Tunable intrinsic strain in two-dimensional transition metal electrocatalysts

See allHide authors and affiliations

Science  22 Feb 2019:
Vol. 363, Issue 6429, pp. 870-874
DOI: 10.1126/science.aat8051

Harnessing self-tuned strain

Strain can modify the electronic properties of a metal and has provided a method for enhancing electrocatalytic activity. For practical catalysts, nanomaterials with high surface areas are needed. However, for nanoparticles, strain is often induced with overlayers (adsorbates or heteroatoms) that can undergo reconstruction during operation that releases the induced strain. Wang et al. show that freestanding palladium nanosheets (three to five monolayers thick) form with an internal compressive strain of 1 to 2% and can be much more active for both the oxygen and hydrogen evolution reactions under alkaline conditions compared with nanoparticles.

Science, this issue p. 870


Tuning surface strain is a powerful strategy for tailoring the reactivity of metal catalysts. Traditionally, surface strain is imposed by external stress from a heterogeneous substrate, but the effect is often obscured by interfacial reconstructions and nanocatalyst geometries. Here, we report on a strategy to resolve these problems by exploiting intrinsic surface stresses in two-dimensional transition metal nanosheets. Density functional theory calculations indicate that attractive interactions between surface atoms lead to tensile surface stresses that exert a pressure on the order of 105 atmospheres on the surface atoms and impart up to 10% compressive strain, with the exact magnitude inversely proportional to the nanosheet thickness. Atomic-level control of thickness thus enables generation and fine-tuning of intrinsic strain to optimize catalytic reactivity, which was confirmed experimentally on Pd(110) nanosheets for the oxygen reduction and hydrogen evolution reactions, with activity enhancements that were more than an order of magnitude greater than those of their nanoparticle counterparts.

Recent advances in controlled growth of nanomaterials (16), combined with elucidation of governing principles of catalysis from fundamental research on well-defined model systems, have led to a blossoming of the field of rational catalyst design (7, 8). One particularly successful strategy has been to tune surface strain (911) and thereby modulate the surface electronic structure and energetics of reaction intermediates, leading to a fundamental understanding and identification of catalysts with improved activity (1223). For the well-known example of the oxygen reduction reaction (ORR) on platinum, previous studies indicated that a 1% compressive strain could improve the activity by more than 300% (13, 15, 19, 20).

Although strain tuning of catalysts has shown promise for catalytic enhancement, its application has been limited by both practical and fundamental considerations. First, the effect of strain on high–surface area nanoparticle catalysts is often obscured by nanoparticle shape and by the presence of undercoordinated defects (13, 19, 21, 24, 25), whereas model single-crystal and polycrystalline catalysts, for which high specific activities are achievable by strain tuning (15, 16, 19, 20), have prohibitively small mass activities and are not of practical interest. It is therefore desirable to develop strain tuning strategies for nanomaterials that preserve the intrinsic activity of single crystals and also possess surface areas and mass activities that exceed those of nanoparticles. Second, existing methods for strain tuning of catalysts often rely on overlaying a catalyst layer on a substrate, using one of a large number of techniques that have been developed, yielding overlayers where strain may be tuned either continuously or discretely (10, 13, 1522, 2628). These approaches result in the formation of overlayer-substrate interfaces that can be subject to mechanical instabilities and strain relaxation (13, 17, 20, 21). Additionally, during prolonged usage, dissolution may occur around the interfaces (15, 19), and for catalysis in alkaline, base metals from substrates may form surface (hydroxy)oxide species that block active sites (29). Further, the presence of heterointerfaces convolutes multiple physical factors, such as substrate ligand effects and production of surface alloys, that may complicate the optimization of catalyst properties by strain tuning. Finally, if the substrates are composed of precious metals, they may limit the reactivity of the catalyst per mass of precious metal, which is undesirable for cost-effectiveness. These considerations point to the importance of developing strain tuning strategies that do not rely on manipulation of a catalyst-substrate interface.

Here, we report on the utilization of surface stress to drive intrinsic strain in freestanding metal nanosheets. These two-dimensional (2D) nanostructures do not contain catalyst-substrate interfaces, and they integrate the high specific activity of single-crystal catalysts with the high surface areas of traditional nanoparticles. First, we determined the relationship between the surface stress, the resulting strain, and the nanosheet thickness through density functional theory (DFT) analysis. We demonstrated that nanosheets with thicknesses of 1 to 12 monolayers (ML) exhibit tunable strains that are up to 10% and are inversely proportional to the thickness, with the exact magnitudes also depending on the elemental identity of the transition metal and the surface orientation. We then validated the predicted intrinsic strain levels by synthesizing ultrathin Pd nanosheets with atomic-level control over the thickness and by using aberration-corrected high-resolution transmission electron microscopy (HRTEM) imaging to characterize the lattice spacing. We next exploited the strain to enhance the rates of the electrochemical ORR and hydrogen evolution reaction (HER). We demonstrated that in both alkaline and acidic environments Pd nanosheets can enhance reaction rates by more than an order of magnitude above rates with Pd and Pt nanoparticles, suggesting that generating and tuning intrinsic strain in 2D nanosheets can be a powerful strategy for the design and development of advanced catalytic materials.

Cleavage of bulk metal atoms to create a surface often leads to charge redistribution and attractive interactions between surface atoms. These attractive interactions are manifested as a tensile surface stress that, in turn, depends linearly on the bulk modulus of the metal, with notable exceptions for Pt and Au (Fig. 1). For platinum group metals and coinage metals, intrinsic tensile surface stresses generally induce a pressure on the order of 105 atm on the surface atoms (Fig. 1C and figs. S1 to S4), which provides a strong driving force for surface contraction and reduction of the surface energy. Although this surface pressure has little impact on the lattice of single crystals with macroscopic thicknesses, it can markedly affect the structure of 2D ultrathin films and lead to compressive in-plane strain (Fig. 1, A and D), accompanied by a corresponding change in the distance between metal layers, which can be approximated by the Poisson ratio for the metals (fig. S5). The compressive strain strongly depends on the slab thickness and on the nature of the transition metal, as shown in Fig. 1, E to G, and figs. S6 and S7 for freestanding transition metal surfaces with face-centered cubic (111) [fcc(111)], hexagonally close-packed (0001) [hcp(0001)], fcc(100), and fcc(110) symmetries and thicknesses varying from 1 ML (~0.2 nm) to 12 ML (~2.5 nm).

Fig. 1 The relationship between bulk modulus (B), surface stress (τ), and lattice strain (ε) of platinum group and coinage metal slabs.

(A) Mechanism of the generation of intrinsic strain in 2D transition metal nanosheets. h is the height of an atomic layer. (B) Surface stress of fcc(111) and hcp(111) surfaces compared to the bulk modulus. (C) Ratio of the pressure (p) of close-packed surfaces, induced by surface stress, and the bulk modulus for 11 transition metals. (D) Potential energy profile of strained versus unstrained Pd(111) slabs with thicknesses of 8 ML, 4 ML, and 2 ML. (E to G) Intrinsic in-plane strain of fcc(111), fcc(100), and the Embedded Image direction of unreconstructed fcc(110) slabs, with thicknesses from 1 ML to 12 ML. See the supplementary materials for results for the reconstructed fcc(110) surface.

For a given element, the overall strain generally increases with decreasing slab thickness, and the magnitude is inversely proportional to the slab thickness, except for the thinnest slabs (~1 to 2 ML), for which quantum size effects are important. For a given thickness, differences by a factor between two and five result from changes in the transition metal element. For example, for hcp(0001) and fcc(111) surfaces (Fig. 1E), slabs with thicknesses of 2 nm (9 to 10 ML), 1 nm (4 to 5 ML), and 0.5 nm (2 ML) exhibit compressive strains with magnitudes of 0.3 to 1.4%, 0.8 to 3.1%, and 2.2 to 5.7%, respectively. For (100) surfaces, the corresponding magnitudes are 0.5 to 2.0%, 1.0 to 3.8%, and 2.7 to 6.3%, respectively (Fig. 1F). Finally, for unreconstructed (110) surfaces, compressive strain magnitudes of 0.1 to 1.4%, 0.4 to 2.6%, and 0.4 to 5.1% were determined in the Embedded Image direction, which is parallel to step edges (Fig. 1G), with similar results for (110) surfaces with missing row reconstructions (fig. S7).

In general, Pt and Au exhibit the largest strains, whereas the values for Pd, which was chosen for further experimental study, are close to the average. The elemental dependence can be explained by considering the ratio between the intrinsic surface pressure and the corresponding bulk modulus (Fig. 1C). On close-packed surfaces, Pt and Au have an average ratio of 8.3%, leading to high surface strains, whereas the rest of the metals have an average value of 5.5%. The larger ratios for Pt and Au are, in turn, related to their stronger relativistic effects than observed with lighter elements (fig. S8) (30, 31). Due to increased orbital contraction (fig. S8), these effects have a larger impact on undercoordinated surface atoms than on bulk atoms. Similar trends in the dependence of the strain on elemental composition and relativistic effects were also found for (100) and (110) surfaces (Fig. 1, E to G, and fig. S7).

To verify the predicted intrinsic strain tuning in 2D metal nanostructures, Pd nanosheets were experimentally synthesized by using CO as both the reducing agent and stabilizing ligand (2, 32). By varying the CO sources and the synthesis temperature, nanosheets were synthesized with thicknesses of 3 ± 1 ML, 5 ± 1 ML, and 8 ± 1 ML (± standard deviations; these sheet thicknesses are denoted 3 ML, 5 ML, and 8 ML below) and with corresponding edge lengths of 120 to 260 nm, 50 to 150 nm, and ~20 nm, respectively (see Fig. 2D and the supplementary materials for additional details). Aberration-corrected HRTEM imaging with spatial resolution of 0.7 Å was used to analyze the in-plane strain in these nanosheets (Fig. 2, E to K). Fourier transform analysis of the image reveals that the nanosheets adopt a (110) basal plane (Fig. 2, E to G, and fig. S18), distinguishing these nanosheets from previously reported structures that were grown using strong stabilizing ligands such as tetrabutylammonium bromide (6, 33), which is difficult to remove and can be detrimental to electrocatalytic performance (2, 32). A 2D template-matching method (figs. S20 to S26) further permitted determination of the lattice constants and the lateral strain (34, 35). The lattice parameters in the 3-ML Pd nanosheets, shown in Fig. 2I, were determined to be a = 0.270 nm and b = 0.235 nm, with an angle of 124.7° between them, which correspond to the unit vectors along the norm vector of Embedded Image and Embedded Image planes projected along the (110) orientation of the fcc Pd lattice. The lattice parameters were measured to be a = 0.272 nm and b = 0.235 nm for 5-ML nanosheets (Fig. 2J) and a = 0.274 nm and b = 0.237 nm for 8-ML sheets (Fig. 2K).

Compared to the bulk values (a = 0.275, b = 0.238), the 8-, 5-, and 3-ML nanosheets possess average compressive strains of 0.3, 1.2, and 1.5%, respectively (Fig. 2L). These results clearly show that the extent of compressive strain increases as the thickness of the 2D nanosheets decreases. Moreover, the trend of the measured strain values matches the predicted values from calculations for Pd, for which the average strains are 0.35%, 0.92%, and 2.50% for 8-ML, 5-ML, and 3-ML slabs, respectively (Fig. 1G and fig. S6). We emphasize that these Pd nanosheets have high surface areas [for example, ~74 m2/g for 5-ML nanosheets (table S3)] but nevertheless expose extended, low–Miller index surfaces, as is the case for single crystals. The demonstrated tuning of intrinsic strain thus represents a promising approach toward modulation of surface reactivity to produce active, practical catalytic materials.

Fig. 2 Structural characterization of Pd nanosheets.

(A to C) TEM images of as-prepared Pd nanosheets (NSs) with average thicknesses of 3 ML (A), 5 ML (B), and 8 ML (C), with insets depicting typical structures. (D) Size and thickness distribution of Pd nanosheets (see figs. S15 to S17 for more details). Error bars indicate SD. Before electrochemical testing and lattice measurements, nanosheets are dispersed on high–surface area carbon, followed by low-temperature annealing to remove organic ligands, which leads to loss of stacked structures (fig. S14). Panels (E) to (L) show results of the dispersed nanosheets. (E to G) Aberration-corrected HRTEM images of Pd nanosheets with average thicknesses of 3 ML (E), 5 ML (F), and 8 ML (G) on a carbon support. (H) The intensity profile and calculated average d-spacing of (111) planes from the nanosheets in (E) to (G) and Pd nanoparticles (fig. S19). The area and the direction of the measurement are indicated by yellow, green, magenta, and red lines in (E), (F), and (G) and fig. S19, respectively. Two adjacent (111) planes, from which the first two peaks in the intensity profile are obtained, are indicated by two parallel lines. a. u., arbitrary units. (I to K) Illustration of the template-matching method used in the measurement of the lattice strain, which is effected by superimposing the Pd bulk lattice (yellow dots) on the HRTEM images of Pd nanosheets of 3 ML (I), 5 ML (J), and 8 ML (K). The yellow circles, in turn, are reference points that indicate the measured atomic positions along Embedded Image and Embedded Image directions on the nanosheets. These points also superimpose directly on the bulk lattice points. The adjacent nanosheet atoms are indicated with blue circles. The mismatch between the blue circles and the yellow dots shows the degree of lattice contraction. (L) The measured strain of Pd nanosheets using the template-matching method on the HRTEM images in (I) to (K). To enable a direct comparison to be made with the DFT calculations, the strain in the Embedded Image direction is projected onto the [001] direction. Error bars indicate SD.

It has been well established that strain modifies the electronic structure and reactivity of catalyst surfaces (9, 10, 13, 15, 17, 19, 20, 22, 28, 36). To probe these effects in 2D nanosheets, the d-band center and the surface energy of the strained slabs were calculated and compared with corresponding unstrained slabs. There is generally a downshift of the d-band center and a decrease of the surface energy for thinner slabs with larger compressive in-plane strain, regardless of the surface structure [fcc(111), hcp(0001), fcc(110), or fcc(100)] or elemental identity of the metal (figs. S27 to S29). These results suggest, in turn, that the surfaces interact less strongly with adsorbates as the thickness decreases and the in-plane compression increases (9, 10). This prediction is consistent with calculated weakening of the adsorption energy of atomic hydrogen and oxygen (Fig. 3A and fig. S30), which are widely used as descriptors for hydrogen and oxygen redox reactions, respectively (37, 38). For example, for O adsorption on fcc(111) and hcp(0001) surfaces, the average weakening is 0.05, 0.10, and 0.25 eV per oxygen atom for slabs with thicknesses of around 2, 1, and 0.5 nm, respectively (Fig. 3A); for a few metals (in particular, Pt and Au), notable adsorption energy oscillations due to quantum size effects can also be seen for slabs of fewer than ~6 ML (39). As with the trends in surface strain, the weakening of adsorption is dependent on the nature of the substrates, with the effect on Pt and Au being generally three to six times as large as that on the other transition metals and the effect on Pd being close to the average. The trend in the weakening of the adsorption therefore holds for a broad spectrum of metals and different surfaces. The continuously tunable strain and adsorption properties thus allow for the design of optimal catalytic surfaces for given reactions based on the freestanding 2D nanosheets.

Fig. 3 The dependence of adsorption energies and ORR activity on the thickness of nanosheets with intrinsic strain.

(A) Adsorption energy shift (ΔEad) of atomic oxygen on strained fcc(111) and hcp(0001) slabs, compared with adsorption energies on corresponding close-packed single-crystal surfaces. (B) Predicted ORR overpotential of Pd(110) single-crystal and strained nanosheets with various thicknesses. The dashed line in (B) is fit based on a strained single-crystal surface (see the supplementary materials for details).

According to the Sabatier principle, the interactions between an optimal catalyst and reaction intermediates should be neither too strong nor too weak, suggesting the presence of a maximum in catalytic activity at some point in the thickness-strain continuum. Because transition metal surfaces often bind too strongly to reaction intermediates for common electrocatalytic reactions, such as HER and ORR (13, 15, 19, 21, 22, 40, 41), weakening the adsorbate interactions with the surface via tuning of strain is a useful strategy to improve the activity of transition metals for these reactions. To illustrate these relationships for the strained 2D nanosheets, we have performed both experimental and theoretical investigations of Pd nanosheets as alkaline ORR electrocatalysts, because Pd is of particular interest for anion exchange membrane fuel cells, where it is considerably more stable than in acidic environments (42).

Experimentally, Pd nanosheets with (110) terminations are found to be considerably more active than Pd/C (palladium nanoparticles on a carbon support) and Pt/C (Pd/C serves as a practical, high–surface area reference material for the catalytic performance of nanosheets), with the ORR activity exhibiting a classic volcano relationship as the thickness decreases (Fig. 4, tables S3 to S7, and figs. S41 to S47). From the polarization curves (Fig. 4A), the half-wave potential is determined to be 0.90, 0.95, and 0.93 V for the 8-, 5-, and 3-ML nanosheets, respectively, compared to 0.86 V for Pd/C. The most active 5-ML nanosheets achieve a specific activity of 0.70 mA/cm2 at 0.95 V [10.91 mA/cm2 at 0.9 V (table S3)], representing an improvement factor of >18, versus Pd/C with a specific activity of 0.04 mA/cm2 at 0.95 V (0.41 mA/cm2 at 0.9 V) (Fig. 4B and tables S3 and S5). The 3-ML nanosheets are the next most active, with 0.60 mA/cm2 at 0.95 V (10.42 mA/cm2 at 0.9 V), whereas the 8-ML nanosheets have considerably lower specific activities of 0.07 mA/cm2 at 0.95 V (1.07 mA/cm2 at 0.9 V). Similar trends were also observed in mass activity, with enhancement factors of up to 26 and 47 for 5-ML nanosheets at 0.95 V (0.52 mA/mg of Pd) and 0.9 V (8.02 mA/mg of Pd), respectively, compared to Pd/C (0.02 A/mg of Pd at 0.95 V and 0.17 A/mg of Pd at 0.9 V).

Fig. 4 Electrochemical activity of Pd nanosheet catalysts on carbon.

(A) ORR polarization curves of Pd nanoparticles, as well as Pd nanosheets with average thickness of 3 ML, 5 ML, and 8 ML in 0.1 M KOH (inset shows the halfwave potential). RHE, reversible hydrogen electrode. (B and C) Specific activity and mass activity of ORR at 0.95 V (versus RHE) in 0.1 M KOH and the corresponding improvement factors compared to those for Pd nanoparticles. (D) HER overpotential of Pd nanoparticles, as well as Pd nanosheets with average thickness of 3 ML, 5 ML, and 8 ML at 5 mA/cm2 in 0.1 M KOH or 0.1 M HClO4. (E and F) Specific activity and mass activity of ORR at 0.95 V (versus RHE) in 0.1 M HClO4 (E) and the corresponding improvement factors compared to those for Pd nanoparticles (F).

The relationship between the thickness of Pd nanosheets and the ORR activity is further revealed by calculation of the free-energy barriers of the ORR on (110)-terminated Pd nanosheets with thicknesses of 3 to 8 ML. As discussed above, each such surface has a different extent of compressive strain, with the thinner slabs experiencing the greatest compression (fig. S7). The unstrained Pd surface binds to oxygen too strongly, which limits the ORR kinetics due to the slow oxidative desorption of hydroxide (OHad); this result is very similar to the case with Pt surfaces (4144). As the O* binding is weakened on the thinner, more compressed Pd nanosheets, the predicted activity gradually increases with decreasing slab thickness and increasing compressive strain, with maximal reactivity predicted for nanosheets with thicknesses of ~4 ML (Fig. 3B). These predictions provide a compact explanation for the enhanced alkaline ORR activities observed in the electrocatalytic studies (Fig. 4). Further, as also predicted by the DFT analyses, we expect that these structure-property relationships will be applicable to 2D nanosheets of other metals and with different lattice orientations.

In addition to alkaline electrolytes, we have conducted measurements of the ORR on the Pd nanosheets in acid. We note that, although Pd is insufficiently stable in acid and may have limited potential for practical use in proton exchange membrane fuel cells, the predictions suggest that similar activity enhancements will be observed in acidic media. The measured enhancements of specific activity in 0.1 M HClO4 at 0.95 V are 14, 10, and 5 for 5-ML, 3-ML, and 8-ML nanosheets (Fig. 4C and table S7), whereas the corresponding improvement factors in 0.1 M KOH are 18, 15, and 2 (table S5), confirming that trends in ORR activities are consistent for Pd nanosheets in acid and in base. We further note that similar trends in activity improvement should be observable for other electrocatalytic reactions that are sensitive to surface strain effects. As confirmation of this principle, we found similar enhancements in reactivity trends for HER in both acid and base, with the activity following the order 5-ML nanosheets > 3-ML nanosheets > 8-ML nanosheets > nanoparticles (Fig. 4, figs. S48 and S49, and table S8).

The considerations discussed above demonstrate that metal nanosheet catalytic properties may be tuned via manipulation of intrinsic strain for multiple electrocatalytic reactions in both alkaline and acidic media. These results, in turn, suggest that generating and tuning of intrinsic surface strain represent a powerful and general strategy to engineer nanocatalysts with simultaneously enhanced specific activity and mass activity for a spectrum of electrocatalytic processes.

Supplementary Materials

Materials and Methods

Figs. S1 to S49

Tables S1 to S9

References (4562)

References and Notes

Acknowledgments: Funding: Work at Purdue was supported through the Office of Science, Office of Basic Energy Sciences, Chemical, Biological, and Geosciences Division under DE-SC0010379 (J.G.) and by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, under DE-EE0007270 (Z.Z.). Z.Z. and J.G. also gratefully acknowledge the computing resources provided by the Center for Nanoscale Materials, which is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under E-AC02-06CH11357, as well as the computational resources through the National Energy Research Scientific Computing Center (NERSC). C.W. acknowledges support from the National Science Foundation (CBET-1437219) and the JHU Catalyst Award. W.G. and X.P. are supported by the National Science Foundation under grant numbers CBET 1159240, DMR-1420620, and DMR-1506535. TEM work was conducted using the facilities in the Irvine Materials Research Institute (IMRI) at the University of California, Irvine. Author contributions: Z.Z. and J.G. developed the strain tuning strategy. L.W. and C.W. conceived the idea of Pd nanosheet synthesis. Z.Z. performed strain and strain-property calculations. T.M. performed relativistic effect calculations. L.W., D.R., and M.G. performed the synthesis and electrochemical tests. W.G. and X.P. performed TEM measurements. Z.Z. and J.G. wrote the manuscript with input from L.W., W.G., and C.W. Competing interests: None declared. Data and materials availability: The data presented in this paper are available in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article