In situ synthesized low-PtCo@porous carbon catalyst for highly efficient hydrogen evolution

Department of Chemistry, University Coll WC1H 0AJ, UK. E-mail: z.x.guo@ucl.ac.uk; Semiconductor Measurement Solutions, K California 95051, USA Department of Chemical Engineering, Univ London, WC1E 7JE, UK UK Catalysis Hub, Research Complex at Laboratory, Harwell Oxon, OX11 0FA, UK Departments of Chemistry andMechanical E and Innovation, University of Hong Kong, H † Electronic supplementary informa 10.1039/c8ta12263f Cite this: J. Mater. Chem. A, 2019, 7, 6543


Introduction
Clean and sustainable energy sources are highly desirable for both reduction of carbon dioxide emission and reliance on pollutive fossil fuels. Hydrogen is an ideal choice due to its elemental abundance, high energy density per unit mass (143 kJ g À1 ) and zero carbon emission at point of use. 1,2 Industrial hydrogen is manufactured by steam methane reforming and coal gasication, in which CO 2 emission persists unless effective CO 2 capture is implemented. In contrast, electrocatalytic water-splitting, using off-peak electricity, offers a cost-effective path to produce hydrogen (Table S1 †). 3 Commercial Vulcan XC72 platinum on carbon (20 wt% Pt/C) is the benchmark catalyst for the hydrogen evolution reaction (HER), as it requires a very small overpotential. However, the noble-metal catalyst is costly and prone to detachment from the carbon support during cycling. Thus, it is desirable to develop low-cost, highly efficient and stable alternatives for HER. [4][5][6] One straightforward strategy is to lower the Pt loading but preserve its performance. Several approaches have been proposed in the past, one of which is to dope Pt into ne structures to maximise Pt exposure or active sites. [7][8][9] Tang et al. studied platinum nanowires grown on single-layered nickel hydroxide, Pt NWs/SL-Ni(OH) 2 , and showed HER performance comparable with commercial 20 wt% Pt/C. 7 However, the synthesis method is arduous and time-consuming. Another approach is to adopt alternative transition metals to Pt. A nanocrystalline Ni 5 P 4 and its related phosphorous compounds have been reported to exhibit good HER performance. 10 However, they utilize hydrogen in the annealing procedure, which conicts with the initial purpose of hydrogen production from water. Recently, nitrogen-doped graphene with a small amount of transition metal clusters [11][12][13] has also exhibited HER activity, as nitrogen can induce active sites, but the HER activity is much lower than that of 20 wt% Pt/C, as the Tafel slope is more than 2.6 times larger than its commercial counterpart. 13 Inspired by the above studies, zeolitic imidazolate framework (ZIF-67) (Co-2-methylimidazole) was selected as the precursor for a nitrogen doped carbon skeleton, 14,15 as well as a transition metal provider for anchoring Pt to minimize metal clustering via direct annealing under N 2 atmosphere (Scheme 1). [16][17][18][19] The resulting encapsulated-Pt and nitrogen-doped carbon framework catalysts demonstrated excellent HER activity and long-term stability, outperforming the commercial counterpart, particularly at low potential in an acidic electrolyte, as ZIF-67 provided a skeletal structure for Pt to form an alloy and prevent clustering during the annealing process. Moreover, Hubbard-corrected density functional theory (DFT+U) simulations revealed that the free energy of the HER reaction could be dramatically tuned by use of different ratios of the metal alloy inside the carbon cage. The Pt atom tended to dope at the edge positions of the Co clusters at a low concentration, but inhabited bulk positions with increasing Pt loading, indicating that not all Pt atoms will act as active sites in the alloy clusters. Hence, HER performance does not increase proportionately with Pt content in the system and dilute Pt doping can be very cost-effective in future catalyst designs for HER. Our results provide not only a strategy of efficient utilization of noble metal elements in HER, but also a practical route for lowcost synthesis of such alloy catalysts. The simple synthesis procedure also offers strong potential for industrial scale-up.

Experiments
All chemicals were purchased from Sigma-Aldrich and used without any purication unless otherwise noted.

Synthesis ZIF-67
First, cobalt nitrate hexahydrate (7.185 g) was dissolved in 500 mL methanol. Then, 2-methylimidazole (8.106 g) was dissolved in another 500 mL methanol. The latter solution was slowly added to the former in a 1 L wide mouth jar while stirring at room temperature. Stirring proceeded for 2 h, aer which the solution was le to settle for 24 h. The clear top solution was drained and the purple precipitates were collected by centrifugation with further methanol washing. The samples were dried in an oven at 80 C for 24 h.

Synthesis of Pt@ZIF-67 and CPt@ZIF-67
As-synthesized ZIF-67 crystal powder (100 mg) and chloroplatinic acid hexahydrate H 2 PtCl 6 $6H 2 O (26 mg) were added into methanol (20 mL). Sonication and stirring were carried out for 1 h to achieve a homogenous inky solution. Pt-coated ZIF-67 (Pt@ZIF-67) was obtained by grinding the purple solid into ne powder aer evaporating the methanol in the oven at 80 C for 24 hours. Then, the product was carbonized at 900 C (heating rate of 5 C min À1 ) in an alumina boat (1 Â 1.5 Â 5 cm) in a horizontal tube furnace to form carbonized ZIF-67 with Pt doping, termed CPt@ZIF-67. The powders were annealed at 900 C for 6 hours under nitrogen and cooled to room temperature. Aer carbonization, black solids were collected and ground into ne powder. To optimize the carbonization procedure, sample synthesis was carried out at different temperatures and durations, termed CPt@ZIF-67-x C-y h, where x and y represent the temperature and duration of the carbonization.

Characterizations
Powder X-ray diffraction (PXRD, Stoe Stadi-P, Cu-K-alpha and Mo-K-alpha) was carried out by encapsulation of samples in a 0.5 mm diameter borosilicate glass capillary under ambient conditions. X-ray photoemission spectroscopy (XPS, Al-K-alpha, Thermo Scientic) and scanning electron microscopy (SEM, Jeol) were carried out on carbon tape. Transmission electron microscopy (TEM, Jeol) measurements were completed on a carbon-coated copper TEM grid support. Raman spectroscopy (514.5 nm laser, Renishaw) was carried out with powder samples on a glass slide. Atomic force microscopy measurements were carried out in AC mode using silicon rectangular-shaped cantilevers with a tip curvature radius of 6 nm and a force constant of 40 N m À1 . The test samples were prepared by dip-coating the materials on clean silicon substrate. (AFM, Keysight 5600LS). The Pt content was measured by microwave plasma atomic emission spectroscopy (Agilent 4210 MP-AES). method with a glassy carbon rotating disk as the working electrode and carbon rod and Ag/AgCl/saturated KCl as counter and reference electrode, respectively, in an acidic electrolyte (0.5 M H 2 SO 4 ) at room temperature. All measurements were carried out with a xed catalyst deposition of $0.26 mg cm À2 on a 3 mm diameter (area of 0.0707 cm 2 ) glassy carbon disk electrode. The catalyst was prepared as follows: 2 mg of sample was dispersed in a 540 mL solution consisting 500 mL of 4 : 1 v/v water/ethanol and 40 mL of Naon (5% solution) under sonication. The sonication was carried out up to an hour to get uniform dispersion of catalyst ink, of which 5 mL was micropipetted and dropped on a glassy carbon disk electrode followed by drying at 60 C in the oven. All electrochemical test results are reported with respect to the reference Ag/AgCl. The linear sweep voltammetry (LSV) curves were recorded with voltage sweeping at 10 mV s À1 in the potential range of +0.0 V to À0.7 V with disk rotating speed of 1600 rpm in 0.5 M H 2 SO 4 . The data were collected aer stable CVs were obtained. The measured potentials against Ag/AgCl were converted to RHE using the following relation: in 0

Density functional theory (DFT) simulations
All calculations were carried out based on density functional theory, in the Vienna ab initio package (VASP). 20,21 The Perdew-Burke-Ernzerhof (PBE) functional 22 was used for the exchangecorrelation term with the projector augmented wave method. 23,24 A plane wave cutoff of 500 eV with forces converged to 0.01 eVÅ À1 was employed. Considering the cobalt carbon interaction, on-site Hubbard U corrections (3.3) were selected for the Co delocalized d electrons, as has been assessed in the past. 25,26 Spin polarization was counted in all calculations. Gamma only was selected for the K-point sampling.
The free energies of the intermediates were obtained by where DE(H*) is the adsorption energy calculated by DZPE is the zero-point energy change, and DS is the entropy change of adsorption H. All analysis within DFT scope used normal-mode analysis, as proposed by Nørskov. 27 where ZPE(H 2 ) is 0.27 eV. 28 Since the vibration frequency of H is less than 1 meV, . For H 2 at 300 K and 1 atm, TS(H 2 ) ¼ 0.41 eV. Then TDS z À0.2 eV.

Results and discussion
Carbon-encapsulated Pt-alloy carbon network, hereaer referred to as CPt@ZIF-67, was synthesized by a two-step approach, modied from previous work of Gadipelli et al. 19 First, ZIF precursors and platinum salt were completely dissolved in methanol by sonication and stirring; then, the solvent was evaporated and the as-synthesized material was annealed in an inert N 2 atmosphere at a given temperature to yield the nal sample. For comparison, carbonized ZIF without the Pt salt was also prepared, denoted CZIF-67. The procedures are illustrated in Scheme 1 and were detailed in the Experiments section. First, carbonization temperature and duration were optimized from 600 to 1000 C for 2 or 6 hours. These samples were denoted as CPt@ZIF-67-x-y, where the carbonization process is at x C for y hours. The morphology of the pre-annealing samples was studied by AFM and SEM. Before annealing, the samples show clear polyhedral structures (Fig. S2a, b, S3b and S4a †). The AFM height prole measurements (Fig. S2c †) indicate that the particle sizes of Pt@ZIF-67 vary from 0.8 to 1.5 mm. Aer annealing, the samples became amorphous (Fig. S4 †). The degree of amorphization gradually increased with the increasing carbonization temperature and duration, which is in line with previous observations. 29 For CPt@ZIF-67-600-2 ( Fig. S4b †) and CPt@ZIF-67-700-2 ( Fig. S4c †), the crystalline cubic ZIF morphology became spherical on the edges. When the annealing temperature increased to 900 C and above, the original structure fully collapsed into an amorphous structure ( Fig. S4f-h †). Hence, the optimum carbonization temperature and duration are 900 C and 6 hours. The optimized CPt@ZIF-67-900-6 also showed effective catalysis performance, with the lowest onset potential and the highest current density compared with those from other carbonization conditions, as will be illustrated later. Hence, all the samples used were 900 C-6 h, unless specied otherwise.
TEM images (Fig. 1a-c) show that the size of the PtCo alloy clusters in CPt@ZIF-67 is 2.7 nm on average, almost half the particle size (5.2 nm on average) of the commercial 20 wt% Pt/C (Fig. S6e †). Meanwhile, the densities of the metal clusters in CPt@ZIF-67 and 20 wt% Pt/C were at a comparable level ( Fig. 1b and S6d † or Fig. S6a and S6e †). Hence, the metal particles of our samples are one eighth scale of the commercial ones.
By tailoring the carbonization conditions, the graphitic or graphene shell may still be structurally porous or of variable thickness, i.e. not completely covering the whole cluster or with only one or two graphene layers, allowing easy access of electrolyte to the catalytic metal sites while protecting the clusters from further growth and aggregation. 30,31 As a result, the catalyst is expected to be effective and stable over a long period of service.
The high crystallinity of the PtCo alloy is shown in Fig. 1a, with a lattice spacing of 0.23 nm corresponding to the PtCo(111) plane. 32 Elemental mapping (Fig. 1e-h) reveals that both Pt and Co elements distribute homogenously in the alloy. Meanwhile, the elemental mapping of Pt@ZIF-67 (Fig. S7 †) indicates that these elements are uniformly distributed even before annealing.
Further, structural analysis was carried out using PXRD patterns. Before annealing (Fig. S8a †) the peaks of samples with and without Pt doping appeared at the same positions in strong contrast with the baseline. This indicates the high crystallinity of the pre-annealed samples and that Pt was well distributed without disturbing the ZIF structure. 33,34 Aer annealing, peaks at 44.2 and 51.3 of CZIF-67 could be assigned to Co (111) and (200) (JCPDS 15-0806) (Fig. S8c †). 35 Blue shis of those peaks were observed in the CPt@ZIF-67 sample, due to the Pt atoms alloyed in the Co cluster structure leading to lattice expansion. 32,36 Meanwhile, the closeness of peaks when compared to those of CZIF-67 indicates that a limited amount of Pt atoms is alloyed into the Co lattice. The Mo-radiation PXRD of CPt@ZIF-67 is shown in Fig. S8e; † the peaks can be indexed as C (002), 19 Co 3 O 4 (311), 37 PtCo (111) 38 and PtCo (200), 32 which indicates that the catalyst consists of carbon, cobalt oxide and platinum cobalt alloys.
Raman spectroscopy measurements were carried out and are shown in Fig. S9. † The D band peak (signature disorder-induced sp 2 -hybridized graphitic carbon oen activated by defects) and the G band peak (the ordered graphitic carbon with stretching vibrations) are located at 1340 and 1596 cm À1 , respectively. Moreover, the high I D /I G ratio of CPt@ZIF-67 indicated that the carbonized ZIFs possessed graphitic characteristics with abundant defect concentration (Fig. S9a †). 39 Another relatively weak 2D broad peak located at 2652 cm À1 indicates that the metal core contact doped electrons to the carbon cages.
The surface composition and chemical and electronic states of CPt@ZIF-67 were characterised by XPS. As shown in Fig. S10a, † comparison of CPt@ZIF-67 and CZIF-67 proved the formation of Pt alloy. In Fig. 2a, three distinct peaks at 398.4, 400.6 and 402.3 eV can be assigned to pyridinic N, graphitic N and quaternary N, respectively. 40,41 The formation of pyridinic N in the annealed sample indicates that the carbonization led to more stable six-membered rings (Table S2 †). The Co and Pt spectra of CPt@ZIF-67 are shown in Fig. 2b and c, respectively. The two peaks in each spectrum could be assigned to Co 2p 3/2 and Co 2p 1/ 2 and Pt 4f 7/2 and Pt 4f 5/2 , respectively. Compared with the pure Pt metal peaks in 71.4 and 74.5 eV, notable shis of Pt in CPt@ZIF-67 to 71.9 and 75.1 eV were observed. 42 This positive shi was due to the charge transfer from transition metal Co to noble metal Pt in the alloy, which is in line with our PXRD analysis. 43 The Brunauer-Emmett-Teller (BET) surface area and pore size distribution were obtained by N 2 sorption isotherms. The calculated surface areas of CPt@ZIF-67, CZIF-67 and Pt@ZIF-67 were 90.1 m 2 g À1 , 330.2 m 2 g À1 and 1516.2 m 2 g À1 , respectively (Fig. S11a-c †). The pore size distributions were also investigated. As shown in Fig. S11d-f, † the CPt@ZIF-67 possessed larger pore size than CZIF-67 and Pt@ZIF-67, mainly concentrated in the range of 4 to 6 nm (Fig. S11d †). In contrast, the pore size ranges of CZIF-67 and Pt@ZIF-67 were focused within 1-4 nm and 1-2 nm, respectively. (Fig. S11e and f †). Hence, the CPt@ZIF-67, with mesoporous characteristics, effectively promotes electrolyte penetration and charge transfer, even while possessing much less surface area.
Both MP-AES and XPS were employed to determine the Pt loading and near-surface composition of the Pt in the alloy cluster, respectively. The MP-AES results show that the average Pt loading in the catalysts is $5 wt%, which is comparable with the level of an atomic layer deposition (ALD) synthesis depositing single Pt atoms on nitrogen doped graphene. 4 To eliminate error, six points were selected from the XPS analysis to give a mean value of 0.93 wt% Pt in the near-surface region of the catalyst, as detailed in Table S3. † Collective results show that CPt@ZIF-67 is of much lower Pt content than the commercial 20 wt% Pt/C.
The electrocatalytic HER capability of the catalysts was studied using steady-state LSV on a glassy carbon disk electrode in 0.5 M H 2 SO 4 . The LSV curves of CPt@ZIF-67 are shown in Fig. 3a with those of CZIF-67, carbon black and 20 wt% Pt/C for comparison. Overall, the catalysts with Pt loading outperform those without. To achieve a reference current density of 10 mA cm À2 , the alloy CPt@ZIF-67 required a 50 mV overpotential, which was 5 mV lower than 20 wt% Pt/C (55 mV). Tafel slopes of the four samples are plotted in Fig. 3b. The values are 167.9, 27.1, 331.8 and 35.5 mV dec À1 for CZIF-67, CPt@ZIF-67, carbon black and 20 wt% Pt/C, respectively (Fig. 3b). The results indicate that the HER follows the Volmer-Heyrovsky mechanism. 44,45 It is worth mentioning that additional tests were carried out on 20 wt% Pt/C (Fig. S12a †); the Tafel slopes calculated from these three new tests were identically 35.5 mV dec À1 , which is slightly poorer than previous literature values (31 mV dec À1 ). 46,47 This data demonstrates the same trend as the polarization curves. CPt@ZIF-67 outperformed the commercial 20 wt% Pt/C. The exceptionally low Tafel slope of 27.1 mV dec À1 at a low potential for CPt@ZIF-67 represents nearly 30% improvement over commercial 20 wt% Pt/C. A 24 hour stability test with CPt@ZIF-67 and 20 wt% Pt/C was carried out, as shown in Fig. 3c, where the activity decays are virtually negligible for CPt@ZIF-67 in comparison with 20 wt% Pt/C. This could be attributed to the catalyst being protected by the carbon shell from erosion by the acidic electrolyte. The morphology and structure of the catalyst are well-retained, since Fig. S6b, c, S9c and S10b † show that there is imperceptible change in CPt@ZIF-67 before and aer stability tests. Assuming all Pt atoms are active sites to the acidic electrolyte, approximately 4.01 Â 10 16 sites cm À2 are accessible to the electrolyte, which is more than in Pt(111) (1.5 Â 10 15 sites cm À2 ). 48 Then, the turnover frequency (TOF) of the HER was calculated (ESI note 1 †). The TOFs of CPt@ZIF-67 are 2.94 and 0.80 s À1 at h ¼ 100 and 50 mV, respectively, much better than other published noble-metal or non-noble metal-based catalysts (Table S4 †). Likewise, compared with other reported HER catalysts, including Pt alloy (Table 1), the present Tafel slope was a record low for HER electrocatalysts.
The electrochemically active surface area (ECSA) was calculated by measuring the total charge of the H upd  adsorption/desorption region. 63 The specic ECSA (ECSA per unit weight of metal) of CPt@ZIF-67 is 64.3 m 2 g À1 pt (ESI note 2 †) is slightly higher than that of commercial 20 wt% Pt/C (63 m 2 g À1 pt). 64 To further understand the electrocatalytic activity of CPt@ZIF-67 for HER, we performed electrochemical impedance spectroscopy (EIS). The Nyquist plots of the EIS results are demonstrated in Fig. S13. † The Nyquist plot of CPt@ZIF-67 gave a similar semicircle to that of commercial 20 wt% Pt/C. This result is another proof that CPt@ZIF-67 affords similar kinetics as the commercial catalyst (with only a quarter of the Pt amount).
The mass density comparison is shown in Fig. S14. † At an overpotential of 10 mV, CPt@ZIF-67 exhibited a mass density of 0.87 A mg À1 Pt, which is about 4 times higher than that obtained by commercial 20 wt% Pt/C (0.22 A mg À1 Pt). This outstanding electrocatalytic activity of CPt@ZIF-67 may be ascribed to the uniformly dispersed PtCo clusters spread over a relatively large surface area for electrochemical reactions, even at a relatively low overall loading (5 wt%). More importantly, a downshi of the Pt D-band may exist for Pt electronic structure when PtCo alloy is formed, due to the charge localization between Co and Pt. 65 This should reduce the desorption energy of protons around the Pt site, facilitating formation of H 2 gas molecules.
Density functional theory (DFT) calculations were carried out to further clarify the effects of alloying on the metal cluster embedded nitrogen-doped carbon catalysts. Predominately, |DG(H*)| is a good descriptor for HER activity and should be close to zero to balance H* adsorption and desorption within the reaction steps. 66 The chosen models included the metal clusters, e.g. Co 55 cluster, and Pt-doped Co clusters with different Pt/Co ratios, as shown in Fig. S15. † The C 240 fullerene ball was adopted to mimic the outer carbon coating for the metal clusters, as in previous studies. 67,68 Furthermore, to avoid DFT description failure on the open shell d-electrons of the transition metal, DFT+U, originally proposed by Anisimov et al., 69 was adopted to appropriately treat the strongly correlated electron intra-atomic Coulomb (U) and exchange (J) interactions within a HF-like theory, while the rest of the system was treated with pure DFT.
Unlike the simple cubic Co 55 metal clusters used in previous studies, the model here was adopted from the global minima with the Gupta potential, as shown in Fig. S15a. † The energy difference of those two clusters was 1.05 eV per atom, which is too great to be ignored. Furthermore, Pt doping positions were also considered, including corner, edge and body positions. As shown in Table S5, † for PtCo 54 , a single Pt atom tended to occupy the edge positions (Fig. S15b †). This conclusion is consistent with a recent publication where Pt atoms enrich the edge positions of dodecahedral Pt/Ni clusters. 70 It should also be noted that both doping at the corner and edge positions was exothermic, while replacement at the body position was endothermic. For Pt 2 Co 53 (Fig. S15c †), the formation energy is listed in Table S6. † Dual Pt atoms preferred to dope the edge and body positions (Fig. S15c †). Clearly, only surface doped Pt could act as an effective active site for HER. Bader analysis was also carried out to study electron transfer within the metal alloy cluster. 71 The results show that the average valence electrons on Pt increased by 0.77e and 0.87e for PtC 54 and Pt 2 Co 53 , respectively. This is also in agreement with the observations in our XPS and PXRD analyses. According to the D-band theory, this charge transfer to the active sites can further reduce the overpotential.
The details of calculation of the HER free energy are provided in the Methods section, and different adsorption/desorption sites of H* were also considered, as shown in Fig. S16. † The calculated free energy diagrams of a single-layer carbon shell and a carbon shell within different PtCo alloy clusters are shown in Fig. 4b. Compared with bare C 240 , the free energy was dramatically reduced, by 48.6%, if the metal cluster was inside the carbon shell. For different levels of Pt doping, the energy barrier of the structure with 2 Pt atoms was 90.1 meV higher than that of single Pt doped. Hence, the "extra" Pt atom in Pt 2 Co 53 @C 240 does not provide an additional active site or lower the energy barrier for activation. Therefore, catalyst performance did not linearly increase with Pt doping level. The free energy of the metal-only clusters was also calculated for comparison, as seen in Fig. S17. † Though the metal-only clusters possess a lower free energy (0.12 eV), they were difficult to stabilize in the practical environment, due to the lack of a ligand stabilizer. The electron density difference with/without the C-shell, PtCo 54 @C 240 , was also calculated, shown in Fig. 4a. The electron density seems to accumulate at the outside of the carbon cage, which is in line with the Raman shis in our experiment section. This charge redistribution could greatly facilitate the attraction/sorption of H + for the HER reaction.

Conclusions
PtCo cluster-embedded porous carbon catalysts were effectively synthesized via Pt-doped ZIF-67 formation and subsequent tailored carbonization, leading to nitrogen-doped and nely dispersed PtCo clusters rmly anchored over the derived porous carbon. The resulting catalyst (CPt@ZIF-67) exhibits extraordinary performance for HER in 0.5 M H 2 SO 4 , achieving a Tafel slope of 27.1 mV dec À1 with an overpotential of only 50 mV at the reference current density of 10 mA cm À2 , which is four times the mass-specic energy density of the commercial counterpart (20 wt% Pt/C), even at a much lower Pt loading of 5 wt%. The improvement is partly attributed to the ease of charge transfer from the metal cluster to lower the D-band centre, the high electron conductivity of the well-integrated carbon substrate, and the richness of N-functional groups, along with the good dispersion of ne clusters over large surface area. Meanwhile, the catalysts also possessed excellent stability, with negligible loss of activity for up to 24 hours in an acidic electrolyte. DFT calculations revealed that Pt doping in the Co cluster does not always favour the surface sites to impart direct benet to HER. Hence, additional Pt doping does not proportionally enhance HER activity. The carbon cage outside the metal clusters is more electronegative, which also enhances H + sorption for HER. Overall, this approach harnessed several benecial effects to generate a very effective HER catalyst as a low cost and highly stable alternative to commercial 20 wt% Pt/C for hydrogen production. It also provides a design strategy for low Pt loading over surfaces to impart effectiveness at a lower material cost. The catalyst could also be a strong candidate for an overall water splitting process in the future.

Conflicts of interest
There are no conicts to declare.