Ligand-free Ni nanocluster formation at atmospheric pressure via rapid quenching in a microplasma process

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 134.84.228.26

This content was downloaded on 03/09/2014 at 01:37

Please note that terms and conditions apply.

Ligand-free Ni nanocluster formation at atmospheric pressure via rapid quenching in a

microplasma process

View the table of contents for this issue, or go to the journal homepage for more

2014 Nanotechnology 25 385601

(http://iopscience.iop.org/0957-4484/25/38/385601)

Home Search Collections Journals About Contact us My IOPscience

Ligand-free Ni nanocluster formation atatmospheric pressure via rapid quenching ina microplasma process

Ajay Kumar1, Seungkoo Kang2, Carlos Larriba-Andaluz2, Hui Ouyang2,Christopher J Hogan2 and R Mohan Sankaran1

1Department of Chemical Engineering, Case Western Reserve University, USA2Department of Mechanical Engineering, University of Minnesota, 111 Church St. S.E., Minneapolis, MN55455, USA

E-mail: [email protected] and [email protected]

Received 15 May 2014, revised 16 July 2014Accepted for publication 31 July 2014Published 2 September 2014

AbstractThe production of metal nanoclusters composed of less than 103 atoms is important forapplications in energy conversion and medicine, and for fundamental studies of nanomaterialnucleation and growth. Unfortunately, existing synthesis methods do not enable adequate controlof cluster formation, particularly at atmospheric pressure wherein formation typically occurs onsub-millisecond timescales. Here, we demonstrate that ligand-free, unagglomerated nickelnanoclusters can be continuously synthesized at atmospheric pressure via the decomposition of bis(cyclopentadienyl)nickel(II) (nickelocene) in a spatially-confined microplasma process that rapidlyquenches particle growth and agglomeration. The clusters were measured on line by ion mobilityspectrometry (IMS) and further analyzed by atomic force microscopy (AFM). Our results revealthat stable clusters with spherical equivalent mean diameters below 10 ̇A are produced, and bycontrolling the nickelocene concentration, the mean diameter can be tuned up to ∼50 ̇A. Althoughdiameter is often the sole metric used in nanocluster and nanoparticle characterization, to infer thenumber of atoms in AFM and IMS detected clusters, we compare measured AFM heights andIMS inferred collision cross sections to theoretical predictions based on both bulk matterapproximations and density functional theory and Hartree–Fock calculated Ni nanoclusterstructures (composed of 2–15 atoms for the latter). The calculations suggest that Ni nanoclusterscomposed of less than 102 atoms can be produced repeatably with simple microplasma reactors.

S Online supplementary data available from stacks.iop.org/NANO/25/385601/mmedia

Keywords: cluster, nanoparticles, nanocluster, microplasma, aerosol, ion mobility measurement,atomic force microscopy

(Some figures may appear in colour only in the online journal)

Introduction

The formation and stability of metal nanoclusters, i.e., con-densed phases composed of <103 atoms and characteristicsizes below ∼3 nm, is of technological and scientific interest.Compared to their bulk or larger nanostructured counterparts,metal nanoclusters possess structures [1, 2], melting tem-peratures [3–5], and catalytic and photocatalytic activities[6–10] that vary non-monotonically, and often drastically,

with the number of atoms in each cluster. These uniqueproperties are presently under examination for emergingapplications in energy conversion [11] and medical imaging[12]. Further, in bottom-up synthesis of larger nanomaterials,such clusters must initially nucleate from vapor or liquidphase precursors, with the overall nanomaterial formation ratelargely dependent on cluster stability and reactivity [13–15].

Both practical applications and fundamental studies ofclusters require their stable formation by simple and

Nanotechnology

Nanotechnology 25 (2014) 385601 (10pp) doi:10.1088/0957-4484/25/38/385601

0957-4484/14/385601+10$33.00 © 2014 IOP Publishing Ltd Printed in the UK1

inexpensive processes. Regrettably, stable metal nanoclusterformation (i.e., the formation of clusters which do not con-tinue to grow into larger nanoparticles or aggregates) remainsa challenge. In prior studies, stable metal nanoclusters havebeen produced in the gas phase at reduced pressures by a two-step process consisting of laser ablation of a solid metal fol-lowed by mass-selection using a quadrupole mass filter[16–19]. These cluster beams are well-controlled and highpurity, but the complexity of these systems makes it difficultto scale up and may result in nanoclusters distinct in structureand physicochemical properties from those produced inhigher pressure environments. Conversely, at higher gaspressures, gas-phase synthesis by, for example, flame synth-esis, typically results in larger particles and aggregates [20].Metal nanoclusters have alternatively been prepared in theliquid phase by reduction of metal cations in the presence of awide variety of large molecule ligands, including dendrimers[21], DNA templates [22], metal-organic frameworks (MOFs)[23], polymers [24], and other organics [25, 26]. These ligandmolecules effectively mitigate cluster growth and aggrega-tion. Liquid-phase methods can be much simpler and lesscostly than cluster beams. However, the presence of ligandson the surface of the clusters poses a significant problem inboth the characterization of nanoclusters as well as theirapplication; removal of the surface ligands leads to coarsen-ing of nanoclusters into larger particles or aggregates [27].

Recently, low-temperature plasma-based processes havebeen developed for the continuous synthesis of metal [28] andsemiconductor [29] nanoparticles (with diameters >∼3 nm).In general, plasmas are a source of energetic species,including electrons, ions, and metastables, which can dis-sociate molecular precursors near room temperature throughgas-phase collisions. Among these, microplasmas are uniquebecause of their operating regime that allows nanoparticles tobe produced at atmospheric pressure conditions [17], withuniquely short residence times (millisecond to sub-milli-seconds). Here, we show that isolated (unagglomerated),ligand-free metal nanoclusters with diameters as small asfractions of nanometers and corresponding to less than 101

atoms can be stably produced in a microplasma process.While the microplasma synthesis procedure is quite general,we focused on the formation of nickel (Ni) nanoclusters frombis(cyclopentadienyl)nickel(II) or nickelocene, an organo-metallic precursor for Ni. The precursor was continuouslyintroduced as a vapor in a gas flow into the microplasma anddissociated to Ni atoms. As the microplasma operates nearroom temperature, low kinetic energy collisions between Niatoms can effectively lead to nucleation of Ni nanoclusters.Rapid quenching by the combination of short residence times(∼1 ms) afforded by the nanoliter-sized volume and the highdilution flow limited cluster growth. Thus, nanoclusters areproduced by spatial control, without the need for stabilizingligands or mass selection, and clusters can be produced athigh pressures without aggregation. The as-grown nanoclus-ters were deposited on atomically-smooth mica and evaluatedby atomic force microscopy (AFM). In addition, since thenanoclusters are continuously produced at atmospheric pres-sure, online monitoring of cluster growth by ion mobility

spectrometry (IMS) is facilitated, and was performed using adifferential mobility analyzer of modest-to-high resolvingpower [30]. In support of the experimental measurements, gasmolecule scattering calculations [31, 32] were carried out oncandidate structures to estimate the number of atoms in thenanoclusters. Our analysis confirms that clusters containingless than 102 atoms are produced by this process. Further, itindicates that microplasmas are uniquely capable of produ-cing nanoclusters at atmospheric pressure, which may enablefundamental studies of nucleation, as these clusters couldserve as precursors for gas-phase reactions in a subsequentprocess, and technological applications, as the method ishigh-purity, simple, and low cost.

Experimental methods

Microplasma reactor operation

The operation of microplasma reactors for gas-phase nano-particle synthesis has been previously reported in detail[33, 34]. Briefly, in this work, a microplasma was formed in abackground of argon (Ar) gas with a direct current (dc) powersupply (Gamma High Voltage Model RR10-30P). Afterignition, the microplasma was sustained at a steady-statecurrent of 4.0 mA with an inline ballast resistor. Bis(cyclo-pentadienyl)nickel (II) [Ni(Cp)2, nickelocene] (STREM che-micals, >99% purity) was used as a precursor for Ninanocluster formation. The solid powder was loaded andsealed inside a stainless steel tube using a dry glove box, andconnected by a valve system to the inlet of the microplasmareactor. To form nanoclusters, the precursor was sublimed ina flow of Ar and diluted with an additional flow of pure Ar tocontrol the final vapor concentration. The total argon flowratethrough the reactor was ∼0.1 l min−1 under all operatingconditions, and the Ni precursor concentration entering thereactor ranged from 0–10 ppm. Nitrogen (N2) gas was addedat the exit of the microplasma reactor to limit furthernanocluster collisional growth prior to collection and analysis.

Cluster collection and materials analysis

As-grown nanoclusters were deposited at the reactor outletwith an electrostatic precipitator [35] (TSI, Model 3089) ontocarbon-coated Cu grids for transmission electron microscopy(TEM) characterization, or freshly-cleaved mica substrates(SPI supplies) for AFM characterization. A N2 dilution flowof 1.4 l min−1 was used following the microplasma reactor.TEM was performed at high-resolution with a Philips TecnaiF30 field-emission electron microscope operated at 300 kV.AFM was performed at ambient conditions using a scanningprobe microscope system (Agilent Technologies Model 5500)operated in tapping mode with an n-type silicon tip (NSC 15,Mikromasch USA).

Ion mobility spectrometry

Nanocluster formation was monitored online with a differ-ential mobility analyzer (DMA, 1/2 mini model,

2

Nanotechnology 25 (2014) 385601 A Kumar et al

Nanoengineering) [30] coupled to a Faraday cage electro-meter, with a N2 dilution flow of 5 l min−1 following themicroplasma reactor employed during DMA analysis. DMAsoperate as ion mobility filters, transmitting charged species ofa given ion mobility in the low field limit, and their appli-cation in the detection of nanometer scale species is describedelsewhere [36]. Prior to entering the DMA, nanoclusters werebrought to a near steady-state charge distribution via diffusioncharging (collisions) with ions produced in a Kr85 source. Atthis steady-state, charge distribution calculations [37] revealthat the majority of nanoclusters (sub 10 nm entities) areneutral, with a small percentage of non-neutral clusters havingcharge states of +1 or −1. Therefore, the nanoclusters trans-mitted through the DMA under prescribed operating condi-tions (sheath flowrate and voltage applied across electrodes)had known mobilities, K, linked to their collision crosssections, Ω, by the Mason–Schamp equation [38]:

⎛

⎝

⎜⎜⎜⎜⎜

⎛⎝⎜

⎞⎠⎟

⎞

⎠

⎟⎟⎟⎟⎟

π

Ω=

+K

ze

N

m

m n

m kT

3

16

2 11

, (1)gas

gas

Ni Ni

gas

1/2

where Νgas is the gas number density, mgas is the average gasmolecule mass (both calculated assuming the gas compositionwas 0.1:5 Ar:N2), mNi is the molecular mass of Ni, nNi is thenumber of Ni atoms in the nanocluster, kT is the product ofBoltzmann’s constant and the gas temperature, e is the unitelectron charge, and z is the cluster charge state (+1 or −1). Toacquire ion mobility spectra, the voltage across DMA elec-trodes was stepped in increments of 10 V between 0 and4000 V at a fixed sheath flowrate. Both positive and negativeion mobility spectra were collected. DMAs are linear mobilityspectrometers (i.e., 1/K for the transmitted clusters is linearlyproportional to the ‘scanned’ parameter, the applied voltage)and therefore require only a single point calibration for a fixedsheath flowrate. As previously reported [39, 40], DMA cali-bration was carried out via measurement of thetetraheptylammonium+ (THA+) ion, whose mobility wasmeasured by Ude and Fernandez de la Mora [41] in air andadjusted for the gas composition used in experiments(assuming the collision cross section of THA+ is relativelyinsensitive to the change to an Ar–N2 mixture from air).

Results and discussion

AFM measurements

We have previously reported the synthesis of crystalline,unagglomerated, narrowly-dispersed Ni nanoparticles withtunable diameters between 2.5 and 5 nm by an atmospheric-pressure microplasma process [42–44]. Here, the same pro-cess and procedure was used to synthesize Ni nanoclusters,except that the nickelocene precursor was limited to sig-nificantly lower vapor concentrations in order to control thefinal cluster size by reaction limitation (see table S1, sup-porting information). The clusters were initially collected and

characterized by AFM, which is becoming an increasinglyimportant tool for the characterization of nanoclusters[45, 46]. Although TEM can provide more information aboutthe material (e.g., crystallinity), energy transfer from theelectron beam can compromise as-synthesized nanoclusterstructures [47], sizes, and densities [48], particularly forsubnanometer clusters where an electron beam is focused to<1 nm. Additionally, AFM can offer better resolution, with atypical signal-to-noise of ∼0.1 nm in step height for atom-ically-smooth substrates. We used freshly-cleaved mica as thesubstrate, which was measured to have a root mean squaredsurface roughness of 0.07 nm (figure S1, supporting infor-mation). Because AFM characterization was carried out atambient conditions, oxidation of the Ni clusters from expo-sure to air and water could not be avoided; we attempted tominimize oxidation by collecting and imaging samples withinthe same day. To verify the presence of Ni, larger nano-particles were deposited on carbon-coated Cu grids andanalyzed by HRTEM (figure S2, supporting information).AFM height profiles were obtained by measuring each clusterindividually; the diameter of each nanocluster was assumed tobe equivalent to its height. Figure 1(a) shows a representativetapping mode AFM image of Ni clusters synthesized at anickelocene vapor concentration of 0.89 ppm. The clusterswere deposited at a constant flow rate for 15 min; the col-lection time was optimized to avoid excessive surface cov-erage. A representative HRTEM image of a Ni clustersynthesized at the same nickocelene vapor concentration isshown in the inset of figure 1(a), indicating that the clustersgrown by this method are crystalline. AFM images of samplesprepared at nickelocene vapor concentrations of 0.47 ppm and0.30 ppm are similarly shown in figures 1(b) and (c),respectively. At these concentrations, we were unable toobtain TEM images. The line profiles show representativenanoclusters with heights of 12 ̇A at 0.47 ppm, and 5 ̇A at0.30 ppm.

A more comprehensive analysis was performed by ima-ging and sizing clusters as a function of the nickelocene vaporconcentration (see figures S3 and S4, supporting information).Histograms of nanocluster heights from AFM image analysisare shown in figures 2(a)–(c) for 0.89 ppm, 0.47 ppm, and0.3 ppm nickelocene vapor concentrations, respectively. Thecorresponding mean diameters and standard deviations of thenanoclusters obtained by fitting the distributions at eachprecursor concentration are indicated on each histogram. Wenote that at 0.3 ppm, the size distribution is extremely narrowwith over 90% of the clusters between 2.5 and 5 ̇A. Thisfinding is unique when compared to alternative gas phasemethods, in which broader size distributions necessitate size(mass) selection or the smaller nanoclusters are lost to thewalls or aggregation [49].

Ion mobility spectra

Ion mobility spectra measured at the microplasma reactoroutlet as a function of varying nickelocene precursor con-centrations are shown in figure 3, represented by the rela-tive electrometer signal detected as a function of

3

Nanotechnology 25 (2014) 385601 A Kumar et al

Ω +−( )1

m

m n

1/2gas

Ni Ni, which is inferred from the mobility of

nanoclusters transmitted through the DMA. Both positiveand negative spectra were acquired, and are found to besimilar for a given nickelocene vapor concentration, withslightly more negatively-charged nanoclusters detected. Thissignal difference is attributable to the slightly lower averagemass and higher mobilities of negative ions produced bythe Kr-85 source as compared to positive ions; this leads tomore efficient negative diffusion charging of nanoclusters[38]. The difference in ion mobilities between Kr-85 gen-erated positive and negative ions is further confirmed bycomparing positive and negative mode spectra in theabsence of nickelocene vapor (see figure 3); both spectra

contain a peak near 100 ̇A2, with the peak in the negativemode slightly shifted to the left as compared to the positivemode. While the detection of such ions demonstratesclearly that the DMA employed can detect ions with

collision cross sections below approximately 100 ̇A2, theirpresence also hinders direct observation of nanoclusters inthis collision cross section range (∼20 or fewer atoms). Inexperiments in which the Kr-85 bipolar ion source wasremoved, we were unable to detect either positively- ornegatively-charged nanoclusters, indicating that the gener-ated nanoclusters are predominantly neutral upon exitingthe microplasma reactor. Therefore, an ionization scheme isnecessary for IMS. In future measurements, the use of

Figure 1. AFM images of Ni nanoclusters synthesized in amicroplasma process at nickelocene vapor concentrations of (a)0.89 ppm, (b) 0.47 ppm, and (c) 0.30 ppm. Below (b) and (c) are theobtained height profiles as the AFM tip is moved long the depictedlines. Inset of (a) shows a HRTEM image of a larger Ni nanoclustersynthesized at 0.89 ppm. The lattice spacing confirms face-centeredcubic (fcc) crystalline structure.

Figure 2. Histograms of nanocluster heights obtained via AFManalysis of Ni nanoclusters synthesized in a microplasma process atnickelocene vapor concentrations of (a) 0.89 ppm, (b) 0.47 ppm, and(c) 0.30 ppm. Measured heights are binned by 1 ̇A.

4

Nanotechnology 25 (2014) 385601 A Kumar et al

ultraviolet photoionization [50] may mitigate the influenceof charging ions on nanocluster ion mobility spectra.

With increasing nickelocene vapor concentration, a sec-ond peak appears in both the positive and negative ion

mobility spectra, below 103 ̇A2 at 0.2 ppm, and extending

beyond 104 ̇A2 when the nickelocene vapor concentration wasincreased to 4 ppm. This trend agrees qualitatively well withAFM measurements of nanoclusters. Furthermore, whiledetection of nanoclusters via ion mobility is hindered by theKr-85 generated ion signal, the Ni nanocluster distribution iscontinuous, suggesting that at the lowest examined nick-elocene vapor concentration of 0.2 ppm, clusters less than

100 ̇A2 in collision cross section, which corresponds to adiameter below 10 ̇A, are produced.

Cluster diameter and collision cross section calculations

AFM and IMS provide information on the nanocluster sizedistribution in different manners, in terms of height in the caseof AFM and in terms of collision cross sections in the case of

IMS. Although height is related to cluster diameter andmobility measurements can be used to infer an effectivemobility diameter [51], the underlying parameter that trulylinks these two measurements is the number of atoms percluster (which governs the structure of each cluster). For thisreason, we estimate the number of atoms in the Ninanoclusters via two approaches, enabling comparison to bothAFM and IMS measurements. First, approximating Ninanoclusters as spheres with the properties of bulk matter(i.e., a solid density of ρNi = 8908 kg m

−3), nanocluster dia-meters (dnc) were approximated as:

⎛⎝⎜⎜

⎞⎠⎟⎟πρ

=dn m6

(2)ncNi Ni

Ni

1/3

providing a correlation between nanocluster diameter andnumber of atoms for AFM measurements. For ion mobility

measurements, we assume again that Ω +−( )1

m

m n

1/2gas

Ni Ni,

which is the unknown parameter in equation (1), is linked todnc and hence nNi via the approximation [32, 41]:

⎛⎝⎜

⎞⎠⎟ ⎛

⎝⎜⎞⎠⎟

Ω ψπξ

+ = ϒ+

+

−

( ) ( )m

m n

d d

m

m n

a1

4 1

, (3 )gas

Ni Ni

1/2

pol

nc g2

gas

Ni Ni

1/2

where dg is the effective gas molecule diameter (approxi-mated as 3 ̇A based on prior measurements), ξ is the gas

molecule momentum transfer factor, and ψϒ( )pol , which

corrects for the influence of ion-induced dipole potential ongas molecule motion about clusters, is approximated as [32]:

⎞⎠⎟

ψ ψ

ξψ

ϒ = +

+ +

( )( )

b

1 (0.322

10.0625 0.1212

. (3 )pol pol

pol

ψpol, the polarization energy to thermal energy ratio, isdefined as [51]:

ψα

πε=

+( )z e

kT d dc

2, (3 )pol

pol2 2

0 nc g4

where ε0 is the permittivity of free space, αpol is the polar-izability of the background gas and kT is the product ofBoltzmann’s constant and the gas temperature. In imple-menting equations (3a)–(3c), we use the valuesαpol = 1.7 × 10

−30 m3 (the approximate value for both Ar andN2), T= 30 K, and ξ = 1.36. The value of ξ is based on aneffectively diffuse gas molecule scattering model, which hasbeen validated experimentally [32], and can result because ofboth (1) multiple collision events between gas molecules andclusters upon close approach and (2) exchange betweentranslational, vibrational, and rotational degrees of energywithin both a nanocluster and non-monoatomic gas moleculewhen the two entities collide. Equation (3b) is a regression fitto numerical simulations and applies only in instances whereψ < 1pol . Well above this value (ψ → ∞pol ), the collision

Figure 3. Ion mobility spectra of (a) positively-charged and (b)negatively-charged species synthesized in a microplasma process atthe indicated varying nickelocene vapor concentrations. Spectra areexpressed as the detected electrometer signal intensity (arbitrary

units) as a function of the parameter Ω +−( )1

m

m n

1/2gas

Ni Ni, which is

inferred via DMA calibration.

5

Nanotechnology 25 (2014) 385601 A Kumar et al

cross section approaches the polarization limit [52]:

⎛⎝⎜

⎞⎠⎟

⎛

⎝

⎜⎜⎜⎜⎜⎛⎝⎜

⎞⎠⎟

⎞

⎠

⎟⎟⎟⎟⎟

Ω

α

ε

+

=+

−m

m n

e

kTm

n m

d

1

0.9026

1

. (3 )

gas

Ni Ni

1/2

pol

0gas

Ni Ni

1/2

However, for nanoclusters, bulk approximations are notnecessarily valid, particularly in instances where dnc approa-ches the subnanometer length scale [53]. Therefore, as asecond method to link nanocluster diameter to atomic struc-ture, we carried out density functional theory calculations toinfer local energy minimum structures for positively chargedNi nanoclusters with nNi = 2–10, and Hartree–Fock levelcalculations to infer local energy minimum structures forneutral nanoclusters with nNi = 12 and 15. Calculations werelimited to small numbers of atoms, because of known diffi-culties in probing larger transition metal cluster structurescomputationally [54]. Larger Ni cluster structures have beenpreviously examined using molecular dynamics (MD), but thepotentials that are invoked are semi-empirical/phenomen-ological in nature [55–57]. Therefore, such simulations mayonly provide qualitative information about cluster structure(i.e., cluster geometries may match those observed experi-mentally, but the interatomic bond distances and clusterdensities obtained are not necessarily accurate). As the goal ofour calculations is to provide estimates of cluster structuresfor subsequent use in gas molecule scattering calculations anddetermination of collision cross sections, we opt not to utilizeMD calculations here. Further, as the collision cross section istypically insensitive to small changes in cluster structure, it isnot necessary to probe the entire energy landscape of clustersin DFT and HF calculations; rather, local energy minimumstructures are sufficient for estimating collision cross sections.All DFT and HF calculations were performed with GAUS-SIAN (Gaussian, Wallingford, CT). For nNi = 2–10, theB3LYP density functional [58] was employed and the basisset LANL2DZ was used [59–61]. For nNi = 12 & 15, the basisset LANL2DZ was used with Hartree–Fock (HF) level cal-culations, while basis set 3-21G was used for the secondisomer of nNi = 15. Each structure presented represents astationary point on the energy surface. The coordinates of thecenters of atoms in all obtained clusters are provided in thesupporting information, and depictions of each of the clusterstructures obtained are shown in figure 4. For the nNi = 4, 9,and 15 clusters, two local energy minimum structures areobtained and displayed. However, the subsequently describedcalculations differ by less than 1% in both inferred diametersand collision cross section for different isomers, hence for theremaining clusters only one structure for each examinednumber of Ni atoms is utilized.

To compare the cluster structure calculations with AFMmeasurements, we calculated the orientationally-averagedprojected area [31, 32] of each cluster (Ap), and then used the

relationship dnc = (4Ap/π)1/2− dg (subtracted because the gas

molecule was used as a probe in projected area calculations).Each Ni atom was approximated as a sphere with a diameterof 2.48 ̇A, and the inferred diameter was compared to theheights measured by AFM. Similarly, to compare DFT cal-culations to ion mobility measurements, collision crosssections were calculated based on a procedure previouslydeveloped by Larriba and Hogan [31, 32]. Briefly, collisioncross sections are determined by directly monitoring the tra-jectories of individual gas molecules about a cluster (releasedfrom the surface of a control volume), and from these tra-jectories the rate of momentum transfer from gas molecules toa cluster are determined, as the cluster migrates through thebackground gas at low speed (relative to the mean thermalspeed). Gas molecule trajectories were monitored consideringthe ion-induced dipole potential between all gas moleculesand the cluster (with the net charge placed at the clustercenter), at a temperature of 300 K in a background gas with apolarizability of 1.71 × 10−30 m3, and (approximating gasmolecules as spheres) with a gas molecule diameter of 3 ̇A.All clusters were assumed to be singly charged and composedof spherical Ni atoms with diameters of 2.48 ̇A. Crucial tocollision cross section calculations is the manner by whichgas molecule-cluster collisions, i.e., events wherein gasmolecule impinge on the cluster surface and are reemitted, aremodeled. Here, two different gas molecule-impingement andreemission rules were employed: (1) diffuse hard spherescattering (DHSS), in which impinging gas molecules arereemitted at randomly selected angles with speeds resampledfrom a Maxwell–Boltzmann distribution (with a slightlyreduced mean speed, as described by Larriba and Hogan [32])or (2) elastic hard sphere scattering (EHSS), in whichimpinging gas molecules are reemitted from cluster surfaceswith their translational kinetic energy conserved during col-lision at the specular reemission angle. In both DHSS andEHSS calculations, reemitted gas molecules were free toagain collide with the cluster if their trajectory upon reemis-sion leads them to do so. When applied to determine thecollision cross sections of structures with atomically roughsurfaces, and those which are considerably larger than gasmolecules, DHSS and EHSS collision rules can give rise tosimilar results (and can reproduce experimental measure-ments); if a gas molecule collides multiple times with astructure, then the net momentum change to the gas moleculeonce it leaves the structure surface is roughly independent ofcollision model. However, as noted previously [32], use ofdifferent collision rules for collision cross section calculationsfor entities similar in dimension to gas molecules themselvestypically gives rise to appreciably different results (i.e. DHSSand EHSS calculations may disagree with one another by∼30%), and at present a reliable method to model gasmolecule impingement-reemission from surfaces withoutmodeling both gas molecule and cluster internal degrees offreedom remains elusive. We therefore report the results ofboth calculation procedures here.

The inferred diameters and collision cross sections for allmodeled cluster structures are summarized in table 1. Thesevalues enable estimation of the number of atoms in clusters

6

Nanotechnology 25 (2014) 385601 A Kumar et al

observed in AFM and ion mobility measurements. Figure 5(a)

shows the parameter Ω +−( )1

m

m n

1/2gas

Ni Nias a function of nNi,

as determined from equation (3a), calculating ψϒ( )pol via

equation (3b), and assuming ψϒ =( ) 1pol (i.e., neglecting the

influence of the ion-induced dipole potential, which has oftenbeen done when examining larger nanoclusters [52]). Thisparameter is also plotted as it is determined fromequation (3d), and from calculations with both scatteringmodels. For the largest nanoclusters (nNi > 100), we expectequation (3a) to be reasonably accurate, irrespective ofwhether the influence of the ion-induced dipole potential isconsidered. For intermediate nanocluster sizes

(10 < nNi < 100), we additionally expect equation (3a) to holdvalid when accounting for ion-induced dipole potentialinfluences and provided that gas-molecule impingement andreemission from nanocluster surfaces is effectively diffuse innature. The validity of this equation for intermediate sizedclusters is supported by the reasonable agreement in predictedcollision cross sections with equation (3a) and for structuresusing the DHSS collision model (within 11% of one another).However, below nNi = 10, the relationship between collisioncross section and number of atoms per nanocluster is some-what ambiguous. With decreasing nNi, the bulk matterapproximation is expected to breakdown, collisions pre-sumably become less diffuse and more specular as the numberof internal degrees of freedom in a cluster decrease, and the

Figure 4. Depictions of local energy minimum Ni nanoclusters as determined via density functional theory for nNi = 2–10 and Hartree–Fockcalculations for nNi = 12 and 15. The number of atoms per nanocluster is noted. For density functional theory calculations clusters had a netpositive charge, while Hartree–Fock calculations clusters were neutral. Roman numerals I and II are displayed in instances where two isomerswere obtained.

7

Nanotechnology 25 (2014) 385601 A Kumar et al

ion-induced dipole influence on the collision cross sectionbecomes increasingly important (leading to little change inthe collision cross section with change in nNi). Nonetheless, inthis instance, the collision cross sections of clusters are likelybounded by DHSS (upper limit) and EHSS (lower limit)collision model predictions, hence detection of nanoclusters

with Ω +−( )1

m

m n

1/2gas

Ni Niin the 100 ̇A2 range indicates that

nanoclusters with less than ten atoms are in fact present.To compare calculations with experimental measure-

ments, the locations of the modes of the first and secondpeaks from mobility spectra acquired for clusters synthesizedat 0.20 ppm nickelocene (see figure 3) are shown as hor-izontal lines on figure 5(a). The points at which these linesintersect calculation results gives the approximate number ofatoms in measured clusters. The results indicate that clusterscomposed of <10–102 atoms (depending on the collision

cross section model) were detected by ion mobility mea-surements. Since ion mobility measurements can only detectcharged clusters and clusters with larger collision crosssections charge more efficiently, ion mobility measurementsmay underestimate the number of small collision cross sectionclusters. To correlate these results with AFM measurements,the diameter of clusters as a function of nNi from bothequation (2) and for computed structures is shown infigure 5(b). For reference, mean cluster diameters obtainedfrom AFM measurements at nickelocene vapor concentrationsof 0.30 ppm and 0.47 ppm, respectively, are shown as hor-izontal lines. The results are consistent with ion mobilitymeasurements and collision cross section calculations, con-firming the production of clusters composed of <10–102

atoms by our process.

Conclusions

We have demonstrated that a relatively simple microplasmaprocess can be used to generate stable Ni nanoclusters withspherical equivalent diameters below 10 ̇A at atmosphericpressure. AFM and IMS are used to independently char-acterize the produced nanoclusters. A combination of bulkmatter approximations and calculations based upon structuralmodels are used to correlate the AFM and IMS measurementsto the number of atoms in the produced nanoclusters. Overall,the analysis suggests that nanoclusters with less than 102

atoms can be produced. In future studies, the formation of

Figure 5. (a) The parameter Ω +−( )1

m

m n

1/2gas

Ni Nias a function of the

number of atoms in a nanocluster as determined from bulk matterapproximations (lines) and gas molecule trajectory calculations(symbols). The locations of the first and second peak modes from ionmobility spectra acquired for Ni nanoclusters synthesized in amicroplasma process at 0.20 ppm nickelocene are indicated. (b) Theeffective diameter of nanoclusters as determined from bulk matterapproximations (lines) and for density functional theory/Hartree–-Fock calculated structures (symbols). The locations of the meandiameters measured by AFM (assumed to be equivalent to theheights) for Ni nanoclusters synthesized in a microplasma process at0.30 and 0.47 ppm are indicated.

Table 1. Summary of the parameter, Ω +−( )1

m

m n

1/2gas

Ni Ni, estimated

by B3LYP density functional theory and Hartree–Foch calculations,for the proposed Ni cluster structures. The results are reported in ̇A2

using both diffuse hard sphere scattering (DHSS) and elastic hardsphere scattering (EHSS) models. Also shown are the projected areaequivalent diameters of clusters (in ̇A). B3LYP-Density functionaltheory calculated structures; HF-Hartree–Foch calculations. Thecharge state of the cluster is additionally noted via the superscript‘+’, ‘2+’ or ‘3+’ under the ‘Method’ column.

Ω +−( )1

m

m n

1/2gas

Ni Ni

#Ni atoms Method DHSS EHSS Diameter (Å)

2 B3LYP+ 100.73 80.16 3.403 B3LYP+ 107.46 82.85 4.134 B3LYP+ 113.76 84.18 4.696 B3LYP+ 120.93 85.97 5.374 B3LYP2+ 112.78 86.07 4.715 B3LYP3+ 118.12 87.19 5.268 B3LYP+ 128.79 87.67 5.987 B3LYP+ 124.59 89.78 5.739 B3LYP+ 131.37 93.90 6.279 B3LYP2+ 130.81 93.93 6.2510 B3LYP+ 135.32 97.49 6.5612 HF 149.45 107.21 7.3715 HF 152.76 110.79 7.7215 HF 155.26 111.86 7.59

8

Nanotechnology 25 (2014) 385601 A Kumar et al

nanoclusters by this novel technique may facilitate studies ofnanocluster gas phase reactions/heterogeneous uptake [62],enabling new insights into size dependent behavior in thenano- and subnanometer size range. Further, we suggest thatthe modeling approach introduced here, linking measure-ments to the number of atoms per cluster, be adopted insubsequent studies, as the number of atoms per nanocluster isthe most unambiguous metric of size.

Acknowledgements

Microplasma reactor construction and AFM measurementswere supported by the Air Force Office of Scientific Researchunder AFOSR Award No. FA9550-10-1-0160. Ion mobilityspectrometry measurements and density functional theorycalculations were made through the support of NSF-CHE-1011810. CLA acknowledges support from a Ramon ArecesFellowship. Finally, we thank the Minnesota SupercomputingInstitute for providing the high performance computing plat-form used in cluster structure calculations.

References

[1] Hakkinen H, Yoon B, Landman U, Li X, Zhai H J andWang L S 2003 On the electronic and atomic structures ofsmall Au-N(-) (N = 4–14) clusters: a photoelectronspectroscopy and density-functional study J. Phys. Chem. A107 6168–75

[2] Fernandez E M, Soler J M, Garzon I L and Balbas L C 2004Trends in the structure and bonding of noble metal clustersPhys. Rev. B 70 165403

[3] Krishnamurty S, Chacko S, Kanhere D G, Breaux G A,Neal C M and Jarrold M F 2006 Size-sensitive meltingcharacteristics of gallium clusters: comparison of experimentand theory for Ga17

+ and Ga20+ Phys. Rev. B 73 045406

[4] Neal C M, Starace A K and Jarrold M F 2007 Meltingtransitions in aluminum clusters: the role of partially meltedintermediates Phys. Rev. B 76 054113

[5] Starace A K, Neal C M, Cao B, Jarrold M F, Aguado A andLopez J M 2009 Electronic effects on melting: comparisonof aluminum cluster anions and cations J. Chem. Phys. 131044307

[6] Valden M, Lai X and Goodman D W 1998 Onset of catalyticactivity of gold clusters on titania with the appearance ofnonmetallic properties Science 281 1647–50

[7] Lei Y et al 2010 Increased silver activity for direct propyleneepoxidation via subnanometer size effects Science 328224–8

[8] Aiken J D and Finke R G 1999 A review of modern transition-metal nanoclusters: their synthesis, characterization, andapplications in catalysis J. Mol. Catal. A: Chem. 145 1–44

[9] Zheng J, Nicovich P R and Dickson R M 2007 Highlyfluorescent noble-metal quantum dots Annu. Rev. Phys.Chem. 58 409–31

[10] Shena P, Zhaoa S, Sub D, Lic Y and Orlov A 2012Outstanding activity of sub-nm Au clusters forphotocatalytic hydrogen production Appl. Catal. B: Environ.126 153– 160

[11] Shena P, Zhaoa S, Sub D, Lic Y and Orlov A 2012 Outstandingactivity of sub-nm Au clusters for photocatalytic hydrogenproduction Appl. Catal. B: Environ. 126 153– 160

[12] Tan X and Jin R 2013 Ultrasmall metal nanoclusters for bio-related applications WIRES Nanomed. Nanobiotechnol. 5569–81

[13] Zheng H, Smith R K, Jun Y-W, Kisielowski C, Dahmen U andAlivisatos A P 2009 Observation of single colloidal platinumnanocrystal growth trajectories Science 324 1309–12

[14] Girshick S L, Agarwal P and Truhlar D G 2009 Homogenousnucleation with magic numbers: aluminum J. Chem. Phys.131 134305

[15] Peng Z A and Peng X G 1993 Nearly monodisperse and shape-controlled CdSe nanocrystals via alternative routes:nucleation and growth J. Am. Chem. Soc. 124 3343–53

[16] Duncan M A 2012 Invited review article: laser vaporizationcluster sources Rev. Sci. Instrum. 83 041101

[17] Jarrold M F and Bower J E 1988 Chemisorption on themicrosurface of metal-clusters—activation barriers andchemical-reactions for CO, N2, O2, and CH4 on Al25

+ J. Am.Chem. Soc. 110 6706–16

[18] Benz L, Tong X, Kemper P, Lilach Y, Kolmalov A, Metiu H,Bowers M T and Buratto S K 2005 Landing of size-selectedAgn

+ clusters on single crystal TiO2 surfaces at roomtemperature J. Chem. Phys. 122 081102

[19] Winans R E, Vajda S, Ballentine G E, Elam J W, Lee B,Pellin M J, Seifert S, Tikhanov G Y and Tomcyzk N A 2006Reactivity of supported platinum nanoclusters studied byin situ GISAXS: clusters stability under hydrogen Top.Catal. 39 145–9

[20] Pratsinis S E 1998 Flame aerosol synthesis of ceramic powdersProg. Energy Combust. Sci. 24 197–219

[21] Zheng J and Dickson R M 2002 Individual water-solubledendrimer-encapsulated silver nanodot fluorescence J. Am.Chem. Soc. 124 13982–3

[22] Petty J T, Zheng J, Hud N V and Dickson R M 2004 DNA-templated Ag nanocluster formation J. Am. Chem. Soc. 1265207–12

[23] Houk R J T, Jacobs B W, Gabaly F E, Chang N N, Talin A A,Grahan D D, House S D, Robertson I M and Allendord M D2009 Silver cluster formation, dynamics, and chemistry inmetal-organic frameworks Nano. Lett. 9 3413–8

[24] Wang Y, Ren J, Deng K, Gui L and Tang Y 2000 Preparationof tractable platinum, rhodium, and ruthenium nanoclusterswith small particle size in organic media Chem. Mater. 121622–7

[25] Wu Z, Lanni E, Chen W, Bier M E, Ly D and Jin R 2009 Highyield, large scale synthesis of thiolate-protected Ag7 clustersJ. Am. Chem. Soc. 131 16672–4

[26] Metin O, Duman S, Dinç M and Özkar S 2011 Oleylamine-stabilized palladium(0) nanoclusters as highly activeheterogeneous catalyst for the dehydrogenation of ammoniaborane J. Phys. Chem. C 115 10736

[27] Menard L D, Xu F, Nuzzo R G and Yang J C 2006 Preparationof TiO2-supported Au nanoparticle catalysts from a Au13cluster precursor: ligand removal using ozone exposureversus a rapid thermal treatment J. Catal. 243 64–73

[28] Mariotti D and Sankaran R M 2010 Microplasmas fornanomaterials synthesis J. Phys. D 43 323001

[29] Kortshagen U 2009 Nonthermal plasma synthesis ofsemiconductor nanocrystals J. Phys. D 42 113001

[30] Fernandez de la Mora J and Kozlowski J 2013 Hand-helddifferential mobility analyzers of high resolution for1–30 nm particles: design and fabrication considerationsJ. Aerosol Sci. 57 45–53

[31] Larriba C and Hogan C J 2013 Free molecular collision crosssection calculation methods for nanoparticles and complexions with energy accommodation J. Comput. Phys. 251344–63

[32] Larriba C and Hogan C J 2013 Ion Mobilities in diatomicgases: measurement vs prediction with non-specularscattering models J. Phys. Chem. A 117 3887–901

9

Nanotechnology 25 (2014) 385601 A Kumar et al

[33] Sankaran R M, Holunga D, Flagan R C and Giapis K P 2005Synthesis of blue luminescent Si nanoparticles usingatmospheric-pressure microdischarges Nano Lett. 5 537–41

[34] Lin P A and Sankaran R M 2011 Plasma-assisted dissociationof organometallic vapors for continuous, gas-phasepreparation of multimetallic nanoparticles Angew. Chem.,Int. Ed. Engl. 50 10953–6

[35] Dixkens J and Fissan H 1999 Development of an electrostaticprecipitator for off-line particle analysis Aerosol Sci.Technol. 30 438–53

[36] Fernandez de la Mora J, de Juan L, Eichler T and Rosell J 1998Differential mobility analysis of molecular ions andnanometer particles Trend. Anal. Chem. 17 328–39

[37] Gopalakrishnan R, Meredith M R, Larriba-Andaluz C andHogan C J 2013 Brownian dynamics determination of thebipolar steady state charge distribution on spheres and non-spheres in the transition regime J. Aerosol Sci. 63 126–45

[38] Mason E A and McDaniel E W 1988 Transport Properties ofIons in Gases (New York: Wiley)

[39] Maiβer A, Premnath V, Ghosh A, Nguyen T A, Attoui M andHogan C J 2011 Determination of gas phase protein iondensities via ion mobility analysis with charge reductionPhys. Chem. Chem. Phys. 13 21630–41

[40] Ouyang H, Larriba-Andaluz C, Oberreit D R and Hogan C J2013 The collision cross sections of iodide salt cluster ionsin air via differential mobility analysis-mass spectrometryJ. Am. Soc. Mass Spectrom. 24 1833–47

[41] Ude S and Fernandez de la Mora J 2005 Molecularmonodisperse mobility and mass standards fromelectrosprays of tetra-alkyl ammonium halides J. AerosolSci. 36 1224–37

[42] Chiang W-H and Sankaran R M 2008 In-flight dimensionaltuning of metal nanoparticles by microplasma synthesis forselective production of diameter-controlled carbonnanotubes J. Phys. Chem. C 112 17920

[43] Chiang W-H and Sankaran R M 2008 Synergistic effects inbimetallic nanoparticles for low temperature carbonnanotube growth Adv. Mater. 20 4857

[44] Lin P A, Kumar A and Sankaran R M 2012 New insights intoplasma-assisted dissociation of organometallic vapors forgas-phase synthesis of metal nanoparticles Plasma Process.Polym. 9 1184–93 Invited to Special Issue,

[45] Love O, Kidwell D A and Epshteyn A 2013 The atomic metron—a basis for synthesis of metal sub-nanometer particles witha discrete number of atoms MRS Online ProceedingsLibrary 1548

[46] Brege J J, Hamilton C E, Crouse C A and Barron A R 2009Ultrasmall copper nanoparticles from a hydrophobicallyimmobilized surfactant template Nano Lett. 9 2239–42

[47] Li Z Y, Young N P, Di Vece M, Palomba S, Palmer R E,Bleloch A L, Curley B C, Johnston R L, Jiang J and Yuan J

2008 Three dimensional atomic-scale structure of size-selected gold nanoclusters Nature 451 46–8

[48] Ramalingam B, Mukherjee S, Mathai C J,Gangopadhyay K and Gangopadhyay S 2013 Sub-2 nm sizeand density tunable platinum nanoparticles using roomtemperature tilted-target sputtering Nanotechnology 24205602

[49] Ullmann M, Friedlander S K and Schmidt-Ott A 2002Nanoparticle formation by laser ablation J. NanoparticleRes. 4 499

[50] Jung T, Burtscher H and Schmidt-Ott A 1988 Multiplecharging of ultrafine aerosol-particles by aerosolphotoemission (APE) J. Aerosol Sci. 19 485–90

[51] Larriba C, Hogan C J, Attoui M, Borrajo R,Fernandez-Garcia J and Fernandez de la Mora J 2011 Themobility-volume relationship below 3.0 nm examined bytandem mobility-mass measurement Aerosol Sci. Technol.45 453–67

[52] Tammet H 1995 Size and mobility of nanometer particles,clusters and ions J. Aerosol Sci. 26 459–75

[53] Baletto F and Ferrando R 2005 Structural properties ofnanoclusters: energetic, thermodynamic, and kinetic effectsRev. Mod. Phys. 77 371–423

[54] Barden C J, Rienstra-Kiracofe J C and Schaefer H F 2000Homonuclear 3D transition-metal diatomics: a systematicdensity functional theory study J. Chem. Phys. 113690–700

[55] Cleveland C L and Landman U 1991 The energetics andstructure of nickel clusters: size dependence J. Chem. Phys.94 7376–96

[56] Michaelian K, Rendón N and Garzón I L 1999 Structure andenergetics of Ni, Ag, and Au nanoclusters Phys. Rev. B 602000–10

[57] Baletto F, Ferrando R, Fortunelli A, Montalenti F and Mottet C2002 Crossover among structural motifs in transition andnoble-metal clusters J. Chem. Phys. 116 3856–63

[58] Becke A D 1993 Density-functional thermochemistry. 3. Therole of exact exchange J. Chem. Phys. 98 5648–52

[59] Hay P J and Wadt W R 1985 Ab initio effective core potentialsfor molecular calculations—potentials for the transition-metal atoms Sc to Hg J. Chem. Phys. 82 270–83

[60] Hay P J and Wadt W R 1985 Ab initio effective core potentialsfor molecular calculations—potentials for K to Au Includingthe outermost core orbitals J. Chem. Phys. 82 299–310

[61] Wadt W R and Hay P J 1985 Ab initio effective core potentialsfor molecular calculations—potentials for main groupelements Na to Bi J. Chem. Phys. 82 284–98

[62] Oberreit D R, McMurry P H and Hogan C J 2014 Analysis ofheterogeneous uptake by nanoparticles via differentialmobility analysis-drift tube ion mobility spectrometry Phys.Chem. Chem. Phys. 16 6968–79

10

Nanotechnology 25 (2014) 385601 A Kumar et al