Highly Active, Nonprecious Electrocatalyst Comprising Borophene

Jul 7, 2017 - Electrochemical water splitting holds great promise for scalable commercial production of the clean fuel hydrogen from water, especially...
2 downloads 14 Views 843KB Size
Communication pubs.acs.org/JACS

Highly Active, Nonprecious Electrocatalyst Comprising Borophene Subunits for the Hydrogen Evolution Reaction Yanli Chen,†,∇ Guangtao Yu,‡,∇ Wei Chen,*,‡ Yipu Liu,† Guo-Dong Li,† Pinwen Zhu,*,§ Qiang Tao,§ Qiuju Li,† Jingwei Liu,‡ Xiaopeng Shen,‡ Hui Li,‡ Xuri Huang,‡ Dejun Wang,∥ Tewodros Asefa,*,⊥ and Xiaoxin Zou*,† †

State Key Laboratory of Inorganic Synthesis and Preparative Chemistry, International Joint Research Laboratory of Nano-Micro Architecture Chemistry, College of Chemistry, Jilin University, Changchun 130012, P. R. China ‡ Laboratory of Theoretical and Computational Chemistry, Institute of Theoretical Chemistry, International Joint Research Laboratory of Nano-Micro Architecture Chemistry, Jilin University, Changchun 130023, P. R. China § State Key Laboratory of Superhard Materials, Jilin University, Changchun 130012, P. R. China ∥ Department of Chemistry, Tsinghua University, Beijing 100084, P. R. China ⊥ Department of Chemistry and Chemical Biology & Department of Chemical and Biochemical Engineering, Rutgers, The State University of New Jersey, Piscataway, New Jersey 08854, United States S Supporting Information *

HER electrocatalysts developed (e.g., MoS2) usually operate well only at small current densities (e.g., at 10 mA/cm2), but they are inefficient and/or become unstable to produce large current densities, in the order of 1000 mA/cm2 or more: current densities that the systems are required to generate for large-scale industrial applications.2,3 The major causes of the poor catalytic activity of many nonprecious electrocatalysts for HER, especially at large current densities, are their poor electronic conductivity and low density of catalytic sites. Transition metal dichalcogenides such as MoS2 are some of the most prominent non-Pt hydrogen-evolving electrocatalysts studied.4 MoS2 possesses a two-dimensional layered structure consisting of sheets held together by van der Waals force (see its crystal structure in Figure 1a). Moreover, MoS2 has semiconducting property (or relatively low conductivity) and large electrical resistance in the direction

ABSTRACT: Developing nonprecious hydrogen evolution electrocatalysts that can work well at large current densities (e.g., at 1000 mA/cm2: a value that is relevant for practical, large-scale applications) is of great importance for realizing a viable water-splitting technology. Herein we present a combined theoretical and experimental study that leads to the identification of α-phase molybdenum diboride (α-MoB2) comprising borophene subunits as a noble metal-free, superefficient electrocatalyst for the hydrogen evolution reaction (HER). Our theoretical finding indicates, unlike the surfaces of Pt- and MoS2based catalysts, those of α-MoB2 can maintain high catalytic activity for HER even at very high hydrogen coverage and attain a high density of efficient catalytic active sites. Experiments confirm α-MoB2 can deliver large current densities in the order of 1000 mA/cm2, and also has excellent catalytic stability during HER. The theoretical and experimental results show α-MoB2’s catalytic activity, especially at large current densities, is due to its high conductivity, large density of efficient catalytic active sites and good mass transport property.

E

lectrochemical water splitting holds great promise for scalable commercial production of the clean fuel hydrogen from water, especially when the reaction is driven by electricity generated by wind, solar, or other renewable energy resources. However, splitting water to generate hydrogen is difficult unless the process is assisted by active and robust catalyst. Pt-based materials are regarded as the most efficient electrocatalysts for the hydrogen evolution reaction (HER) during water splitting, but the high cost and low earth abundance of these noble metalbased catalysts make their use unfeasible. Hence, much effort has been devoted to developing suitable nonprecious alternatives to these noble metal-based catalysts.1,2 Despite advances in nonnoble metal electrocatalysts for HER, they are not viable for water-splitting technology.2,3 These promising nonprecious © 2017 American Chemical Society

Figure 1. (a) Crystal structure of MoS2. (b) Crystal structure of α-MoB2, in which Mo−Mo metallic bonds not shown for clarity. (c) Threedimensional metallic Mo−Mo framework in α-MoB2. (d) Twodimensional graphene-like borophene units in α-MoB2. (e) Density of states of α-MoB2, in which Fermi level is 0 eV. Received: June 18, 2017 Published: July 7, 2017 12370

DOI: 10.1021/jacs.7b06337 J. Am. Chem. Soc. 2017, 139, 12370−12373

Communication

Journal of the American Chemical Society

Furthermore, we calculated the ΔGH* values of the Moterminated (001) and (100) surfaces of α-MoB2 in a range of H coverage (θH*), from 0 to 2 ML. The calculations suggest that (1) in the case the α-MoB2(001) surface, hydrogen occupies first the Mo−Mo−Mo 3-fold hollow sites up to θH* of 1 ML, and then the top of the Mo atoms at θH* ranging from 1 to 2 ML and (2) in the case the α-MoB2(100) surface, hydrogen adsorbs onto the Mo− Mo bridging sites at θH* all the way up to 2 ML (see detailed discussion and Figure S4 in SI). Figure 2 presents ΔGH* values as a function of θH* for the αMoB2(001) and α-MoB2(100) surfaces. The graph of ΔGH*

perpendicular to its sheets. In addition, although its edges are catalytically active, its basal planes are catalytically inert for HER. As a result, the electrocatalytic activity of MoS2 is limited. Thus, researchers are in pursuit of nonprecious electrocatalysts with good electronic conductivity, large density of catalytic active sites, and can deliver large catalytic current densities for HER. Herein we report theoretical and experimental studies that identify α-MoB2 as an active and stable electrocatalyst for HER. In contrast to MoS2, α-MoB2 is a nonlayered, three-dimensional material that possesses metallic properties and high electrical conductivity as well as is rich with catalytic active sites (because not just its “edges”, like those in MoS2, possess the catalytically active sites). This is among the reasons why α-MoB2 affords a high current density of 1000 mA/cm2 during HER at a small overpotantial. α-MoB2 has a hexagonal structure (Figure 1b) and possesses three types of bonds: ionic Mo−B, metallic Mo−Mo and covalent B−B bonds (see Figure S1 and discussion in Supporting Information (SI)). The Mo atoms in α-MoB2 are metallically bonded into 3D frameworks (Figure 1c) whereas the B atoms are covalently bonded into unusual 2D graphene-like borophene structures (Figure 1d) that do not exist in isolated forms.5 The structure of α-MoB2 can, thus, be viewed as an interconnected, sandwich-like configuration comprising borophene-inserted Mobased 3D frameworks. α-MoB2 should be conductive and have metallic property, as evident that its Fermi level is crossed by the density of states (DOS) stemming from both the Mo-frameworks and borophene units (Figure 1e). Experiments show α-MoB2 has a high conductivity or a small electrical resistivity of 6.6 ± 1.7 × 10−7 Ω·m, which is significantly larger than that of Pt (1.1 × 10−7 Ω· m). Its conductivity must be the result of its unusual Moframeworks and borophene subunits. This property, together with its unique structures, should enable α-MoB2 to transfer electrons in all directions throughout its structure. Next, we carried out detailed density functional theory (DFT) calculations to predict α-MoB2’s electrocatalytic activity toward HER. An obtained value of hydrogen adsorption free energy (ΔGH*) for a given catalytic site is often a reliable indicator of the catalytic activity of the site toward HER, with a smaller absolute value of ΔGH* generally implying a better catalytic activity.4b The sites in MoS2 responsible to catalyze HER were previously determined through catalytic activities of HER on the typical (001) and (100) facets, corresponding to the basal plane and the edge plane, respectively, of MoS2 (Figure 1a).4b For compison, in the work herein, the (001) and (100) facets of α-MoB2 are also sampled, and their catalytic activities for HER investigated. The (001) and (100) surfaces of α-MoB2 with different terminations, namely Mo- and B-terminations, are obtained upon cleaving the bulk α-MoB2 structure through the corresponding planes (Figure S2). The results show the Mo-terminated (001) surface is more stable in terms of energy than the B-terminated (001) surface of α-MoB2, and the Mo-terminated (100) surface has comparable structural stability with the B-terminated (100) surface of αMoB2 (see discussions in SI). We assessed the ΔGH* values on the energetically favorable Mo-terminated (001) surface as well as Mo-terminated and Bterminated (100) surfaces of α-MoB2 for H* directly adsorbed at all of possible adsorption sites on these surfaces. The results suggest the Mo-terminated (100) and (001) surfaces of α-MoB2 should possess good catalytic activity toward HER (see Figure S3 in SI).

Figure 2. Hydrogen adsorption free energy (ΔGH*) on α-MoB2(001), α-MoB2(100) and Pt(111) surfaces as function of hydrogen coverage (ranging from 0 to 2). ΔGH* values for the edges of MoS2 are taken from ref 4b for comparison.

versus θH* on the Pt(111) surface is included, for comparison. Note that Pt(111) is the most representative and commonly studied model surface of Pt for such types of calculations.6 Additionally, the ΔGH* values for the edges of MoS2 are from Nørskov et al., included in Figure 2 for comparison.4b As shown in Figure 2, with the increase in θH* from 0 to 1 ML, the ΔGH* value for the α-MoB2(001) surface linearly varies from −0.665 to −0.518 eV whereas the ΔGH* value for the α-MoB2(100) surface fluctuates within a range of −0.603 to −0.292 eV. This result indicates, in a low θH* range (0 to 1 ML), both α-MoB2(100) and α-MoB2(001) surfaces have fewer active catalytic sites for HER than Pt(111) surface, because the ΔGH* values of the latter are closer to zero than those of the former. However, our results show that Pt(111) surface finds it difficult to adsorb more hydrogen when the θH* exceeds 1 ML, reflected by its positive ΔGH* values (0.341−1.495 eV) at higher hydrogen coverage. The main reason is the strong steric effect present at high hydrogen coverage in the Pt(111) surface (see Figure S5 and detailed discussion in SI). Some other metals and nonmetallic compounds have also large positive ΔGH* values and unfavorable hydrogen adsorption properties at large hydrogen coverages.7 In contrast, α-MoB2(100) and α-MoB2(001) surfaces can both be fully covered by hydrogen up to a θH* of 2 ML while their ΔGH* values remain near-zero, even in such high θH* (ranging from 1 to 2 ML) (Figure 2). In addition, the basal plane (i.e., (001) facet) of MoS2 is catalytically inert, and even the edge (i.e., (100) facet) of MoS2 is active for HER only when θH* is ∼0.5 (Figure 2). The above results thus indicate that both α-MoB2(100) and αMoB2(001) surfaces should have large density of highly efficient catalytic sites for HER. Encouraged by these results, including high density of catalytically active sites as well as its high conductivity, we synthesized α-MoB2 using boron and molybdenum as starting materials and investigated its actual catalytic activity for HER (see experimental details in SI). The successful synthesis of highpurity α-MoB2 is confirmed by the powder X-ray diffraction (XRD) (Figure S6a in SI) and transmission electron microscopy 12371

DOI: 10.1021/jacs.7b06337 J. Am. Chem. Soc. 2017, 139, 12370−12373

Communication

Journal of the American Chemical Society (TEM) (Figure S7 in SI). Scanning electron microscopy (SEM) images (Figures S6b−d in SI) show the α-MoB2 has many closely packed, microsized particles. The electrocatalytic activity of α-MoB2 toward HER in acidic solution was evaluated using a standard three-electrode cell comprising a carbon rod as a counter electrode (see Figures S8 and S9 and experimental details in SI). For comparison, the catalytic activity of a commercially available Pt/C (20 wt %) was also measured. Pt/C’s catalytic activity in acidic solution at both low and high current densities is comparable with those reported by others (Table S1). As shown in Figure 3a, α-MoB2 exhibits catalytic activity toward HER, giving current densities of 10 and 1000 mA/cm2 at

exhibits better catalytic activity for HER than nanostructured/ nanosized MoS2 materials, both at low and high current densities. Figure 3a also shows α-MoB2 exhibits significantly higher electrocatalytic activity toward HER than Pt/C in a wide current density range above ∼250 mA/cm2, although α-MoB2’s catalytic activity is lower than that of the latter for current densities smaller than 250 mA/cm2. For example, α-MoB2 requires a much lower overpotential, by ca. 564 mV, than Pt/C to achieve a current density of 1000 mA/cm2, but the former needs a larger overpotential, by ca. 120 mV, than the latter to generate a current density of 10 mA/cm2. It is well-known the commercially available benchmark electrocatalyst Pt/C has high intrinsic catalytic activity for HER. This can also be seen in our control experiment, where an extremely low overpotential is required to promote HER over it, especially at small current densities (e.g., 10 mA/cm2). However, Pt/C cannot deliver large current densities (several hundreds of mA/cm2 or larger) in 0.5 M H2SO4 solution as efficiently as it can give small current densities (Figure 3a), primarily due to the very strong “bubble effect” at large current densities.7 This is supported by the study conducted by Durst et al., in which Pt/C was reported to give a current density of 500 mA/cm2 at an overpotantial of ∼50 mV in a bubble effect-free (or mass transport-free) proton exchange membrane (PEM) fuel cell.8 The results also suggest α-MoB2’s good mass transport property, besides its high conductivity and large density of catalytic active sites, should be among the main reasons it can efficiently deliver such large current densities. To assess the catalytic stability of α-MoB2 during HER, first, a polarization curve over it between 0 and −0.6 V for 15 000 cycles was measured. As shown in Figure 3b, the material gives almost the same polarization curve as the original one even after such many cycles, demonstrating its excellent stability during the electrocatalytic reaction. Second, a multistep chronoamperometric curve was recorded over the material in a wide overpotential range (from 200 to 550 mV) with an increment of 50 mV (Figure 3c). The result shows the current density remains stable at each overpotential in the test range, confirming α-MoB2’s high catalytic activity as well as stability in a current density range (from 0 to 2000 mA/cm2). Third, the current density versus time (I−t) curve for the α-MoB2-catalyzed HER at a large current density of ca. 1500 mA/cm2 (Figure 3d) reveals the catalyst retains its electrocatalytic activity for, at least, 3600 min. Furthermore, the crystal structure of the α-MoB2 remains intact even after such a long electrocatalytic reaction, as revealed by XRD and TEM (Figure S11 in SI). The results demonstrate αMoB2 is not only a highly active but also a very stable electrocatalyst for HER. Figure S12 shows plots of overpotential versus log(j) for HER over α-MoB2 and Pt/C in a current density range of 0 to 100 mA/cm2. From the extrapolation of the linear region of the graph, the exchange current densities for α-MoB2 and Pt/C are 0.132 and 0.989 mA/cm2, respectively. However, the exchange current density of α-MoB2 is 1−2 orders of magnitude larger than some notable nanostructured MoS2 materials (Table S2 and Figure S13 in SI), suggesting α-MoB2 can be a promising, active nonprecious HER electrocatalyst. Moreover, the Tafel slope value of 74.2 mV/dec for α-MoB2 indicates the HER process in the presence of α-MoB2 most likely goes sequentially through the Volmer (H+ + e− → H*) and Heyrovsky (H+ + e− + H* → H2) steps. To deduce further how α-MoB2’s structure renders superior catalytic activity to the material, we prepared several other relevant molybdenum boride phases with different structures or

Figure 3. (a) Linear sweep voltammetry (LSV) curves for HER over αMoB2 and Pt/C (20 wt %) in 0.5 M H2SO4 solution, with inset showing LSV curves at low potentials. (b) LSV curves for HER over α-MoB2 before and after potential sweeps between 0 and −0.6 V for 15 000 cycles. (c) Multistep chronoamperometric curves obtained with αMoB2 at different overpotentials, starting at 200 mV and ending at 550 mV with an increment of 50 mV every 1000 s. (d) Current density versus time curve obtained over α-MoB2 at a current density of 1500 mA/cm2 over 3600 min long catalytic HER. Note the current density is normalized with the electrode’s geometric surface area.

overpotentials of ca. 149 and 334 mV (after iR correction4c), respectively. During electrocatalysis, the large volume of H2 gas generated on α-MoB2 quickly comes off of the electrode as soon as it forms even at 1000 mA/cm2 (see video in SI). This observation suggests α-MoB2 has surfaces conducive for the H2 gas forming by the HER to quickly leave the reaction mixture. Additionally, α-MoB2 gives nearly 100% Faradaic yield during HER (Figure S10 in SI). This result indicates the obtained current is exclusively due to the catalytic HER. α-MoB2’s electrocatalytic activity at both low and high current densities can be appreciated by comparing its catalytic properties with those of MoS2 (Table S2). Reports on the electrocatalytic activity of MoS2 show (i) microsized (or bulk) MoS2 particles are inactive for HER, or only nanosized MoS2 materials exhibit good activity for the reaction; (ii) even notable MoS2 nanomaterials usually require overpotentials as high as 150−250 mV to give a current density of 10 mA/cm2; and (iii) no MoS2 nanomaterial that can deliver a large current density (>1000 mA/cm2), even at large overpotential, has been reported. Even the reported nanostructured MoS2 thin film by Jiang et al.,7 which has favorable surfaces that can help with the removal of the as-formed H2 bubbles, just like those on α-MoB2, gives a current density of only 110 mA/cm2 at an overpotential of 400 mV. The above comparisons reveal, despite its microsized particle, α-MoB2 12372

DOI: 10.1021/jacs.7b06337 J. Am. Chem. Soc. 2017, 139, 12370−12373

Journal of the American Chemical Society



compositions, including β-MoB2, MoB and Mo2B, and then compared their catalytic properties with that of α-MoB2. The details of β-MoB2, MoB and Mo2B as well as characterization results of their structures and crystal structures are provided in SI, Figure S14 and Tables S3 and S4. The most significant structural difference between the materials is the presence of different types of covalent linkage patterns between the B atoms in them (Figures S15 and S16 in SI). In the case of α-MoB2, the B atoms are arranged into planar, graphene-like borophene structures, whereas in the case of β-MoB2, the B atoms have two bonding patterns: planar borophene and puckered boron sheets. Although the B atoms in MoB are arranged into some onedimensional, zigzag chains, those in Mo2B are all isolated, and not covalently bonded. We then compared the electrocatalytic activities toward HER of all the molybdenum boride phases among each other. Figures S16 and S17 show the catalytic activities of the materials decrease in the order of α-MoB2 ≫ β-MoB2 > MoB > Mo2B at both low and high current densities. Electrochemical impedance spectra (Figure S18 in SI) also reveal α-MoB2 has much smaller electrontransfer resistance than the other three materials, further justifying α-MoB2 has the highest catalytic activity. It is worth noting the electrocatalytic activity of α-MoB2 is about 10.3 times higher than that of β-MoB2, whereas the electrochemical surface area of α-MoB2 is ca. 3.4 times higher than that of β-MoB2 (Figure S19 in SI). This result indicates the higher active area present in α-MoB2 is not the only reason αMoB2 shows superior catalytic activity compared with β-MoB2. The high intrinsic catalytic activity of the active sites in α-MoB2 must have been the major contributor to the superior electrocatalytic activity exhibited by this material in HER. These results, especially the difference in terms of catalytic activity between α-MoB2 and β-MoB2, coupled with the differences in structures of the materials, demonstrate the presence of borophene subunits in α-MoB2 is crucial for the realization of ultrahigh catalytic activity at large current densities in this material. Because puckered boron sheets, besides planar borophenes, exist in β-MoB2, the 3D Mo-metal framework, which is found throughout the crystal structure of α-MoB2, cannot form in β-MoB2 (Figure S20 in SI). This is likely the structural basis for the difference in catalytic activity between the two materials. Furthermore, despite hydrogen interacts strongly with the metallic Mo surface, making the subsequent protonation or recombination step difficult during HER,9 the presence of borophene subunits helps with optimization of the electronic structure of Mo-framework and the reduction of Mo−H bond energy in α-MoB2, resulting in more active catalytic sites for HER in this material. In summary, we demonstrated α-MoB2 efficiently catalyzes the hydrogen evolution reaction, even at large current densities. We also demonstrated α-MoB2 was able to do so because it fulfills the following three important criteria that are desirable for efficient electrocatalysis of HER: (1) it has a good electronic conductivity, enabling its catalytically active sites to be easily accessible to the electrons coming from the external circuit; (2) it has a large density of highly active electrocatalytic sites; and (3) it has good mass transport properties, and thus allows a sufficient supply of reactants to reach its catalytic sites and gaseous products forming there due to HER to diffuse out quickly. The results presented can give a fresh impetus to the search or design of non-noble metal catalytic materials with high catalytic performances for practical water electrolysis.

Communication

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b06337. Experimental details and supporting results (PDF) Video of H2 gas generated on α-MoB2 off electrode (AVI)



AUTHOR INFORMATION

Corresponding Authors

*[email protected] *[email protected] *[email protected] *[email protected] ORCID

Tewodros Asefa: 0000-0001-8634-5437 Xiaoxin Zou: 0000-0003-4143-9274 Author Contributions ∇

These authors contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS X. Zou thanks the support of NSFC 21401066 and National Key R&D Program of China, Grant No. 2017YFA0207800. W. Chen and G. Yu thank the financial support from NSFC 21673093 and 21673094, Jilin Province S&T Development Plan 20170101175JC and S&T Research Program of Education Department JJKH20170780KJ. Y. Chen thanks the support of NSFC 21601062. D. Wang acknowledges the support of NBRP 2013CB632403. T. Asefa acknowledges the financial assistance of the US NSF, Grant No. DMR-1508611.



REFERENCES

(1) For example: (a) Vrubel, H.; Hu, X. Angew. Chem., Int. Ed. 2012, 51, 12703−12706. (b) Cao, B.; Veith, G. M.; Neuefeind, J. C.; Adzic, R. R.; Khalifah, P. G. J. Am. Chem. Soc. 2013, 135, 19186−19192. (2) Morales-Guio, C. G.; Stern, L.-A.; Hu, X. Chem. Soc. Rev. 2014, 43, 6555−6569. (3) Zhu, L.; Lin, H.; Li, Y.; Liao, F.; Lifshitz, Y.; Sheng, M.; Lee, S.-T.; Shao, M. Nat. Commun. 2016, 7, 12272. (4) For example: (a) Fan, X.; Xu, P.; Li, Y. C.; Zhou, D.; Sun, Y.; Nguyen, M. A. T.; Terrones, M.; Mallouk, T. E. J. Am. Chem. Soc. 2016, 138, 5143−5149. (b) Hinnemann, B.; Moses, P. G.; Bonde, J.; Jørgensen, K. P.; Nielsen, J. H.; Horch, S.; Chorkendorff, I.; Nørskov, J. K. J. Am. Chem. Soc. 2005, 127, 5308−5309. (c) Lukowski, M. A.; Daniel, A. S.; Meng, F.; Forticaux, A.; Li, L.; Jin, S. J. Am. Chem. Soc. 2013, 135, 10274−10277. (d) Li, H.; Tsai, C.; Koh, A. L.; Cai, L.; Contryman, A. W.; Fragapane, A. H.; Zhao, J.; Han, H. S.; Manoharan, H. C.; Abild-Pedersen, F.; Nørskov, J. K.; Zheng, X. Nat. Mater. 2016, 15, 48−53. (5) Piazza, Z. A.; Hu, H.-S.; Li, W.-L.; Zhao, Y.-F.; Li, J.; Wang, L.-S. Nat. Commun. 2014, 5, 3113 and references therein.. (6) Skúlason, E.; Tripkovic, V.; Björketun, M. E.; Gudmundsdóttir, S.; Karlberg, G.; Rossmeisl, J.; Bligaard, T.; Jónsson, H.; Nørskov, J. K. J. Phys. Chem. C 2010, 114, 18182−18197. (7) Lu, Z.; Zhu, W.; Yu, X.; Zhang, H.; Li, Y.; Sun, X.; Wang, X.; Wang, H.; Wang, J.; Luo, J.; Lei, X.; Jiang, L. Adv. Mater. 2014, 26, 2683−2687. (8) Durst, J.; Siebel, A.; Simon, C.; Hasché, F.; Herranz, J.; Gasteiger, H. A. Energy Environ. Sci. 2014, 7, 2255−2260. (9) Cook, T. R.; Dogutan, D. K.; Reece, S. Y.; Surendranath, Y.; Teets, T. S.; Nocera, D. G. Chem. Rev. 2010, 110, 6474−6502.

12373

DOI: 10.1021/jacs.7b06337 J. Am. Chem. Soc. 2017, 139, 12370−12373