High Density Cobalt Nanostructures by Ion Beam Induced Dewetting

Ion beam induced dewetting of thin metallic films has emerged as a promising way to grow metallic nanoparticles in a controlled manner. Metal films tend to dewet from non-reactive substrates upon heating, forming islands, due to the high energies of metal surfaces and interfaces. Ion beam irradiation triggers the self organized formation of nanostructures, and allows control over nanoparticle size distributions, since dewetting is initiated by highly localized (in space and time) thermal spikes along the ion track. From 25 nm cobalt films grown by e-beam evaporation, followed by irradiation with 100 keV Ar ions, a high density array of nanostructures is formed. AFM scans show their morphological evolution as a function of ion fluence; optimum fluences for narrow particles size distribution have been identified. Glancing angle XRD and Rutherford Backscattering Spectroscopy show evidence of increased substrate exposure as dewetting proceeds. RBS indicates that within the optimal fluence range for nanostructure formation, sputtering and ion beam mixing are not the significant effects in this system. We present ion beam induced dewetting as a well controlled method of forming nanostructured catalysts from metal films. [DOI: 10.1380/ejssnt.2013.99]


I. INTRODUCTION
Metallic nanoparticles are of interest for a variety of scientific reasons and a wide range of applications, from magnetic memory arrays and plasmonic waveguides, to catalysts for the growth of CNT's and semiconductor nanowires [1][2][3].Metal films on non-reactive substrates tend to 'dewet' upon heating, as the high surface energies of metals drive surface morphological changes such as increased roughness, which lead to the eventual break-up of the continuous film into globules [4].This process has been recognized as analogous to the Rayleigh-Plateau hydrodynamic instability that results in the beading up of a liquid film on a surface and the conversion of a cylindrical stream of liquid into droplets [5].Thus dewetting of thin films has emerged as a promising 'bottom up' way to grow nanoparticles by using differences in surface energy to drive the morphological instability.While the droplets formed upon dewetting typically vary in size, efforts to influence the size distribution have focused on controlling parameters such as temperature, film thickness, and the use of templates to 'pre-structure' the substrates before the deposition of the thin film [6].Over the past decade, formation of nanocrystalline catalysts due to dewetting by thermal annealing of Pt films [7] and laser induced dewetting of Au, Co and Ni films have been reported [8][9][10].In the case of thermal annealing or laser heating, dewetting is initiated by spinodal nucleation of bare sites [6,11].
More recently, ion bombardment with beams in the tens to hundreds of keV energy range has been used to induce dewetting of metal thin films, to create nanostructures on insulating substrates, at room temperature [2,7].Here the structural change is initiated by a different mechanism, since the samples are at room temperature, but the ion beam induces the formation of highly localized 'thermal spikes' or 'nanometer sized high temperature zones' along its path [13,14].
Molecular Dynamics simulations of ion bombardment * Corresponding author: lnair@jmi.ac.in of materials have shown that in a very narrow (a few nm), roughly cylindrical region along the track of the ion beam, temperatures can go up to several thousand Kelvin for time scales of the order of less than a pico second [14][15][16].These delta-function shaped thermal perturbations which dot the film, initiate local melt zones that extend through the film and are manifested on the surface as accumulations centered at the point of incidence of the ions [17].This is evident in our AFM scans, initially manifesting as randomly spaced projections at low ion fluences, and upon further irradiation (at fluences ranges of 10-30 ions per surface atom), the entire film has been transformed into nanoclusters.
In this study we report the formation of cobalt nanostuctures following ion bombardment of 25 nm cobalt films with 100 keV Ar + ions.We have investigated the development of the film morphology as a function of increasing ion fluence using AFM, grazing incidence XRD and Rutherford Backscattering Spectroscopy.The motivation is to determine the optimal fluence and other conditions necessary for the creation of well controlled, high density nanostructures from thin films.

II. EXPERIMENTAL DETAILS
The 25 nm Co films were deposited using an electron beam evaporation source in a stainless steel evaporation chamber of the Target Lab at the Inter University Accelerator Centre (IUAC), New Delhi [18].The chamber was pumped with cryo and turbomolecular pumps and the base pressure was 1.5×10 −8 Torr before evaporation and 2-3×10 −8 Torr during evaporation of the films.The source metal is heated by a 4-pocket electron gun with a 6 kW power supply, interfaced to a controller.The output power of the e-gun is controlled by the PID method, according to the programmed deposition schedule.Online thickness measurement is done by a dual crystal thickness monitor, with film thickness defined within an accuracy of 1 Å, while the rate of deposition can be controlled with accuracy of 0.1 Å/s.
The substrates were of size 1 cm×0.5 cm, cut from com-mercially available Si(001)wafer, with the native oxide intact.The prepared substrates were degreased by sonicating in warm acetone and methanol and mounted on to glass slides before being placed in the evaporation chamber.The films were deposited with the substrates at room temperature at a deposition rate of 0.4 Å/s as monitored by the quartz crystal thickness monitor.After being kept in vacuum overnight as the source cooled, the films were removed from the chamber and stored in a dessicator at a vacuum of 10 −3 Torr.Ion beam irradiation was done in another chamber located at the Low Energy Ion Beam Facility beam line of the IUAC, with Ar + ions of 100 keV energy.The ion current for the Ar + beam was measured in the beam line by a Faraday cup, and again at the sample, with a suppression voltage for secondary electrons, and was maintained at 1.0 µA.The samples were attached on to a stainless steel mounting rod using silver paint at the back of the substrate, and irradiated until the desired fluence (as monitored by an ion counter) was achieved.The ion beam was rastered over an area larger than the sample size, so as to have uniform irradiation over the area of the film.The vacuum in the chamber at the time of irradiation was 8×10 −7 Torr.Ion fluences of 1×10 15 , 5×10 15 , 1×10 16 , 5×10 16 and 1×10 17 ions/cm 2 were used in our experiment and hence irradiation times ranged from a few seconds to several hours.
The films were characterized using AFM and XRD facilities provided by the Materials Sciences Group at the Centre.The samples were transferred in the dessicator to the AFM facility, and scanned using a Nanoscope IIIa Instrument (Digital/Veeco), in non-contact mode, with lateral resolution of 3 nm.Glancing incidence XRD was also done on these samples using a Bruker AXS D-8 xray diffractometer, with Cu Kα conventional x-ray source, Göbel mirror, LiF monochromator and scintillator detector.RBS measurements for surface composition were done using 2 MeV He ions from the Pelletron accelerator at IUAC.He + ions were bombarded at an angle of 9 • to the surface normal and the back scattered ions were detected at an angle of 170 • to the beam direction.Details of the RBS system are available at Ref. [19].

A. AFM
The AFM images obtained for various ion fluences are shown in Fig. 1.The scale of the scans is 1 µm in each case.These images are representative of a series of AFM scans taken at various locations on each sample.The features observed are also present on much larger area scans (up to 10 µm).The initially plain, random surface (Fig. 1(a)) starts to show the development of short range order after an ion fluence of 1×10 15 ions/cm 2 , in Fig. 1(b).The structures get more pronounced with increased ion fluence until, at a value of 5×10 16 ions/cm 2 (Fig. 1(d)), the film has been transformed into an array of well developed nanostructures of fairly uniform lateral size of ∼35 nm.Upon further irradiation, up to 1×10 17 ions/cm 2 , which was the highest fluence we used, we find that the structures formed agglomerate into larger  The structural evolution can be analyzed through the profiles of the AFM scans, which are shown in Fig. 2. The pristine, or as deposited film showed a random surface profile with no characteristic length scale.Upon irradia- tion with the Ar ion beam, at our initial dose of 1×10 15 ions/cm 2 , the surface roughness decreased and distinct structures emerged on the surface, presumably, at those points at which the ions have impinged on the surface.Using the atomic spacing of metallic cobalt as 0.250 nm, the atom density of Co is 1.6×10 15 /cm 2 ; hence at this fluence value, the probability is that about 62.5% of the atoms on the surface would have undergone a collision.
At the ion current of 1 µA that we use, the time interval between the impinging ions is less than a picosecond, but as these ions are distributed over the sample, at a fluence of 1×10 15 ions/cm 2 , each protrusion is likely to have been created by a single incident ion.At a fluence of 5×10 15 ions/cm 2 , (Figs. 1(c) and 2(c), more than three times the surface atom density) the features seen are much more well developed, of about 7 nm in height above the background and 30-40 nm in width at the base.There is clear evidence of short range order, with their spacing being fairly regular, of 80-90 nm, irrespective of which direction the surface height profiles have been taken.Figure 1(d) shows the scan of a sample which has ten times the fluence of Fig. 1(c), i.e., 5×10 16 ions/cm 2 , where the sample is now covered with a well developed array of nanostructures, all of approximately the same size lateral size at the base, of 30-40 nm as in Fig. 1(c), 5×10 15 ions/cm 2 , but the spacing has now reduced drastically to about 45 nm, to give a high density array of nanostructures on the surface.At this fluence, each surface atom has undergone multiple collisions (about 30 from the above surface atom density values) and the features seem to be the cumulative effect of the individual ion impingements.
Further irradiation beyond this fluence value results in a larger scale change of the surface profile, as seen in Fig. 1(e) and in the profile of Fig. 2(e), with broad island like structures of about 500 nm at the base being formed, with the 30 nm features superimposed on them.
The profiles in Fig. 2 show that the nanostructures formed have a characteristic feature size of about 30 nm at the base.This has been analyzed in detail and the width histograms are presented in Fig. 3.The black histogram in Fig. 3 has been obtained from the pristine surface of Fig. 1(a), which shows a random surface profile, with a spread of nanoparticle widths from 5 to 115 nm.The red histogram in Fig. 3 has been obtained from the high density array of nanostructures (Fig. 1(d)), and indicated the narrowing of the width range (15-60 nm), as compared to the pristine film.The characteristic size of the high density array of nanostructures that we observe appears to be an intrinsic feature of the metal film and efforts to determine the correlation with the surface free energy and interfacial energy of the metals and substrate as well as the incident ion energy are underway.
The Si(001) substrates we used were cut from commercially available single crystal wafer with the native oxide intact and were flat well within the 3 nm lateral resolution of the AFM used.The surface roughness of the initial pristine film is comparable even for films of different thickness which we had tried, of 10 nm and 50 nm.However, the size of the nanostructures increased for the 50 nm film (not presented here), to about 80 nm lateral size for a comparable ion fluence of 1×10 16 ions/cm 2 , the fluence at which the 25 nm films shows 30-35 nm structures.For the 50 nm film, at the 1×10 17 ions/cm 2 fluence value, the scale of the features increases, as the perturbations seem to coalesce over a larger distance, with the 80 nm features 'riding on' larger, 600-700 nm features present on the film.
We also studied 10 nm films and the features seen are on a similar scale as the 25 nm film.As this is the lowest thickness we have studied, we cannot conclusively state whether there is an intrinsic lower limit in the size of the nanostructures formed by the dewetting process, and whether the particle size can be decreased merely by reducing film thickness, without exploring this effect further, for a set of films of lower thickness.To summarize, we would like to say that the size and density of the nanostructured dots can be controlled by film thickness and ion fluence, but we are still working to determine the precise relationship.We have repeated this experiment three times now and are confident of the structures that we are able to obtain for the 25 nm films at the Ar + ion energy and fluences used.

B. XRD results
The glancing incidence X-ray diffraction scans are shown in Fig. 4. XRD scans were taken using the Bruker instrument mentioned shows a broad amorphous feature at scattering angle 2Θ of about 12 • , and a trace of a peak from the substrate at 56.5 • degrees, at about twice the background level.Upon ion beam irradiation, however, the peak from the substrate is enhanced significantly, and in the fluence range where the uniform nanostructures are formed, from 5×10 15 up to 5×10 16 , it is up to five orders of magnitude above the background, and has been identified as the (002) peak from the Si substrate [20].
The emergence of the substrate peak provides evidence for dewetting as irradiation proceeds.At the highest fluence we have used, the substrate peak decreases in magnitude, due to the degradation in the surface crystallinity of the Si wafer from continued ion impact, as some part of it is now exposed due to the large scale modification of the film, as seen in Fig. 1(e), Fig. 2(e) and Fig. 5 below.
At the fluence ranges that we use, SRIM calculations show that except for the highest fluences used, sputter- ing is not sufficient to remove more than ten percent of the cobalt film, as the sputtering yield for Ar + is 4.6 atoms/incident ion for a 100 keV Ar + beam.This is also shown in the RBS data presented below.

C. RBS Results
Rutherford Backscattering Spectroscopy was conducted using 2 MeV helium ions from a Pelletron accelerator at IUAC.He 1+ ions were bombarded at an angle of 9 • to the surface normal and the backscattered ions were detected at an angle of 170 • to the beam direction.The pristine film showed the Co peak at 1.5 MeV and the edge of the Si substrate starting at 1.15 MeV.For fluence ranges from 1×10 −15 to 1×10 −16 ions/cm 2 , where clear structural changes are observed in the AFM and XRD data, RBS showed no significant changes in the shape of the Co peak, indicating the Co stays on the surface.This suggests that sputtering is not a substantial effect in this system, at least at the initial fluences where the film gets restructured.SRIM calculations for this system show that for a fluence of 1×10 −16 ions/cm 2 , only 4.6% of the total number of Co atoms are sputtered.This has also been independently calculated from the RBS peak area, where it is of the same order ∼5%.
Careful observation of the Si edge shows a shift forward (towards higher energies) indicating that Si substrate is being exposed.And while the Co peak itself does not change in area up to a coverage of 1×10 −16 ions/cm 2 , there is a slight increase in peak height and reduction in peak width on the lower energy side.This indicates that more Co atoms are being seen by the undeviated incident He + ions, i.e. from being a continuous film, with the large majority of atoms being found below others, we now have a nanostructured morphology, with the numbers of 'surface' to 'bulk' or 'exposed to buried' atoms, increased.This is consistent with our XRD data, which shows the silicon substrate peak increasing at these fluences.
At the highest fluences we used, 5×10 −16 ions/cm 2 and 1×10 −17 ions/cm 2 , the Co peak height decreases as sputtering starts becoming significant (area is decreased by a factor of 23% and 46% respectively).The peak at 1.3 MeV is due to the embedded Ar atoms from the incident beam.The Si edge is also shifted forward considerably, indicating the exposure of the substrate from dewetting and sputtering.Our results are consistent with the work of Lian et al., who have formed ordered Co nanostructures from dewetting and sputtering of nanowires of varying thicknesses using an FIB system [1].

D. SRIM Calculations
We performed systematic SRIM calculations on this system for incident energies of Ar + ranging from 1 keV to 1 MeV [21].The idea was to estimate the kinetic energy of the target atoms in this system as the incident ion energy was increased.This would allow us to determine the energies at which the region of the film that has increased kinetic energy due to collisions could be considered to extend throughout the thickness of the film, so as to cause mass movement of the scale required at the interface.As expected, with a 1 keV beam all the energy is deposited in the first few atomic layers of the film, in a narrow distribution, but as the incident ion energy is increased, the beam penetrates into the film and beyond, into the substrate, with increasing amounts of kinetic energy transferred to the targets, further into the film.We found that the amount of KE transferred to the targets peaked at about 140 eV, and at incident energy of about 80 keV, the energy loss was uniform throughout the film.Thus, at our incident energy of 100 keV, the Ar + ions have transferred about 140 eV transferred to atoms in the cobalt film along the ion track.The SRIM calculations use a Monte Carlo based Binary Collision Model, where the energy imparted occurs in one-on-one collisions between projectile ion and target atom and subsequently between target atoms, with the impact parameter randomly assigned.This method is applicable for amorphous materials and is most effective for lighter projectiles and less dense targets, since many body collisions and channeling effects due to crystallinity are not considered.This process has also been modeled in detail through molecular dynamics simulations by Mookerjee et al. [22].The crucial difference between the binary collision method of SRIM and the Molecular Dynamics simulations is in the way that the transfer of energy from the electronic system to the lattice is considered.In the SRIM calculations, the electronic energy loss is a parameter based on an average value determined from experimental data available in the literature, while in the MD simulations, the fast electronlattice energy transfer in the immediate aftermath of the ions passage is explicitly modeled and its impact on the sputtering yield is calculated.One model considers that the electron-lattice energy transfer is uniform in a cylindrical region around the ion transit path and then the system evolves from there.In the other model, radial energy distribution in time and space based on the thermal spike model is used to calculate sputtering yields.The choice of the energy distribution model has a strong impact on the sputtering yields as shown in the paper.This is also suggested from experimental results on ion beam mixing in metals, where metals of comparable size like Cu and Ni have very different mixing behavior [23,24].These earlier papers suggested that the loss of energy to the electrons of the target and the subsequent electron-phonon coupling plays a crucial role in energy transfer when ion beams interact with solids, and this has continued to motivate efforts to model this process [25,26].
In such a transient process, it is hardly possible to define a temperature, but if we consider the KE of the target atoms as representative of the temperature, this corresponds to a 'temperature spike' of several thousands of Kelvin, (well above metal melt temperatures) for a fraction of a picosecond.The result of this transient spike is the restructuring of the film into droplets on the surface, resulting in a randomly distributed high density array of nanostructures.In the paper by Mookerjee et al. [22], where the peak energy transfer of 1-2 thousand eV is calculated for the cylindrical track region which contains several hundreds to a thousand target atoms.Assuming even distribution of kinetic energy within the region, and considering that thermal energy at 300 K corresponds to about 40 meV, the transient temperature may be estimated to be of the order of 10 3 -10 4 K.The SRIM simulations for a 100 keV Ar ions in Co give a figure of about 1400 eV/nm as the energy loss to target atoms throughout the thickness of the film.There are also estimates from the literature of the thermal spike in metals, ranging from several hundred K to several thousand K [25][26][27], depend-ing on the value of the electron-phonon coupling parameter, which is involved in the transfer of energy from the electronic to the lattice subsystems and on whether the model of lattice relaxation used was linear or non-linear.At this point, a more accurate estimate awaits full molecular dynamics calculations for more substantial periods or the development of experimental techniques which allow access to phenomena at shorter time scales.

IV. SUMMARY
We have used ion beam irradiation as a self-organized method to influence size uniformity in nanostructure fabrication from thin films.The structural changes in the film are analyzed using AFM, XRD and RBS and indicate that dewetting is the dominant effect, driven by intrinsic mismatch in surface energy.Ultrashort scale temperature spikes formed upon irradiation ensure that the ion beam fluence becomes an important parameter to control particle size.The impact of the surface free energy, interfacial energy, ion energy, charge state, ion species and fluence needs to be studied in detail to determine their influence on nanostructure formation from thin films of different materials.This method also suggests the possibility of creating ordered arrays of metal nanoparticles by growing metal films on substrates which have been prestructured by ion beam irradiation.
Figure 1(f) shows the scan of Fig. 1(c) as seen in 3-d view.