Structural Characterization of Cobalt Thin Films ... - Wiley Online Library

Full Paper

Structural Characterization of Cobalt Thin Films Grown by Metal-Organic CVD** By Mariana F. Chioncel and Peter W. Haycock* Cobalt thin films were produced by metal-organic CVD from C5H5Co(CO)2, at various temperatures and for various deposition times. The films have been grown onto glass substrates with no buffer. The crystalline structure, morphology, and composition of the films were analyzed by X-ray diffractometry (XRD), field-emission scanning electron microscopy (FESEM), atomic force microscopy (AFM), and Auger electron spectroscopy (AES). Routine XRD patterns were collected in symmetric geometry for phase identification and the sin2w diffraction technique was employed to calculate the average in-plane stress. Structural studies indicate that the films tend to grow in island mode, as predicted by theory, and have a structure between that of face-centered cubic (fcc) and hexagonal close-packed (hcp) cobalt. There is significant in-plane tensile stress at the interface with the substrate, which relaxes to a compressive stress an order of magnitude lower at the surface. The films have a relatively low impurity content, as determined by AES, except near the surface. Keywords: Cobalt thin films, Crystalline structure, MOCVD, Surface morphology

1. Introduction Thin films of cobalt, cobalt-based alloys, and Co/X multilayers have been the subject of significant scientific research since the early 20th century. A variety of magnetic domain structures and a wide range of magnetic properties[1±3] can be produced through controlled structural and morphological design of the environment of the Co atoms. In recent years, there has been increased interest in Co-based thin films and multilayers, in part because their magnetic properties have made these structures useful for high density magneto-optic recording media,[4±6] and because of their giant magnetoresistance.[7] Cobalt thin films have been deposited by various means, including traditional inorganic CVD,[8,9] sputtering,[10] electron-beam evaporation,[11,12] and metal-organic (MO)CVD.[13±16] CVD is a technique which offers potential for producing films with high uniformity of thickness and composition, high purity, conformal step coverage, minimal substrate damage, high deposition rates, ± [*] Dr. P. W. Haycock School of Chemistry and Physics, Lennard-Jones Laboratories, Keele University Staffordshire, ST5 5BG (UK) E-mail: [email protected] Dr. M. F. Chioncel Physics Department, Faculty of Chemistry, University of Bucharest Bd. Elisabeta 4-12, 70346 Bucharest (Romania) [**] The authors thanks go to Dr. C.C. Tang at the Daresbury SRS for his help with X-ray experiments, Dr. M. Rotov (Keele) for help provided with the AFM study and Dr. G.W. Critchlow of the Institute for Surface Science and Technology at Loughborough University for Auger measurements. Dr. Chioncel thanks the UK Committee of Vice Chancellors and Principals for the award of an Overseas Research Scholarship.

Chem. Vap. Deposition 2005, 11, 235±243

DOI: 10.1002/cvde.200406341

and the possibility for selected area growth. In addition, MOCVD often provides lower kinetic energy routes to desired materials than corresponding inorganic techniques. However, the number of reports concerning CVD of Co thin films is small, mainly due to the limited number of volatile precursors to transport the Co atoms to the reaction zone. There has, therefore, been an increasing interest in organometallic precursors, where the metal is made volatile by bonding it to organic ligands. It is well known that many physical properties of thin films depend to a considerable degree on the grain size and morphology, as well as the crystallographic structure and orientation (texture). It is, therefore, important to know how particular deposition conditions influence the structure, in order to produce films with potentially useful properties. Additionally, internal stresses induced during thin film preparation by deviations from perfect crystal structure,[17,18] or interfacial stress generated by either lattice mismatch or differing thermal expansion coefficients,[19] can affect to a considerable extent the magnetic behavior. Studies of the structure and composition of the surfaces also aid interpretation of the mechanisms by which the films grow. The deposition of pure Co by MOCVD has been carried out previously using sources including Co2(CO)8,[13,14,16,20] Co(C5H5)2,[16], Co(C5H5)(CO)2,[14,16] CoCF3(CO)4,[16] cobalt tricarbonyl nitrosyl,[21] and cobalt complexes of the type [Co(CO)2(NO)L] (L = PEt3, TeMe2 or TeEt2) or [(eta(3)C3H3R2)Co(CO)2L] (R = H or Me; L = CO, PEt3, TeEt2 or CNR¢ with R¢ = iPr, tBu or cyclohexyl).[22] In this paper we report the results of a detailed investigation of the crystalline structure, grain morphology and composition of Co thin films grown by MOCVD, using CpCo(CO)2 as the precursor, in order to determine the correlation between the deposition 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

235

Full Paper

conditions, film microstructure, and stages of the film growth. The ultimate motivation for this work was to examine the prospect of tailoring the microstructure of thin films with specific magnetic properties. CpCo(CO)2 was chosen for its compatibility with methylcyclopentadienylplatinum trimethyl for the production of Co/Pt multilayers.

2. Results The thin films were grown by MOCVD, using an MO350 system built by Metals Research Semiconductors Ltd., at a range of substrate temperatures and with a flow of 70 sccm or 100 sccm of carrier gas through the bubbler containing the precursor. They were characterized by means of SEM, AFM, XRD, and AES.

2.1. Electron Microscopy The change of grain morphology with deposition temperature is clearly observed in the micrographs. Some representative results will be presented here. The sample deposited at 350 C for 1 h, with a flow of 70 sccm through the bubbler, exhibited a mirror finish to the naked eye. Under the electron microscope (Fig. 1a) the surface can be seen to be a uniform distribution of grains with different shapes and dimensions, varying from about 55 nm to 280 nm; grains with elongated finger' type shapes are predominant. Intergranular interstices are present and have approximately the same size as the grains. The micrograph of the sample deposited at 250 C for 1 h under the same conditions (Fig. 1b) indicates the existence of better defined polygonal grains, which are in general less elongated and some of them hexagonal in shape; these are orientated both parallel and oblique relative to the substrate, together with a few fingertype grains, similar in shape and dimensions to those observed at 350 C. Compared to this latter sample, the size distribution has narrowed (here 55 nm to 85 nm), while the intergranular interstices are larger. The micrograph (Fig. 1c) of the sample deposited under the same conditions at 140 C for 1 h reveals a very different morphology. Here, crystalline grains, some of very well-defined hexagonal shapes and

relatively large size, averaging 140 nm across, conglomerate in clusters, packing close together and leaving large voids between the nucleation centers. The clusters have dimensions varying from 200 nm to 1100 nm and are oriented randomly on the glass substrate, but the grains within a cluster tend to lie parallel to each other. This morphology reflects the low nucleation rate at this temperature. The microstructural study of the films deposited at 350 C for different times provides information concerning the nature of the film growth and the stages thereof. These samples were deposited using the same deposition conditions as the previous set of films, apart from the higher flow through the bubbler (100 sccm) and the variation of the deposition time. The SEM image of the sample deposited for 0.5 h (Fig. 2a) displays grains of various shapes with lateral dimensions varying from 23 nm to 90 nm, and a very few, small intergranular voids. Enlargement of the grains is observed in the micrographs of the sample deposited for one hour (Fig. 2b), which is reflected in a wider distribution of particle sizes, from 30 nm to 180 nm, with the majority of the grains measuring ~160 nm across. The micrograph of the sample deposited for 1.5 h (Fig. 2c) shows a different surface morphology with greater uniformity of grain size and shape, as well as an increase of the average size to 230 nm. Further increase in the regularity of the size and shape of the grains is observed in the micrographs of the samples deposited for 2 h (Fig. 2d). Larger grains, 350±600 nm across, with welldefined hexagonal habits, are aligned with their basal planes parallel to the substrate surface. Some of the hexagons are proud of the surface, growing probably as individual crystals on the top layer. Comparison of Figures 1a and 2b shows the effect of changing the H2 flow through the bubbler (the only deposition parameter that was different between these two samples). Clearly the higher precursor flow led to a more regular crystal morphology and better defined habit.

2.2. AFM Measurements Representative data from these measurements are summarized in Table 1, where it should be noted that the grain sizes quoted are in rough agreement with those obtained

Fig. 1. SEM images of Co thin films grown using a flow rate of 70 sccm, 1 h deposition time and at a substrate temperature of a) 350 C (thickness ~86 nm), b) 250 C (thickness ~14 nm), and c) 140 C.

236

2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.cvd-journal.de


Full Paper

(a)

(b)

(b)

1 µm

1 µm

(d)

(c)

1 µm

1 µm

Fig. 2. SEM images of Co thin films grown at 350 C, using 100 sccm flow rate through the bubbler, for a) 0.5 h (thickness ~43 nm), b) 1 h (thickness ~85 nm), c) 1.5 h (thickness ~127 nm), and d) 2 h (thickness ~170 nm). Table 1. Grain sizes, fractal dimension, and root mean square (rms) surface roughness as measured using AFM. Dep. temp [C] 400 350 350 350 350 300 250

Dep. time [min]

Average Grain size [nm]

Fractal Dimension

Rms roughness [nm]

60 120 90 60 30 60 60

± 450 290 207 90 ± ±

2.124 2.259 2.255 2.254 2.263 2.24 2.232

13 15 7 5 6 19 23

from the electron micrographs, although there are slight discrepancies due to the different areas of the samples analyzed in the two cases, as well as the different grain sizing techniques used (manual for the FESEM images and computerized in the case of the AFM data). The 3D AFM image of the sample deposited at 400 C for 1 h depicts conglomerations of grains with some of these proud of the surface, leading to the formation of pyramidal clusters (Fig. 3a). In the group of films deposited at 350 C and 100 sccm flow through the bubbler, the computed calculation indicates an increase of the mean grain size with the deposition time. The average grain width rises from 90 nm for the 0.5 h sample (Fig. 3b) to 450 nm in the 2 h sample (Fig. 3c). The AFM images reveal grains with ill-defined boundaries for the 0.5 h and 1 h samples, which are replaced by very well-contoured shapes, mainly hexagonal, in the 1.5 h and 2 h samples, in agreement with the SEM images. The fractal factor is 2.3 for all these samples (compared to a Chem. Vap. Deposition 2005, 11, 235±243

www.cvd-journal.de

measured value of 2.0 for 2D control samples in the form of clean glass substrates). The smoothest sample in this group was the sample deposited for 1 h. A longer deposition time generally leads to rougher surfaces, an exception being the 0.5 h deposition.

2.3. XRD Measurements Representative symmetric XRD patterns are shown in Figure 4, while the results of data analysis are summarized in Table 2. Phases were identified using database standards. In all the samples investigated, the diffraction patterns yield a diffraction peak centered at around 44.4 in 2h, the details of the line profile varying with the deposition parameters. This diffraction peak is positioned between the fcc (111) of b-Co (face centered cubic) at 44.21 and the (002) reflection of a-Co (hexagonal close-packed) at 44.76, indicating a strained lattice. The peak is somewhat closer to the theoretical position for the fcc peak. Hence, although the atomic arrangement will not correspond to either an ideal hcp or fcc lattice, for the purposes of identification only, the XRD line in question will be referred to as fcc(111). Two other peaks, at 47.26 and 42.53 in 2h, are observed in the film deposited at 140 C for 1 h. These correspond to the hcp Co (10.1) and fcc CoO (100) lines, respectively. The analysis of the XRD line profiles provides additional information about crystallite thickness perpendicular to the plane of the film. Calculations employing the Sherrer formula indicate a variation of the crystallite 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

237

Full Paper

70

(a)

Intensity (a.u.)

60 50

(111) fcc Co

40

(111) CoO

30

(10.1) hcp Co

20 10 0 0

20

40

60

80

100

2Theta (degrees) 8000

(b)

7000

(a) Intensity (a.u.)

6000 5000 4000 3000 2000 1000 0 0

10

20

30

40

50

60

70

80

90

2Theta (degrees)

Fig. 4. XRD patterns recorded from thin films deposited at a) 140 C for 1 h, and b) 350 C for 0.5 h (dashed line) and 2 h (solid line); the very broad, low 2h peak, very clear in the 140 C pattern, is due to the amorphous glass substrate.

(b)

Table 2. Summary of data analysis of XRD patterns measured with monochromatic CuKa radiation. Hmax is the background-corrected peak intensity value, t[] is the crystallite size calculated using the Sherrer formula, and d is the d-spacing value corresponding to measured h values Deposition Bubbler flow temp [C] (sccm) 400 400 400 350 350 350 350 300 300 350 250 140

(c) Fig. 3. AFM images of thin films grown at a) 400 C, b) 350 C, 0.5 h, and c) 350 C, 2 h, showing surface morphology with the corresponding heightdistance profile.

thickness with temperature and deposition time. The largest grain width, as reported above, was measured for the film deposited at 350 C for 2 h. This same sample also has the largest coherence length perpendicular to 238


100 100 100 100 100 100 100 70 70 70 70 70

Time [min]

FWHM []

2h []

d []

Hmax [counts/s]

t []

120 60 30 120 90 60 30 60 40 60 60 60

0.338 0.244 0.284 0.121 0.313 0.181 0.27 0.269 0.401 0.257 0.35 0.277

44.31 44.46 44.39 44.49 44.42 44.4 44.48 44.55 44.32 44.43 44.52 44.38

2.042 2.035 2.038 2.034 2.037 2.038 2.034 2.031 2.041 2.037 2.033 2.039

970 831 225 7601 1322 1104 753 395 142 681 87 14

253 351 301 707 273 473 317 318 213 333 244 309

the film plane, with the lowest (0.121) value of FWHM in 2h. From the position of the peak alone it is not possible to work out the stress within the film, since the ideal relaxed structure is not known and there will be stress relief mechanisms operating at various points inside it. However, quantitative measurements of the stress present in magnetic thin films are relevant when correlating structure with magnetic behavior, as it is well known that stress can induce www.cvd-journal.de


Full Paper

Concentration (%)

Intensity (counts/10s)

magnetic anisotropy. Hence more de60000 tailed measurements were made on one of the samples at the SRS X-ray source at 50000 the CCLRC Daresbury Laboratory. The sample deposited at 350 C for 2 h has Ome 4.65 40000 been analyzed using the asymmetric sin2w Ome 8.65 diffraction technique, at a wavelength of Ome 12.65 30000 2.2 . At this wavelength, measurement in Ome 16.65 symmetric h/2h geometry shows a central Ome 20.65 Ome 24.65 diffraction peak at 32.6 in h, indicating a 20000 Ome 28.65 d-spacing of 2.042 (c.f. 2.036 as Ome 32.65 measured on the diffractometer at Keele, Ome 36.65 10000 quoted in Table 2), with the theoretical position for the fcc(111) line being 0 32.504. In order to investigate the strain, 64 64.5 65 65.5 66 66.5 the angle x was varied from 4.65 to 40.65 2Theta (degrees) in 4 steps, and for each x setting intensity versus 2h diffractograms were collected. Fig. 5. 2h scans at different tilt angles (w = x - h), for the 350 C, 2 h sample. The reduction in peak intensity with increasing w is due to the decrease in the volume of the sample probed by the X-ray beam; distinct The scans for x = 36.65 and x = 40.65 shoulders can be noticed at deeper penetration of the radiation into the film. were identical to those for x = 4.65 and x = 8.65, respectively, and so have not been considered further. The reduction in peak intensity with stage process during growth, leading to a relatively relaxed increasing w is due to the decrease in the volume of the surface represented by the central diffraction peak. sample probed by the X-ray beam and the preferred orientation. The Lorentz-polarization and absorption correction factors, which should generally be considered in such 2.4. AES Measurements calculations, were neglected, as the estimated sample thickness of about 170 nm is sufficiently thin not to affect AES depth profiling was employed to investigate the the results. Distortion of the line profile is manifest in the impurity distribution in the as-deposited films. Figure 6 appearance of a low-angle shoulder which becomes well shows the Auger profile for the film deposited at 350 C for resolved at the lower values of ½w½, while a second, less1 h with a flow of 70 sccm through the bubbler. Surface and prominent shoulder, on the high-angle side relative to the sub-surface compositions were determined in a single area of main peak, is present at all values of ½w½, although only a sample considered to be representative of the whole. resolved at ½w½ = 0 because of counting statistics (Fig. 5). A Relatively high levels (~ 20 atm.-%) of carbon were present perceived broadening of the composite diffraction peak is on the surface of the sample; the carbon concentration largely due to the varying relative intensity of the low-angle reduces in the bulk' of the sample and is insignificant at the side peak, but could also have contributions from the size of substrate-film interface. A film thickness of 85±90 nm could the diffracting domains, by non-uniformly strained grains, or be estimated, based on the depth at which silicon is first by a combination of both.[23] observed coupled with a sudden rise in the oxygen The appearance of clear shoulders of the peak in the X-ray 120 pattern can be attributed to variation of the crystallography through the depth of the film due to stress relief mechanisms. 100 A set of fitted peak centers was obtained using the pseudo80 Voigt function which is characteristic for the diffraction [24] 2 Co % geometry used. Analysis of the data in terms of sin w has 60 O% been performed, although the difficulty of obtaining precise C% 40 Si% fits for the positions of the three peaks limits the accuracy of 20 the results of this. However, the best fit lines to graphs of 2h versus sin2w suggest compressive in-plane stresses of 0 6 107 Pa and 8 107 Pa for the middle and left hand peaks, 0 20 40 60 80 100 120 -20 respectively, whilst the corresponding value for the right Depth (nm) hand peak is a tensile stress of 8 108 Pa. This last value is the Fig. 6. AES depth profiling of the 350 C, 1 h, 70 sccm sample; the curves give least accurate, since the right hand peak is the least wellthe compositions atom percent as a function of depth. Depth profiling was continued until silicon and oxygen were observed in the ratio 1:2, indicating resolved. However, this is consistent with a high tensile stress that the glass substrate had been reached; the first value (depth 0 nm) has not in the film at the interface with the substrate, caused by been considered because of the inevitable contamination of the sample surface. differential expansion coefficients, which is relieved in a two Chem. Vap. Deposition 2005, 11, 235±243

www.cvd-journal.de


239

Full Paper

concentration (indicating that the glass substrate has been reached) together with the theoretical sputter-etch rate for cobalt. The cobalt concentration is very uniform with depth into the film over the range 20 nm to 80 nm. The films have low contamination levels beyond 20 nm, with oxygen and carbon levels of about 1 % and 1.2 %, respectively, in the bulk of the film. Thickness values based on the Auger spectroscopy indicate a Co growth rate of approximately 0.236 s±1 for this sample.

2.5. Growth Rate The growth rate was determined at four temperatures (250 C, 300 C, 350 C, 400 C) from the X-ray data presented above, calibrated against the thickness calculated from the AES data. Data could not be obtained at 140 C because there was not a strong enough preferred orientation at that temperature and also a significant amount of the X-ray scattering was due to cobalt oxide in that case. Since the width of all measured rocking curves in h was 11 ± 1, the product of the height of the peak and its width in 2h can be taken as a good indicative measure of the cobalt content when there is a very strong preferred orientation, and the material can be taken to be predominantly crystalline cobalt. A general increase of integrated peak intensity with deposition time was observed in all sets of samples deposited at the same temperature, in accord with the growth in film thickness. Previous studies in the same reactor confirm that under nominally identical conditions to those used here (with bubbler flow of 70 sccm), the growth rate of Co films from CpCo(CO)2 is constant over at least the first two hours.[25] The growth rate varies with temperature as shown in Figure 7a, where it can be seen that the growth rate drops off between 350 C and 400 C. An Arrhenius plot of the same data gives a straight line, the slope of which implies an activation energy of 68 kJ mol±1 (Fig. 7b). It was observed that the growth rate was not significantly dependent on precursor delivery rate at 350 C.

3. Discussion The experiments undertaken in this work show a temperature dependence of the film growth, as expected.[26] The growth is an activated process in the temperature range 250±350 C, and starts decreasing at higher (> 400 C) temperatures, because of an increased desorption rate and decomposition of the reactants on the reactor walls. Only very slight deposition was observed for temperatures lower than 150 C and none below 140 C, as expected considering that the decomposition temperature for CpCo(CO)2 is around 140 C. AFM, FESEM, and XRD led to data which are in general agreement. The measured values of the grain sizes as determined by AFM and FESEM from different areas of 240


the same samples are comparable. The XRD patterns show an intense Bragg peak positioned between the fcc(111) and hcp(00.2) reflections of crystalline Cobalt. Conventional theory predicts that for temperatures below 417 C, hcp will be the stable phase for relatively large Co grains and fcc for small grains (< 150 nm), while for a range of grain size that traverses this boundary a mixture of these two phases can exist,[27] although there is much experimental evidence that manufacture of Co products under non-equilibrium conditions can violate the conventional theory.[10,28,29] In the current case, most micrographs (other than those for the samples deposited at 140 C) show a grain size distribution with an average between 90 nm and 450 nm, suggesting, according to the conventional theory, a mixture of fcc and hcp crystalline structure. However, most diffractograms show strong crystalline texture with only one reflection positioned between the unstrained positions of the fcc(111) and hcp(00.2) lines. As the morphology of the 140 C deposition for 1 h is very different, with large hexagonal grains conglomerating in clusters and extensive intergranular spaces, the crystalline structure would be expected to change and the X-ray diffractometry confirms this, indicating a mixture of hcp(10.1) and fcc(111)-textured Co and fcc(111) CoO. The existence of CoO is in agreement with the AES evidence of oxygen incorporation near the surface. The analysis of the morphology and crystalline structure of thin films deposited using the same deposition conditions but for different times is relevant when trying to understand the growth stages and the evolution of the parameters during the growth. The FESEM, AFM, and XRD measurements of the films deposited at 350 C provide a good data set for this. The SEM image of the 0.5 h film indicates that at this growth stage both nucleation and pre-coalescence have begun, while the AFM data in showing a relatively rough surface confirm that the initial stage of coalescence is not complete. Recrystallization has not started at this point, as it is known that grain regrowth in metallic films cannot take place as long the nucleated grains are separated by significant voids; however, once these are filled the recrystallization proceeds and the mean size increases.[30] The coalescence stage has finished after 1 h, hence the lower roughness value at this stage. The increase of the roughness for longer deposition times (1.5 h and 2 h) is an indication of grain coarsening during coalescence and post-coalescence growth. Hence, the roughest sample as determined by AFM is the one deposited for 2 h. The AFM data indicate that the grain boundaries have the same degree of complexity throughout the growth. By comparison, after 1 h the surface was the flattest, consistent with the onset of recrystallization. However, the micrographs do not necessarily confirm that recrystallization has occurred; it is possible that the lower layers of the films have not recrystallized but are being covered by a layer of larger, more uniform crystallites as the growth progresses. It is also worthy of note that a large difference exists between the morphology of films grown at same temperature for the same deposition period, but using different carrier gas flow www.cvd-journal.de


Full Paper

0.03

(a)

Growth rate (nm/s)

0.025

0.02

0.015

0.01

0.005

0 500

520

540

560

580

600

620

640

660

680

700

Temperature (K)

0 0.00155

0.0016

0.00165

0.0017

0.00175

0.0018

0.00185

0.0019

0.00195

-0.5

(b) -1

Ln(growth rate)

-1.5 -2 -2.5 -3 -3.5 -4 -4.5

1/temperature (/K) Fig. 7. a) Growth rate versus temperature for Co film depositions; b) Arrhenius plot of the same data over the intermediate temperature range

rates, FESEM data showing a much less regular growth and smaller grains with the lower gas flow.

4. Conclusions Activated growth of cobalt was achieved through MOCVD, using CpCo(CO)2 as the precursor, with an activation energy of 68 kJ mol±1 over the temperature range 250±350 C. The film growth starts to diminish at higher (³ 350 C) temperatures, being probably caused by rapid desorption from the film and depletion of the reactants on reactor walls. Very slight deposition was observed for Chem. Vap. Deposition 2005, 11, 235±243

www.cvd-journal.de

temperatures lower than 150 C and no deposition below 140 C, as expected considering that the decomposition temperature for this precursor is around 140 C. Changes in the deposition temperature and time led to changes in the microstructure of the film. AFM, FESEM, and XRD measurements led to results which can offer a consistent theory about the crystalline structure and morphology of the films. In all the samples, a diffraction peak positioned at around 44.4 in 2h was measured; the diffraction maximum is characterized by different line profile parameters, dependent on the deposition conditions. Corroborative XRD, FESEM, and AFM data show that most deposited films consist of strongly (111) textured fcc' Co, 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

241

Full Paper

although the d-spacing is between that of fcc and hcp cobalt and a definitive crystal structure cannot be assigned. The (111) diffraction peak is split, with different d-spacings normal to the film plane occurring at different depths into the film, presumably through stress relief mechanisms. At the interface tensile in-plane stress will be present because of the differential thermal expansion of the glass substrate and the cobalt film as the sample cools after deposition. Tentative agreement with this is provided by the sin2w data, which also indicates a much more relaxed (but somewhat compressed) structure closer to the film surface. Although growth of cobalt films in the fcc phase has been reported previously for depositions on crystalline substrates, there are just two previous reports of cobalt growth on amorphous substrates.[10,31] All the fcc' samples exhibit a (111) texture; hence the conclusion that the grain growth is very likely to be controlled by surface and interface energy minimization, which promotes this texture.[32] The dominance of grains with a strong (111) texture but with random orientation with respect to rotation about the film normal suggests that grain growth takes place by abnormal growth mechanisms. In cobalt films grown on glass, the cobalt atoms are more likely to be strongly bound to each other than to the substrate; therefore theory predicts island growth, a common growth mode for metals on insulators.[33] The current micrographs support the idea of island growth, the films exhibiting granular morphology and the grain sizes, shapes and interstitial spaces depending on the deposition parameters. Dynamic AES shows a low contamination level in the films, with a lower impurity content than that found in some sputtered Co films.[20]

5. Experimental Growth of Co Thin Films: The thin films were grown by MOCVD, using an MO350 system built by Metals Research Semiconductors Ltd, specifically engineered for the deposition of magnetic films and multilayers. This CVD apparatus with a horizontal, IR heated reactor allows a high degree of deposition uniformity and very good operational reproducibility. To eliminate any impurities, the cell was cleaned before each new deposition with concentrated nitric acid, acetone and deionized water and a pump/purge procedure was carried out to ensure that all atmospheric gases were purged from the reactor by a flow of argon gas, prior to and post deposition. The thin films were deposited onto standard glass substrates. The amorphous and non-porous nature of glass leads to minimal influence of the substrate on the structure of the deposited film other than through the development of interfacial stress. The physical parameters of glass substrate relevant for this study are: anneal point 512 C; thermal expansion 9.2 ppm deg±1; strain point 470 C; rms 4.2 nm. Prior to deposition, a surface cleaning procedure was performed, which included washing in nitric acid followed by rinsing in deionized water, acetone, and again deionized water. The precursor, cyclopentadienylcobalt dicarbonyl, was provided by Epichem Ltd and used without further purification. CoCp(CO)2 is a red liquid with a vapor pressure at room temperature of 0.5 torr, soluble in water, with a boiling point of 37±38 C. The higher (~ 140 C) decomposition temperature [16] than those of other carbonyl compounds used for MOCVD reduces the risk of the premature decomposition of the precursor before reaching the substrate. The molecule appears to have a tenfold barrier to rotation of the Co(CO)2 fragment about the C5 axis of the Cp group and the value of the OC±Co±CO angle was found to be 98.5 [34]. The valency of Co in this precursor is +1; hence clean decomposition must be conducted in a

242


reducing atmosphere, in this case hydrogen gas. The chemical reaction leading to the deposition of Co is represented generically by Equation 1 [16]. C5H5Co(CO)2(g) +

1 nH2(g) ® Co(s) + 2CO(g) + C5H5+n(g) 2

(1)

The precursor was held in a bubbler through which part of the hydrogen flow was passed, the remainder entering directly into the reaction cell. The bubbler temperature was set at 12.7 C and the cell was under atmospheric pressure. To improve the substrate exposure to the precursor flux, the substrate was positioned slightly inclined on the susceptor (inclination angle of around 3). In a first set of experiments, using a constant deposition time of 1 h, thin films were deposited over the temperature range 120±400 C, in steps of 10 C. All other deposition parameters were maintained at the optimum values, as established in a previous study undertaken at Keele University [25]. Hydrogen was used as the carrier gas, with a flow of 70 sccm through the bubbler and 2000 sccm through the reactor. These films were continuous and uniform over the central 1 cm2 of the sample. Attempts to deposit at higher (³ 400 C) deposition temperatures led to films which could be seen by eye to be thinner and less uniform, probably owing to rapid desorption from the film surface and depletion of the reactants on reactor walls. Only very slight deposition was observed at temperatures lower than 150 C and no deposition occurred below 140 C, as expected considering that the decomposition temperature of CpCo(CO)2 is around 140 C. In a second set of experiments, at a substrate temperature of 350 C, the deposition was carried out for 10 min, 20 min, 0.5 h, 1 h, 1.5 h, and 2 h, in order to observe the evolution of the grain morphology and crystalline structure during the growth process. The H2 flow through the bubbler was increased to 100 sccm for these depositions. Microstructural Characterization: The surfaces of the films, the size and the shape of the grains and intergranular interstices were analyzed by FESEM and AFM. While the electron micrographs give information about the two dimensional morphology of the films, i.e., the shape and the size of the grains and the dimensions of interstices, AFM was used to provide additional information, including the 3D profile of the film surface, its roughness and geometric complexity. A Hitachi S-4500 cold field emission SEM with single crystal tungsten tip cathode was used for the electron microscopy. This was operated at accelerating voltages up to 30 keVand under low pressure (10±7 Pa in the electron gun and 10-4 Pa in the specimen chamber). A Digital Instruments NanoScope III was utilized for the AFM measurements. All samples were analyzed using 256 data points per scan line, with a vertical scan range and scan rate varied in the ranges 5±440 V and 5±9 Hz, respectively, in order to optimize the quality of the image. Roughness parameters were calculated for distinct areas, as well as for profile lines selected both along and across the growth directions, while the mean grain size and fractal factor calculations were performed over larger areas. Crystallographic Characterization: Extensive XRD characterization was performed in order to analyze the crystallinity and the stress in the films. Routine patterns for phase identification were collected using a coupled h/2h Phillips X-ray Powder Diffractometer, operating with Cu Ka radiation, over an angular range of 10 to 90. These patterns were analyzed using an in-house fitting program (SPLOT') for a rigorous determination of the diffraction line parameters, assuming Voigt line profiles. The particle size in the direction normal to the film plane was calculated from the width of the diffraction peak at half-height, according to the Sherrer formula. However, it must be considered that the individual grains might be non-uniformly strained and this strain can contribute to the diffraction line broadening, hence limiting the interpretation of the line broadening in terms of the Sherrer equation for determining the size of the individual crystals in a thin film. The high-resolution powder X-ray diffractometer at Station 2.3 of the CCLRC synchrotron radiation source at Daresbury was used to perform complementary high-resolution measurements, with the intention of determining the stress in the films. This diffractometer, equipped with an Eulerian cradle, has high precision sample x (0.2 mdeg) and detector 2h (0.1 mdeg) circles, v and u circles (orthogonal to x and v, respectively) with a precision of 0.5, and allows the operation of several scattering geometries. It also offers the advantage of a well-collimated white beam from the synchrotron (~ 1 mrad vertical divergence) and wavelength tunability (0.7±2.5 ). When performing stress measurements, there are distinct benefits to using such an instrument [35]. Several methods are presently used to measure strains and stresses in thin films [36±39]. The stress measurement technique employed in this work was the sin2w method, where w = x ± h, with x the angle between the sample surface and the incident beam and h the Bragg angle. This kind of measurement has been performed previously on thin films [38]. The beam size was set to 2 mm (vertical) 10 mm (horizontal), and the wavelength to k = 2.2 . The relatively large beam was necessary at this wavelength to www.cvd-journal.de


Full Paper

improve photon flux and to provide a bigger w range. An in-plane stress will cause a shift in the peak position when a symmetric h/2h scan is repeated with the sample offset by an angle w about the diffractometer axis. In the absence of shear stress or a stress gradient over film thickness, a linear variation of 2h with sin2w is expected, as shown in Equation 2 2

D2h 360=p tan h sin wr1 m=E

(2)

E is Young's modulus and m is Poisson's ratio (E = 209.6 GPa [27], and m = 0.32 [40], for cobalt). However, there are practical difficulties caused by large grain size and preferred orientation: a moderate degree of preferred orientation causes no difficulty, but a sharp texture leads a very strong diffraction line at w = 0 which is almost absent at w = 45 [41]. Compositional Analysis: Dynamic AES was used at Loughborough University for compositional analysis and determination of the film thickness, using a Varian Scanning Auger electron spectrometer with a multiplex ionbeam depth profiling system, cylindrical mirror analyzer with five micrometer resolution, Auger chemical mapping system and multiple sample handling facilities. Growth Rate Calibration: No direct measurements were made of film thickness. The most reliable calibration was obtained from the time taken to sputter through the film during dynamic AES measurements. The film thickness can then be inferred from the known sputtering rate of pure cobalt. Since the sputtering rate of the films can be expected to be higher than that of bulk cobalt, this gives an upper limit to an estimate of the film thickness. This information was only available for one film, although multiple regions of this were studied to provide a measure of the thickness and compositional uniformity. Comparisons of thickness between films have been carried out approximately by measuring the integrated intensity of the main diffraction line in the cases where there is a very strong preferred orientation. This has been corrected for X-ray absorption by the film itself and calibrated against the value obtained by AES. This is an indirect measure of the film thickness, relies on a similar degree of preferred orientation in the films, and assumes that the X-ray scattering is only from crystalline cobalt and that there is not a significant amount of amorphous cobalt present. The growth rates were calculated using all the data available and averaging for a particular temperature. For the films deposited in the middle of the deposition temperature range these are reasonable assumptions and allow a rough calculation of growth rate and activation energy. This calculation relies on the experimental proof that in the same reactor, under nominally identical conditions to those used here, the thickness of Co films from CpCo(CO)2 increases linearly with time over at least the first two hours [25]. The individual values for thickness are not completely accurate since measured indirectly. Received: September 7, 2004 Final version: February 7, 2005 ± [1] L. Szunyogh, B. Újfalussy, P. Bruno, P. Weinberger, J. Magn. Magn. Mater. 1997, 165, 254. [2] F. López-Urías, J. Dorantes-Dµvila, H. DreyssØ, J. Magn. Magn. Mater. 1997, 165, 262. [3] P. Chubing, D. Daosheng, F. Ruiyi, Phys. Rev. B 1992, 46, 12022. [4] S. Rüegg, G. Schütz, P. Fischer, R. Wienke, J. Appl. Phys. 1991, 69, 5655. [5] H. W. van Kesteren, A. J. den Boef, W. B. Zeper, J. H. M. Spruit, B. A. J. Jacobs, P. F. Garcia, J. Appl. Phys. 1991, 70, 2413.

[6] T. P. Nolan, R. Sinclair, R. Ranjan, T. Yamashita, IEEE Trans. Magn. 1993, 21, 299. [7] T. Lucinski, G. Reiss, H. Brückl, J. Magn. Magn. Mater. 1999, 193, 484. [8] T. Maruyama, Jpn. J. Appl. Phys., Part 2 1997, 36, L705. [9] T. Maruyama, T. Nakai, Appl. Phys. Lett. 1991, 59, 1433. [10] B. X. Fu, H. Wang, J. Magn. Magn. Mater. 1998, 187, 47. [11] M. Hehn, K. Cherifi-Khodjaoui, K. Ounadjela, J. P. Bucher, J. Arabski, J. Magn. Magn. Mater. 1997, 165, 520. [12] J. M. Alameda, F. Carmona, F. H. Salas, L. M. Alvarez Prado, R. Morales, G. T. Perez, J. Magn. Magn. Mater. 1996, 154, 249. [13] M. E. Gross, K. J. Schnoes, in Chemical Vapor Deposition (Ed:. W. Cullen), Proc. 10th Int. Conf., Electrochem. Soc. Proc. Vol. 87-8, The Electrochemical Society, Pennington, NJ 1987, p. 759. [14] F. Y. Ogrin, P. W. Haycock, J. Magn. Magn. Mater. 1996, 155, 199. [15] M. A. Paranjape, A. U. Mane, A. K. Raychaudhuri, K. Shalini, S. A. Shivashankar, B. R. Chakravarty, Thin Solid Films 2002, 413, 8. [16] G. J. M. Dormans, G. J. B. Meekes, E. G. J. Staring, J. Cryst. Growth 1991, 114, 364. [17] R. W. Hoffman, Thin Solid Films 1976, 34, 185. [18] J. W. Matthews, Surf. Sci. 1972, 31, 241. [19] R. Koch, R. Aberman, Thin Solid Films 1985, 129, 63. [20] Y. K. Ko, D. S. Park, B. S. Seo, N. I. Yang, H. J. Yang, H. J. Shin, J. Y. Kim, J. H. Lee, W. H. Lee, P. J. Reucroft, J. G. Lee, Mater. Chem. Phys. 2003, 80, 560. [21] P. A. Lane, P. E. Oliver, P. J. Wright, C. L. Reeves, A. D. Pitt, B. Cockayne, Chem. Vap. Deposition. 1998, 4, 183. [22] R. S. Dickson, P. Yin, M. Z. Ke, J. Johnson, G. B. Deacon, Polyhedron 1996, 15, 2237. [23] E. Hurtós, J. Rodríguez-Viejo, J. Appl. Phys. 2000, 87, 1748. [24] R. J. Cernik, P. K. Murray, P. Pattison, A. N. Fitch, J. Appl. Cryst. 1990, 23, 292. [25] F. Y. Ogrin, Ph.D. Thesis, Keele University, 1996. [26] M. L. Hitchman, K. F. Jensen, Chemical Vapor Deposition, Academic Press, London, 1993. [27] R. S. Young, Chemical and Physical Properties of Cobalt, Reinhold Publishing Corporation, Chapman & Hall, Ltd., London, 1960, p. 69. [28] K. Hara, K. Itoh, M. Kamiya, H. Fujiwara, K. Okamoto, T. Hashimoto, Jpn. J. Appl. Phys., Part 1 1994, 33, 3448. [29] M. Donnet, K. M. Krishnan, Y. Yajima, J. Appl. Phys. 1995, 28, 1942. [30] J. V. Sanders in Chemisorption and Reactions on Metallic Films, (Ed: J. R. Anderson) Academic Press, London, 1971 Vol. 1, p. 7. [31] K. Sculmeister, J. G. Lunney, B. Bucley, J. Appl. Phys. 1992, 72, 3480. [32] C. V. Thompson, R. Carel, J. Mech. Phys. Solids 1996, 44, 657. [33] J. Bloem, J. Cryst. Growth 1980, 50, 581. [34] M. A. Roehrig, Q. Q. Chen, S. T. Haubrich, S. G. Kukolich, Chem. Phys. Lett. 1991, 183, 84. [35] A. N. Fitch, C. R. A. Catlow, A. Atkinson, J. Mater. Sci. 1991, 26, 2300. [36] B. D. Cullity, P. Zorobedian, F. Adar, Appl. Phys. Lett. 1983, 43, 177. [37] J. T. Pan, I. Blech, J. Appl. Phys. 1984, 55, 2874. [38] C. C. Tang, E. J. MacLean, N. D. Telling, J. Magn. Magn. Mater. 1999, 199, 653. [39] M. Levit, E. Zolotoyabko, B. Z. Weis, Appl. Phys. Lett. 1996, 69, 4242. [40] E. Fletcher (co-ordinator), Cobalt Monograph, Centre d'Information du Cobalt, Brussels, 1960, p. 109. [41] B. D. Cullity, Elements of X-Ray Diffraction, Addison-Wesley, Reading, MA 1978.

______________________


www.cvd-journal.de


243