On the role of surface diffusion in determining theshape or morphology of noble-metal nanocrystalsXiaohu Xiaa,1, Shuifen Xiea,1, Maochang Liua,1, Hsin-Chieh Pengb, Ning Luc, Jinguo Wangc, Moon J. Kimc,d,and Younan Xiaa,b,e,2

aThe Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, GA 30332; bSchool ofChemistry and Biochemistry and eSchool of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332; cDepartment ofMaterials Science and Engineering, University of Texas at Dallas, Richardson, TX 75080; and dDepartment of Nanobio Materials and Electronics, World ClassUniversity, Gwangju Institute of Science and Technology, Gwangju 500-712, Korea

Edited by Gabor A. Somorjai, University of California, Berkeley, CA, and approved March 19, 2013 (received for review December 18, 2012)

Controlling the shapeormorphologyofmetal nanocrystals is centralto the realization of their many applications in catalysis, plasmonics,and electronics. In one of the approaches, themetal nanocrystals aregrown from seeds of certain crystallinity through the addition ofatomic species. In this case, manipulating the rates at which theatomic species are added onto different crystallographic planes ofa seed has been actively explored to control the growth pattern ofa seed and thereby the shape or morphology taken by the finalproduct. Upon deposition, however, the adsorbed atoms (adatoms)may not stay at the same siteswhere the depositions occur. Instead,they can migrate to other sites on the seed owing to the involve-ment of surface diffusion, and this could lead to unexpecteddeviations from a desired growth pathway. Herein, we demon-strated that the growth pathway of a seed is indeed determined bythe ratio between the rates for atom deposition and surfacediffusion. Our result suggests that surface diffusion needs to betaken into account when controlling the shape or morphology ofmetal nanocrystals.

seeded growth | shape control | noble metals

Surface diffusion is a general process that involves themotion ofadsorbed atoms (adatoms), molecules, or atomic clusters on

the surface of a solid material (1, 2). Over the past decades, it hasemerged as an important concept in many areas of surface science,including catalysis, epitaxial growth, and electromigration of voids(3–7). Here, we demonstrated that surface diffusion also playsa pivotal role in determining the growth pathway of a seed and thusthe shape or morphology taken by the final product in a solution-phase synthesis of metal nanocrystals. Fig. 1 schematically illus-trates four possible pathways for the growth of a cubic seed. Asamodel system, we focused on Pd nanocubes with slight truncationat corners and edges, together with six side faces passivated bychemisorbedBr– ions. In the following discussion, we refer to themas “Pd cubic seeds” for simplicity. We chose them as seeds for twomajor reasons: (i) they had a well-defined shape, together witha set of low-index facets on the surface (8, 9); and (ii) their sidefaces are blocked by Br– ions to ensure selective deposition ofatoms onto the corner sites during seed-mediated growth (10–12).These two distinctive features allowed us to easily track the de-position of atoms and their surface diffusion during a growthprocess by analyzing the shape or morphology of the final product.The newly formed Pd atoms resulting from the reduction of a Pd

precursor are expected to deposit at the corners of a cubic seedbecause the side faces are blocked by the chemisorbed Br– ions(Fig. 1A, 1). Upon deposition, there will be two different optionsfor these adatoms: staying at the corner sites or migrating to othersites, including edges and side faces, through surface diffusion(Fig. 1A, 2 and 3). It should be pointed out that only surface dif-fusion was allowed here to move atoms from corners to edges andside faces of a seed during growth. Other mechanisms such asOstwald ripening (13) were not considered because the sidefaces of a seed were blocked by Br– ions. The growth pathway of

a cubic seed is determined by the ratio between the rates foratom deposition and surface diffusion (Vdeposition/Vdiffusion).WhenVdeposition/Vdiffusion >> 1, surface diffusion can be ignored andthereby the growth will be largely confined to the corner sites alongthe <111> directions, resulting in the formation of Pd octapods(Fig. 1B, i). On the contrary, when Vdeposition/Vdiffusion << 1, thegrowth will be dominated by surface diffusion and be switched tothe <100> and <110> directions as most of the adatoms at thecorners can quickly migrate to edges and side faces of a cubic seed,promoting the formation of a cuboctahedron as the final product(Fig. 1B, iv). Similar arguments can also be applied to the situationswhere the ratios of Vdeposition/Vdiffusion are between these twoextremes. For example, when Vdeposition/Vdiffusion is slightly largerthan 1, a small portion of the adatoms at the corners will migrateto the edges (which are relatively more active than the side facesdue to a lower coverage density for the Br– ions) of a seed, leadingto the formation of Pd concave nanocubes (Fig. 1B, ii). WhenVdeposition/Vdiffusion is slightly less than 1, some of the adatoms willstay at corners while the rest can diffuse to both edges and sidefaces of a seed. As a result, the final product will be an enlarged Pdnanocube with slight truncations at the corners (Fig. 1B, iii).We conducted a set of experiments based on seed-mediated

growth to validate the proposedmechanisms. The growth involvedthe use of Pd nanocubes as seeds in an aqueous solution, withL-ascorbic acid (AA), Na2PdCl4, and poly(vinyl pyrrolidone) (PVP)serving as the reductant, Pd precursor, and stabilizer, respectively.In a standard synthesis, an aqueous Na2PdCl4 solution was injectedusing a syringe pump into an aqueous suspension containing AA,PVP, and Pd seeds that were hosted in a glass vial at room tem-perature (∼22 °C) under magnetic stirring (see Materials andMethods for experimental details). For this synthesis, Na2PdCl4 issupposed to be immediately reduced into Pd atoms by AA uponaddition into the reaction solution due to the strong reductionpower of AA (14, 15). As such, the concentration of the newlyformedPd atoms in the reaction solution and therebyVdeposition willbe mainly determined by the injection rate for Na2PdCl4 solutionthat can be readily controlled through the use of a syringe pump.Because surface diffusion is a thermally promoted process with itsrate increasing with temperature (2), Vdiffusion can be adjusted bypresetting the oil bath to a specific temperature. Collectively, theroles played by Vdiffusion and Vdeposition can be separated from eachother for investigation by varying the reaction conditions.

Author contributions: X.X., S.X., and Y.X. designed research; X.X., S.X., and M.L. per-formed research; N.L., J.W., and M.J.K. contributed new reagents/analytic tools; X.X.,S.X., M.L., H.-C.P., and Y.X. analyzed data; and X.X. and Y.X. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1X.X., S.X., and M.L. contributed equally to this work.2To whom correspondence should be addressed. E-mail: younan.xia@bme.gatech.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1222109110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1222109110 PNAS | April 23, 2013 | vol. 110 | no. 17 | 6669–6673

CHEM

ISTR

Y

Results and DiscussionValidation of the Proposed Mechanism. As shown by the trans-mission electron microscopy (TEM) images in Fig. S1, the Pdseeds were nanocubes with an average edge length of 15 nm, whichwere prepared in the presence of Br– as a capping agent asreported previously (16). A close examination indicates thata small portion of the cubes were slightly elongated to become barswith an average aspect-ratio of 1.1. Because their side faces werestill bound by {100} facets and covered by Br− ions, they shouldnot affect our discussion on surface diffusion as long as their aspectratios were close to 1. In our discussion, we also refer to them as“Pd cubic seeds” for simplicity. As indicated by the X-ray photo-electron spectroscopy (XPS) spectra shown in Fig. S2 for Pd cubesand octahedrons, only Pd{100} facets or the side faces of the Pdcubic seeds were covered by chemisorbed Br– ions.In the first set of experiments, we varied the reaction tempera-

ture while the injection rate for Na2PdCl4 solution was kept thesame as the standard synthesis (0.5 mL/h). In this case, the ratio ofVdeposition/Vdiffusion is expected to decrease with the increase ofreaction temperature because Vdeposition is fixed. If the proposedmechanisms are correct, Pd nanocrystals with all four differentshapes or morphologies shown in Fig. 1B, i–iv, should be obtainedas the reaction temperature is increased. As shown in Fig. 2, weindeed obtained Pd octapods (Fig. 2A), concave nanocubes (Fig.2B), nanocubes with slightly truncation at corners (Fig. 2C), andcuboctahedrons (Fig. 2D) from the same batch of Pd cubic seedswhen the reaction temperature was set to 0, 22, 50, and 75 °C,

respectively. The sizes (defined as the distance between two op-posite {100} facets of a nanocrystal as illustrated in Fig. S3) ofthese four different types of Pd nanocrystals were measured to be15.5, 18, 20.5, and 23 nm, respectively. Therefore, the growth of15-nm cubic seeds was mainly confined to the <111> directionsat 0 °C (Fig. 2A) and almost no deposition was found alongthe <100> directions. When the reaction temperature was in-creased from 0 to 75 °C (Fig. 2D), however, the growth wasswitched to the<100> and<110> directions because the size of thecuboctahedrons (i.e., 23 nm) was consistent with that of a cuboc-tahedron produced through exclusive growth on the {100} and{110} facets of a 15-nm cubic seed (Fig. S4) (17, 18). These resultsimply that, for the current system, the critical reaction temper-atures for all of the adatoms to stay at corners sites or diffuse to theedges and side faces of a cubic seed are 0 and 75 °C, respectively.In the second set of experiments, we adjusted the injection rate

for Na2PdCl4 solution while keeping the reaction temperaturethe same as the standard synthesis (22 °C). In these cases, the ratioof Vdeposition/Vdiffusion should increase as the injection rate forNa2PdCl4 solution increases because Vdiffusion is fixed. Accordingto the proposedmechanisms (Fig. 1), it is not difficult to argue thatPd nanocrystals with shapes and sizes similar to those shown inFig. 2 would also be obtained by decreasing the injection rate forNa2PdCl4 solution. As expected, Pd nanocrystals with shapesranging from octapods to concave nanocubes, nanocubes withslight truncation at corners, and cuboctahedrons were indeedobtained when the injection rate for Na2PdCl4 solution was de-creased from 1.5 to 0.75, 0.5, and 0.25 mL/h, respectively (Fig. S5).These results again supported our proposed mechanisms. Inter-estingly, the Pd octapods shown in Fig. 2Awere found to transforminto concave nanocubes (Fig. S6B) and then truncated nanocubes(Fig. S6C) after annealing in an aqueous solution at 80 °C for10 and 20 d, respectively. Similar shape transformations fromcubes to cuboctahedrons have also been observed in our previousstudies when Pd nanocubes were annealed in an electron micro-

Fig. 1. Effect of surface diffusion on the growth pattern of a Pd cubic seed.Schematics of (A) three different options for the Pd atoms added to the cornersite of a Pd cubic seed whose side faces are capped by Br− ions, and (B) dif-ferent pathways and the corresponding shapes or morphologies expected forthe growth of a Pd cubic seed under four different conditions. The size of Br−

ions was reduced relative to Pd atoms to show the surface structure clearly.

Fig. 2. Four distinctive types of Pd nanocrystals that were obtained at dif-ferent reaction temperatures. TEM images of Pd nanocrystals prepared usingthe standard procedure except for the variation in reaction temperature: (A)0, (B) 22, (C) 50, and (D) 75 °C. [Scale bar (applies to all images), 50 nm.] Insetsshow TEM images of individual nanocrystals at a higher magnification. [Scalebar (applies to all Insets), 5 nm.]

6670 | www.pnas.org/cgi/doi/10.1073/pnas.1222109110 Xia et al.

scope chamber at 500 °C for 1 h or when silver (Ag) nanocubeswere annealed in ethylene glycol at 145 °C for 2 h (19, 20). Theseresults were also consistent with the mechanisms described inFig. 1, where Vdeposition was essentially zero whereas Vdiffusion wasrelatively large.

Kinetics of Surface Diffusion. To better understand the role playedby surface diffusion in the shape evolution of a Pd cubic seed duringgrowth, we studied the kinetics of surface diffusion. By comparingthe TEM images of Pd nanocrystals obtained at different reactiontemperatures (Fig. 2) with that of the initial cubic seed (Fig. S1), wecould roughly estimate the total number (Ndiffusion; see Fig. S3 fordetailed calculation) of adatoms that had diffused to the edges andside faces of an individual cubic seed in the course of growth. Fig. 3shows a curve generated by plotting Ndiffusion as a function of re-action temperature (Temp, in degrees Celsius). It can be seen thatNdiffusion increased with the reaction temperature while the slope ofthe curve decreased as Temp was increased. From surface chem-istry studies, it is known that the kinetics of surface diffusion can beunderstood in terms of adatoms residing at adsorption sites ona surface and moving between adsorption sites by a hopping orjumping process (2). The diffusion coefficient (D) that measuresthe rate of spreading of an adatom across a surface can beexpressed as an Arrhenius-like equation as follows:

D=D0 exp�−Ediff=RT

�; [1]

where D0 is the diffusion preexponential factor, Ediff is the poten-tial energy barrier to diffusion, R is the ideal gas constant, and T isthe absolute temperature (in kelvin). At lower T or short diffusiontime (t), the number of adatoms that escape from the (111) face(i.e., Ndiffusion) will be proportional to sqrt(D), a quantity that ismore or less linear with T for certain values of Ediff. However,Ndiffusion will be saturated as T becomes large and/or the dimen-sions of the (111) face become much less than sqrt(2Dt). Remark-ably, such a trend for the calculated Ndiffusion was also reflectedexperimentally in Fig. 3: Ndiffusion increased with Temp at lowtemperatures (0–50 °C) and became saturated at high temper-atures (>75 °C). This observation implies that the atoms depos-ited on the edges and side faces of a cubic seed during growthmainly came from the corner sites through surface diffusion.

Effect of Seed Size on Surface Diffusion. Besides the reaction tem-perature, seed size was also found to have an impact on the

timescale of surface diffusion. To examine the effect of seed size, weconducted a set of experiments by replacing the 15-nm Pd cubeswith 6-nm Pd cubes as the seeds, with all other parameters beingkept the same as in the standard synthesis except that the injectionvolume ofNa2PdCl4 solution was reduced from 3.0 to 0.5 mL. Fig. 4shows TEM images of the Pd nanocubes of 6 nm in edge length thatwere used as the seeds and the corresponding Pd cuboctahedronsgrown from these 6-nm cubic seeds, respectively. The size of thecuboctahedrons was consistent with the size of an octahedronproduced via exclusive growth along <100> directions of a 6-nmcubic seed (Fig. S4). This result indicates that most of the adatomsat the corners of a cubic seed had diffused to the edges and sidefaces during growth. In contrast, the final product took a concavecubic morphology when 15-nm cubes were used as the seeds (Fig.2B), suggesting that only a portion of the adatoms at the corners wasable to diffuse to the edges and side faces of larger seeds. AlthoughVdiffusion was roughly the same for both the 6- and 15-nm cubicseeds, the time needed to diffuse from the corners to side faces wasmuch shorter in the former case due to a shorter diffusion distance.

Extension from Pd–Pd to Pd–Pt Nanocrystals. We also extended themechanistic understanding of surface diffusion to the synthesis ofPd–Pt bimetallic nanocrystals. Platinum (Pt) is an invaluablematerial due to its excellent performance as catalysts in a myriadof industrial processes (21–25). However, Pt is extremely expen-sive due to its low content in the Earth’s crust and the continuousgrowth of demand in the automobile industry. In general, such

Fig. 3. Relationship between surface diffusion and reaction temperature.The curve was generated by plotting the total amount of Pd atoms (Ndiffusion)that had diffused to the edges and side faces of individual Pd cubic seeds(shown in Fig. 2) as a function of reaction temperature (Temp). Error bars areSEs with n = 20.

Fig. 4. Effect of seed size on surface diffusion. TEM images of (A) Pd cubicseeds with an average edge length of 6 nm and (B) the correspondingcuboctahedrons grown from these seeds. All of the reaction conditions werekept the same as those in the standard synthesis except for the differences insize (6 vs. 15 nm) for the seeds and injection volume (0.5 vs. 3.0 mL) for theNa2PdCl4 solution. [Scale bar (applies to both images), 20 nm.] Insets show TEMimages of individual nanocrystals at a higher magnification. (Scale bar, 2 nm.)

Xia et al. PNAS | April 23, 2013 | vol. 110 | no. 17 | 6671

CHEM

ISTR

Y

a dilemma can be overcome by using two different approaches (ora combination of them): (i) engineering the surface structure ofcatalytic Pt nanocrystals to maximize the specific activity; and (ii)depositing a thin layer of Pt on the surface of a less expensivemetal such as Pd, which is only one-third of the price for Pt, tominimize the use of Pt. Our previous studies showed that thegrowth of Pt on Pd cubic seeds in an aqueous system at 60–100 °Ctypically yielded Pd–Pt bimetallic nanodendrites with irregular Ptbranches randomly positioned at the corners and edges of Pdcubic seeds (26–28). We believe that the relatively low reactiontemperature and thus small Vdiffusion might be responsible for theformation of Pt branches. In this regard, we expect that Pd–Ptcore–shell nanocrystals with well-defined surface can be obtainedif Vdiffusion is greatly enhanced. To this end, we conducted a syn-thesis at elevated reaction temperatures by injecting Na2PtCl6 (aprecursor to elemental Pt) solution at a rate of 8.0 mL/h intoethylene glycol containing the 15-nm Pd nanocubes, AA, and Br−

(serving as seeds, reductant, and capping agents, respectively) thatwas held in an oil bath preset to 160–200 °C (see Materials andMethods for experimental details). It was found that Pd–Pt bi-metallic multipods (Fig. 5A), concave nanocubes (Fig. 5B), andnanocubes (Fig. 5C) were obtained when the reaction tempera-ture was set to 160, 180, and 200 °C, respectively.Because the Pd–Pt concave nanocubes (Fig. 5B) have high-index

facets that usually provide an enhanced catalytic activity (29–31),we specifically characterized the individual Pd–Pt concave nano-cubes. As shown by the high-angle annular dark-field scanningTEM (HAADF-STEM) image in Fig. 5D, the brighter surfacelayer can be attributed to Pt, most of which is concentrated at thecorners and edges of the Pd cubic seed. The selected area electrondiffraction (SAED) pattern from the same particle (Fig. 5D, Inset)indicates that it was a piece of single crystal sitting against a plane

perpendicular to the [001] zone axis, implying an epitaxial re-lationship between the two metals. The energy dispersive X-ray(EDX)mapping (Fig. 5E) clearly shows a color difference betweenthe core (red, Pd) and shell (green, Pt), confirming a bimetalliccore–shell structure and indicates that the Pd cubic seed was intactduring the growth of Pt shell. The magnifiedHAADF-STEM (Fig.5F) clearly shows the atomic steps at the edge of a core–shellconcave nanocube, implying the presence of high-index facets onthe surface. Line-scan EDX profiles along the edge-to-edge andcorner-to-corner directions of an individual Pd–Pt concave nano-cube (Fig. 5 G and H) confirm that the out-extending corners andedges were dominated by Pt, whereas the cubic core was essentiallymade of pure Pd. Notably, such Pd–Pt core–shell nanocrystals withconcave structures and high-index facets were rarely reported inprevious studies. It is worth pointing out that deposition of Ptatoms on Pd cubic seeds was found to mainly occur at the cornersites as indicated by the HAADF-STEM image in Fig. S7. Thisimage was taken from a sample collected in the very early stage ofa synthesis of Pd–Pt concave nanocubes (after the addition of only3.0mL of Na2PtCl6 solution). This observation provides additionalevidence to support our proposedmechanisms illustrated in Fig. 1.Fig. S8 shows HAADF-STEM images of the Pd–Pt multipods andnanocubes shown in Fig. 5 A and C, respectively.Obviously, the key to coating a thin layer of Pt on Pd cubic seeds

was to increase Vdiffusion and thus decrease the ratio of Vdeposition/Vdiffusion by elevating the reaction temperature. This strategy wasalso consistent with our proposed mechanism (Fig. 1). Comparedwith the case of growth Pd on Pd cubic seeds, the temperaturerequired for the Pt adatom to diffuse from corners to side faces wassignificantly higher. This difference might be related to the fol-lowing order in bonding energies:EPt–Pt (307 kJ/mol)>EPt–Pd (191kJ/mol) > EPd–Pd (136 kJ/mol) (26, 32). An additional energy

Fig. 5. Extension from Pd–Pd to Pd–Pt bimetallic nanocrystals. (A–C) TEM images of Pd–Pt bimetallic nanocrystals that were obtained by depositing Pt on Pdcubic seeds at: (A) 160 °C, (B) 180 °C, and (C) 200 °C. [Scale bar: in C (A–C), 50 nm.] Insets show TEM images at a higher magnification. (Scale bar: 5 nm.) (D–H)Characterizations of an individual Pd–Pt concave nanocube in B: (D) HAADF-STEM image and SAED pattern (Inset) recorded from the same particle; (E) EDXmapping of the same particle shown in D; (F) magnified HAADF-STEM image of the region marked by a blue box in D; (G and H) line-scan EDX spectra ofelemental Pd and Pt that were recorded along the edge-to-edge and corner-to-corner directions, respectively, of the same particle (see Insets for illustrations).

6672 | www.pnas.org/cgi/doi/10.1073/pnas.1222109110 Xia et al.

barrier, that is, Ediff in Eq. 1, would involve in surface diffusionbecause the newly formed Pt atoms prefer to stay at corners toform more stable Pt–Pt bonds rather than to diffuse to side facesand form less stable Pt–Pt bonds. Therefore, to facilitate surfacediffusion, a higher reaction temperature is required to overcomethe relatively higher energy barrier to surface diffusion.

ConclusionWe have demonstrated that surface diffusion plays an importantrole in shape-controlled synthesis of metal nanocrystals. It can beconcluded that the relative rate of atom deposition over surfacediffusion determines the growth pathway of a seed and thus theshape taken by the final product. Using the overgrowth of Pd cubicseeds as a model example, we found that the rate of surface dif-fusion increased with reaction temperature. The bonding energybetween metal atoms was also found to have an impact on surfacediffusion. Based on the mechanistic understanding, a number ofPd nanocrystals with controlled shapes could be easily preparedfrom Pd cubic seeds by controlling the reaction conditions thataffect the rates of atom deposition and surface diffusion. As shownin the present work, the growth mechanisms could be extendedfrom Pd–Pd to a bimetallic system such as Pd–Pt. The mechanismscan also be used to explain the growth habits of other types of seedssuch as Rh and Pt cubic seeds observed in previous studies (10).We believe that the mechanistic understanding can be potentiallyextended to other systems involve various metal nanocrystals andother inorganic nanomaterials.

Materials and MethodsPreparation of 15- and 6-nm Pd Cubes to Be Used as the Seeds. The 15-nm cubicPd seeds were prepared using a previously reported protocol with someminor modifications (16). In a typical synthesis, 11 mL of an aqueous solutioncontaining 105 mg of PVP (molecular weight of ∼55,000), 60 mg of AA,500 mg of KBr, and 57 mg of Na2PdCl4 was heated at 80 °C under magneticstirring for 3 h and then cooled down to room temperature. After centri-fugation and being washed thrice with water, the seeds were dispersed in11 mL of water for further use. The protocol for synthesizing the 6-nm cubicPd seeds was similar to the aforementioned procedure, except that theamount of KBr was reduced from 500 to 5 mg.

Standard Procedure for the Growth of Pd Nanocrystals. In a standard synthesis,3.0 mL of aqueous Na2PdCl4 solution (1.2 mg/mL) was injected at a rate of0.5 mL/h using a syringe pump into an 8.0-mL aqueous suspension con-taining the 15-nm Pd cubic seeds (∼2.3 × 1012 particles per mL), 60 mg of AA,and 50 mg of PVP that was hosted in a 30-mL glass vial at room temperature(∼22 °C) under magnetic stirring. After injection of Na2PdCl4, the solutionwas maintained with stirring for another 5 min to allow the reaction to com-plete. The product was collected by centrifugation at 15,000 × g for 10 minand washed twice with water.

Synthesis of Pd–Pt Bimetallic Nanocrystals. In a typical synthesis, 13 mL ofethylene glycol containing the 15-nm Pd cubic seeds (∼4.6 × 1012 particles permL), 54 mg of KBr, 100 mg of AA, and 66.6 mg of PVP, was preheated at110 °C for 60 min in a three-neck flask with magnetic stirring and thenramped to the desired temperature (160, 180, and 200 °C). Then, 12 mL ofNa2PtCl6·6H2O solution in ethylene glycol (0.5 mg/mL) was injected into thepreheated solution at a rate of 8 mL/h using a syringe pump. After the ad-dition of Na2PtCl6, the solution was maintained with magnetic stirring foranother 5 min to allow the reaction to complete. The product was collectedby centrifugation at 15,000 × g for 10 min and washed twice with ethanoland thrice with water.

Characterizations. The samples were characterized by TEM using a JEOL mi-croscope (JEM-1400) operated at 120 kV. The concentration of Pd ions wasdetermined using inductively coupled plasma mass spectrometry (Perkin-Elmer Elan DRC II), which could be converted to the particle concentration ofPd nanocrystals once the particle size and morphology had been resolved byTEM imaging. The XPS data were recorded using a Thermo K-Alpha spec-trometer with an Al Kα source (eV). HAADF-STEM and EDX analyses wereperformed using a JEOL ARM200F with STEM Cs corrector operated at200 kV.

ACKNOWLEDGMENTS. This work was supported in part by National ScienceFoundation Grant DMR-1215034 and start-up funds from Georgia Instituteof Technology. Y.X. was also supported by the World Class University (WCU)Program from Ministry of Education, Science and Technology (MEST)through National Research Foundation (NRF) Grant R32-20031. As jointlysupervised PhD students from Xiamen University and Xi’an Jiaotong Univer-sity, respectively, S.X. and M.L. were partially supported by Fellowships fromthe China Scholarship Council. M.J.K. was supported by the WCU Programfrom MEST through NRF Grant R31-10026.

1. Oura K, Lifshits VG, Saranin AA, Zotov AV, Katayama M (2003) Surface Science: AnIntroduction (Springer, Heidelberg), 1st Ed, p 324.

2. Kolasinski KW (2008) Surface Science: Foundations of Catalysis and Nanoscience(Wiley, New York), 2nd Ed, pp 127–132.

3. Shenoy VB, Ramasubramaniam A, Freund LB (2003) A variational approach to non-linear dynamics of nanoscale surface modulations. Surf Sci 529(3):365–383.

4. Averbuch A, Israeli M, Ravve I (2003) Electromigration of intergranular voids in metalfilms for microelectronic interconnects. J Comput Phys 186(2):481–502.

5. Rigby SP (2003) A model for the surface diffusion of molecules on a heterogeneoussurface. Langmuir 19(2):364–376.

6. Weber D, Sederman AJ, Mantle MD, Mitchell J, Gladden LF (2010) Surface diffusion inporous catalysts. Phys Chem Chem Phys 12(11):2619–2624.

7. Zhan D, Velmurugan J, Mirkin MV (2009) Adsorption/desorption of hydrogen on Pt nano-electrodes: Evidenceof surfacediffusionand spillover. JAmChemSoc131(41):14756–14760.

8. Xiong Y, Xia Y (2007) Shape-controlled synthesis of metal nanostructures: The case ofpalladium. Adv Mater 19(20):3385–3391.

9. Xia Y, Xiong Y, Lim B, Skrabalak SE (2009) Shape-controlled synthesis of metal nano-crystals: Simple chemistrymeets complex physics?AngewChem Int EdEngl 48(1):60–103.

10. Zhang H, et al. (2011) Controlling the morphology of rhodium nanocrystals by ma-nipulating the growth kinetics with a syringe pump. Nano Lett 11(2):898–903.

11. He G, et al. (2012) A mechanistic study on then and growth of Au on Pd seeds witha cubic or octahedral shape. ChemCatChem 4(10):1668–1674.

12. Xie S, et al. (2012) Synthesis of Pd-Rh core-frame concave nanocubes and their con-version to Rh cubic nanoframes by selective etching of the Pd cores. Angew Chem IntEd Engl 51(41):10266–10270.

13. Ratke L, Voorhees PW (2002) Growth and Coarsening: Ostwald Ripening in MaterialsProcessing (Springer, Heidelberg), pp 117–118.

14. Habas SE, Lee H, Radmilovic V, Somorjai GA, Yang P (2007) Shaping binary metalnanocrystals through epitaxial seeded growth. Nat Mater 6(9):692–697.

15. Xia X, et al. (2011) Silver nanocrystals with concave surfaces and their optical and surface-enhanced Raman scattering properties. Angew Chem Int Ed Engl 50(52):12542–12546.

16. Lim B, et al. (2009) Shape-controlled synthesis of Pd nanocrystals in aqueous solutions.Adv Funct Mater 19(2):189–200.

17. Tsuji M, et al. (2009) Shape evolution of octahedral and triangular platelike silver nano-crystals fromcubic and right bipyramidal seeds inDMF.CrystGrowthDes 9(11):4700–4705.

18. Xia X, Zeng J, Oetjen LK, Li Q, Xia Y (2012) Quantitative analysis of the role played bypoly(vinylpyrrolidone) in seed-mediated growth of Ag nanocrystals. J Am Chem Soc134(3):1793–1801.

19. Lim B, et al. (2010) New insights into the growth mechanism and surface structure ofpalladium nanocrystals. Nano Res 3(3):180–188.

20. McLellan JM, SiekkinenA, Chen J, XiaY (2006) Comparisonof the surface-enhancedRamanscattering on sharp and truncated silver nanocubes. Chem Phys Lett 427(7):122–126.

21. Greeley J, Nørskov JK, Mavrikakis M (2002) Electronic structure and catalysis on metalsurfaces. Annu Rev Phys Chem 53:319–348.

22. Teng X, Liang X, Maksimuk S, Yang H (2006) Synthesis of porous platinum nano-particles. Small 2(2):249–253.

23. Lee H, et al. (2006) Morphological control of catalytically active platinum nano-crystals. Angew Chem Int Ed Engl 45(46):7824–7828.

24. Wang C, Daimon H, Onodera T, Koda T, Sun S (2008) A general approach to the size-and shape-controlled synthesis of platinum nanoparticles and their catalytic re-duction of oxygen. Angew Chem Int Ed Engl 47(19):3588–3591.

25. Chen J, Lim B, Lee EP, Xia Y (2009) Shape-controlled synthesis of platinum nano-crystals for catalytic and electrocatalytic applications. Nano Today 4(1):81–95.

26. Lim B, et al. (2009) Pd-Pt bimetallic nanodendrites with high activity for oxygen re-duction. Science 324(5932):1302–1305.

27. Jiang M, et al. (2010) Epitaxial overgrowth of platinum on palladium nanocrystals.Nanoscale 2(11):2406–2411.

28. LimB, JiangM, Yu T, Camargo PHC, Xia Y (2010) Nucleation andgrowthmechanisms forPd-Pt bimetallic nanodenrites and their electrocatalytic properties.Nano Res 3(2):69–80.

29. Somorjai GA, Blakely DW (1975) Mechanism of catalysis of hydrocarbon reactions byplatinum surfaces. Nature 258(5536):580–583.

30. Tian N, Zhou ZY, Sun SG, Ding Y, Wang ZL (2007) Synthesis of tetrahexahedralplatinum nanocrystals with high-index facets and high electro-oxidation activity.Science 316(5825):732–735.

31. Huang X, Zhao Z, Fan J, Tan Y, Zheng N (2011) Amine-assisted synthesis of concavepolyhedral platinum nanocrystals having 411 high-index facets. J Am Chem Soc133(13):4718–4721.

32. Zhang H, et al. (2011) Nanocrystals composed of alternating shells of Pd and Pt can beobtainedby sequentially addingdifferentprecursors. JAmChemSoc133(27):10422–10425.

Xia et al. PNAS | April 23, 2013 | vol. 110 | no. 17 | 6673

CHEM

ISTR

Y