Autocatalytic surface reduction and its role incontrolling seed-mediated growth of colloidalmetal nanocrystalsTung-Han Yanga,b, Shan Zhouc, Kyle D. Gilroya, Legna Figueroa-Cosmec, Yi-Hsien Leeb, Jenn-Ming Wub,and Younan Xiaa,c,d,1

aThe Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, GA 30332; bDepartment ofMaterials Science and Engineering, National Tsing Hua University, Hsinchu, 30013, Taiwan; cSchool of Chemistry and Biochemistry, Georgia Institute ofTechnology, Atlanta, GA 30332; and dSchool of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332

Edited by Thomas E. Mallouk, The Pennsylvania State University, University Park, PA, and approved November 10, 2017 (received for review August 6, 2017)

The growth of colloidal metal nanocrystals typically involves anautocatalytic process, in which the salt precursor adsorbs onto thesurface of a growing nanocrystal, followed by chemical reductionto atoms for their incorporation into the nanocrystal. Despite itsuniversal role in the synthesis of colloidal nanocrystals, it is stillpoorly understood and controlled in terms of kinetics. Through theuse of well-defined nanocrystals as seeds, including those withdifferent types of facets, sizes, and internal twin structure, herewe quantitatively analyze the kinetics of autocatalytic surfacereduction in an effort to control the evolution of nanocrystals intopredictable shapes. Our kinetic measurements demonstrate thatthe activation energy barrier to autocatalytic surface reduction ishighly dependent on both the type of facet and the presence oftwin boundary, corresponding to distinctive growth patterns andproducts. Interestingly, the autocatalytic process is effective notonly in eliminating homogeneous nucleation but also in activatingand sustaining the growth of octahedral nanocrystals. This workrepresents a major step forward toward achieving a quantitativeunderstanding and control of the autocatalytic process involved inthe synthesis of colloidal metal nanocrystals.

metal nanocrystals | autocatalytic surface reduction | seed-mediatedgrowth | kinetic model | shape control

Autocatalysis, in which at least one of the products is also areactant and thus a catalyst for the same or a coupled re-

action, is a ubiquitous process in nature (1–3). It has drawn ever-increasing attention owing to the impacts of its unique kineticfeatures on the evolution of both biological and nonbiologicalsystems. Different from a conventional catalytic system involvinga fixed amount of catalyst, an autocatalytic process is pro-gressively accelerated as the quantity of catalyst is increasedduring the reaction. The reaction kinetics is not supposed to slowdown until one or more reactants has been depleted in the sys-tem. In the context of nanomaterials science, autocatalytic sur-face reduction of salt precursor is involved in the synthesis of Ag,Au, Cu, Ir, Ru, Pd, and Pt colloidal nanocrystals, as well as in thepreparation of supported metal nanocatalysts (e.g., Ir/Al2O3 andPt/TiO2) (4–13). Specifically, Yang and coworkers (13) recentlydemonstrated that certain types of intermediates from a saltprecursor can induce autocatalytic surface reduction in the syn-thesis of Pt nanocrystals. In addition, it has been reported thatthe self-accelerated oxidation of Ru, Pb, or InAs surface alsobecomes autocatalytic upon the initiation of nucleation (14–16).Despite the pivotal role played by autocatalytic surface re-

duction in the synthesis of colloidal metal nanocrystals, there isstill no quantitative understanding or account of its kineticcharacters. This is because all of the syntheses reported in lit-erature were conducted in the setting of a one-pot approach,where the salt precursor is initially reduced in the solution phaseto generate zero-valent atoms, followed by their aggregation intonuclei via homogeneous nucleation. Afterward, the precursor

will be reduced on the surface of the just-formed nuclei throughan autocatalytic mechanism (4–9, 13). Although the reductionkinetics of the precursor can be fitted to the Finke–Watzkymodel that involves these two pseudoelementary steps (4), it hasbeen very difficult or impossible to differentiate the type andquantify the number of nuclei formed in the initial stage of aone-pot synthesis. These parameters can also change drasticallywhen any one of the thermodynamic or kinetic conditions is al-tered, resulting in major modifications to the products (17–19).These issues can be addressed by switching to seed-mediated

synthesis, in which a precursor is introduced into a reaction so-lution containing the seed with a well-defined size and shape(19–21). In this case, the precursor can be reduced to atoms inthe solution, followed by their heterogeneous nucleation andgrowth via atomic addition onto the surface of the seed. Theatoms can also undergo homogeneous nucleation to generate adifferent population of nanocrystals. This is especially true whenthe atoms derived from the reduction of precursor are in largeexcess relative to the number of deposition sites provided by theseed. Alternatively, the precursor can adsorb onto the surface ofthe seed and be reduced through an autocatalytic process. Dif-ferent from solution reduction, surface reduction is often auto-catalytic, together with a much smaller activation energy. Assuch, autocatalytic surface reduction on the preformed seed caneasily surpass solution-phase reduction to become a dominantprocess. Through the use of preformed seeds with different

Significance

Controlling the shape of colloidal metal nanocrystals is centralto the realization of their diverse applications in catalysis,photonics, electronics, and medicine. Here, we demonstratethat autocatalytic surface reduction can be employed to enablethe formation of metal nanocrystals with well-controlled andpredictable shapes through seed-mediated growth. Our quan-titative analysis suggests that the kinetics of autocatalyticsurface reduction is highly sensitive to the atomic structures onthe surface of the seed, leading to different growth rates fordifferent sites on the seed and eventually resulting in theevolution of nanocrystals into different shapes. The mecha-nistic insights into autocatalytic surface reduction obtained inthis work can be extended to other systems involving nano-crystals with different compositions, facets, and structures.

Author contributions: Y.X. designed research; T.-H.Y. performed research; S.Z., K.D.G.,L.F.-C., Y.-H.L., and J.-M.W. analyzed data; and T.-H.Y., K.D.G., and Y.X. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence should be addressed. Email: younan.xia@bme.gatech.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1713907114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1713907114 PNAS | December 26, 2017 | vol. 114 | no. 52 | 13619–13624

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

6, 2

020

shapes (and thus diverse types of facets on the surface), sizes, andtwin structures (single crystal vs. multiply twinned), we are able toachieve a quantitative understanding of how these experimentalparameters affect the kinetics of autocatalytic growth and thus theoutcome of a synthesis. This quantitative understanding of auto-catalytic surface reduction will shed light on the rational designand experimental control of colloidal nanocrystal syntheses.

Results and DiscussionAutocatalytic Surface Reduction and Its Dependence on the Type ofFacet. Our recent quantitative analysis suggests that the reductionpathway undertaken by a salt precursor is largely determined by thekinetics involved (9). The precursor is reduced on the surface of aseed through an autocatalytic process under slow kinetics, whereasit is reduced in the solution phase when the kinetics is fast. Basedon this finding, here we chose to focus on PdBr4

2−, a complexknown for its relatively slow reduction kinetics, to systematicallyinvestigate how the Pd(II) precursor would be reduced in thepresence of Pd seeds enclosed by different types of facets. A typicalexperiment involves the one-shot injection of aqueous PdBr4

2−

solution into an aqueous mixture containing Pd nanocrystals(seeds), ascorbic acid (AA, reducing agent), and poly(vinyl pyrro-lidone) (PVP, colloidal stabilizer). For the seeds with cubic andoctahedral shapes, their surfaces are enclosed by six {100} andeight {111} facets, respectively, allowing us to examine the explicitdependence of surface autocatalytic reduction on the type of facet.It should be pointed out that the PdBr4

2− complex is supposedto immediately hydrolyze upon its dissolution in water at a rel-atively low concentration of 0.34 mM (used in all of the synthesesreported here), leading to the formation of a series of hydratedspecies, PdBrn(H2O)4−n

2−n (n < 4), with a specific distribution atequilibrium (22, 23). SI Appendix, Fig. S1A shows the UV-visspectrum taken from an aqueous solution of freshly preparedPdBr4

2−, which has a concentration identical to the reactionmixture used for the standard seed-mediated growth except forthe absence of AA, PVP, and seeds. The Pd(II) species in thesolution showed a broad peak at 269 nm, which is different fromthe absorption peak at 332 nm expected for PdBr4

2−. Based on theresult from computational calculation, the broad peak at 269 nm

can be assigned to the absorption of PdBr2(H2O)2 and PdBr3(H2O)−

(23). This result implies that the hydrolysis of PdBr42− has to be

taken into account. Furthermore, when PVP was introduced, thepeak intensified and shifted to a longer wavelength, suggestingthat PVP could also coordinate with the Pd(II) ions (SI Appendix,Fig. S1A). Therefore, under the conditions used in the presentwork, the actual Pd(II) precursor likely comprised H2O mole-cules, Br− ions, and PVP. To avoid misunderstanding and con-fusion, we use “the precursor based on PdBr4

2−” or simply “the

Pd(II) precursor” throughout this paper, with an intention to reflectthe initial Pd(II) complex added into the reaction system.Fig. 1 A and B shows transmission electron microscopy (TEM)

images of the Pd cubic and octahedral seeds, respectively, as wellas their 3D atomic models. They had a purity approaching 100%,together with sizes (see SI Appendix, Fig. S2 for the definition ofsize) of 18.0 ± 2.1 nm and 25.0 ± 2.9 nm, respectively. As shownin Fig. 1 C–F, the products of seed-mediated growth containedno particles smaller than the original seeds, implying the absenceof homogeneous nucleation. In the case of Pd cubic seed, thetruncation at corners, that is, the appearance of eight small{111} facets, became more significant with the increase in re-action time, while the distance between the opposite {100} facetsgradually increased from 19.4 ± 2.3 nm at t = 2 h (Fig. 1C) to20.7 ± 2.4 nm at t = 4 h (Fig. 1E), respectively. This data clearlyindicate that the growth rate along the <100> direction wasgreater than that along the <111> direction.When seed-mediated growth was applied to the Pd octahedral

seed, octahedra with a slight truncation at the {100} facets wereobserved at t = 2 h, as shown in Fig. 1D. When the reaction wasprolonged to t = 4 h, truncation became more visible at the cornersites (Fig. 1F). This result confirms that the growth mainly oc-curred on the {111} facets, which is opposite to what was observedin the case of Pd cubic seed. It should be pointed out that theintroduction of PdBr4

2− as a precursor could release Br− ions intothe reaction solution, and some of the Br− ions might chemisorbonto the {100} facets of the seed and thereby retard the growth atthese sites (24–26). For the cubic seed, the Br− ions released fromthe precursor could chemisorb onto the {100} faces, but thecoverage density at equilibrium should be relatively low due to the

Fig. 1. Autocatalytic surface reduction on Pd cubic and octahedral seeds enclosed by {100} and {111} facets, respectively. (A, C, and E) TEM images of (A) thePd cubic seed and the products obtained at (C) t = 2 h and (E) t = 4 h using the standard protocol. (B, D, and F) TEM images of (B) the Pd octahedral seed andthe products obtained at (D) t = 2 h and (F) t = 4 h using the standard protocol. (Scale bar in E applies to all images in A, C, and E. Scale bar in F applies to allimages in B, D, and F.) (Insets) Three-dimensional atomic models of the corresponding particles at different stages of the synthesis.

13620 | www.pnas.org/cgi/doi/10.1073/pnas.1713907114 Yang et al.

Dow

nloa

ded

by g

uest

on

July

6, 2

020

large surface area of {100} facets on the cubic seed. For the oc-tahedral seed, however, the number of Br− ions released from theprecursor was adequate for a higher coverage due to the relativelysmall surface area of the {100} facets on the octahedral seed. Asconfirmed by inductively coupled plasma mass spectrometry (ICP-MS) (24), the ratios between the numbers of Br− ions and surfacePd{100} atoms were 0.27 and 0.77, respectively, for the Pd cubicand octahedral seeds (see SI Appendix for detailed analysis).For the seed-mediated growth of Ag and Pd nanocrystals,

recent experimental observations indicated that, once a completeoctahedron had been formed, the growth of the close-packed{111} facets would be automatically shut down and the atomssubsequently formed in the solution would undergo homoge-neous nucleation instead of being deposited onto the surface ofthe octahedral nanocrystals via heterogeneous nucleation (19,27–29). This phenomenon, known as self-termination of growth,

can be attributed to the closest packing of atoms in the {111} facetsof a face-centered cubic (fcc) metal and thus a greater energybarrier to heterogeneous nucleation relative to homogeneous nu-cleation. We observed the self-termination phenomenon when wecarried out a control experiment with the use of aqueous PdCl4

2− asa precursor (SI Appendix, Fig. S3), while all other experimentalconditions were kept the same as the standard protocol. In contrast,when PdBr4

2− was added as a precursor, the limitation imposed byself-terminated growth was lifted for the Pd octahedral seed,resulting in continuous evolution in terms of size and shape (Fig. 1D and F). Taken together, it can be concluded that the use ofdistinct precursors (e.g., PdBr4

2− vs. PdCl42−) can lead to different

products depending on the reduction kinetics. The same argumentcan also be applied to other types of seeds with different facets andinternal twin structures (SI Appendix, Fig. S4). In addition to theprecursor, both the type of reducing agent and reaction tempera-ture could also affect the reduction kinetics and thus significantlyalter the outcome of a synthesis (9) (SI Appendix, Fig. S5).To quantitatively understand how the reduction kinetics of the

precursor correlates with the type of seed involved, we measuredthe concentrations of Pd(II) ions remaining in the reaction

Fig. 2. Quantitative analyses of the kinetic parameters for autocatalyticsurface reduction on Pd cubic and octahedral seeds, respectively. (A) Theconcentrations of the Pd(II) precursor remaining in the reaction solutions asa function of time in the presence of Pd cubic or octahedral seed at roomtemperature. The rate constants k′

2 for the autocatalytic surface reductionon Pd cubic and octahedral seeds could be derived through curve fitting.(B) Plots showing lnk′

2 as a function of 1/T for the autocatalytic surface re-duction on Pd cubic and octahedral seeds, respectively, where the slope ofthe linear regression line can be used to calculate the corresponding acti-vation energy (Ea) using the Arrhenius equation.

Fig. 3. Effect of seed size on autocatalytic surface reduction. (A) The rateconstants k′

2 for the autocatalytic surface reduction on 7-, 12-, 18-, and23-nm Pd cubic seeds, respectively, at room temperature. (B) Plots showinglnk′

2 as a function of 1/T for the autocatalytic surface reduction on differentlysized Pd cubic seeds.

Yang et al. PNAS | December 26, 2017 | vol. 114 | no. 52 | 13621

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

6, 2

020

solution at different time points using UV-vis spectroscopy (SIAppendix, Fig. S6). The experimental procedures for kineticanalysis can be found in SI Appendix. Fig. 2A shows the con-centrations of Pd(II) remaining in the reaction solution in thepresence of Pd cubic and octahedral seeds, respectively, as afunction of reaction time. Clearly, it can be seen that the re-duction of the Pd(II) precursor by AA had a strong dependenceon the type of seed involved. The reduction in the presence ofcubic seed was much faster than the case of octahedral seed.We also focused on the elementary reactions involved in the

synthesis and estimated their corresponding kinetic parameters. Inour recent work (9), we found that the reduction kinetics of Pd(II)in the absence of preformed seeds could be fitted to the Finke–Watzky two-step growth mechanism (4), by which Pd(II) was re-duced to generate atoms [solution reduction: PdðIIÞ+ 2e−!k1 Pd0],which then aggregated into nuclei (defined as Pd0n) via homogeneousnucleation, followed by autocatalytic surface reduction enabled by thejust-formed nuclei [surface reduction: Pd0n +PdðIIÞ+ 2e−!k2 Pd0n+ 1].The rate constants (k1 and k2) for the reduction of Pd(II) in theabsence of seeds could thus be extracted from curve fitting basedon the Finke–Watzky mechanism (SI Appendix, Eq. S7). Here wefurther extended this quantitative analysis to seed-mediatedgrowth. With the introduction of preformed seed, denotedPd0nðseedÞ, the seed would provide additional nucleation sites forthe autocatalytic surface reduction of Pd(II) precursor [surface

reduction: Pd0nðseedÞ+PdðIIÞ+ 2e−!k′2 Pd0n+ 1ðseedÞ]. Actually, the

reduction of the precursor was greatly accelerated in the presenceof Pd cubic or octahedral seed (SI Appendix, Figs. S7–S9), sug-gesting that the added seed not only participated in but alsodominated the reduction of the Pd(II) precursor. Due to the ab-sence of homogeneous nucleation (as confirmed by the TEMimage in Fig. 1), we expected that autocatalytic surface reductionshould occur on the added seed only. Thus, by setting k1 to thevalue obtained from a control experiment, in which the samesynthesis was carried out in the absence of preformed seed andthus k2 = 0 (SI Appendix, Figs. S7–S9 and Table S1), the rateconstants (k′2) for the autocatalytic surface reduction of Pd(II) byAA on the Pd cubic and octahedral seeds were derived as 2.61 ×

10−1 and 1.35 × 10−1 min−1·mM−1, respectively, through curvefitting (Fig. 2A and SI Appendix, Eq. S12). The detailed mathe-matical models for analyzing the reduction kinetics can be foundin SI Appendix,Materials and Methods (SI Appendix, Eqs. S3–S12).From the kinetic parameters in SI Appendix, Table S2, we

further calculated the percentages of solution reduction andsurface reduction for the precursor in the presence of preformedseed as a function of reaction time by integrating reduction ratesover time (SI Appendix, Fig. S10). We found that the percentageof surface reduction approached 86% and 77% in the cases ofcubic and octahedral seeds, respectively. The kinetic analyses,along with the TEM images shown in Fig. 1, suggest that thereduction of the precursor based on PdBr4

2− followed the au-tocatalytic surface growth pathway (SI Appendix, Eq. S10), inwhich the reduction of precursor mainly occurred on the surfaceof preformed seed rather than in the solution phase. These re-sults also imply that only a limited number of Pd atoms wereformed in the solution phase through solution reduction (SIAppendix, Eq. S8) and their concentration could be kept belowsupersaturation (the lowest concentration of atoms needed forinitiating homogeneous nucleation) (18) throughout the syn-thesis. These atoms could only undergo heterogeneous nucle-ation and growth (SI Appendix, Eq. S9) via atom deposition ontothe surface of the preformed seed.From the rate constants (k′2) at different reaction tempera-

tures (T), we obtained the activation energy (Ea) for autocata-lytic surface reduction involving the precursor based on PdBr4

2−,AA, and preformed seeds using the Arrhenius plot. As shown inFig. 2B for the plot of ln k′2 as a function of 1/T, activation en-ergies of 37.6 and 65.1 kJ/mol, respectively, were derived for theautocatalytic surface reduction on the Pd cubic and octahedralseeds. The energy barrier to the autocatalytic surface reductionon the cubic seed was ∼1.7× lower than that on the octahedralseed, suggesting that the {100} facets require a lower energy toinitiate autocatalytic surface growth relative to the {111} facets.This observation explains why the {100} facets grew at a rela-tively faster rate than the {111} facets, which is consistent withthe TEM observations in Fig. 1. Taken together, our quantitativeanalysis clearly indicates that the kinetics of autocatalytic surface

Fig. 4. Extension from single-crystal to multiply twinned seeds. (A, C, and E) TEM images of (A) the Pd decahedral seed and the Pd nanocrystals obtained at(C) t = 0.5 h and (E) t = 1 h using the standard protocol. (B, D, and F) TEM images of (B) the Pd icosahedral seed and the Pd nanocrystals obtained at (D) t =0.5 h and (F) t = 1 h using the standard protocol. (Scale bar in E applies to all of the images in A, C, and E. Scale bar in F applies to all of the images in B, D, andF.) (Insets) Three-dimensional atomic models of the corresponding nanocrystals at different time points; red lines represent the twin boundaries.

13622 | www.pnas.org/cgi/doi/10.1073/pnas.1713907114 Yang et al.

Dow

nloa

ded

by g

uest

on

July

6, 2

020

reduction is highly dependent on the type of facet expressed onthe surface of the seed. Similar results were also observed for thecases of Pd cuboctahedral and truncated octahedral seeds withdifferent proportions of {100} and {111} facets (SI Appendix,Figs. S11 and S12). Although the exact surface exposed on theseed undergoes constant remodeling as growth proceeds, thisapproach still allows for a good estimate of the kinetic param-eters, including both rate constant and activation energy, in-volved in the autocatalytic growth process.

Dependence of Autocatalytic Surface Reduction on the Size of theSeed. In addition to the type of facet, we also examined how thesize of the seed affects autocatalytic surface reduction. Whenvarying the size of a seed, the ratio between different types (e.g.,vertex, edge, and plane) of surface atoms that differ in coordinationnumber will also change accordingly (30–33). To examine the sizeeffect, we conducted a set of experiments with the use of Pd cubicseeds with four different edge lengths (7, 12, 18, and 23 nm). Basedon the statistics of surface atoms for fcc metals (SI Appendix, Fig.S13), about 70% of the surface atoms of a 7-nm cube are locatedon the side faces (i.e., plane atoms), while the percentage ofplane-based atoms increases beyond 90% for a 23-nm cube (30,31). SI Appendix, Fig. S14 shows TEM images of the truncatedcubes grown from these seeds when PdBr4

2− was added as theprecursor, suggesting that the growth still occurred preferentiallyon the {100} facets regardless of the difference in edge length.Again, we did not observe homogeneously nucleated particles inthe final products. These observations were consistent with thegrowth mechanism based on surface autocatalytic reduction.The corresponding kinetic parameters were then obtained from

the data acquired using the spectroscopic method (SI Appendix, Fig.S15 and Table S3). Importantly, the rate constant k′2 for the auto-catalytic surface reduction on Pd cubic seeds exhibited a negligibledependence on the size of the seed, as shown in Fig. 3A. We furtherestimated the activation energies for autocatalytic surface reductionon the differently sized seeds by measuring the rate constant over arange of temperatures. The autocatalytic reduction of PdBr4

2− onthe Pd cubic seeds with different sizes had an almost identical ac-tivation energy (Fig. 3B) although the seeds are characterized bydifferent ratios between different types of surface atoms. The ki-netic data (Fig. 3), along with the TEM images shown in SI Ap-pendix, Fig. S14, suggest that the precursor based on PdBr4

2−

tended to undergo surface reduction on the {100} facets with alower activation energy than other types of facets (e.g., corners) forthe single-crystal Pd cubic seed, resulting in the formation ofnanocubes with significant truncation at the corners.

Extension from Single-Crystal to Multiply Twinned Seeds. We alsoextended the mechanistic analysis of autocatalytic surface re-duction to multiply twinned seeds, including those with decahedraland icosahedral structures. Fig. 4 A and B shows TEM images andthe corresponding 3D atomic models of the 19.2 ± 2.4-nm Pddecahedral and 12.3 ± 2.1-nm Pd icosahedral seeds, respectively.

A decahedral nanocrystal is made of 5 single-crystal tetrahedralunits connected by 5 twin boundaries and enclosed by 10 {111}facets, while an icosahedron consists of 20 tetrahedral units and30 twin boundaries, and enclosed by 20 triangular {111} facets.Upon seed-mediated growth with the introduction of PdBr4

2−, theresultant Pd decahedral (Fig. 4 C and E) and icosahedral (Fig. 4 Dand F) nanocrystals clearly showed concave structures on theirsurfaces. At the early stage of growth, the deposition of Pd atomspreferentially took place on the low-coordination vertex sitesintersected by multiple-twin boundaries on each Pd decahedral(Fig. 4C) or icosahedral seed (Fig. 4D) relative to the {111} faces.With the increase in reaction time, the Pd atoms were continu-ously deposited on the vertices of the seeds at a rate greater thanthe surface diffusion of adatoms (26), generating a tip at the pointof deposition, finally resulting in the formation of Pd concavedecahedral (Fig. 4E) and icosahedral (Fig. 4F) nanocrystals with ahigh density of low-coordination sites (e.g., steps, edges, andkinks) and high-index facets on the surface (34).Again, we obtained the corresponding activation energies from

the Arrhenius plots (SI Appendix, Figs. S16 and S17 and Table S4),as summarized in Fig. 5 and Table 1. Interestingly, the energy bar-riers associated with the autocatalytic surface reduction on multiplytwinned seeds (Ea = 25.8 and 21.1 kJ/mol for decahedral and ico-sahedral seeds, respectively) are significantly lower than those on thesingle-crystal, octahedral seeds enclosed by {111} facets. These re-sults are consistent with our TEM observations in that the growth ofdecahedral and icosahedral seeds preferentially occurred on thelow-coordination vertex sites instead of (111) planes (Fig. 4 and SIAppendix, Fig. S18). In other words, the precursor based on PdBr4

2−

tended to be reduced to Pd atoms on the vertex sites with smalleractivation energy barriers through autocatalytic surface reduction,transforming the decahedral and icosahedral seeds into concavedstructures. In addition to the low coordination of vertex sites, thesignificant surface strain associated with twin boundaries is alsobelieved to be instrumental to adsorption and surface reduction ofthe precursor because they alter the electronic structures of surfaceatoms by shifting their d-band center (35, 36). This effect is partic-ularly pronounced for a decahedral or icosahedral nanocrystal owingto the high density of twin boundaries. As a result, the strain-induced electronic effect may be another factor responsible forthe formation of decahedral and icosahedral nanocrystals withconcaved surface structures. Further study along this line will greatlyadvance our understanding of surface strain on autocatalytic surfacereduction involved in seed-mediated growth.

Fig. 5. Comparison of the activation energy (Ea) for autocatalytic surfacereduction on Pd seeds with different but well-defined facets and twinstructures. (Insets) Preferential sites for the nucleation and deposition of Pdatoms through autocatalytic surface reduction, in which the precursor ad-sorbs onto these sites, followed by chemical reduction to atoms for theirincorporation into the surface layer of the seed.

Table 1. Summary of the activation energies and prefactors forthe various types of seeds

Type ofseed

Seedsize, nm

Activation energy,kJ mol−1

Prefactor,M−1·s−1

Cube 7 35.2 ± 2.1 7.71 × 106

Cube 12 36.3 ± 2.3 1.69 × 107

Cube 18 37.6 ± 2.3 1.98 × 107

Cube 23 38.3 ± 2.4 2.60 × 107

Octahedron 25 65.1 ± 4.7 7.16 × 1011

Decahedron 19 25.8 ± 2.1 2.87 × 105

Icosahedron 12 21.1 ± 2.0 5.69 × 104

Yang et al. PNAS | December 26, 2017 | vol. 114 | no. 52 | 13623

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

July

6, 2

020

It is worth emphasizing the importance of knowing the acti-vation energies of autocatalytic surface reduction on differenttypes of facets or twin boundaries presented on the surface ofwell-defined seeds used in the present study (Fig. 5 and Table 1).Such quantitative information can be used to predict the growthpattern of a more complicated seed with a unique energy land-scape and the shape or morphology taken by the final metalnanocrystals. Ultimately, such a quantitative understanding andcontrol will push the synthesis of colloidal metal nanocrystalsaway from the conventional trial-and-error approach, and to-ward a predictable, deterministic setting.

ConclusionsUsing Pd as a typical example, we have demonstrated that auto-catalytic surface reduction can be employed to enable the forma-tion of metal nanocrystals with predictable and well-controlledshapes through seed-mediated growth. In addition to the elimina-tion of homogeneous nucleation, autocatalytic surface reductionallows one to lift the limitation imposed by self-terminated growthon the {111} facets of octahedral seed. From the quantitativeanalysis, it can be concluded that the kinetics of autocatalytic sur-face reduction is sensitive to the atomic structures on the surface ofthe seed, leading to different growth rates for different sites on theseed and eventually resulting in the formation of nanocrystals withthe diverse but well-controlled shapes. It is anticipated that the

mechanistic understanding of autocatalytic surface reductionobtained in this work can be extended to other systems involvingnanocrystals with different compositions, facets, and structures.

Materials and MethodsAdditional details with regard to the materials and methods can be found inSI Appendix.

Seed-Mediated Growth of Pd Nanocrystals Grown from the Capping-Free PdSeeds. In a standard protocol, 2.0 mL of PdBr4

2− (1.71 mg of K2PdBr4) precursorwas injected in one shot into 8 mL of an aqueous suspension containing0.36 mg of the capping-free Pd seeds (cubic, cuboctahedral, truncated octa-hedral, octahedral, decahedral, or icosahedral seeds), 60mg of AA, and 100 mgof PVP that was hosted in a 20-mL glass vial at a certain temperature undermagnetic stirring. The reaction was then capped, and maintained at a certaintemperature for various reaction times. After the reactions, the final productwas collected by centrifugation at 17,500 rpm (Beckman Coulter Optima MAX-XP ultracentrifuge; TLA-55 rotor) for 30 min and washed three times withwater. For the ICP-MS analysis, the particles were dissolved in aqua regia andfurther diluted with 1% (vol/vol) aqueous HNO3 solution to a level of 100 ppb.

ACKNOWLEDGMENTS. This work was supported in part by a research grantfrom the NSF (CHE 1505441) and startup funds from Georgia Institute ofTechnology (Georgia Tech). The electron microscopy studies were performedat Georgia Tech’s Institute for Electronics and Nanotechnology, a member ofthe National Nanotechnology Coordinated Infrastructure supported by theNSF (ECCS-1542174).

1. Mata-Perez F, Perez-Benitol J (1987) The kinetic rate law for autocatalytic reactions.J Chem Educ 64:925–927.

2. Lavabre D, Pimienta V, Levy G, Micbea JC (1993) Reversible, mixed first- and second-order and autocatalytic reactions as particular cases of a single kinetic rate law. J PhysChem 97:5321–5326.

3. Pasternack RF, et al. (1998) A nonconventional approach to supramolecular formationdynamics. The kinetics of assembly of DNA-bound porphyrins. J Am Chem Soc 120:5873–5878.

4. Watzky F, Finke RG (1997) Transition metal nanocluster formation kinetic andmechanistic studies. A new mechanism when hydrogen is the reductant: Slow, con-tinuous nucleation and fast autocatalytic surface growth. J Am Chem Soc 119:10382–10400.

5. Harada M, Kizaki S (2016) Formation mechanism of gold nanoparticles synthesized byphotoreduction in aqueous ethanol solutions of polymers using in situ quick scanningx-ray absorption fine structure and small-angle X-ray scattering. Cryst Growth Des 16:1200–1212.

6. Widegren JA, Bennett MA, Finke RG (2003) Is it homogeneous or heterogeneouscatalysis? Identification of bulk ruthenium metal as the true catalyst in benzenehydrogenations starting with the monometallic precursor, Ru(II)(η6-C6Me6)(OAc)2, pluskinetic characterization of the heterogeneous nucleation, then autocata-lytic surface-growth mechanism of metal film formation. J Am Chem Soc 125:10301–10310.

7. Liu Q, et al. (2016) Quantifying the nucleation and growth kinetics of microwavenanochemistry enabled by in situ high-energy X-ray scattering. Nano Lett 16:715–720.

8. Besson G, Finney EE, Finke RG (2005) Nanocluster nucleation, growth, and then ag-glomeration kinetic and mechanistic studies: A more general, four-step mechanisminvolving double autocatalysis. Chem Mater 17:4925–4938.

9. Yang TH, et al. (2017) Toward a quantitative understanding of the reduction path-ways of a salt precursor in the synthesis of metal nanocrystals. Nano Lett 17:334–340.

10. Mondloch JE, Yan X, Finke RG (2009) Monitoring supported-nanocluster heteroge-neous catalyst formation: Product and kinetic evidence for a 2-step, nucleation andautocatalytic growth mechanism of Pt(0)n formation from H2PtCl6 on Al2O3 or TiO2.J Am Chem Soc 131:6389–6396.

11. van den Berg R, et al. (2016) Revealing the formation of copper nanoparticles from ahomogeneous solid precursor by electron microscopy. J Am Chem Soc 138:3433–3442.

12. Kent P, Mondloch JE, Finke RG (2016) Synthesis of heterogeneous Ir0∼600-900/γ-Al2O3 inone pot from the precatalyst Ir(1,5-COD)Cl/γ-Al2O3: Discovery of two competing trace“ethyl acetate effects” on the nucleation step and resultant product. ACS Catal 6:5449–5461.

13. Yin X, et al. (2017) Quantitative analysis of different formation modes of platinumnanocrystals controlled by ligand chemistry. Nano Lett 17:6146–6150.

14. Thürmer K, Williams E, Reutt-Robey J (2002) Autocatalytic oxidation of lead crystallitesurfaces. Science 297:2033–2035.

15. Over H, Seitsonen AP (2002) Surface chemistry. Oxidation of metal surfaces. Science297:2003–2005.

16. Clemens JB, Bishop SR, Feldwinn DL, Droopad R, Kummel AC (2009) Initial stages ofthe autocatalytic oxidation of the InAs(0 0 1)-(4 x 2)/c(8 x 2) surface by molecularoxygen. Surf Sci 603:2230–2239.

17. Xia Y, Xia X, Peng HC (2015) Shape-controlled synthesis of colloidal metal nano-

crystals: Thermodynamic versus kinetic products. J Am Chem Soc 137:7947–7966.18. Xia Y, Xiong Y, Lim B, Skrabalak SE (2009) Shape-controlled synthesis of metal

nanocrystals: Simple chemistry meets complex physics? Angew Chem Int Ed Engl 48:

60–103.19. Xia Y, Gilroy KD, Peng HC, Xia X (2017) Seed-mediated growth of colloidal metal

nanocrystals. Angew Chem Int Ed Engl 56:60–95.20. Fan Z, Zhang H (2016) Template synthesis of noble metal nanocrystals with unusual

crystal structures and their catalytic applications. Acc Chem Res 49:2841–2850.21. Zhang Z, et al. (2016) Submonolayered Ru deposited on ultrathin Pd nanosheets used

for enhanced catalytic applications. Adv Mater 28:10282–10286.22. Elding LI, Olsson LF (1978) Electronic absorption spectra of square-planar chloro-aqua

and bromo-aqua complexes of palladium(II) and platinum(II). J Phys Chem 82:69–74.23. Roux CJL, Kriek RJ (2017) A detailed spectrophotometric investigation of the com-

plexation of palladium(II) with chloride and bromide. Hydrometallurgy 169:447–455.24. Peng HC, Xie S, Park J, Xia X, Xia Y (2013) Quantitative analysis of the coverage

density of Br- ions on Pd{100} facets and its role in controlling the shape of Pd

nanocrystals. J Am Chem Soc 135:3780–3783.25. Lim B, et al. (2009) Shape-controlled synthesis of Pd nanocrystals in aqueous solutions.

Adv Funct Mater 19:189–200.26. Xia X, et al. (2013) On the role of surface diffusion in determining the shape or

morphology of noble-metal nanocrystals. Proc Natl Acad Sci USA 110:6669–6673.27. Jin M, Zhang H, Xie Z, Xia Y (2012) Palladium nanocrystals enclosed by {100} and {111}

facets in controlled proportions and their catalytic activities for formic acid oxidation.

Energy Environ Sci 5:6352–6357.28. Wang Y, et al. (2013) Synthesis of silver octahedra with controlled sizes and optical

properties via seed-mediated growth. ACS Nano 7:4586–4594.29. Zhang Z, et al. (2015) Redox reaction induced Ostwald ripening for size- and shape-

focusing of palladium nanocrystals. Chem Sci 6:5197–5203.30. Hardeveld RV, Hartog F (1969) The statistics of surface atoms and surface sites on

metal crystals. Surf Sci 15:189–230.31. Crespo-Quesada M, Yarulin A, Jin M, Xia Y, Kiwi-Minsker L (2011) Structure sensitivity

of alkynol hydrogenation on shape- and size-controlled palladium nanocrystals:

Which sites are most active and selective? J Am Chem Soc 133:12787–12794.32. Shao M, Peles A, Shoemaker K (2011) Electrocatalysis on platinum nanoparticles:

Particle size effect on oxygen reduction reaction activity. Nano Lett 11:3714–3719.33. Deng L, Deng H, Xiao S, Tang J, Hu W (2013) Morphology, dimension, and compo-

sition dependence of thermodynamically preferred atomic arrangements in Ag-Pt

nanoalloys. Faraday Discuss 162:293–306.34. Gilroy KD, et al. (2017) Thermal stability of metal nanocrystals: An investigation of the

surface and bulk reconstructions of Pd concave icosahedra. Nano Lett 17:3655–3661.35. Wu J, et al. (2012) Icosahedral platinum alloy nanocrystals with enhanced electro-

catalytic activities. J Am Chem Soc 134:11880–11883.36. Wang L, et al. (2015) Aerobic oxidation of cyclohexane on catalysts based on twinned

and single-crystal Au75Pd25 bimetallic nanocrystals. Nano Lett 15:2875–2880.

13624 | www.pnas.org/cgi/doi/10.1073/pnas.1713907114 Yang et al.

Dow

nloa

ded

by g

uest

on

July

6, 2

020