Structure and Giant Magnetoresistance of Co-Fe/Cu Multilayer Films Electrodeposited from Various Bath Formulations

For multilayer deposition, two-pulse plating from a single-bath was applied, i.e., the salts of all metals to be deposited were present in a single electrolyte. For each bath formulation investigated, the same preparation procedure was applied.

To get rid of the effect of precipitates appearing in the solution with time due to the formation of ferric ions and their hydrolysis, a single iron-free stock solution was used for each major bath types. The Mohr's salt was only added immediately before the electrodeposition experiments. The resulting electrolyte will be denoted by the relative Fe²⁺ ion concentration c_ion,Fe in the bath which was defined by the ionic ratio Fe²⁺/[Fe²⁺ + Co²⁺]. Two different Fe²⁺ ion concentrations were used (5 and 20 mol%, as defined above). The composition of the two stock solutions for the four bath formulations are summarized in Table I. Cyclic voltammetric measurements performed for the test of the baths were carried out with Pt sheet working electrodes in the same cell that was used for the multilayer preparation.

The Co-Fe/Cu multilayers were deposited on a [100]-oriented, 0.26 mm thick silicon wafer covered with a 5 nm thick Cr and a 20 nm thick Cu layer, both made by evaporation. The purpose of the chromium layer was to assure adhesion whereas the Cu layer ensured a good electrical conductivity of the cathode surface. The deposition was performed in a tubular cell of 8 mm × 20 mm cross section with an upward-facing cathode at the bottom of the cell.^21,22 The galvanostatic-potentiostatic (G/P) pulse combination^19,21 was used to deposit the multilayered samples. For the deposition of the magnetic layer, galvanostatic (G) mode was used with two different current densities (−20.7 mA/cm² and −34.5 mA/cm²). For the Cu-layer, potentiostatic (P) mode was used. The potential was measured and will be referred to hereinafter with respect to a saturated calomel electrode (SCE). By varying the deposition time in the G mode, the magnetic layer thickness could be set to a predetermined value. For controlling the thickness of the Cu layer, the charge flowing through the system was measured during the P pulse. Then, one can calculate the charge necessary to get the preset nominal layer thickness from Faraday's law by assuming 100% current efficiency.

The current efficiency of Cu deposition at the optimized potential can be taken as 100% since hydrogen evolution is negligible here. Furthermore, recent detailed XRD studies^23,24 demonstrated that for electrodeposited Co/Cu multilayers prepared with optimized Cu deposition potential, both the magnetic and the non-magnetic layers have actual layer thicknesses within about 10% of the preset nominal values obtained from Faraday's law with 100% current efficiency. Since we have to deal here with the Co-Fe-Cu system electrodeposited from various bath formulations, we have carried out a direct measurement of the current efficiencies for the magnetic layers. For this purpose, from each bath formulation at both Fe²⁺ ion concentrations and at both current densities used for magnetic layer deposition in the multilayers, we prepared d.c. plated deposits and by weighing the substrates before deposition and after deposition together with the deposits, we determined the current efficiency by using an average molar mass for the Co-Fe deposit. This analysis yielded current efficiencies in most cases within about 5% of the ideal current efficiency (100%), the few larger deviations coming certainly from random errors. Therefore, in all our multilayers we can expect that the actual layer thicknesses correspond well to the nominal preset values for both kinds of constituent layers.

Several sample series were produced with the common goal of investigating the effect of both the Fe/Co ratio in the magnetic layer and the bath formulation on the magnetotransport properties of the samples. The nominal layer thicknesses were fixed at d_FeCo  =  5 nm and d_Cu  =  5 nm with a total multilayer thickness of 1000 nm. We have used here fixed thicknesses for both the magnetic and non-magnetic layers since the major goal was a comparative study of the effect of various bath formulations on structure and GMR. The actual values of the layer thicknesses were chosen since in our previous study of electrodeposited Co-Fe/Cu multilayers,¹¹ the GMR maximum was found to be around d_Cu  =  5 nm and in multilayered nanowires usually even higher layer thicknesses have been used for the magnetic layers than in multilayer films. Furthermore, we have demonstrated recently²⁵ that a very regular layered structure can form in electrodeposited Ni-Co(5nm)/Cu(5nm) multilayered nanowires fabricated with an electrochemically optimized Cu deposition potential.

The overall multilayer composition was determined in a TESCAN MIRA3 scanning electron microscope (SEM) equipped with an EDAX Element EDS analyzer. The typical maximum error for compositional analysis of metallic elements in alloys is about 2 at.%.

The structure of the Co-Fe/Cu multilayers was studied by XRD, using a Philips X'pert powder diffractometer with CuKα radiation (wavelength: λ = 0.15418 nm), Bragg-Brentano geometry and a secondary graphite monochromator. The applied voltage and current were 40 kV and 30 mA, respectively. The step size in the measurement of the diffraction angle (2θ) was 0.02°. The lattice constants of the different phases were determined from the peak positions (peak 111 for the fcc phase and peak 110 for the bcc phase) using the Bragg equation. For the samples with high Fe content where both fcc and bcc phases were present, the 111 and 110 peaks of the two phases strongly overlapped. Therefore, the peak positions of the individual phases were obtained by fitting two Lorentzian functions on the overlapping XRD profiles. For some samples with an fcc structure, faint shoulders appeared on both sides of the main 111 Bragg reflection which seemed to correspond to multilayer satellites. However, due to the relatively large multilayer periodicity (Λ_nom = 10 nm), the satellite peaks strongly overlapped with the main peak that itself was also broadened due to various reasons. Therefore, in these cases the main and satellite peaks were separated by fitting three Lorentzian functions on the profile consisting of the main fcc 111 peak and the two satellite reflections. The multilayer periodicity (Λ) was obtained from the Bragg angles of the two satellite peaks (2θ_s1 and 2θ_s2) around the reflection 111 of the fcc phase using the following equation: Λ = λ/[sinθ_s1 – sinθ_s2].

The magnetotransport parameters were determined at room temperature by using four-point-in-line probes. To obtain the magnetoresistance (MR), the resistance was measured as a function of the external magnetic field (H) up to 9 kOe. The MR ratio was defined with the formula MR(H)  =  [R(H) - R₀ ]/R₀ where R₀ is the resistance maximum of the sample in a magnetic field close to zero and R(H) is the resistance in an external magnetic field H. The magnetoresistance data were determined in the field-in-plane/current-in-plane geometry in both the longitudinal (LMR, magnetic field parallel to the current) and the transverse (TMR, field perpendicular to the current) configurations. If one takes the difference between the longitudinal and the transverse component, the anisotropic magnetoresistance can be obtained: AMR  =  LMR – TMR.

The measured MR(H) curves were decomposed according to a procedure described previously²⁶ in order to establish the ferromagnetic (GMR_FM) and superparamagnetic (GMR_SPM) contributions to the GMR. In this process, a Langevin function is fitted to the high-field section of the MR(H) curves (beyond the saturation of the FM regions, typically above about 2–3 kOe) which yielded the parameters of the GMR_SPM contribution and, then, this term was subtracted from the experimental data which procedure provided us the FM contribution to the magnetoresistance. In most cases, a linear term should also be considered in the fitting procedure which accounted for the nearly linearly decreasing resistivity of a ferromagnet with increasing magnetic field at finite temperatures.^27,28

For multilayer deposition with reliable layer thicknesses, the electrochemically optimized Cu deposition potential E^EC_Cu has to be established¹⁸ where neither the dissolution of the magnetic layer nor the codeposition of either of the magnetic metals can take place during the Cu deposition pulse.

In order to obtain preliminary information for the optimization of the Cu deposition potential, the polarization curves of the stock solutions without Fe²⁺ ions and with two different Fe²⁺ ion concentrations were measured for each bath formulation investigated (see Fig. 1).

Zoom In Zoom Out Reset image size

An inspection of the cyclic voltammetry curves of the two Fe-containing solutions shown in Fig. 1 suggests that E^EC_Cushould lie somewhere in the potential range of the extended plateau since the optimal potential, at which neither a dissolution of the previously deposited magnetic material, nor a codeposition of the magnetic atoms with Cu will occur, can be expected only in this range. However, this is still a wide potential window; therefore, the final optimization can only be carried out on the basis of the current transient studies.^18,21 It is also apparent from the curves in Fig. 1 that the onset of the dissolution of the magnetic layer is shifted toward negative potentials for all bath types studied as the Fe²⁺ concentration in the solution increases, and, in parallel, the initial anodic current gets larger. This is related to the facts that the onset potential of the alloy dissolution shifts to the negative direction with the increase of the solution Fe content and the density of defects on the Cu layer growing on the Fe-containing Co layer increases.

To determine the exact value of E^EC_Cu, the current transients have to be measured over the plateau region of the CV curve.¹⁸ The potential value at which the current transient curve decays the fastest without the appearance of a current more negative than the current value specific for the Cu deposition with constant rate (diffusion-limited current) corresponds to the optimum E^EC_Cu value. Similarly as in our previous study,¹¹ the potential ranges determined from the polarization curves for mapping out the current transients were chosen for all baths as follows: −0.760 V to −0.520 V for the solution with c_ion,Fe  =  5% and  −0.700 V to −0.600 V for c_ion,Fe  =  20 %. The current transients measured for all baths studied here with either low or high Fe²⁺ ion concentration were very similar to those reported in our previous work.¹¹ The present E^EC_Cu values determined from the current transients are collected in Table II for all investigated baths.

According to the above results (Table II), it is found for the Co-Fe-Cu system that the value of E^EC_Cu depends on the bath formulation. Additionally, for a given bath formulation, it also depends on the relative concentration of Fe²⁺ ions in the solution. This comes from the circumstance that Fe (or a Fe-Co alloy) starts to dissolve from the cathode surface at a more negative potential than Co. Because of this difference, if both Fe and Co metals are simultaneously present on the surface with different ratios, the potential value at which the given alloy is neither deposited nor dissolved back into the electrolyte will also depend on the ratio of Fe and Co in the alloy. A further technical problem arises because only the total current flowing through the surface can be measured. Even if the total current is zero, it is possible that Fe is dissolved selectively and, in parallel, Co is deposited at the same rate. However, both the dissolution and the deposition are slow enough for the magnetic layer to be covered with Cu without significant change in the magnetic layer thickness.

The observed variation of E^EC_Cu with c_ion,Fe, i.e., that E^EC_Cu is more negative for higher c_ion,Fe, is in agreement with the results of our previous study on the Co-Fe/Cu system¹¹ and this observation is valid for all four bath formulations.

The XRD patterns for the two multilayer samples deposited from the sulfate bath at the lower current density value (j_magn = –20.7 mA/cm²) are shown in Fig. 3 (left panel: sample 1, c_ion,Fe = 5 mol%; right panel: sample 3, c_ion,Fe = 20 mol%). According to the left panel, satellite reflections were observed for sample 1. A qualitatively very similar XRD pattern was obtained also on sample 2 deposited with j_magn = –34.5 mA/cm². The appearance of multilayer satellite reflections requires that the amplitudes of the reflected X-rays are added sufficiently coherently²⁹ which can only occur for a fairly periodic repetition of the bilayer stacks. Therefore, the results indicate that samples 1 and 2 exhibit a well-defined multilayered structure consisting of an alternating sequence of Co₈₉Fe₁₁ magnetic layers and pure Cu layers.

For sample 1 (left pattern in Fig. 3), the bilayer period deduced from the positions of the satellite reflections was Λ_XRD = 14 nm (for sample 2, also Λ_XRD = 14 nm was obtained). By considering that the nominal multilayer periodicity was Λ_nom = 10 nm, this means that the Λ_XRD values exceed the nominal value by 40%. This result conforms qualitatively to our previous experience in that the repeat periods deduced from XRD for electrodeposited Co/Cu and Co-Fe/Cu multilayers^{11,23,24,30,31} was typically 10 to 30% higher than the nominal values.

According to Michaelsen,²⁹ for a coherent multilayer with layer thicknesses corresponding to our ones, a single main XRD peak should appear the position of which is intermediate between the positions corresponding to the constituent bulk materials. We can see in Fig. 3 that this is well fulfilled for the Co-Fe/Cu multilayer sample 1 (and also for sample 2, not shown). It is implied, furthermore, by this result that the Co₈₉Fe₁₁ magnetic layer adopts an fcc structure as observed previously for Co-Fe with low Fe-content¹¹ as well as for Co in Co/Cu multilayers.^{11,23,24,30,31} By considering the small energy difference between the hcp and fcc lattices, the formation of a metastable fcc-Co₈₉Fe₁₁ phase is possible since the equilibrium phase diagram of the Co-Fe system indicates the borderline between the fcc and bcc phases around this composition.³²

From the peak position of the observed fcc(111) peak, we have calculated the fcc lattice parameters which are collected in Table IV. For the two multilayers with satellite reflections, we obtained a = 0.3580 nm (sample 1) and a = 0.3578 nm (sample 2). The standard lattice parameter for pure fcc-Cu is a = 0.36148 nm³³ and for pure fcc-Co is a = 0.35446.^33,34 Thus, we can see that the lattice parameters for both samples are between the pure Co and Cu values as it should be for a multilayer with layer thicknesses like in our case.²⁹ We can make a more correct estimate for the expected multilayer lattice parameter by taking into account the composition analysis results and using Vegard's law with an fcc-Fe lattice parameter of a = 0.36017 nm.³⁴ These estimates are given in the uppermost box above the XRD pattern for sample 1 in Fig. 3: first, we estimated the expected lattice parameter for the Co₈₉Fe₁₁ magnetic layer and, then, for a (Co-Fe)₄₄Cu₅₆ alloy corresponding to the analyzed overall multilayer composition. Comparing the multilayer lattice parameter estimated with the help of Vegard's law with the measured one, the agreement is very good for sample 1: Δa = +0.0007 nm (and even better for sample 2: Δa = +0.0002 nm). This feature and the clear satellites demonstrate that both samples 1 and 2 can be considered as exhibiting a fairly regular superlattice structure.

The XRD pattern of sample 3 which was deposited from the sulfate bath with a higher Fe²⁺ ion concentration of 20 mol% is shown in the right panel of Fig. 3 (very similar results was obtained also for sample 4 deposited with the higher current density of j_magn −34.5 mA/cm²). We can observe that for a Co:Fe atomic ratio of about 2:1 in the magnetic layer, in addition to an fcc(111) peak between the fcc-Cu(111) and fcc-Co(111) positions, also a large bcc(110) peak appears, corresponding to the Co-Fe magnetic layer. Again using Vegard's law, we have estimated the lattice parameter of this bcc magnetic layer by using a bcc-Fe lattice parameter of a = 0.28665 nm³³ and a bcc-Co lattice parameter of a = 0.2827 nm.³⁵ The uppermost box above the XRD pattern in the right panel indicates a fairly good agreement of the estimated lattice parameter of the bcc Co-Fe alloy with the experimental value of the bcc-phase lattice parameter (deviation: Δa(bcc) = +0.0001 nm for sample 3; for sample 4: Δa(bcc) = +0.0003 nm. It is noted finally that the lattice parameter of the fcc-Cu phase is larger than the value of the bulk fcc-Cu phase (Δa(fcc) = +0.0009 nm, see uppermost box in the right panel of Fig. 3) and this indicates a significant stretching of the fcc-Cu regions in the two-phase deposit which may arise due to the matching of the lattice planes at the interfaces between the fcc and bcc crystallites. Interestingly, the difference Δa(bcc) = +0.0001 nm between the lattice parameters between the Vegard's law value and the experimental value of the bcc phase is much smaller which should mean that the stresses in the bcc regions are much smaller. This may be possible if the bcc phase has a lattice plane for matching with the fcc phase that is more conform to the joint lattice plane distances.

It is worth noting that the ratio of the amplitudes of the first XRD peaks of the fcc and bcc phases shown in this study does not reflect necessarily the volume ratio of the two phases since the intensity of the peaks are influenced by the texture, the multiplicity and the different chemical compositions of the two phases. The texture effect can be handled if the sum of the areas under the peaks in the whole diffractograms are used for the comparison. Indeed, for instance for sample 3 the ratio of the sums of the peak intensities for the fcc and bcc phases was 1.04, although the (110) peak for the bcc phase was much stronger than the reflection (111) for the fcc phase in Fig. 3.

Even if the fcc and bcc phases in samples 3 and 4 are present in the form of an alternating sequence of continuous fcc and bcc layers, we cannot expect the occurrence of satellite reflections since the layer thickness non-uniformities and the strong dilatation and/or compression at the interfaces between the two kinds of layer certainly destroy the coherency of the X-rays which would be the prerequisite for satellite reflections.²⁹ Furthermore, according to the considerations of Michaelsen,²⁹ we cannot decide from the present XRD results whether a bcc-[Co-Fe]/fcc-Cu layered structure was formed during deposition or the deposit consists of a random mixture of bcc-(Co-Fe) alloy and fcc-Cu grains.

The XRD patterns for the two multilayer samples deposited from the sulfate bath at the lower current density value (j_magn = –20.7 mA/cm²) are shown in Fig. 4 (left panel: sample 5, c_ion,Fe = 5 mol%; right panel: sample 7, c_ion,Fe = 20 mol%). According to the left panel, satellite reflections were observed for sample 5. A qualitatively very similar XRD pattern was obtained also on sample 6 deposited with j_magn = –34.5 mA/cm². Similarly to samples 1 and 2 deposited from the sulfamate bath, these results indicate that samples 5 and 6 exhibit a well-defined multilayered structure consisting of an alternating sequence of Co₈₈Fe₁₂ magnetic layers and pure Cu layers.

Zoom In Zoom Out Reset image size

For sample 5 (left pattern in Fig. 4), the bilayer period deduced from the positions of the satellite reflections was Λ_XRD = 11 nm (for sample 6, it was Λ_XRD = 13 nm). By considering that the nominal multilayer periodicity was Λ_nom = 10 nm, this means that the Λ_XRD values exceed the nominal value by 10 and 30%, respectively, again in good agreement with our previous experience on the relation of the nominal and measured repeat periods of electrodeposited multilayers.^{11,23,24,30,31}

We can see in Fig. 4 for the multilayer sample 5 (and also for sample 6, not shown) that a single main XRD peak appears the position of which is intermediate between the positions corresponding to the constituent fcc bulk materials. This implies again that the Co₈₈Fe₁₂ magnetic layer adopts an fcc structure.

From the peak position of the observed fcc(111) peak, we have calculated the fcc lattice parameters also for samples 5 and 6 produced from the sulfate bath and the results are given in Table IV. Comparing the multilayer lattice parameters estimated again with the help of Vegard's law with the measured ones, the agreement was surprisingly good for sample 5: Δa = +0.0007 nm (and very similar result was obtained for sample 6: Δa = +0.0012 nm). This feature and the clear satellites demonstrate that both samples 5 and 6 can also be considered as exhibiting a fairly regular superlattice structure.

The XRD pattern of sample 7 which was deposited from the sulfate bath with a higher Fe²⁺ ion concentration of 20 mol% is shown in the right panel of Fig. 4 (very similar results was obtained also for sample 8 deposited with the higher current density of j_magn = −34.5 mA/cm²). Again, similarly to the sulfamate bath samples with higher Fe-content, in addition to an fcc(111) peak between the fcc-Cu(111) and fcc-Co(111) positions, also a large bcc(110) peak appears, corresponding to the Co-Fe magnetic layer. The same analysis on the basis of Vegard's law was carried out also here as for sample 3 from the sulfamate bath and very similar conclusions could be drawn both qualitatively and quantitatively (see the uppermost boxes in the right panel of Fig. 4.

The absence of satellite reflections for samples 7 and 8 from sulfate bath can be explained in a similar fashion as was done for samples 3 and 4 from the sulfamate bath.

The XRD pattern of sample 10 deposited from the chloride bath with a low Fe²⁺ ion concentration of 5 mol% is shown in the left panel of Fig. 5. A single fcc (111) Bragg peak can be observed, indicating that the magnetic and non-magnetic regions in the deposit adopt a common fcc lattice, but no satellite reflections can be identified. Estimating again the lattice parameter of the deposit with the known composition by Vegard's law, we get a very good agreement with the experimental lattice parameter given in Table IV. In the absence of any satellite reflections, we cannot decide whether we have to deal here with either a layered structure consisting of alternating layers of fcc-Cu and fcc-Co₉₀Fe₁₀, both with a layer thickness of about 5 nm or nanoscale regions of these two phases are intermixed randomly since both arrangements would yield the same XRD pattern.²⁹

Zoom In Zoom Out Reset image size

The XRD pattern of sample 12 deposited from the chloride bath with a high Fe²⁺ ion concentration of 20 mol% is shown in the right panel of Fig. 5 (a very similar pattern with somewhat stronger bcc peak with respect to the fcc peak was obtained also for sample 11 deposited at the lower value of j_magn). Similarly to the sulfate bath, we can observe that for a Co:Fe atomic ratio of about 2:1 in the magnetic layer, in addition to an fcc(111) peak between the fcc-Cu(111) and fcc-Co(111) positions), also a bcc(110) peak appears, corresponding to the Co-Fe magnetic layer. Interestingly, whereas the compositions of samples 7 and 8 (sulfate bath) as well as of samples 11 and 12 (chloride bath) are fairly similar, the ratios of the fcc and bcc peaks are just opposite in the chloride bath samples and the sulfate bath samples. However, when we calculated the sums of the areas under the peaks detected in the whole diffractogram for each of the two phases, similar values were obtained for the fcc and bcc phases, indicating that the stronger fcc(111) peak is a texture effect only.

Again using Vegard's law with the analyzed compositions, we have estimated the lattice parameter of the bcc magnetic layer in samples 11 and 12 from the chloride bath. The uppermost box above the XRD pattern of sample 12 indicates a fairly good agreement of the estimated lattice parameter of the bcc Co-Fe alloy with the experimental value of the bcc phase lattice parameter (a very similar result was obtained also for sample 11).

The XRD pattern of sample 13 deposited from the citrate bath with a low Fe²⁺ ion concentration of 5 mol% is shown in the left panel of Fig. 6 (a very similar XRD pattern was obtained also for sample 14 from the same bath). Similarly to the chloride bath with low Fe²⁺ ion concentration, a single fcc (111) Bragg peak can be observed, indicating that the magnetic and non-magnetic regions in the deposit adopt a common fcc lattice. Estimating the lattice parameter of the deposit with the known composition by Vegard's law, we get a fairly good agreement with the experimental lattice parameter (see Table IV). Similarly to the case of the chloride bath samples, the XRD patterns do not allow to decide whether we have to deal here with a layered structure or two kinds of nanoscale regions are intermixed in the sample.

Zoom In Zoom Out Reset image size

The XRD pattern of sample 15 deposited from the citrate bath with a high Fe²⁺ ion concentration of 20 mol% are shown in the right panel of Fig. 6 (a very similar pattern with somewhat stronger bcc peak with respect to the fcc peak was obtained also for sample 16 deposited from the same bath at higher magnetic layer deposition current density). Similarly to the sulfate bath, we can observe that for a Co:Fe atomic ratio of about 2:1 in the magnetic layer, in addition to an fcc(111) peak between the fcc-Cu(111) and fcc-Co(111) positions), also a large bcc(110) peak appears, corresponding to the Co-Fe magnetic layer. As opposed to the Fe-rich samples from the sulfate and chloride baths where the fcc peak amplitude was much smaller and larger, respectively, than the bcc peak, here in the case of the citrate bath, the fcc and bcc peaks have comparable magnitudes with a slightly larger bcc peak.

Again using Vegard's law, we have estimated the lattice parameter of this bcc magnetic layer. The uppermost boxes above the XRD patterns indicate a fairly good agreement of the estimated lattice parameter of the bcc Co-Fe alloy with the experimental value of the bcc phase lattice parameter (see Table IV).