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

Download details:

IP Address: 138.37.211.113

This content was downloaded on 13/07/2014 at 00:28

Please note that terms and conditions apply.

Diffusion phenomena in atomic layer deposition

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

2012 Semicond. Sci. Technol. 27 074001

(http://iopscience.iop.org/0268-1242/27/7/074001)

Home Search Collections Journals About Contact us My IOPscience

IOP PUBLISHING SEMICONDUCTOR SCIENCE AND TECHNOLOGY

Semicond. Sci. Technol. 27 (2012) 074001 (8pp) doi:10.1088/0268-1242/27/7/074001

Diffusion phenomena in atomic layerdepositionMato Knez1,2,3

1 Max Planck Institute of Microstructure Physics, Weinberg 2, D-06120 Halle, Germany2 CIC nanoGUNE Consolider, Tolosa Hiribidea 76, 20018 Donostia-San Sebastian, Spain3 Ikerbasque, Basque Foundation for Science, Alameda Urquijo 36-5, 48011 Bilbao, Spain

E-mail: m.knez@nanogune.eu

Received 21 December 2011, in final form 19 January 2012Published 22 June 2012Online at stacks.iop.org/SST/27/074001

AbstractAtomic layer deposition (ALD) is a mature technology for the deposition of conformal thinfilms. During the ALD process or in a post-treatment, a variety of diffusion phenomena canoccur which can not only deteriorate the desired product, but also can be used to fabricatematerials or structures in a novel way. This special issue reviews some of the observeddiffusion processes and strategies to make use of those.

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

1. Introduction

The design of nanomaterials with various shapes anddimensions has seen many diverse strategies in the pastyears. Although the common goal is to manufacture nanoscalestructures, particles, wires, etc., the anticipated applicationdetermines the quality and the way in which the structuresare built. The ways include not only self-organization orwet-chemical growth as cost-efficient strategies, but alsolithographic methods for large scale structuring as moredemanding ones, particularly if electron beam lithography isapplied.

The enormous variety of strategies and synthesizedmaterials make it challenging to overlook those and onecan easily get the impression that for every aimed purposethe properly fitting nanostructure can be synthesized. Inmany cases, however, a rational combination of the desiredshape and functionality cannot be achieved. In such cases,often template-based approaches are applied, where a givenstructure is modulated in its functionality either by means ofcoating with another functional material of choice or evenby structural replication. Among the ways to functionalizesuch nanomaterials, especially atomic layer deposition (ALD)plays an increasingly important role as this method enablesthe deposition of extremely thin coatings, still maintainingstructural fidelity.

ALD is a vacuum-based coating technology, which isderived from chemical vapour deposition (CVD), with onedistinct technical difference: the occurring chemical reaction

is separated into two half-reactions in which initially thefirst chemical precursor (usually a reactive metal compound)chemisorbs to any available surface site of the substrateand subsequently, after a purging step, the second precursor(water, oxygen, diverse plasmas, etc.) is injected to react withthe adsorbed species. In this way, line-of-sight depositionis avoided and a controlled growth of the coating on theA-scale is achieved simply by repeating the depositioncycle. The precision in coating thickness and conformalitymade ALD meanwhile an important method-of-choice forfunctionalization and even replication of nanomaterials,especially in those cases where the materials have extremelycomplex shapes and aspect ratios [1, 2].

The coating of materials to add or replace functionalitiesappears straightforward for most cases in which ALD isapplied. Nevertheless, frequently side reactions can occur,intentionally or unintentionally, which alter the structural orphysical properties of the resulting material. Such alterationsmay not always be desired, but if they are chemically andphysically understood, they offer an additional degree offreedom to manipulate materials in a controlled way.

This non-exhaustive review will show some exampleson how diffusion phenomena can be beneficially used tomanufacture shapes and materials by ALD coating or by post-coating processing. Examples regarding interfacial diffusion,surface diffusion, diffusion/migration through coatings anddiffusion of precursors into the substrates during coatingprocesses (see figure 1) will be presented and discussed.

0268-1242/12/074001+08$33.00 1 © 2012 IOP Publishing Ltd Printed in the UK & the USA

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

Figure 1. Schematic overview of the various possible diffusionphenomena occurring during ALD coatings or duringpost-processing.

2. Interfacial diffusion

Most ALD-deposited materials require processingtemperatures between 150 ◦C and 250 ◦C and thus arefar below temperatures required for inducing solid-statediffusion reactions from a thermal point of view. Therefore,coating of a template will in most cases result in a core-shellstructure with a relatively sharp interface between the twomaterials. Nevertheless, even at those temperatures a slightintermixing of the coating with the substrate is possible. Thesolid-state diffusion will become more pronounced if energyis applied to the system, for example by means of heating. Inthis case, the mobility of the constituting parts of substrateand coating will induce interpenetration in both directions,eventually leading to a new compound. This diffusion isdependent on several factors, most importantly the chemicaland physical nature of the materials, the annealing temperatureand the annealing duration. In a simple case, the diffusionof material at the interface will be mutual, meaning that thematerials diffuse in both directions with similar velocity andquantity.

ALD allows coating of materials in a very precise way.Most of the ALD processes deposit binary oxides like Al2O3,ZnO, TiO2, etc. If needed for specific purposes, ternarycompounds or even higher order compounds can be deposited[3], but the selection of materials is rather sparse and theprocesses in many cases become complicated. Often, suchprocesses combine two or more individual ALD processes intoone. Since multiple precursors are required and in many casesthe thermal deposition windows of the individual processesmay not be compatible with each other, it becomes verydifficult to find parameters good enough for a reasonablegrowth control. Material diffusion during the deposition orinduced by post-process thermal treatment can be beneficiallyused to obtain such compounds in an alternative way. The keybenefits of ALD are maintained, since individual (e.g. binary)compounds can be grown very conformally with controlledthickness and the resulting mixed compound still shows acoating with good structural fidelity.

A number of such interfacial diffusion reactions havebeen reported to take place either at the boundary of theALD coating to the substrate or between two subsequentALD coatings. An early study [4] showed that ALD coatingand subsequent annealing of Al2O3 on Si induces inter-diffusion of the Al and Si species. A follow-up work [5]described that at the interface between Al2O3 and Si a mixedphase of Al2O3 and SiO2 forms with variable composition.

Further materials with a variable composition have beenobtained from Al2O3/HfO2 [6] and Ru/TaN [7] systems,although it is difficult to assign the intermixing to diffusionphenomena as in both cases the coating thickness was verysmall. Considering that an individual ALD cycle is usually notcovering a complete monolayer [3] so that some surface siteswill remain unoccupied, it is natural to assume that those siteswill possibly be occupied in the next half-reaction, resultingin a mixed compound driven by the surface reactivity ratherthan diffusion.

A better evidence for interfacial diffusion of materialsis given by controlled post-coating thermal treatment. Agood example is the formation of MgAl2O4 spinel nanotubes,square shaped in their cross-section, after annealing of MgOnanowires coated with Al2O3 (see figure 2) [8]. The inter-diffusion of Mg2+ and Al3+ at elevated temperatures leads tothe formation of the spinel. The remaining MgO core wasremoved wet chemically. The uniqueness of this procedure isreflected in the resulting single crystalline nanotubes, althoughthe initial Al2O3 coating was amorphous. Individual processesfor MgO and Al2O3 deposition are available and a combinationthereof may be achieved. Nevertheless, it is hardly imaginablethat such a combined process would lead to single crystallinestructures, even after annealing. Such an attempt was triedwith the combination of a SrCO3 and TiO2 process toobtain a SrTiO3 film [9]. Indeed, crystalline SrTiO3 formedafter annealing at 650–900 ◦C, although not being singlecrystalline.

However, not all diffusion processes result in singlecrystalline structures. In another case, a solid-state interfacialdiffusion reaction after ALD coating was used to produce amixed phase of CoO/ZnCo2O4/ZnO [10]. The intention ofthis experiment was to synthesize thin layers of ferromagneticmaterials, which appeared to be unsuccessful for this materialcombination. Nevertheless the experiment shows anothercouple of binary oxides which undergo interfacial diffusion.

The lack of single crystallinity may be considered adrawback, but in fact it is not. Frequently, just the materialcomposition is of importance and grain boundaries, especiallyif the film thickness is very low, may even improve somephysical characteristics like the conductivity. This is, forexample, the case of yttria-stabilized zirconia (YSZ). Thiscompound is used as an electrolyte in solid oxide fuel cells(SOFCs). Ginestra et al [11] have shown that after depositingmultilayer structures of Y2O3 and ZrO2 and annealing upto 950 ◦C in an oxygen atmosphere, a mixed compoundcan be produced which outperforms bulk YSZ in terms ofconductivity. As a reason for the good conductivity, the smallgrain size and the multitude of grain boundaries are proposed.

Some material combinations show a very special case ofinterfacial diffusion. If a material flux in one direction acrossthe interface cannot be compensated with an equal materialflux in the counter direction, this can be compensated with aunidirectional flux of vacancies which accumulate to voids onone side of the interface. This unequal material flux is calledthe Kirkendall effect [12, 13] and has been known for decadesin metallurgy. In nanoscience, this effect was often used forthe synthesis of hollow nanomaterials [14]. On ALD-based

2

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

Figure 2. Upper image: schematics of the fabrication process of MgAl2O4 spinel tubes. Lower image: high magnification TEM image of apart of the single crystalline spinel tube. Figures adapted with permission from [8]. Copyright 2006 by the Institute of Physics Publishing.

systems, the Kirkendall effect was observed on three materialcombinations: Fe2O3/ZnO [10], SiO2/ZnO [15] (mediated byAu migration) and Al2O3/ZnO [16].

In the former case, at the interface of ALD-depositedFe2O3 and ZnO upon annealing at 600 ◦C and 700 ◦C, voidsare observed as well as the formation of a ZnFe2O4 spinel.The resulting film exhibits ferromagnetism after the spinelformation, which was not observed in the as-depositedbilayer. Most research effort has been, however, investedin the Al2O3/ZnO bi- or multilayer system, which results inthe formation of a ZnAl2O4 spinel [10, 16–24] and voids.In all those cases, upon annealing, Zn2+ diffusion into theAl2O3 layer occurred, leaving the hollow area behind. Theinteresting aspect of making use of the Kirkendall effect isthe possibility of converting bulk to shell. It becomes possibleto produce nanochannels in various shapes, dependent onthe morphology of the ZnO. In the very simple case ofZnO nanowires, the transformation leads to spinel nanotubes[16, 23], which can even have higher order hierarchy [24].Rippled ZnO nanowires will result in peapod-like structuresof ZnO embedded in ZnAl2O4 [17], and coiled nanofibres inhelical nanotubes [22].

A very interesting aspect of the Kirkendall-basedtransformation is the possibility of delaminating films initiallybound to each other. Formation of interconnected voids in 2Dwas observed on planar and V-shaped multilayer structures[17], but this was even more expressed in the case of multilayercoating of fibres [21]. The thickness control of the ZnO layersis of critical importance as the ZnO needs to be completelyconsumed. In multilayered alternating alumina/ZnO films, theZn2+ can diffuse bi-directionally and result in isolated spinellayers forming tube-in-tube structures (see figure 3). With thethickness precision of the ALD layers and the resulting high

Figure 3. Typical TEM image of Al2O3/ZnO/Al2O3 microtubes(∼18/10/44 nm) annealed at 700 ◦C for 12 h. Broken tubes clearlyreveal physical separation and delineation between the inner andouter spinel tubes. The image confirms increased roughness of thespinel layers along the length of the tube, on the surfaces adjacent tothe Kirkendall voids. Image adapted with permission from [21].Copyright 2009 by the American Chemical Society.

crystallinity, one may envision this procedure being attractivefor numerous applications. Freestanding thin films are, forexample, very interesting for application as a support for TEMinvestigation.

One should, however, be aware that not every voidformation relies on the Kirkendall effect. ZnO or CuOnanowires coated with a titania film also form voids uponexposure to UV light either in aqueous environment [25] or

3

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

simply by storage under ambient laboratory conditions [26].The voids are a result of photocatalytic etching of the bulknanowires by the titania shell and not by the diffusion of ZnOinto the shell as can be deduced from the fact that the resultingnanotube does not contain Zn after completion of the ZnOremoval.

3. Surface diffusion

The ALD process is often seen in a static model: precursorchemisorbs at the substrate surface and reacts with the counterprecursor in the next half reaction. However, the reality isin most cases more complicated. The adsorption of the firstprecursor will initially be of physical nature (physisorption)which goes hand in hand with surface diffusion until a reactivesurface site for chemisorption is reached. The diffusiontendency of the precursor is dependent on several factors,among them are the chemical nature of the substrate which isoften related to the process temperature, the dissociation abilityand the binding energy of the resulting precursor/substratecomplex, and the differences in the surface free energy ofthe two materials, the substrate and the coating [27–29]. Thesurface diffusion may occur in two stages of the ALD process:(i) the first precursor may diffuse on the substrate during thefirst pulse/purge period, or (ii) the product may diffuse afterthe pulse of the second precursor. An exact judgement of thediffusion tendency of either of those species, the dissociatedprecursor or the product, is sometimes difficult. However, arough estimation can be made if one considers the Paulingelectronegativity (XM) of the molecules/atoms. Fu and Wagner[30] explained this in more detail on the example of metalatom adsorption on TiO2 surfaces. Metals with XM > 1.9 (e.g.noble metals) will show a tendency to adsorb to Ti sites of thesurface, while metals with lower XM will preferentially bindto oxygen sites. Keeping this in mind, it is not astonishingthat noble metal ALD on dielectric substrates will lead tocluster and island formation as the metals upon reaction withthe second precursor may easily diffuse on the substrate tofind defective sites, kinks, or steps where an interaction withthe metal site of the substrate instead of the oxygen site canoccur. Consequently, the initial metal seed keeps on growingwith each ALD cycle, since this surface site is preferential forfurther growth rather than the dielectric substrate. The islandgrowth can be beneficially used for the formation of metalnanoparticles by ALD [31]. The particle size can even becontrolled with the number of ALD cycles with very highprecision. Figure 4 shows an example of Pt nanoparticlesgrown on SrTiO3 nanocubes by ALD [32]. Within the initialfive ALD cycles one can observe that surface diffusion leadsto nanoparticle coalescence (figure 4(c)) and to a decrease ofthe surface-to-volume ratio (figure 4(d)). Such access to finetuning of the particle sizes opens new ways for the synthesisof supported catalysts.

Surface diffusion can also be used for post-processnanostructure synthesis. Considering CuO nanowires coatedwith metal oxides by ALD, one initially observes a veryconformal coating. Upon reduction of the coated nanowires atelevated temperatures in a H2 atmosphere, the CuO is reduced

(a) (b)

(c ) (d )

Figure 4. SEM images of SrTiO3 (STO) nanocubes before (a) andafter (b) coating with Pt nanoparticles using three Pt ALD growthcycles. (c) Centre-to-centre interparticle spacing (D). The lineardependence of D on the ALD cycle shows that the number densityof nanoparticles is decreasing due to nanoparticle coalescence.(d) Comparison of the surface area to volume ratio (S/V) with theXANES ratio of Pt–O to Pt–Pt bonding. Images reprinted withpermission from [32]. Copyright 2009 by Wiley & Sons.

Figure 5. Schematics of the formation process of Cu nanoparticlesin metal oxide nanotubes. CuO nanowires are coated with Al2O3 andsubsequently reduced at elevated temperatures. The Rayleighinstability induces nanoparticle chain formation. Initial and finalstages of the process are shown with typical SEM images. Figureadapted with permission from [33]. Copyright 2008 by theAmerican Chemical Society.

to Cu. This reduction to metallic Cu causes a shrinkage involume and in parallel a de-wetting due to the high mobility ofthe Cu on Al2O3 [33]. The de-wetting at elevated temperaturestogether with the Rayleigh instability eventually results in theformation of Cu nanoparticle chains embedded in a metal oxideshell (figure 5).

4

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

Figure 6. Schematics of the process to produce CuxO nanoparticles by diffusion of Cu through an ALD coating. Initially, CuO nanowiresare coated with alucone. Subsequent annealing in a H2 atmosphere converts the alucone into porous alumina and reduces CuO to Cu. Furtherannealing in an O2 atmosphere induces diffusion of Cu through the pores and a formation of CuxO on the pore mouths at the outer shell.Figures adapted with permission from [38]. Copyright 2011 by the American Chemical Society.

4. Diffusion/migration through coatings

ALD is commonly referred to as an effective means ofproducing diffusion barriers. Most of the work in this fieldis focused on inhibiting diffusion, relevant for electronicapplications to prevent the diffusion of a conductor intothe dielectric (see e.g. [34]) or as moisture barrier forencapsulation of OLEDs. The conformal and pinhole-freenature of an ALD coating furthermore promises seriousbenefits over commonly used methods for blocking gaspermeation or metal diffusion [2, 35].

In some cases, however, diffusion or migration throughALD coatings may be desired. This is primarily the casewhen a post-coating chemical reaction beneath the ALDlayer is required. Diffusion through ALD coatings is stronglydependent on the chemical composition of the film, itscrystallinity, thickness and the nature of the molecules whichare supposed to diffuse through the coating (e.g. gases orwater vapour). Amorphous coatings with a thickness of afew to a few tens of nanometres are very promising for gooddiffusivity of small molecules, considering the results reviewedin [35]. However, Fe particles coated with 8 nm alumina in afluidized bed reactor showed exceptional oxidation resistance,indicating an effective gas diffusion barrier [36]. The questionarises whether or not the barrier is the driving force forhindering an oxidation or whether further physical effects, likethe confinement and the restriction of volume expansion of Feif oxidized to FexOy, play a role. An inverse experiment wasperformed by Qin et al [33], showing that a layer of 5 nm or20 nm thick Al2O3 does not prevent the encapsulated CuO tobe reduced to Cu if annealed in a H2 atmosphere. The reductionis apparent as the system gains space after CuO is transformedto Cu which subsequently undergoes an undulation andparticle chain formation based on Rayleigh instability. Anotherapproach by the same author [37] showed that sacrificial layers,deposited between the substrate nanowires (Au in this case)and the shell, could also be removed. The layers consisted of

Al2O3 or polyimide (deposited by MLD) and were removedwet chemically or simply by annealing in the case of aluminaor polyimide, respectively. It is not clear whether both ways aresolely based on diffusion through the coating, but particularlyin the latter case this is imaginable as a decomposition of thepolymer should lead to a variety of small molecules like CO,CO2, etc., which may easily pass the barrier.

A further way to induce transport through a compact ALDcoating was shown by Yang et al [15]. In this work, the coatingwas performed with SiO2 which at elevated temperaturesshows serious fluidity. The substrate was Au-coated ZnOnanowires. The Au layer disassembled into nanoparticleswhich migrated through the SiO2 to the surface in order tominimize the interface energy. Such a migration, however,will be limited to very few system combinations only.

A completely different way to achieve transport throughan ALD coating was recently shown by Qin et al [38].Coating of CuO nanowires was performed by molecularlayer deposition (MLD) of alucone [39], an organic–inorganic hybrid material, deposited by replacing water vapourwith ethylene glycol or similar organic compounds. Suchinorganic–organic hybrid coatings form nanopores ormesopores upon annealing by decomposition of the organicconstituent of the film [40]. Once the pores form, Cu fromthe core can diffuse through the coating to the outer surface ifthe system is annealed (see figure 6). Nanoparticles of copperoxide form on the outer surface which indicates that the out-diffusion of copper takes place much more rapidly than thein-diffusion of oxygen. This may be related to the size and thecomplicated pathways of the pores. An additional interestingeffect is that with thicker coatings more and smaller particlesare formed. An explanation for this is given with the numberof pore openings which increases if the porous shell becomesthicker. More nucleation spots are available where the out-diffusing copper encounters oxygen and forms CuxO particles.

5

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

(a)

(a) (b) (c )

(b) (c)

(f ) (e) (d )

Figure 7. Upper image: Schematics of a lithographic process. The pattern is produced from diblock-copolymers and one phase reinforcedwith alumina by infiltration. The remaining processing steps are common lithographic procedures. Lower image: three examples of patternstransferred from standing PMMA cylinders, in-plane PMMA cylinders and in-plane PMMA cylinders aligned using graphoepitaxy. Imagesreprinted with permission from [49]. Copyright 2011 by the American Chemical Society.

5. Diffusion of precursors into the substrate

The exposure of a substrate to gaseous precursors can insome cases also result in diffusion of those precursors intothe substrate. This will often happen if the substrate is asoft solid, for example, a polymer. Attempts to coat severaltypes of polymers like polystyrene (PS), polypropylene (PP),polyethylene (PE), etc. with Al2O3 resulted in clear evidencethat the trimethylaluminum (TMA) which was used asprecursor is absorbed by the polymer and induces a nucleationin the subsurface region of the polymer [41] upon exposure towater vapour. The ability of the polymer to absorb and retainTMA is dependent on its porosity or density and the functionalgroups the polymer exhibits. Poly(vinyl-chloride) (PVC) wasshown to absorb less TMA than PP or PE. The proposed modelfor a subsurface nucleation was later on directly observedby TEM investigation of fibres of PP, poly(vinyl-alcohol)(PVA) and polyamide 6 (PA-6) [42]. Due to the thermalexpansion of polymers, the subsurface growth is temperaturedependent [43], which is related to the enhanced diffusivity andincreased free volume fraction of the polymer. The diffusionis also limited by the dose of the ALD precursor and thediffusion time. Pushing those parameters to an extreme resultsin a saturation of the bulk polymer with TMA which uponhydrolysis forms a monolith of porous Al2O3 [44].

The possibility of infiltrating an ALD precursor intoa polymer together with the differences in the absorptionbehaviour of polymers with differing functional groups or

densities opens a completely new route for nanostructuresynthesis. The approach relies on the formation ofdomains of one polymer within the matrix of anotherpolymer by using block copolymers [45]. Nanofibres withinternal helical or doughnut-like structures were producedby confining polystyrene-block-poly(2-vinlypyridine) (PS-b-P2VP) in anodic aluminium oxide [46]. After removal fromthe template and selective swelling, the fibres were subjectedto a ZnO-ALD process with diethyl zinc (DEZ) and watervapour as precursors. The ZnO deposited at the outer shellof the fibres, but also at the internal walls of the polymer. Aremoval of the polymer resulted in a structural replicationby ZnO. In another series of publications, a PS-b-PMMA(poly methyl methacrylate) structured diblock-copolymer wasprocessed with various ALD processes (Al2O3, ZnO, TiO2,SiO2 and W) to produce inorganic patterns resulting from theprecursor diffusion through the polymer block and reaction atthe interface of the polymer domains [47–49]. After removal ofthe polymer, the resulting inorganic nanofeatures can be usedfor further lithographic patterning. The concept was proven towork on a variety of substrates, among them silicon, indium tinoxide and a Ni-Fe permalloy. The resulting structures dependon the features of the block copolymer and may resembleupstanding nanotips or even structured lines (see figure 7).

As a side effect, the precursor diffusing into the polymermay also induce a chemical reaction with the substrate[50, 51]. Such chemical changes will in some cases gohand in hand with changes in the physical properties

6

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

of the material. In various experiments with spider silk,collagen or poly-tetrafluoroethylene (PTFE) as substrates,the mechanical properties of the (bio)polymers have beenshown to significantly change as a direct consequence of thediffusion of precursors into the polymers and the chemicalreaction between those [52–54].

6. Conclusion

Diffusion phenomena related to ALD or to a post-treatmentprocess may have significant impact on the material orstructure which is subjected to coating. Diffusion of materialcan occur at material interfaces, at substrates, throughALD coatings, and into the substrate, strongly dependingon the chemistry of the substrate, the precursors, and thedeposited material. In most cases, diffusion phenomena maybe considered a drawback, but the examples in the manuscriptshow that many beneficial aspects can be found, be it for thesake of nanochannel formation, nanoparticle synthesis or evenfor producing metal-organic hybrid materials. Controlling thediffusion in various aspects may enable new ways of materialsynthesis and functionalization which was already shown on acouple of examples described above. A deeper understandingof the various diffusion processes is still required in order toexploit the processes for the benefit of new functional materialsand advanced technologies. Given the fact that more and moreresearch groups from various research fields and disciplinesare starting to use ALD for their particular purpose, morescientific input and novel ideas will develop in the near futureand diffusion phenomena will inevitably play an importantrole.

References

[1] Knez M, Nielsch K and Niinisto L 2007 Adv. Mater. 19 3425[2] Pinna N and Knez M (ed) 2012 Atomic Layer Deposition of

Nanostructured Materials (Weinheim, Germany:Wiley-VCH)

[3] Puurunen R L 2005 J. Appl. Phys. 97 121301[4] Jakschik S, Schroeder U, Hecht T, Krueger D, Dollinger G,

Bergmaier A, Luhmann C and Bartha J W 2003 Appl. Surf.Sci. 211 352

[5] Shulakov A S, Braiko A P, Bukin S V and Drozd V E 2004Phys. Sol. State 46 1145

[6] Park P K, Cha E S and Kang S W 2007 Appl. Phys. Lett.90 232906

[7] Kumar S, Greenslit D, Chakraborty T and Eisenbraun E T2009 J. Vac. Sci. Technol. A 27 572

[8] Fan H J, Knez M, Scholz R, Nielsch K, Pippel E, Hesse D,Gosele U and Zacharias M 2006 Nanotechnology 17 5157

[9] Kosola A, Putkonen M, Johansson L S and Niinisto L 2003Appl. Surf. Sci. 211 102

[10] Zolotaryov A, Goetze S, Zierold R, Novikov D, Birajdar B,Hesse D and Nielsch K 2010 Adv. Eng. Mater. 12 509

[11] Ginestra C N, Sreenivasan R, Karthikeyan A, Ramanathan Sand McIntyre P C 2007 Electrochem. Solid-State Lett.10 B161

[12] Kirkendall E O 1942 Trans. AIME 147 104[13] Smigelskas A D and Kirkendall E O 1947 Trans. AIME

171 130[14] Fan H J, Gosele U and Zacharias M 2007 Small 3 1660

[15] Yang Y, Yang R B, Fan H J, Scholz R, Huang Z P, Berger A,Qin Y, Knez M and Gosele U 2010 Angew. Chem. Int. Ed.49 1442

[16] Fan H J, Knez M, Scholz R, Nielsch K, Pippel E, Hesse D,Zacharias M and Gosele U 2006 Nat. Mater. 5 627

[17] Guder F, Yang Y, Goetze S, Berger A, Ragmir N, Hesse Dand Zacharias M 2010 Small 6 1603

[18] Guder F, Yang Y, Danhof J, Hartel A, Schwarz U Tand Zacharias M 2011 Appl. Phys. Lett. 99 023105

[19] Guder F, Yang Y, Goetze S, Berger A, Scholz R, Hiller D,Hesse D and Zacharias M 2011 Chem. Mater. 23 4445

[20] Jang Y W, Bang S, Jeon H and Lee J Y 2011 Phys. StatusSolidi B 248 1634

[21] Peng Q, Sun X Y, Spagnola J C, Saquing C, Khan S A,Spontak R J and Parsons G N 2010 ACS Nano 3 546

[22] Qin Y, Kim Y, Zhang L, Lee S M, Yang R B, Pan A,Mathwig K, Alexe M, Gosele U and Knez M 2010 Small6 910

[23] Yang Y, Kim D S, Knez M, Scholz R, Berger A, Pippel E,Hesse D, Gosele U and Zacharias M 2008 J. Phys. Chem. C112 4068

[24] Yang Y, Kim D S, Scholz R, Knez M, Lee S M, Gosele Uand Zacharias M 2008 Chem. Mater. 20 3487

[25] Kim D S, Yang Y, Kim H, Berger A, Knez M, Gosele Uand Schmidt V 2010 Angew. Chem. Int. Ed. 49 210

[26] Yang Y, Kim D S, Qin Y, Berger A, Scholz R, Kim H,Knez M and Gosele U 2010 J. Am. Chem. Soc.131 13920

[27] Huang R and Kitai A H 1992 Appl. Phys. Lett. 61 1450[28] Lee J P, Jang Y J and Sung M M 2003 Adv. Funct. Mater.

13 873[29] Elliott S D and Greer J C 2004 J. Mater. Chem. 14 3246[30] Fu Q and Wagner T 2007 Surf. Sci. Rep. 62 431[31] Yim S S, Lee M S, Kim K S and Kim K B 2006 Appl. Phys.

Lett. 89 093115[32] Christensen S T et al 2009 Small 5 750[33] Qin Y, Lee S M, Pan A, Gosele U and Knez M 2008 Nano

Lett. 8 114[34] Rossnagel S M, Sherman A and Turner F 2000 J. Vac. Sci.

Technol. B 18 2016[35] Park J S, Chae H, Chung H K and Lee S I 2011 Semicond. Sci.

Technol. 26 034001[36] Hakim L F, Vaughn C L, Dunsheath H J, Carney C S, Liang X,

Li P and Weimer A W 2007 Nanotechnology18 345603

[37] Qin Y, Liu L, Yang R, Gosele U and Knez M 2008 Nano Lett.8 3221

[38] Qin Y, Yang Y, Scholz R, Pippel E, Lu X and Knez M 2011Nano Lett. 11 2503

[39] Dameron A A, Seghete D, Burton B B, Davidson S D,Cavanagh A S, Bertrand J A and George S M 2008 Chem.Mater. 20 3315

[40] Liang X H, Yu M, Li J H, Jiang Y B and Weimer A W 2009Chem. Commun. 46 7140

[41] Wilson C A, Grubbs R K and George S M 2005 Chem. Mater.17 5625

[42] Spagnola J C, Gong B, Arvidson S A, Jur J S, Khan S Aand Parsons G N 2010 J. Mater. Chem. 20 4213

[43] Jur J S, Spagnola J C, Lee K, Gong B, Peng Qand Parsons G N 2010 Langmuir 26 8239

[44] Gong B, Peng Q, Jur J S, Devine C K, Lee K and Parsons G N2011 Chem. Mater. 23 3476

[45] Kim H C, Park S M and Hinsberg W D 2010 Chem. Rev.110 146

[46] Wang Y, Qin Y, Berger A, Yau E, He C, Zhang L, Gosele U,Knez M and Steinhart M 2009 Adv. Mater. 21 2763

[47] Peng Q, Tseng Y C, Darling S B and Elam J W 2010 Adv.Mater. 22 5129

7

Semicond. Sci. Technol. 27 (2012) 074001 M Knez

[48] Peng Q, Tseng Y C, Darling S B and Elam J W 2011 ACSNano 5 4600

[49] Tseng Y C, Peng Q, Ocola L E, Elam J W and Darling S B2011 J. Phys. Chem. C 115 17725

[50] Chang C Y, Tsai F Y, Jhuo S J and Chen M J 2008 Org.Electron. 9 667

[51] Zhang L, Patil A J, Li L, Schierhorn A, Mann S, Gosele U andKnez M 2009 Angew. Chem. Int. Ed. 48 4982

[52] Lee S M, Pippel E, Gosele U, Dresbach C, Qin Y,Chandran C V, Brauniger T, Hause G and Knez M 2009Science 324 488

[53] Lee S M, Pippel E, Moutanabbir O, Gunkel I,Thurn-Albrecht T and Knez M 2010 ACS Appl. Mater.Interf. 2 2436

[54] Lee S M, Ischenko V, Pippel E, Masic A, Moutanabbir O,Fratzl P and Knez M 2011 Adv. Funct. Mater. 21 3047

8