the Mg coordinated N atom in Mg-N-Co has ad-block-metal-similarpband,contributing to a strong interaction with oxygen species and the breakage of the O-O bond in∗OOH.This proposed strategy of modulating the local coordination environment viap-block coordination in conventionald-block M-N-C SACs should help pave the way for highly active and selective ORR catalyst design.

2.Experimental procedure

All the electrochemical measurements and calculations are provided in the Section–1 of the Supporting Information (SI).

2.1. Materials

Cobalt(ii) chloride hexahydrate (CoCl2·6H2O,98%),magnesium acetate tetrahydrate ((CH3COO)2Mg·4H2O,99%),sodium hydroxide (NaOH,98%),melamine (C3H6N6,99%)and urea (CH4N2O,99%) were purchased from Sigma Aldrich.Methanol (CH3OH,99.5%) and ethanol (C2H5OH,99.5%) sulfuric acid (H2SO4,95–98%) were purchased from J&K scientific.10 nm SiO2nanosphere was purchased from Shanghai Yuanjiang Chemical Co.,Ltd.All reagents were used without further purification.

2.2. Synthesis of SAC–Co catalyst and SAC-Mg catalyst

First,0.121 g CoCl2·6H2O was dissolved in 50 mL ethanol aqueous solution (50 wt%) to form a red solution.After sonicate for 5 min,1.0 g SiO2nanospheres,1.0 g melamine and 0.5 g urea were subsequently added into the above solution under constant stirring at 60 °C until the mixture was dry.Then,the as-obtained powder was placed in a tube furnace and heated to 500°C for 1 h in an Ar atmosphere(100 sccm).After cooling the sample down naturally,the mixture was subjected to a pyrolysis process in tube furnace under flowing Ar (100 sccm) at 900 °C for 2 h.20 sccm NH3was flowed continuously in the last 15 min during the pyrolysis process.Afterwards,the SiO2template was removed by immersing the powder in 2 M NaOH solution at 60 °C for 1 h and the mixture was centrifuged and washed with deionized water to remove residual NaOH until pH reached 7.SAC–Co was finally obtained after drying the mixture overnight at 60 °C in vacuum oven.SAC-Mg was synthesized by the same method as SAC–Co except that 0.136 g magnesium acetate tetrahydrate was adopted as metal source instead of CoCl2·6H2O.

2.3. Synthesis of Mg-Co catalyst

The synthesis method of various Mg-Co is the same as that of SAC–Co except that after the 500 °C annealing,the obtained powder was dispersed into 20 mL ethanol aqueous solution(50 wt%)and mixed with 0.136 g magnesium acetate tetrahydrate.The mixture was then dried under constant stirring at 60 °C and followed by the second pyrolysis process.

2.4. Characterization

Transmission electron microscopy (TEM) and high-angle annular dark field scanning transmission electron microscopy(HAADF-STEM) images were obtained using an FEI Titan Themis G2 and operated at an accelerating voltage of 300 kV.Energy dispersive spectroscopy (EDS) was performed using FEI Talos 200.The surface chemistry was investigated using X-ray photoelectron spectroscopy (XPS) with an Al Kαradiation source on a Thermo Scientific ESCALA 210 XPS spectrometer instrument.The binding energies were calibrated by the C 1s peak at 284.8 eV from adventitious carbon.The mass fraction of Cu was determined using a PerkinElmer Optima 5300DV inductively coupled plasma-atomic emission spectroscopy (ICP-AES) system.The crystal structure was analyzed by X-ray diffraction (XRD) on a Bruker,D8 ADVANCE with Cu Kαradiation (λ=0.15418 nm).Finally,Co K-edge XAFS was collected at the 1W1B beamline of the Beijing Synchrotron Radiation Facility using a piece of carbon cloth coated with the catalyst.The N2adsorptiondesorption isotherms were measured using a Quadrasorb 2MP.The instrument was equipped with automated surface area measurement capabilities and the measurements were performed at 77 K using Brunauer-Emmett-Teller calculations.The mesopore and micropore size distributions were determined based on the isotherm,using the Barrett-Joyner-Halenda model and non-local density functional theory model,respectively.The Raman scattering spectra were acquired using a high-resolution confocal Raman spectrometer (RENISHAW inVia) with an excitation laser of 532 nm.The inductively coupled plasma-atomic emission spectrometry data were obtained using an iCAP 6300 Duo instrument.

2.5. DMFC tests

The MEA in the assembled DMFC was made up of an anode layer containing 60 wt% commercial PtRu/C and a cathode layer containing 20 wt% commercial Pt/C catalysts or Mg-Co catalyst and a Nafion 117 membrane.The MEA,with a 9 cm2active area and a catalyst loading of 2 mg metal/cm-2,was positioned between two carbon plates that had cross-sectional channels for the flow of methanol or air.The electrical current was collected using a copper plate.The plates were equipped with electrical heaters and thermocouples for temperature control.A pump was used to supply the aqueous methanol solution from a reservoir,without creating back pressure.Air pressure was regulated by a pressure regulator.The performance of the DMFC was evaluated at 80 °C.Before collecting the data,the MEA was activated for 8 h by pumping 1 to 15 M methanol solution through the DMFC anode flow field at a rate of 1 mL/min.The cathode was supplied with air at a pressure of 0.1 MPa and a flow rate of 300 mL/min.The current-voltage and power density curves were obtained stepwise using a Bio-Logic VMP3 potentiostat.

2.6. HAADF simulation

Simulation of the HAADF-STEM images was conducted using the QSTEM code developed by Koch [50],and the parameters used for the simulation are Scherzer focus df=-13.7 nm,spherical aberration Cs=0.05 mm,chromatic aberration coefficient Cc=1.4 mm,and thermal diffuse scattering (TDS) runs=10.The values of the convergence semi-angle of the electron probe,the HAADF collector angles and the probe size were the same as in the experiment.

2.7. DFT calculations

The electronic and geometric properties of SAC–Co,Mg-Co and Pt systems were studied using density functional theory computations.The simulations were performed using the Vienna Ab Initio Simulation Package and the ion-electron interactions were described by the projector-augmented planewave method [50].The exchange-correlation potentials were represented by the Perdew–Burke–Ernzerhof functional with the generalized gradient approximation [51,52].The Co and Mg-Co SACs were constructed on a 6 × 6 graphene supercell and the Pt (111) surface was modeled using a 3 × 3 supercell containing five atomic layers.The energy cutoff for the geometry optimization was set to 500 eV and the Brillouin zone was sampled using a 5 × 5 × 1.To avoid interlayer interaction,a 20 ˚A vacuum layer was added in the perpendicular

direction of the slabs.The systems were relaxed until the energy and forces reached a convergence threshold of 10-5eV and 0.02 eV ˚A-1,respectively.

The adsorption energy is calculated as [53]:

Eads=Etot-(Eslab+Emol)

whereEtotandEslabare the total energy of the system with and without a molecule and adsorbed,respectively,andEmolis the energy of an isolated molecule:

The changes in Gibbs free energy (ΔG) for all intermediates were calculated as:

ΔG=ΔE+ΔZPE-TΔS

whereΔE,ΔZPEandΔSare total energy,zero-point energy and entropy change relative to the initial state.T is the temperature and set to be 27 °C.ZPEwas calculated from the vibrational frequencies [54].Sis obtained directly from Atkins’ Physical Chemistry [55].

3.Results and discussion

Fig.1 shows the preparation process of the Mg-Co catalyst,which involves a pre-polymerization and two-step annealing.Firstly,a mixture of Co salt,urea,melamine and SiO2templates were dispersed in 50 mL ethanol aqueous solution and dried under 60 °C.Later,the above powder went through the first annealing process and then was immersed into Mg salt aqueous solution to adsorb plenty of Mg ions.Subsequently,the second annealing process was carried out in a tubular furnace under Ar and NH3atmosphere,with the purpose of anchoring the single atoms and preventing their aggregation as much as possible due to the strong interaction between single metal atom and nitrogen atom.The Mg–Co catalyst with dual metal sites was finally obtained after removing the templates.As illustrated in the schematic image,the metal species in Mg–Co SAC can be categorized into metal alloys and single atoms.To verify the superior catalytic property of Mg-Co catalyst,SAC–Co and bare porous N-doped carbon (NC) are prepared as reference samples.The XRD patterns (Fig.S1) of these samples reveal that there is clear existence of Co metal phase in Mg–Co catalyst while no Co metal phase in SAC–Co,indicating a larger size of Co species in the Mg–Co catalyst [56].The main peaks of Co metal phase are located at 44.257° and 51.538°,which are slightly larger than the standard references,demonstrating lattice contraction of Co phase in Mg-Co.This should result from the formation of MgCo alloy in Mg-Co catalyst as hexagonal Mg has a smaller lattice parameter (0.321 nm)than that of cubic Co (0.354 nm).Besides,the XRD patterns also show the amorphous property of SiO2and broad carbon (001) peaks at around 24° in NC,SAC–Co and Mg-Co catalyst [57,58].

Fig.1.A schematic illustration for the preparation of Mg-Co catalyst.

The microstructure of pristine template SiO2was firstly clarified by TEM and STEM (Fig.S2).It can be seen that SiO2displays a spherical shape with a diameter of ∼10 nm and no obvious lattice fringes is observed,which is consistent with the XRD pattern.We then examined the microstructures and compositions of the Mg-Co catalyst in different processing steps by TEM and STEM (Figs.S3-S5).An irregular sphere structure is observed in Mg-Co before removing the template,which should stem from the carbon wrapping on SiO2nanospheres.In contrast,a porous structure with welldefined pores can be seen in the final Mg–Co catalyst,making it an ideal structure for providing abundant accessible active sites to greatly enhance the reaction kinetics.In addition,these porous structures enable efficient penetration of acids,resulting in the dissolution of unembedded metal nanoparticles under acidic conditions.The EDS elemental mapping along with ICP–AES further confirms corresponding compositions of the SAC–Co and Mg–Co catalyst (Table S1).Additionally,the Co contents of the two samples are nearly identical to each other.High resolution TEM (Fig.S6) shows well-resolved lattice fringe with graphitic (001) inter-planar spacing of 0.34 nm in Mg-Co [59].

To gain an atomic understanding of the structures of metal sites in Mg–Co,aberration corrected STEM (AC–STEM)measurements were conducted.It is clear in Fig.2a that Mg-Co exhibits a three-dimensional porous structure morphology and there is no large crystallized area,thus indicating the absence of large-size nanoparticles.Fig.2b shows a representative atomic-resolution HAADF–STEM image of Mg–Co.The larger atomic number (Z value) of Co in comparison to Mg results in brighter spots of Co and darker Mg,as indicated by the enlarged image in Fig.2c and the intensity line profiles of Mg–Co atom pairs in Fig.2d and e.It is found that the metal species can be presented as isolated atoms,paired atoms and alloy clusters without a long-range order.To further explore the possible structure of Mg–Co atomic pairs,HAADF-STEM simulations were conducted (Fig.2f and g).By combining the simulated results with experimental results,one can deduce a possible Mg–N2–Co moiety for the Mg–Co atomic pairs,as illustrated in Fig.2h.The elemental mapping in Fig.2i demonstrates the even distribution of hybrid sites across the NC support.Since the SAC–Co,which was prepared without the Mg adsorption process,exhibits a typical single atom catalyst morphology (Fig.S7) and the Raman spectrum (Fig.S8) of both SAC–Co and Mg–Co displays a relatively high ID/IGvalue,i.e.,relatively high ratio disordered sp3(D band) carbon rather than graphitic sp2carbon (G band),it can be inferred that introducing Mg contributes to Co agglomeration and that the Mg–Co has certain number of defects [60].The agglomeration could originate from the intrinsic affinity of Mg and Co.Moreover,textural properties of Mg–Co investigated by N2adsorption–desorption isotherms reveal a hysteresis loop of type IV,affirming the mesoporous structure of the Mg–Co(Fig.S9).The huge Brunauer–Emmett–Teller surface area and total pore volume of Mg–Co (1330.4 m2g-1and 3.3618 cm3g-1;Table S2)imply considerable exposed matrix defects even after twostep annealing process.In addition,the similar BET results of Mg-Co and SAC–Co in Fig.S9 and Table S2 reveal that the incorporation of Mg into SAC–Co catalyst has little influence on textural properties of NC matrix.Therefore,it can be estimated from above characterization that the formation of dual single atom metal sites in NC matrix is most likely to be derived from the adsorption of Mg salts on matrix defects neighboring the Co sites formed in the first annealing step.

Fig.2.(a) TEM image.(b) HAAFF-STEM image.(c) Enlarged HAADF-STEM image from the selected area in b.(d,e) Aquired HAADF-STEM intensity line profiles from the selected areas in c.(f,g) Simulated Mg-Co HAADF-STEM image.(h) Proposed dual-site Mg-N-Co moiety in Mg-Co.(i) STEM-EDS mapping of C,N,Mg and Co of Mg-Co.

To reveal the surface chemical states and local coordination environments of Mg–Co catalyst,X-ray photoelectron spectroscopy (XPS;Fig.S10 and Fig.S11) and Xray absorption spectroscopy (XAS) were carried out.The X-ray photoelectron spectroscopy (XPS) of N 1s spectrum reveals the existence of pyridinic N (∼398.1 eV),Co–N bonding (∼398.8 eV),pyrrolic N (∼400.1 eV),graphitic N(∼401.2 eV) and oxidized N (∼402.1 eV),demonstrating the presence of Co–N moieties in Mg–Co [61].The Co 2p XPS peaks are composed of the Co 2p1/2satellite(∼804.0 eV),Co 2p1/2(∼796.4 eV),Co 2p3/2satellite (∼786.8 eV),and Co 2p3/2peaks (∼781.2 eV),demonstrating that the Co atoms in Mg-Co are positively charged due to the coordination with N [62].The fine structure of the Co species was analyzed by XAS.Fig.3a shows the Co K-edge X-ray absorption nearedge structure spectra of Mg–Co and reference samples.The absorption threshold position of Mg–Co is located between cobalt phthalocyanine (CoPc) and Co foil,implying that the chemical valence of Co is between +2 and 0,agreeing well with the XPS results.The Fourier-transformed extended Xray absorption fine structure (EXAFS;Fig.3b) spectrum of Mg–Co shows a peak at ∼2.0 ˚A in the R space,close to the Co–N peak of CoPc.A shifted peak approaching Co–Co path(2.2 ˚A)at around 2.1 ˚A is also observed,the smaller path and lower peak intensity in contrast with Co foil in Mg–Co suggests the successful incorporation of MgCo alloy clusters.The k negative shift of Mg–Co as shown in EXAFS wavelet transforms plots(Fig.3c-f)compared with the Co–Co bond of Co foil arise from the different coordination numbers caused by Mg-Co coordination and small cluster size.The characterization presented above provides compelling evidence that the Co atoms in the system can exist as both mononuclear and multinuclear centers,and that they are coordinatively and electronically stabilized by the support material.

Fig.3.(a) XANES spectra.(b) EXAFS spectra.(c-f) Wavelet transforms for EXAFS spectra.

The ORR activity of these catalysts was evaluated in O2-saturated 0.5 M H2SO4solution in a standard three–electrode quartz cell.A catalyst-coated glassy carbon was used as rotating-disk electrode with Ag/AgCl as the reference electrode and carbon rod as the counter electrode.Without hetero Mg active species,the SAC–Co catalyst shows a low onset potential (Eonset) of 0.811 V and a half wave potential (E1/2)of 0.737 V versus RHE (Fig.4b).In contrast,theEonsetandE1/2improve to 0.860 V and 0.806 V over Mg–Co catalyst when Mg sites are incorporated into the single-component Co catalyst,leading to performance that is comparable to that of commercial Pt/C (Eonsetof 0.867 V andE1/2of 0.814 V).In addition,the ORR performance of SAC-Mg and its corresponding STEM were also conducted,and the results show aE1/2of 0.787 V for SAC-Mg,indicating the co-enhancement roles of Mg and Co in Mg-Co catalyst (Fig.S12).The electron transport ORR kinetics of all these investigated catalysts were verified by Tafel plots (Fig.S13).The smallest Tafel slope of Mg–Co confirms its superior intrinsic activity as well as fast kinetics towards ORR.In addition,electrochemical impedance spectroscopy demonstrates the smallest arc radius of Mg–Co (Fig.S14),revealing the lowest electron resistance and thus highest electronic conductivity.Moreover,to give a more accurate ORR activity,turnover frequency (TOF) normalized by metal contents characterized via ICP-AES results was calculated (Table S3).The TOF of Mg-Co was found to be 0.182 electrons site-1s-1,which is 2.5 times and 16.5 times of that of SAC–Co and commercial Pt/C catalysts.

Fig.4.(a) A schematic illustration of DMFC.(b) LSV curves.(c) H2O2 selectivity tests.(d) Fuel cell performances of Mg-Co and Pt/C.(e) Duration tests of Mg-Co and Pt/C.

Apart from activity,ORR selectivity is also evaluated by rotating ring-disk electrode tests(Table S4).As can be seen in Fig.4c,the 2 e-byproduct H2O2yield of Mg–Co is less than 2% throughout the working potential range,which is remarkably suppressed compared with than of SAC–Co (4%∼7%)and almost approaches that of Pt/C.The electron transfer number of Mg–Co is around 3.9,indicating a dominant 4 e-reaction pathway.For versatile DMFC application,electrocatalysts must have excellent anti-methanol-poison ability.Accordingly,linear sweep voltammetry (LSV) analysis under CH3OH introduction was carried out,as shown in Fig.S15.Noteworthy is that Mg–Co exhibits high tolerance to CH3OH,while commercial Pt/C shows high methanol oxidation reaction activity in ORR condition.Furthermore,an accelerated durability test of Mg–Co catalyst was conducted using a rectangular wave potential cycling ranging from 0.6 V to 1.0 V for 5000 cycles (Fig.S16).Clearly,the Mg–Co showed negligible performance loss after the durability test,and the results are consistent with the ICP analysis of metal species before and after cycles (Table S1).XRD and HAADF analysis (Fig.S17) reveal that the alloy XRD peaks disappear and the sample delivers a dense single dispersion over the whole atomicresolution image.This suggests the alloy clusters dissolve into single atoms under the working acidic condition.

To demonstrate the viability of the Mg–Co catalyst under realistic conditions,practical DMFC tests were carried out using Mg–Co and commercial 60 wt% PtRu/C as the cathode and anode (Fig.4a),respectively,and the benchmark assembly of commercial 20 wt% Pt/C and 60 wt% PtRu/C serves as a reference device.Fig.4d demonstrates the polarization performance of Mg–Co assembled device and Pt/C assembled device under 10 M methanol aqueous solutions.The cell with Mg–Co as the cathode reaches a peak power density of 0.204 W cm-2and an open circuit voltage (OCV) of 0.45 V,which exceed the performance of 20%Pt/C(82 mW cm-2for peak power density and 0.37 V for OCV).The performance of the Mg–Co assembled cell improves with the increase of methanol concentration while that of Pt/C suffers from severe performance loss when operated at high methanol concentration (Fig.S18).Higher OCV and peak power density at high methanol concentration indicate that Mg–Co has higher ORR activity and methanol tolerance than commercial Pt/C.Moreover,the DMFC stably operates at a constant working voltage of 0.6 V for at least 10 h with a negligible decay of output current density (Fig.4e).These results imply that Mg–Co holds great potential in DMFC practical applications.

In order to gain deep insight into the remarkably advanced ORR performance of the dual-site Mg–N–Co over the single–site Co–N–C,and particularly discern the physical origin of the boosted performance upon incorporation of Mg,first-principles calculations were performed to comparatively study the elementary steps involved in ORR on dualsite Mg–N–Co and single–site Co–N–C (Figs.S19 and S20)according to the HAADF observations.The proposed ORR pathways of SAC–Co and Mg–Co are presented in Fig.5a and b.The calculated free energy diagrams for ORR over the two modeled catalysts are presented in Fig.5c.Clearly,the path of ORR over Mg–N–Co is energetically much smoother than that on Co–N–C.Exactly,the formation of∗OOH is significantly endothermic and is the potential-determined step on Co–N–C.In sharp contrast,the key intermediate∗OOH is unstable over the dual-site Mg–N–Co and will dissociate spontaneously into∗OH and∗O after full geometrical optimization.Subsequent transformation of∗OH and∗O to O∗is found to be most endothermic and thus to be the potentialdetermined step.Accordingly,the overpotential on the dualsite Mg-N-Co was calculated to be 0.27 V,which is significantly lower than that on Co–N–C (0.71 V).These results clearly suggest that Mg–N–Co possesses a remarkably higher intrinsic activity than Co–N–C,rationalizing the experimental observation of a higher ORR kinetics upon introducing Mg.Moreover,it is obvious in Fig.S20 that the three key intermediates alternatively interact with the coordinating N and the Co on the Mg–N–Co moiety,while the Co center in Co–N–C constantly serves as the active site.This distinctive feature aroused by introducing Mg clearly indicates that the dual sites work in concert to properly address different elementary steps of the ORR and hence synergistically provide highly active sites for ORR.

Fig.5.Proposed 4e- and 2e- oxygen reduction mechanism in acidic solution for the ORR on (a) SAC–Co and (b) Mg-Co.Free energy diagrams on (c)SAC–Co and (d) Mg-Co.

In addition,ORR can proceed via either a 2e-or a 4epathway (Fig.5d) and the selectivity is determined by the propensity to break the O–O bond during ORR.It is found that the four consecutive protonation steps over the Co-N4single site follows a typical∗OOH→∗O→∗OH pathway while the Mg-N-Co follows∗O+∗OH→∗O→∗OH.In the first key intermediate (i.e.,∗O+∗OH),the∗OH fragment remains at the Co site,while the dissociated∗O is bonded to a coordinating N between Mg and Co.As a result,the significantly facilitated dissociation of∗OOH would completely suppress the 2e-pathway over Mg–N–Co,thus responsible for the experimentally observed extremely low H2O2yield and thereby the high 4e-selectivity.Besides,further calculations of CO adsorption on Pt and Mg-Co (Fig.S21) reveal that Mg-Co adsorbs CO much more weakly than Pt,indicating a stronger CO poisoning resistance of Mg-Co.

It has been demonstrated that the bonding between∗O and coordinating N between Mg and Co in Mg–N–Co plays a key role in boosting both activity and the selectivity.To uncover the physical origin of the behaviors,a comparative study of the density of states(DOS)of the coordinating N atom in Co–N–C and Mg–N–Co was conducted,as shown in Fig.6a-c.It turns out that the Nporbitals in Co–N–C are delocalized,but become rather localized upon Mg incorporation and exhibit ad-band-ditto feature of conventional transition metals.According to the Newns–Anderson–Grimley model [63,64],which describes the orbital hybridization between adsorbates and active sites,there is remarkable difference in the hybridization when adsorbates interact with delocalized or localized bands,as depicted in Fig.6a.In the former case a broadened adsorbate state forms,whereas the adsorbate state splits into localized bonding and antibonding states in the latter case,resulting in stronger interaction strengths [65].To investigate whether splitting or broadening occurs on the adsorption of O on the coordinating N,the crystal orbital overlap population (COHP) was analyzed (Fig.6d) [66,67].Clearly,the hybridization between O and N splits into distinct bonding and antibonding states,closely resembling that of O adsorption on both Co centers in Co–N–C and Mg–N–Co.Moreover,it is clear that the main peak in the antibonding states between O and N in Mg–N–Co is much higher than those in the Co–O interactions,indicating much less occupation of the antibonding state.As a result,the adsorption of O at the N site is remarkably stronger than that over both Co centers.These features are distinctly ascribed to the dramatic electronic structure modification of the coordinating N atom induced by Mg incorporation.Therefore,the high ORR activity and selectivity of Mg–Co should be intrinsically attributed to the Mg–N–Co moiety.Specifically,the uniqued-band-ditto feature enable to circumvent the linear relation between the energetics of∗OOH and∗OH,leaving sufficient space to improve the ORR catalytic activity and 4e-selectivity.

Fig.6.(a) Proposed D-band-like p band of the coordinating N atom in the Mg–N–Co moiety and (b) its interaction with the states of adsorbates.(c) DOS results.(d) COHP results of the N (or Co)-∗O interactions in dual-site Mg–N–Co and single-site Co–N–C moiety,respectively.The Fermi level is set to zero.

4.Conclusions

In summary,we report the synthesis and analysis of a dual-site Mg-Co single atom catalyst for high-performance acidic ORR and DMFC application with high selectivity and durability.Through pyrolysis-impregnation-pyrolysis and hard template assisted preparation procedure,highly-accessible and dense Mg-N-Co dual sites are obtained in the Mg-Co catalyst.The Mg-Co electrocatalyst exhibits a half wave potential of 0.806 V(vs.RHE)in acid media with good stability,which is remarkably higher than that of SAC-Co and even comparable to that of the commercial Pt/C.Practical DMFCs test using Mg-Co as the cathode achieves a peak power density of over 200 mW cm-2that far exceeds that of the commercial Pt/C counterpart (82 mW cm-2).The durability tests indicate a negligible power loss with 12 h due to its excellent tolerance to CO poisoning and resistance to methanol.Theoretical calculations reveal that the incorporation of Mg could result in dramatic electronic structure modification of the coordinating N atom in the Mg–N–Co moiety,which is responsible for the experimentally observed high ORR activity and selectivity of the Mg–Co catalyst.Our findings suggest the potential of engineering the local coordination environment of catalytically active sites viap-block atom incorporation and provides new insights into rational design of high-performance and durable ORR catalysts in DMFC.

Declaration of competing interest

The authors declare no conflict of interests.

Acknowledgements

The authors acknowledge the funding support from the National Natural Science Fund for Distinguished Young Scholars (52125103),the National Natural Science Foundation of China (52071041,12074048 and 12147102),Chongqing Natural Science Foundation (cstc2020jcyj-msxmX0777 and cstc2020jcyj-msxmX0796),Science Foundation of Donghai Laboratory (DH-2022KF0307) and the Fundamental Research Funds for the Central Universities (106112016CDJZR308808).We also would like to thank JinJing Tang and Bin Zhang at Analytical and Testing Center of Chongqing University for their assistance with XRD and TEM analysis.

Supplementary materials

Supplementary material associated with this article can be found,in the online version,at doi:10.1016/j.jma.2023.04.008.