Journal of Nanoparticle Research

, Volume 13, Issue 1, pp 245–255

Morphology and magnetic properties of island-like Co and Ni films obtained by de-wetting


  • P. Tiberto
    • Electromagnetism DivisionINRIM
  • S. Gupta
    • Politecnico di Torino, DISMIC and DIFIS
  • S. Bianco
    • Italian Institute of Technology—IIT@POLITO, Center for Space Human Robotics
  • F. Celegato
    • Electromagnetism DivisionINRIM
  • P. Martino
    • Politecnico di Torino, DISMIC and DIFIS
  • A. Chiolerio
    • Politecnico di Torino, DISMIC and DIFIS
  • A. Tagliaferro
    • Politecnico di Torino, DISMIC and DIFIS
    • Politecnico di Torino, DISMIC and DIFIS
Research Paper

DOI: 10.1007/s11051-010-0023-2

Cite this article as:
Tiberto, P., Gupta, S., Bianco, S. et al. J Nanopart Res (2011) 13: 245. doi:10.1007/s11051-010-0023-2


The morphological, structural, and magnetic properties of Co and Ni films of different thicknesses grown by RF sputtering on a Si–SiO substrate and submitted to controlled diffusion of atoms on the substrate (de-wetting) are studied through X-ray diffraction (XRD), atomic force microscopy, X-ray photoelectron spectroscopy, and alternating-gradient magnetometry. For both metals, de-wetting treatment leads to the growth of non-percolating, metallic nanoislands characterized by a distribution of sizes and aspect ratios. XRD spectra reveal a polycrystalline multi-component structure evolving by effect of de-wetting and directly affecting the magnetic properties of films. The magnetic response after de-wetting is consistent with the formation of a nanogranular magnetic phase characterized by a complex, thickness-dependent magnetic behavior originating from the simultaneous presence of superparamagnetic and blocked-particle contributions. At intermediate film thickness (around 10 nm), a notable enhancement in magnetic coercivity is observed for both metals with respect to the values measured in precursor films and in their bulk counterparts.


Magnetic nanoparticlesMagnetic thin filmsDe-wetting techniqueCoercive field


De-wetting of thin metallic films deposited on Si–SiO substrates by controlled diffusion of metallic atoms (Oh et al. 2009) has been widely studied from the standpoint of fundamental science and technological applications (Bouville et al. 2007; Gadkari et al. 2005). The combined effects of annealing at constant temperature and differences in surface tension may cause a continuous metallic film to split into an array of non-percolating nanoislands, whose size is in proportion to the original film thickness (Oh et al. 2009). Despite the unquestionable interest in the investigation of the physical and chemical factors which affect the de-wetting of thin metal films (Sieradzki et al. 2001), this effect has been proven detrimental in nanoelectronic device technology, causing failures due to overheating (Srolovitz and Safran 1986). However, de-wetting can be exploited as a self-organization process for nanostructuring.

Magnetic thin films and nanoparticles have become a widely investigated class of materials owing to their potential use in nanoelectronics and data storage (Nalwa 2002). Various techniques have been exploited to obtain nanostructured magnetic thin films, such as for instance conventional top-down lithography methods (Smyth et al. 1988; Martin et al. 2003; Saavedra et al. 2010). However, these techniques are costly, time consuming, and do not allow pattern wide-area patterning.

Bottom-up fabrication methods such as self-assembling may represent a valid alternative to conventional lithographic process. In this way, quite ordered arrangements of magnetic nanostructures may be obtained, whose size and distance can in principle be easily controlled (Goncharov et al. 2005). In this context, the de-wetting process may be exploited as a particular self-assembly technique suitable to create magnetic nanostructures out of precursor thin films grown on Si substrates covered by with a thermally grown SiO2 layer. Arrays of magnetic-metal nanoparticles on insulating substrates obtained by de-wetting are particularly interesting as prospective catalysts in the growth of carbon nanotubes embedding or decorated by magnetic nanoparticles, aimed to technological applications such as orientable structural materials and functional materials (Chiolerio et al. 2008; Soldano et al. 2008).

This study is conducted to investigate ferromagnetic thin films of 3D transition metals submitted to de-wetting to promote the formation of nanoislands. In particular, our study is focused on the morphological, structural and magnetic properties of island-like ferromagnetic Co and Ni films.

A number of contributions aimed to explore the interplay between microstructure and magnetic properties of Co and Ni films grown on planar substrates have been published. Ultrathin Co and Ni films (up to 10 nm) were investigated by surface magneto-optic methods (Chang et al 2007; Shern et al. 2004); thicker Co and Ni films (up to 700 nanometers) were studied in detail through a variety of techniques (Munford et al. 2002; Kumar et al. 2009). This paper, however, is focussed on the evolution from continuous to island-like films by effect of de-wetting.

Tailoring the formation of fcc and/or hcp Co and Ni nanoparticles (i.e., controlling their sizes and distances) may offer great opportunity for studying and exploiting their magnetic behavior (typically ranging from superparamagnetic to magnetically blocked depending on particle size, composition, and temperature).

Experimental: materials and methods

Thin films of Co having thickness ranging from 3.5 to 30 nm were deposited by RF magnetron sputtering at a pressure of 10 mTorr of Ar (base pressure ~10−7 Torr) and microwave power of 75 W at ambient temperature. Likewise, thin films of Ni with varying thickness ranging 1–20 nm were grown using RF magnetron sputtering technique in similar conditions. All films were grown on commercial Si whose surface was previously coated by thermally grown SiO2 of thickness ~120 nm. Prior to metal thin films deposition, Si wafers were ultrasonically cleaned in organic solvents and rinsed in de-ionized water. Clean wafers were then immersed in HF to remove native SiO2. The cleaned wafers were blown dry using N2 gas and used for thermally grow SiO2 and subsequently sputter metal films. Thin film thickness was evaluated by atomic force microscopy (AFM-DME instruments) in non-contact mode on the edge of a step, obtained after the lift-off of a polymeric mask. Each thickness evaluation was done executing a multi-line mean, thus reducing the estimate error to around 5%.

As-deposited films were annealed at constant temperature of 850 °C in 1 mTorr vacuum level to generate island-like structures whose thickness was measured using standard bearing-height routines available from the AFM software.

The films surface morphology, texture, grain size, and domains were analyzed using field-emission scanning electron microscopy (FESEM), X-ray diffraction (XRD), and X-ray photoelectron spectroscopy (XPS) analytical techniques prior to and post de-wetting.

X-ray diffraction technique was used to determine the crystalline structure of the films (Panalytical PW1140–PW3020, Cu Kα X-ray source). The scans were performed in a parallel beam configuration (grazing angle 0.5°), in order to minimize the substrate contribution to the observed diffracted intensities. XPS characterization was performed using a VersaProbe5600 (monochromatic source, Al anode 1486.6 eV).

Magnetic hysteresis loop measurements at room temperature for homogeneous (as-deposited) and island-like (after de-wetting) films were analyzed by using an alternating-gradient field magnetometer (AGFM) in the field range −20 kOe < H < +20 kOe. The diamagnetic contribution of the sample holder and substrates were carefully subtracted from the measured curves. The field was applied in both parallel and perpendicular to the film plane and the magnetization always being measured along a direction parallel to the magnetic field.

Results and discussion


Figure 1 shows typical scanning electron micrographs for as-deposited (a–c) and de-wetted (d–f) Co films with thickness t = 3.5, 9, 30 nm. The as-deposited films appear to be quite smooth and pin-hole free except for those with the smallest thickness. Images taken on the de-wetted films confirm the formation of island-like droplets; the observed granular structure indicates the aggregation of Co particles on the substrate as an effect of thermal treatment. The mean size d of metallic aggregates is generally observed to increase with increasing thickness of the precursor film. In particular, Co particles in Fig. 1d (t = 3.5 nm) are observed to have substantially circular shape with mean diameter in the interval 10–30 nm. In the film with t = 9 nm (Fig. 1e) the mean diameter is around 100 nm, while the thickest film (Fig. 1f) is characterized by aggregates with mean particle size around 300 nm that have lost circular shape. Aggregation and coalescence of adjacent grains may explain the appearance of non-circular particles. We point out that the major axis of non-circular grains is randomly oriented; this observation means that the growth of such particles is not influenced by any in-plane anisotropy.
Fig. 1

SEM of ac as-grown and df de-wetted Co thin films with thicknesses 3.5, 9, and 30 nm, respectively

Figure 2 shows SEM images of Ni films with nominal thickness t = 1, 10, 20 nm prior to and post de-wetting; they are rather similar to those of Fig. 1. Again, Ni droplets or aggregates are formed in de-wetted films, whose mean size increases with increasing thickness.
Fig. 2

SEM of ac as-grown and df de-wetted Ni thin films with thicknesses 1, 9, and 20 nm, respectively

A more detailed statistical analysis of nanoparticle size has been performed on de-wetted samples. Two examples are given in Fig. 3 for the 9 nm thick Co film and the 10 nm thick Ni film. Histograms resulting from analysis of the original images (upper row) are reported in the lower row. The nanoparticles are not monodisperse. In the case of Co, the resulting histogram is well fitted by a superposition of at least three lognormal curves (black lines), centered at 16, 76, and 141 nm. The average nanoparticle diameter d is 105 nm, in agreement with the previous rough estimate. The mean inter-particle distance a obtained from the ratio between the total surface of nanoparticles AP and the whole image area AT using the formula a = 1/2(πAT/AP)0.5d turns out to be a = 170 nm.
Fig. 3

Examples of particle size distribution histograms obtained analyzing the SEM images (upper row) taken on two de-wetted Co and Ni films (9 and 10 nm thick, respectively) (color online)

In the case of Ni, two lognormal curves centered at 41 and 245 nm (black lines) are enough to fit the histogram. In this case, the average nanoparticle size and distance are d = 127 and a = 275 nm.

Morphological data of selected Co and Ni films prior to and after de-wetting are given in Table 1. For both metals, a nearly linear relationship is observed between continuous film thickness t and mean equivalent diameter of nanoislands d, with no substantial difference between Co and Ni data sets. A similar behavior was recently found in both sputtered Co (Bower et al. 2009) and Ni films (Terrado et al. 2009) in comparable thickness ranges. With the exception of the thinnest Co film, where measuring errors can play a greater role, the measured nano-island thickness is systematically lower than their equivalent diameter, although the flattening seems to be less pronounced in Ni than in Co.
Table 1

Morphological data of some selected Co and Ni films prior to and after de-wetting


Continuous films: nominal thicknessa (nm)

De-wetted films: mean equivalent diameter db (nm)

De-wetted films: mean island thickness tc (nm)





















aFrom AFM step measurements

bComputed with a numerical routine from FESEM images

cFrom AFM bearing-height measurements

X-ray diffraction and X-ray photoelectron spectroscopy

The control of crystalline orientation and microstructure is desirable since this directly influences the magnetic properties and corresponding properties such as anisotropy, coercivity, and device functionality (Teranishi et al. 1997; Hirose et al. 1997; Ohsawa et al. 1999).

Figure 4 displays representative X-ray diffractograms of Co thin films of thicknesses 9 and 30 nm prior to and post de-wetting in the range of 10–100°. The presence of several peaks both as-deposited and de-wetted films is suggestive of polycrystalline nature of the material [such as (002), (111), (110), (101), (102) peaks] and evidences a multiphase composition including fcc and hcp textures in addition to occasional oxide forms. Moreover, the peaks become sharper for de-wetted films and there are more peaks for thicker (30 nm) Co films as compared to smaller thickness (9 nm), possibly because of the reduced sensitivity. The experimental limitations of our XRD technique make it difficult to extract detailed information from diffractograms taken on thin films (below 10 nm for both Co and Ni systems). While no evidence was found by XRD for the metastable Co carbides, some minority candidate phases such as ε-Co are observed. With annealing, the (111) peak intensity increases and becomes dominant corresponding to fcc dominant phase for Co films. It is clear that the annealing plays an important role in determining the structure of the films. By using Debye–Scherrer’s formula the crystallite size (Lhkl) follows: Lhkl = /(b cos θ), where K is Debye–Scherrer constant, λ is the X-ray wavelength (1.5405 Å), b is the full-width at half maximum of the (hkl) peak and θ is the Bragg angle for the corresponding (hkl) peak. From the quantitative analysis of fcc Co (111) peak, the crystallite size is about 25 and 33 nm for 9 and 30 nm thick de-wetted samples, respectively. Likewise, the hcp Co crystallite size estimated from (002) peak width is about 28 and 40 nm for 9 and 30 nm thick Co de-wetted samples, respectively. The presence of hcp peaks indicates that c-axes normal to Co (002) planes orient preferentially parallel to the film surface. The a-axes are supposed to orient preferentially perpendicular to the surface because of the presence of (101) and (102) peaks. However, Co crystallites which have a-axes parallel to the surface also exist in the films because Co (110) is detected. The peak width of Co (101) and Co (102) is smaller than those of Co (002) suggesting nanograin structure. Notice that the diffraction intensity of both fcc and hcp peaks is relatively low because of the quantity of Co probed during the measurements. In addition, the Co (111) and Co (002) peak positions shift to smaller angle and the c-axis length becomes larger (i.e., 1.552 Å) for de-wetted films than for the as-deposited samples (i.e., 1.540 Å). Further quantitative analyses are in progress and will be reported subsequently. However, the microstructure of our samples is consistent with those of a study of the crystallization properties of co-sputtered Co and C samples (Wang et al. 2000; Konno and Sinclair 1994; Respaud et al. 1998). The coexistence of multi-crystalline phases of Co, oxygenated Co phases and corresponding Co valence states are demonstrated through XPS. Figure 5 displays results for the as-deposited and de-wetted Co films such that the spectra consist of primary peaks at about 531.0, 778.3, and 794 eV, correspond to the binding energies of O1s, Co2p3/2, and Co2p1/2, respectively. The XPS spectra indicate that there is not any carbide phase formed in the films and the small amount of oxide phases of Co is believed to be originated from surface oxidation as well as SiO2 interface of the film, particularly post de-wetting as the peak intensities are higher.
Fig. 4

Representative XRD spectra of de-wetted 9 nm Co film and as-grown and de-wetted 30 nm Co films, showing assigned multi-component peaks (color online)
Fig. 5

High-resolution XPS spectra of Co2p and O1s peaks for a, c as-deposited and b, d de-wetted Co films, respectively (color online)

X-ray diffraction patterns of Ni thin films of thicknesses 10 and 20 nm prior to and post de-wetting are reported in Fig. 6. The presence of crystalline Ni is observed in the as-prepared films. In both samples, the presence of Ni oxide is indicated by a faint peak occurring at an angle of 43.3° near the reflection of the Ni [200] indicated by a dotted line in the figure. The peak is more evident in the film having thickness t = 10 nm. Such a contribution is seen to disappear in both de-wetted films. Reduction of NiO to metallic Ni occurs in interaction with hydrogen below 300 °C (Richardson et al. 2008) and in our case is probably caused by the interaction with water vapor during the first stages of the thermal process. Using again the Debye–Scherrer formula, the average crystallite size after thermal treatment can be estimated to be 28 and 32 nm for 10 and 20 nm thick films, respectively.
Fig. 6

Representative XRD spectra for Ni films (10 nm, top and 20 nm, bottom), before and after de-wetting (color online)

Magnetic properties

An example of parallel and perpendicular magnetization curves of Co films (after subtraction of spurious diamagnetic signals) is provided in Fig. 7 for the 9 nm thick sample. The different level of noise between the two curves merely reflects the different sensitivity of the AGFM probes used in parallel and perpendicular measurements. For the as-deposited film (upper panel), the measured curves are consistent with the expected response of a homogeneous thin film with in-plane magnetization, dictated by shape anisotropy. While the parallel magnetization curve is step-like with high maximum susceptibility (χmax ≅ 7) and low coercivity (Hc ≅ 70 Oe), the perpendicular magnetization curve is substantially non-hysteretic. The shape of the curves is indicative of the rotation of the magnetization vector out of the film plane against the shape anisotropy torque. The saturation field (H ≅ 12 kOe) is in agreement with that expected from coherent out-of-plane rotation of a magnetization vector of average magnitude of about 1,200 emu/cm3, resulting from contributions of a mixture of fcc Co (Ms ≈ 1,000 emu/cm3 in thin films (Wolf et al. 1994)) and hcp Co (Ms ≈ 1,400 emu/cm3 as in the bulk). On the other hand, de-wetting makes the parallel magnetization curve less steep (χmax ≅ 0.6) and the perpendicular magnetization curve steeper. In both configurations, similar coercivity values were measured (Hc ≅ 430 Oe). This is in agreement with the behavior of an assembly of weakly interacting magnetic islands with almost comparable aspect ratio along the three dimensions. The effect of de-wetting on coercive field of Co films is summarized in Fig. 8 (upper panel) for a magnetizing field applied in the film plane. A comparison between parallel and perpendicular coercivity of de-wetted samples is shown in the lower panel. For both configurations, the highest coercivity is observed in the 7 nm thick film. The maximum coercivity is almost independent of the magnetizing field orientation (Hc,MAX ≅ 475 Oe for parallel and Hc,MAX ≅ 425 Oe for perpendicular configuration). Both values should be compared with the standard Stoner–Wohlfarth model predictions for an assembly of blocked nanoparticles with magnetic easy axes randomly directed in 3D for uniaxial (Hc ≅ Ka/MS) and cubic symmetry (Hc ≅ 0.64Ka/MS), where Ka is the dominant anisotropy constant. In the present case, the magneto-crystalline anisotropy K1 (equal to 6.3 × 105 erg/cm3 for the fcc Co phase (Li et al. 2000) which becomes predominant in de-wetted samples, as evidenced by XRD) is prevailing over shape anisotropy because of the rather low particle ellipticity observed in the thinner Co samples. Using MS ≅ 1 × 103 emu/cm3 (Li et al. 2000), one gets Hc = 630 Oe and 400 Oe for the uniaxial and the cubic case, respectively. Therefore, our result points to a coercivity dominated by cubic magneto-crystalline anisotropy rather than by uniaxial shape anisotropy. This is in agreement with the observation that the Co nanoparticles are disk-shaped. It should be noted that a fraction of superparamagnetic particles reducing the overall coercivity should exist in thin samples by effect of particle size distribution (Fig. 3). This is evidenced by the decrease of HC with decreasing film thickness below 7 nm (lower panel of Fig. 8): in thin samples, the average nano-island size is smaller and the fraction of superparamagnetic particles increases, resulting in a larger decrease of the overall sample coercivity; above 7 nm, the resulting nanoislands are bigger, and finally, coalesce in large aggregates, again exhibiting a low coercivity mostly related to the onset of magnetization processes dominated by magnetic domain wall motion.
Fig. 7

Parallel and perpendicular magnetization curves for a Co thin film (as-deposited film thickness: 7 nm) a before and b after de-wetting (color online)
Fig. 8

Magnetic coercivity of Co films as a function of nominal film thickness: a parallel configuration, before and after de-wetting; b parallel and perpendicular configurations after de-wetting (color online)

A similar effect of de-wetting is observed in the case of Ni films. The magnetic response of two representative samples after de-wetting is reported in Fig. 9. The behavior of coercivity as a function of sample thickness after de-wetting is shown in Fig. 10. The same considerations as for de-wetted Co films apply. A coercivity of 125 Oe for Ni films is quite remarkable and is presumably related to in-plane shape anisotropy of nanoparticles rather than to magneto-crystalline anisotropy, which is very low in this case (K1 = −5 × 104 erg/cm3 (Cullity 1972)). Taking the thickness-to-diameter ratio of Ni particles equal to 1/2 (as previously discussed), the in-plane shape anisotropy KS would range from 0 (perfectly circular particles) to about \( \pi M_{\text{S}}^{2} \) (needle-shaped particles) (Cullity 1972). The Stoner–Wohlfarth model in 3D predicts a maximum coercivity Hc ≅ KS/MS = MS for a dominant contribution from shape anisotropy. The saturation magnetization in Ni films (in the thickness range 0.2–20 nm) stays very close to the accepted value for bulk Ni (MS = 440–480 emu/cm3 (Neugebauer 1959)), so that Hc is expected to vary between very low values up to about 1,500 Oe. Our result is fully compatible with the slightly elongated shape of the nanoislands evidenced by SEM (Fig. 2).
Fig. 9

Parallel magnetization curves of representative Ni thin films after de-wetting; as-deposited film thickness: 5 nm (a) and 20 nm (b) (color online)
Fig. 10

Magnetic coercivity of Ni films as a function of nominal film thickness in parallel configuration (full symbols: as-deposited films; open symbols: after de-wetting) (color online)

De-wetted Ni films exhibit maximum coercivity at an intermediate film thickness; the same considerations as for Co films apply (compare Figs. 8 and 10, upper panel).

It is worth noting that in both Co and Ni de-wetted films, the slight increase in nanoparticle size with increasing film thickness observed in thicker films from X-ray diffractograms (see “X-ray diffraction and X-ray photoelectron spectroscopy” section) has quite negligible effects on the hysteresis properties. The largest magnetic changes occur in thin de-wetted films (nominal thickness below 8 m) where no significant conclusion can be safely drawn from X-ray diffractograms.


In summary, the de-wetting of homogeneous metallic Co and Ni films is effective in producing island-like films exhibiting various magnetic states that can be tailored for a number of subsequent applications, including the growth of CNT arrays.

The morphological and magnetic properties Co and Ni films with varying thickness have been investigated prior to and post de-wetting. The nanoparticle size dispersion evidenced by SEM image analysis brings about a complex magnetic behavior which originates from the superposition of superparamagnetic and blocked-particle contributions, and depends on Co and Ni film thickness. In de-wetted films, the maximum coercivity appears at intermediate thickness. For both metals, XRD spectra revealed polycrystalline multi-component structure directly affecting the magnetic properties.


One of the authors (S.G.) is thankful for a research fellowship from Politecnico di Torino during her visit.

Copyright information

© Springer Science+Business Media B.V. 2010