Characterization of Flow-Induced Compositional Structure in Electrodeposited NiFe ~orn~osition-~odulated Alloys

Steven D. kith,* Weihua Wang, and Daniel T. Schwartz**

Department of Chemical Engineering, University of Washington, Seattle, Washington 981 95-1 750, USA

ABSTRACT

Flow-induced NiFe composition-modulated alloys (CMAs) are plated onto the disk of a rotating ring-disk electrode (RRDE) by oscillating the RRDE rotation rate during galvanostatic deposition. The relationships between processing and compositional structure in the electrodeposited CMAs are explored using an optimized potentiostatic stripping voltam- metry technique, scanning electron microscopy, and energy-dispersive X-ray spectroscopy. Results show that the CMA wavelength scales as the inverse of the flow oscillation frequency and the composition modulation amplitude are strong- ly affected by variations in both electrolyte flow oscillation frequency and amplitude. Fast Fourier transform analysis is used to probe the dynamic time scales of NiFe electrodeposition and to investigate the sensitivity of NiFe deposition to the oscillating electrolyte flow field. Results indicate that critical deposition chemistries occur over time scales much slower than those governing typical mass-transfer processes.

Introduction An emerging area of materials research is the plating of

alloys with a spatially periodic composition in one dimen- sion, referred to as composition-modulated alloys (CMAs). The composition modulation wavelength and the modula- tion amplitude are two parameters often used to define the periodic structure of a CMA. The wavelength represents the repeat length, or period, of the composition modulation, and the amplitude is a measure of the interlayer composi- tion variation. It is well established experimentally that many CMA properties (e.g., magnetostriction, magnetore- sistance, electrical resistivity, and yield strength) are strong- ly affected, directly or indirectly, by the alloy wavelength and amplitude.14 The engineering of alloys with specific properties, therefore, depends in part on the ability to tailor the compositional structure (e.g., CMA wavelength and amplitude) through the control of processing variables.

CMAs with a wide range of composition modulation wavelengths and amplitudes have been made in a variety of ways, including electrodeposition and vapor-phase tech- nique~.~-'' Electrodeposited, short-wavelength CMAs are typically made by periodic pulsing of the plating current during deposition. Recently it was shown that CMAs with nanometer-scale wavelengths can also be made by galvano- static deposition in a sub-Hertz oscillating electrolyte flow field using the rotating ring-disk electrode (RRDE)." In this previous study, potentiostatic stripping voltammetry (PSV) was used to evaluate the wavelength of flow-induced NiFe CMAs and to relate this alloy feature to processing variables such as current density and flow oscillation frequency. While this previous study was the first to illustrate the ex- perimental electrodeposition of flow-induced CMAs and the use of PSV in quantitative assessment of the modulation wavelength, the RRDE stripping technique was not opti- mized, preventing accurate determination of the composi- tion modulation amplitude.

To better explore the relationships between processing variables and compositional characteristics in flow-in- duced CMAs, we have optimized the PSV technique so that it is possible to assess both the CMA wavelength and amplitude, at least semiquantitatively. In this study, com- position profiles of NiFe CMAs deposited in the time peri- odic flow of an RRDE are analyzed using the improved PSV technique. The profiles show periodic composition variation with modest attenuation throughout the depth of the deposit. Analysis of the composition profiles using fast Fourier transforms (FFTs) results in a reciprocal space composition profile which allows quantitative analysis of the average CMA wavelength and semiquantitative char- acterization of the modulation amplitude. The FFT analy- sis is also used to probe the dynamic time scales of NiFe electrodeposition and to investigate the sensitivity of NiFe

* Electrochemical Society Student Member. * * Electrochemical Society Active Member.

deposition to a low-frequency oscillating electrolyte flow field. Comparison to CMA analysis using scanning elec- tron microscopy (SEM) and energy-dispersive X-ray spec- troscopy (EDS) illustrates that the optimized PSV tech- nique is a fast and accurate characterization tool.

Experimental A Pine Instruments RRDE with a platinum disk of radius

r, = 0.229 cm and concentric platinum ring with inner radius r, = 0.246 cm and outer radius r3 = 0.269 cm was used throughout the electroplating and stripping voltam- metry experiments. The RRDE, a large-area platinum mesh counter electrode, and a saturated calomel (SCE) reference electrode were used in single-chambered plating and strip- ping vessels. A Pine Instruments model MSRX rotator dri- ven by an external voltage source was used to control the instantaneous rotation rate of the RRDE. During electrode- position, a Wavetek function generator was used to drive the disk rotation with a well-defined periodic waveform (sinusoidal). A Pine Instruments model AFRDE5 bipoten- tiostat was used to control the disk and ring polarization during plating and stripping. Disk and ring currents were monitored continuously during stripping using a Macintosh Centris 650 personal computer interfaced with the bipoten- tiostat through a National Instruments 12 bit multifunction AID board. Disk and ring current acquisition and calcula- tion of the alloy composition profile were performed using custom software written in the LABVIEW programming environment. FFT analysis of the composition profiles was performed using the Igor Pro software package.

The plating bath was composed of 0.2 M Ni (H2NS03), 4H,O, 0.01 M FeC1,.4H20, 0.40 M H3B03, 1.5 g/L sodium saccharin, 0.2 g/L sodium dodecyl sulfate, and 1.0 g/L ascorbic acid and was operated at room temperature (-23°C). Bath pH was kept constant at 3.00 + 0.01 by addition of HC1 and/or NaOH between plating runs. The stripping bath composition was 0.2 M HC1 and 0.5 M NaCl and was also operated at room temperature. Both plating and stripping baths were made from reagent-grade chem- icals and 18 MR cm deionized water.

Prior to each plating run, the disk and ring electrodes were cleaned by cyclic potential scanning at 100 mV/s between -300 and +I300 mV vs. SCE for approximately 15 min in a 0.2 M HC1/0.5 M NaCl electrolyte. After clean- ing, NiFe thin-film CMAs were plated galvanostatically onto the disk of the RRDE using an applied current densi- ty of -20 mA/cm2. During plating, the ring potential was set to +800 mV vs. SCE to prevent metal deposition on the ring. Unless noted otherwise, plating during PSV experi- ments was terminated after a total charge of roughly 500 mC, resulting in a deposit nearly 1000 nm thick.

During deposition, convective mass transfer to the disk surface was modulated in a time-periodic manner by oscil- lating the disk rotation rate with the sinusoidal waveform

J. Electrochem. Soc., Val. 145, NO. 8, August 1998 0 The Electrochemical Society, Inc. 2827

2828 J. Electrochem. Soc., Vol. 145, No. 8, August 1998 The Electrochemical Society, Inc.

(1(t) = no + Afi cos (2irat)

where 11(t) is the instantaneous rotation rate, no is themean rotation rate, All is the rotation rate oscillationamplitude, and if is the oscillation frequency. The dimen-sionless disk rotation oscillation amplitude is defined asci = AU/fl0. Throughout the study, a constant mean rotationrate of U,, = 26.67 Hz (1600 rpm) was employed. In allexperiments, plating was initiated and terminated at theminimum rotation rate of the waveform, To fully investi-gate the effect of time periodic flow on alloy composition,a range of disk rotation oscillation amplitudes and fre-quencies was used. Compositional characteristics of NiFeCMAs were determined for alloys plated in sinusoidalflows with disk oscillation frequencies of 4, 10, 20, 40, 80,160, and 320 mHz and dimensionless disk oscillationamplitudes of 0.1, 0.3, 0.5, 0.7, 0.9, and 0.95.

After each plating run, the RRDE was removed from theplating bath, immersed in deionized water and rotated toprevent contamination of the stripping bath. The RRDEwas then submerged in the stripping solution and rotatedat a constant rate of 2500 rpm. The disk potential was setto +250 mV vs. SCE to oxidize the alloy, liberating Ni'2and Fe'2 from the film surface. The ring potential was setto +800 mV vs. SCE to detect Fe2 via the mass-transfer-limited reaction Fe'2 —* Fe4' + e . The disk and ring cur-rents were recorded continuously during the stripping toallow determination of the film composition profile, asdetailed elsewhere.1' Preliminary stripping studies wereperformed to determine the best disk stripping potentialto achieve high-quality, reproducible composition profiles.The primary concern in optimization of the PSV techniquewas to determine a stripping potential which resulted inoxidation of successive compositional strata in the CMAfilm. The use of an appropriate oxidation potential wasevident in the stripping voltammograms by the absence ofsignificant attenuation in the modulation amplitude of thealloy as stripping progressed (cf. Fig. 1 here and Fig. 2 inRef. 11). Potentials from 0 to +300 mV vs. SCE were in-vestigated to find the optimum stripping conditions, and+250 mV proved to be the best potential for the NiFe filmsinvestigated here.

Using the same potentiostat, rotator, and plating bathdescribed, flow-induced NiFe CMAs were also plated ontothe removable disk of a rotating disk electrode (RDE). Priorto plating, the removable copper disk electrode (radius0.25 cm) was polished to a mirror finish using 1.0 p.m dia-mond paste and cleaned by immersion in 5% H2S04 for 60 s.After rinsing with deionized water, the copper RDE wasplaced in a single-chambered electroplating vessel withSCE reference electrode and platinum mesh counter elec-trode. NiFe CMAs were plated galvanostatically using anapplied current density of —20 mA/cm' until a charge ofroughly 10,700 mC had passed, resulting in a deposit 15 p.mthick on average. During deposition, the disk rotation ratewas oscillated at a frequency of if= 4 mHz and dimension-less amplitude of a = 0.95 using a constant mean rotationrate 0 = 26.67 Hz. After plating, the RDE was again rinsedwith deionized water and then immersed in a standardWatt's plating bath5 from which a 300 p.m thick nickel layerwas electroplated over the NiFe film.

The plated copper disk electrode was removed from theRDE housing and cut in cross section using a diamondwafering blade. Cross-sectional samples were polished toa mirror finish using 1.0 p.m diamond paste and mountedon standard 1.25 cm diam aluminum SEM sample stubs.NiFe CMA cross sections were studied using a JEOL 5200SEM and EDS instrumentation and software manufac-tured by Link Analytical, Ltd. Quantitative EDS composi-tion maps and profiles were constructed using relativepeak intensities from the Ni Ku line of the Nile sampleand a polished nickel standard, taking into account cor-rections for matrix interactions within the NiFe sample.'2Structural and compositional details of the NiFe CMAswere investigated at a number of radial positions on thecross section, from the center of the disk to the edge,

[11 although an intermediate radial position was used for allanalyses that were compared directly to results from PSV,as described later.

Results and DiscussionPotentiostatic stripping voltammetry.—Figure 1 shows

a typical stripping voltammogram (A) and correspondingcomposition profile (B) for a NiFe CMA electrodepositedat —20 mA/cm' in an oscillating flow field of frequency

= 40 mHz and dimensionless oscillation amplitude a =0.95. As shown in (A), the disk and ring currents oscillateas the CMA is electrochemical stripped from the disk. Thedisk current oscillates due to the enhanced oxidation rateof Fe-enriched strata, and the ring current oscillates inresponse to the periodic flux of liberated ferrous ions fromthe oxidized film. At the end of stripping, the disk andthen the ring current falls to zero.

The corresponding composition profile is shown inFig. lB. Here the mole fraction of iron in the deposit, XF0,is plotted as a function of alloy depth, 5. The alloy compo-sition and depth were determined using the measured diskand ring currents in the manner described elsewhere.1' Thevariation in iron content is evident throughout the depthof the deposit. Six complete composition modulationcycles are present in the deposit, with each cycle beingroughly 140 nm in length (as measured between consecu-tive maxima or minima). This repeat length is the average

I I .- 2.0

Strip Time (s)

30

01.0

U

0.5

E

a.)

c-iadCd)

VIL.

a

C0Uct

ci-a)0

0.0

200 400 600

Deposit Depth, 5 (nm)

Fig. 1. (A) Typical stripping voltammogram for a NiFe CMAdeposited in a sinusoidal oscillating flow with a = 40 mHz anda = 0.95. The measured disk (—) and ring C- - - -) currents areshown as a function of stripping time. (B) Corresponding composi-tion profile determined from the stripping voltammogram in (A).The mole fraction iron in the deposit is plotted as a function ofdeposit depth. The overage CMA wavelength, A, and compositionmodulation amplitude, AXFO, can be estimated as indicated.

J. iectrocflem. SOC., VOl. 145, NO. S, August 199 () The Electrochemical Society, Inc.

wavelength of the CMA, denoted '..The composition mod-ulation amplitude, XF,, for the deposit is defined here bythe difference between the average composition maximaand the mean composition of the alloy. The CMA in Fig. 1exhibits an average composition modulation amplitude ofabout 14 mol % and a mean iron content of 36 mol %.Thus, on average, the alloy has a minimum iron contentnear 22 mol % and a maximum around 50 mol %. Theshape of the PSV composition profile is similar to the com-puted composition profile for an ideal noninteractingbinary pair.'3 Specifically, the profile is characterized bybroad, high concentration regions followed by quick"dips" into the low mole fraction strata. However, as dis-cussed later, the electrodeposition of NiFe displays a num-ber of dynamic traits quite different from an ideal binarypair.

A desirable consequence of using the optimized strip-ping technique is the high spatial fidelity of the composi-tion profile. (Here we use the term high spatial fidelity todescribe a profile in which all composition modulationsare evident and only modest attenuation from the firstperiod to the last is observed.) Using the nonoptimizedstripping conditions in our previous study, one typicallyfound that the amplitude of each successive modulationdecreased appreciably (cf. Fig. 2 in Ref. 11). In general, thespatial fidelity of profiles from the optimized PSV tech-nique is better than one typically finds using destructivecharacterization methods such as Auger spectroscopy withsputtering.' This is presumably due to the matching ofplating and stripping current distributions on the diskelectrode. That is, even though the deposit plates fasterand is thicker at the disk edge (compared to the center ofthe disk), the stripping rate matches this nonuniformityand etches the edge region proportionally faster. The resultis a high-fidelity composition profile with features repre-sentative of the average properties of the film. To assessthe utility of the optimized PSV technique in characteri-zation of NiFe CMAs, we compared the method to depositanalysis using SEM imaging and X-ray spectroscopy.

SEM and EDS analysis—Figure 2A is a secondary elec-tron image of a polished NiFe CMA cross section platedonto a copper substrate. The film was deposited at—20 mA/cm2 in an oscillating flow field of frequency =4 mHz and dimensionless oscillation amplitude a = 0.95.The micrograph clearly shows the substrate, composition-al strata in the NiFe film, and the protective nickel over-layer. As one would expect, given the nonuniform currentdistribution on the RDE, typical CMA cross sections ex-hibited a relatively thin NiFe film (—10 m) near the cen-ter of the disk and a much thicker deposit (—17 m) at thedisk edge. The micrograph in Fig. 2A illustrates the com-positional structure of the film at a radial location midwaybetween the center and edge of the disk at a film thicknessof roughly 13 m. This film location was chosen for char-acterization because analysis at an intermediate radialposition is the closest approximation to sampling the aver-age properties of a the film as is possible using EDS. Sincethe PSV technique provides information about averagefilm properties, compositional analysis at this radial posi-tion allows a more accurate comparison between the EDSand PSV characterization methods.

Figure 2B shows an EDS composition map and typicalX-ray energy spectra of compositional strata from theboxed region in Fig. 2A. Contrast differences in the mapcorrespond to different alloy compositions: Fe-enrichedregions appear dark and Fe-deficient regions are light. Aperiodic composition variation is clearly evident in thefilm. Using Ni Ku peak intensities from the sample and apure nickel standard, quantitative composition analysis ofthe region in Fig. 2B allowed construction of the CMAcomposition profile shown in Fig. 3A. The figure shows thecomposition variation across five successive strata of thefilm and illustrates a CMA with a wavelength of approxi-mately 1400 nm and a modulation amplitude of nearly9 mol %. The resolution of the profile, however, limits the

Energy (eV)Fig. 2. (A) SEM microgroph of a polished NiFe CMA cross sec-

tion. The micrograph shows the copper substrate, compositionalstrata in the NiFe alloy, and the protective nickel overlayer. (B) EDScomposition map and Xray energy spectra of compositional stra-ta from the boxed region of the CMA in (A). Fe-enriched strataappear dark and Fe-deficient regions are lighter.

utility of this technique, as discussed later. Figure 3B is aPSV composition profile of a NiFe CMA plated underidentical polarization and flow oscillation conditions aswere used to plate the alloy shown in Fig. 2. The two pro-files in Fig. 3 illustrate similar features; both techniquescapture the general shape of the composition wave andeither could be used to estimate.\ and The PSV tech-nique, however, measured a modulation amplitude nearly7 mol % greater than the EDS method and the shape of thecomposition wave is more readily apparent, i.e., the broad,high iron regions and the more narrow, low iron regionsare clearly represented in the PSV profile but are not dis-cernible in the EDS analysis.

Accurate determination of.\ and .XXF,. using either char-acterization technique is subject to the spatial and compo-sitional resolution of the given method. In EDS analysis,spatial and compositional resolution is compromised dueto lateral spreading of the X-ray signal from the samplesurface during acquisition.12 Since the detected X-ray sig-nal originates from an interaction volume with a diameterapproaching 1 p.m, high-resolution quantification of suc-cessive strata of this or smaller length scales (such as inthe CMA of Fig. 2 and 3) becomes difficult. The "smear-ing" of signal between successive CMA strata suggeststhat the modulation amplitude determined using EDSshould be viewed as an absolute lower bound to the actu-al composition modulation present in the alloy film.

B

Ni5,Fe4

Ni K

'II-0Fe) kL- iL0 30()0 6000 9000

2830 J. Electrochem. Soc., Vol. 145, No. 8, August 1998 The Electrochemical Society, Inc.

C

0

0

Fig. 3. (A) Composition profile of NiFe CMA constructed usingquantitative EDS analysis. The CMA exhibits a wavelength ofapproximately 1400 nm and a modulation amplitude of nearly9 mol %. Spatial and compositional resolution in the [OS profile iscompromised due to lateral spreading of the X-ray signal from thesample surface. (B) PSV composition profile of a NiFe CMA platedunder identical polarization and flow oscillation conditions as thefilm in (A). The PSV technique measures a CMA wavelength of1400 nm and modulation amplitude of 16 mol %.

Effectively, the EDS method is only reliable in analysis ofCMAs with wavelengths greater than 1 jim, with resolu-tion improving for thicker, micron-scale composition mod-ulations. The spatial and compositional resolution of thePSV technique, however, is tied directly to how well theplating and stripping current distributions are matched. Ifthe current distributions are not well matched, then simul-taneous etching of many compositional strata occurs andthe true composition of any given strata is underestimat-ed, much like the "smearing" of signals in X-ray analysis.Conversely, if the plating and stripping current distribu-tions are well matched, the result is a high-fidelity com-position profile such as those in Fig. 1 and Fig. 3B.

Quantitative composition profiling using FFT analy-sis—We have shown that the optimized stripping voltam-metry technique can be used to construct reasonably highfidelity composition profiles of the flow-induced CMAs.Using Fig. 1, the general characteristics of the spatiallyperiodic waveform can be determined by inspection, i.e.,the CMA wavelength is roughly 140 nm and the composi-tion modulation amplitude is about 14 mol %. Subtle dif-ferences in compositional structure, however, are not al-ways evident when the data is plotted as a spatial profile.Fourier transformation of the original composition profilenot only illustrates details of the alloy composition profilebut also reveals the influence of electrolyte flow dynamicson the resulting alloy composition waveform.

To illustrate the procedure and utility of using FFTs inthe present study, consider Fig. 4. Figure 4A is a composi-tion profile determined from the electrochemical strippingof a film made at —20 mA/cm2 in a sinusoidal flow of oscil-lation frequency ci = 80 mHz and dimensionless amplitudea = 0.95. As in Fig. 1, the variation in iron mole fractionthroughout the depth of the film is clearly evident inFig. 4A, and an estimate of alloy modulation wavelengthand amplitude can be made directly from the figure. Theaverage wavelength is approximately 70 nm and the aver-age modulation amplitude is about 9 mol %.

The FFT of the composition profile Fig. 4A is shown inFig. 4B, and the construction and interpretation of theFFT representation merits brief discussion. The FFT algo-rithm requires a data set in which the number of datapoints is an integral power of 2. We selected 512 datapoints from the composition profile XF,(S), starting fromthe end of the first complete modulation as labeled inFig. 4A. In this example, the 512 data points contain eightcomplete composition modulation periods. Depending onthe wavelength of the CMA, 2—32 complete modulationperiods were typically included in the 512 points. AHanning window was then applied to minimize the effectof aliasing. The FFT of the composition profile X,/5) re-sults in a complex data vector, X,Xk), containing real and

0

00

I)0

Fig. 4. (A) S atial composition profile of a NiFe CMA depositedin a sinusoidarflow with a = 80 mHz and a = 0.95. (B) FF1 of thecomposition profile shown in (A). The characteristic CMA ampli-tude, I I, is planed as a function of red procal space, k. Thefundamental peak at k = 1/A is used to quantify the averageCMA wavelength and characteristic amplitude.

0.6

0.5

0.4 -

A.t#A

LXXEe AAA A AA

AAAASA0.3

0.2 A

0.1

0.0

C)LI,

0

0UctI-,

0

500 1000 1500 2000 2500 3000

Deposit Depth, 3 (nm)

Deposit Depth, 3 (nm)

N-,C). 0.04

Reciprocal Space, k (nm)

J. Electrochem. Soc., Vol. 145, NO. 8, August 1998 O The Electrochemical Society, Inc. 2831

imaginary values representing the composition profile in reciprocal space. For the original data set XF,(6), the FFT was computed using

where x,,(k) represents the mole fraction of iron as a func- tion of CMA reciprocal space (k), A, is the real part of xFe(k), Bk is the imaginary part of x,,(k), and i 2 = -1. The magnitude of each discrete peak in reciprocal space, IxFe(k)i, was calculated using the real and imaginary parts of the transformed data

Figure 4B shows Ix,(k)l as a function of location in rec- iprocal space. Peaks occur at k = n/A, where A is the spa- tial period of the CMA and n is an integer 21. The peak at k = l/A is the fundamental peak of the waveform and is used to determine the average wavelength and character- istic amplitude of the CMA. In Fig. 4B, 1/A = 0.015 nm-', or A = 67 nm, which agrees with the estimated average wavelength determined from Fig. 4A. The characteristic composition modulation amplitude is read directly from Fig. 4B as the magnitude of the fundamental peak at n = 1. In this example, the magnitude is 0.08 (corresponding to a modulation amplitude of 8 mol %), which represents the characteristic amplitude for all modulations in the CMA. This value also agrees with that determined from Fig. 4A.

While both modulation wavelength, A, and modulation amplitude, AX,,, are readily determined in either real space or reciprocal space, subtle traits of the composition- al waves are more apparent in the reciprocal space repre- sentation. The magnitude and location in reciprocal space of each peak relative to the fundamental peak provides detail about the compositional structure in the CMA and the role of nonlinear dynamic response to the sinusoidal disk oscillation. For example, when operated under small- amplitude, linear conditions, disk oscillations at a single frequency (eg., a sinusoidal waveform) produce a response (e.g., the periodic composition variation) containing only a single peak at k = l/A. Conversely, large-amplitude, sinu- soidal disk oscillations produce a nonlinear dynamic response characterized by the presence of harmonic satel- lites in the reciprocal space profile.14 For example, in Fig. 4B, the presence of satellite peaks located at k = n/A, where n 2 2, indicates that the large-amplitude sinusoidal disk oscillation (a = 0.95) results in a nonsinusoidal com- positional variation.

Relationship between processing and compositional structure.-Figure 5 illustrates the effect of varying the waveform oscillation frequency on the alloy composition profile. Shown are real and reciprocal space composition profiles for films plated with a disk rotation oscillation amplitude of a = 0.90 and oscillation frequencies of o = 40, 80, and 160 mHz. These figures show that when the disk rotation oscillation frequency is doubled, the average modulation wavelength is halved. The reduction in wave- length at higher frequencies is represented by shifts in the peak locations in reciprocal space. At higher frequencies, all peaks shift to larger values of reciprocal space, i.e., shorter wavelengths. For example, at 40 mHz, the funda- mental peak occurs at k = 1/A = 0.0075 nm-I, or A =

133 nm. At an oscillation frequency of 80 mHz, A = 67 nm, and at 160 mHz, A = 33 nm. In addition to the peak shifts, it is also seen that as the oscillation frequency is increased, the composition modulation amplitude decreases. For ex- ample, the magnitude of the fundamental peak decreases from 0.12 at 40 mHz to 0.09 at 80 mHz to 0.03 at 160 mHz.

For a given frequency, a change in the oscillation ampli- tude affects the alloy composition profile as well. Figure 6 shows reciprocal space composition profiles from three films made at a frequency of 80 mHz and oscillation amplitudes of 0.70, 0.50, and 0.30. Again, characteristic modulation amplitudes and wavelengths can easily be de-

Deposit Depth, 6 (nm) Reciprocal Space, k (nm-I)

Fig. 5. Real and reciprocal s ce corn osition profiles for NiFe CMAs deposited in a SinwoidaRlow w i t oscillation amplitude a = 0.90 and oscillation frequencies of a = 40, 80, and 160 mHz. At higher flow oscillation frequencies, all peaks shift to higher val- ues of reciprocal space (shorter wavelengths), and the characteris- tic modulation amplitude at k = 1 /A decreases.

termined from the composition profile in reciprocal space. As the disk rotation oscillation amplitude is decreased, the composition modulation amplitude decreases as well. The magnitude of the fundamental peak and all satellite peaks is reduced at lower disk rotation oscillation amplitudes. The composition modulation amplitude decreases from 0.035 at dimensionless oscillation amplitude a = 0.70, to 0.005 at a = 0.30. Since the three films in Fig. 6 were plat- ed with the same oscillating flow frequency, no peak shifts are evident in reciprocal space and the fundamental peak falls at k = 0.015 nm-' for each film, corresponding to an average CMA wavelength of 67 nm. In addition to affect- ing the magnitude of the fundamental and satellite peaks in reciprocal space,the flow oscillation amplitude, a , also has a significant impact on the distribution, or relative magnitude, of the satellite peaks. As shown in Fig. 6, the reciprocal space spectrum becomes dominated by the fun- damental peak at k = l/A as the flow oscillation amplitude decreases. The declining magnitude of the satellite peaks relative to the fundamental peak at low-flow oscillation amplitudes (most clearly seen at n = 2 or k = 2/A) illus- trates the expected nonlinear response of the oscillating RDE system.14

The relationships between processing and compositional structure in NiFe flow-induced CMAs are summarized in Fig. 7 and 8. Figure 7A is a plot of the measured average CMA wavelength as a function of the flow oscillation fre- quency. The expected inverse relationship between oscilla- tion frequency and CMA wavelength is evident, indicating that the PSV-FFT method provides a quantitative picture of the characteristic modulation veriod. Plotted are data points from each stripping voltammetry experiment at even oscillation amplitude and freauencv (excevt a =

4 m ~ z ) and the best fit line through tce daia. The ia ta set includes CMAs with a wide range of wavelengths, from 17 to 540 nm. At all amplitudes and frequencies, the data agree well with the theoretical relationship

J. Electrochem. Soc., Vol. 145, NO. 8, August 1998 0 The Electrochemical Society, Inc.

Reciprocal Space, k (nm-l) Fig. 6. Reciprocal space composition rofiles for NiFe CMAs

deposited in a sinusoidal flow with oscil /' ation frequency of a = 80 mHz and oscillation amplitudes of a = 0.70, 0.50, and 0.30. The characteristic modulati& amplitude at k = 1 /A d&reases as the flow oscillation amplitude is reduced.

h

where V,,, is the mean molar volume of the alloy, r) is the plating current efficiency, j is the plating current density, and F is Faraday's constant."

Figure 7B shows a log-log plot of the normalized com- position modulation amplitude as a function of the flow oscillation frequency. The data points represent results from all experimental runs in which a discernible modula- tion amplitude existed. The wide spread in the data for any given frequency suggests that using PSV-FFT pro-

Fig. 7. (A) Measured average CMA wavelength as a function of disk rotation oscillation frequency. Data from each PSV experi- ment at every flow oscillation am litude and fre uency (exce t o = 4 mHz) is included abng with best fit line %rough the Jata points. The data obe the expected inverse relationship between CMA wavelength andflow oscillation f requen~ &iven by ~ q . 4. (B) Normalized CMA amplitude as a function of 1s rotation oscilla- tion frequency for a = (A) 0.10, (W) 0.30, ( 0 ) 0.50, (A) 0.70, (0) 0.90, and (0) 0.95. The solid line is the frequency response expect- ed for electrodeposition of an ideal binary alloy with no interac- tions between the codepositing species.

vides, at best, a semiquantitative measure of the modula- tion amplitude. The data indicate the presence of a quasi- steady composition modulation plateau at very low oscil- lation frequencies and a sharp decline in the flow-induced composition modulation amplitude as the frequency is increased above 20 mHz. The solid line in Fig. 7B repre- sents the theoretical frequency response for a mass- trans- fer-limited species with Sc = 2300.15 (The kinematic vis- cosity of the electrolyte and diffusivity of Fe" are estimated from lab measurements to be 0.011 cm2/s and 4.7 X cm2/s, respectively.) This curve illustrates the expected behavior of an ideal binary alloy with one species mass-transfer-limited and the other kinetically limited with no interactions between the two codepositing species. The ideal theory predicts deviation from the quasi-steady regime to occur at oscillation frequencies on the order of 1 Hz due to concentration boundary-layer relaxation effects which are normally slower than kinetic time scales.'*-'"ifferences between the theoretical curve and the experimental data suggest that the deposition of NiFe does not behave as an ideal binary pair and illustrate

J. Electrochem. Soc., Vol. 145, No. 8, August 1998 The Electrochemical Society, Inc. 2833

aFig. 8. Magnitude of the nth peak as a function of the disk rota-

tion oscillation amplitude for CMAs made at a frequency if =40 mHz. Best fit lines are drawn through the data for the tunda-mental peak (s) and the k = 2/A rn) and 3/A (A) satellites.Dashed reference lines indicate the expected slope for each peakaccording to Eq. 5.

that the dynamics of NiFe electrodeposition are extremelyslow, even when compared to normal convective—diffusivetime scales. The very slow dynamics of NiFe electrodepo-sition have been noted but not investigated at frequenciesbelow 200 mHz.17 It is possible that the slow dynamics ofNiFe electrodeposition illustrated here may be responsiblefor the observation that CMA structures are seemingly notinduced in the ca. 1 Hz oscillating flow produced by a con-ventional industrial paddle cell.18

The sensitivity of NiFe deposition to electrolyte flowoscillation can be characterized by the magnitude of thefundamental and satellite peaks in the reciprocal spacecomposition profiles. If NiFe were an ideal binary paircharacterized by mass-transfer-limited Fe deposition andsimple kinetically controlled Ni deposition, one wouldfind that the peaks in the FFT composition profile grow inaccordance with a power law relationship

XFe = =

for all n � 1, where K is a constant for the nth peak.'4 Wesee that this relationship does not hold for NiFe depositionbut rather that the composition modulation amplitude ismore sensitive to the flow oscillation amplitude than pre-dicted by Eq. 5. Figure 8 shows the relationship betweenthe measured composition modulation magnitude and thedisk oscillation amplitude used during plating of filmsmade at 40 mHz. The data are plotted on a log—log plot toemphasize the expected power law relationship. The solidlines are best fit plots of the fundamental peak and the n =2 and n = 3 satellites. Dashed reference lines have beenadded above the experimental data to indicate the expect-ed slope for each peak according to Eq. 5. For all peaks, thebest fit lines exhibit a larger slope than predicted from the-ory. The slope of the fundamental peak is approximately2.5, and the slopes for the n = 2 and n = 3 satellites areroughly 3.5 and 4.5, respectively. The larger than expectedslope exhibited by all the peaks in reciprocal space reiter-ates the well-known fact that the deposition of NiFe isespecially sensitive to electrolyte agitation rate.19 This en-hanced dependence on electrolyte mixing suggests that

Conclusion

NiFe deposition exhibits highly nonlinear kinetics with re-spect to the concentration of surface species.

We have shown that optimized PSV, when combinedwith FFT analysis, is a reasonable technique for the com-positional characterization of electrodeposited NiFeCMAs. This method allows fast and reliable estimation ofCMA wavelength and amplitude and compares favorablyto other characterization techniques such as Auger andX-ray spectroscopy. In the characterization of flow-induced NiFe CMAs, PSV was used to illustrate the effectof different electrolyte flow field conditions on the compo-sitional structure of the alloy. The PSV-FFT technique wasalso used to probe the dynamic time scales of NiFe elec-trodeposition and to illustrate the enhanced sensitivity ofNiFe deposition to oscillating electrolyte flows. Results

08 1indicate that critical deposition chemistries occur overtime scales much slower then those governing typicalmass-transport processes. Due to the enhanced sensitivityto low-frequency flow oscillations, electroplating flow-induced NiFe CMAs with novel structural characteristicsmay be possible. We intend to study this aspect of NiFeelectrodeposition in the future.

AcknowledgmentsSupport for this research was provided by the National

Science Foundation through grant no. CTS-9457097.S. D. L. thanks Intel Corporation for support through anIntel Foundation Graduate Fellowship.

Manuscript submitted October 20, 1997; revised manu-script received April 27, 1998.

The University of Washington assisted in meeting thepublication costs of this article.

REFERENCES1. Y. Nagai, M. Senda, and T. Toshima, J. Appl. Phys., 63,

1136 (1988).2. K. D. Bird and M. Schlesinger, J. Electrochem. Soc.,

142, L65 (1995).3. S. Menezes and D. P. Anderson, J. Electrochem. Soc.,

137, 440 (1990).4. D. M. Tench and J. T. White, J. Elect rochem. Soc., 138,

3757 (1991).5. J. Yahalom and 0. Zadok, J. Mater. Sci., 22, 499 (1987).6. U. Cohen, F B. Koch, and R. Sard, 3. Electrochem. Soc.,

130 1987 (1983).[5] 7. D. S. Lashmore and M. P. Dariel, 3. Electrochem. Soc.,

135, 1218 (1988).8. D. Baral, J. B. Ketterson, and J. E. Hilliard, J. Appi.

Phys., 57, 1076 (1985).9. W Tang, C. Gerhards, J. Heise, and H. Zabel, J. Appl.

Phys., 80, 2327 (1996).10. B. J. Thaler, J. B. Ketterson, and J. E. Hilliard, Phys.

Rev. Lett., 41, 336 (1978).11. 5. D. Leith and D. T. Schwartz, J. Electrochem.

Soc.,143, 873 (1996).12. J. I. Goldstein, D. E. Newbury, P. Echlin, D. C. Joy,

A. D. Romig, C. E. Lyman, C. Fiori, and E. Lifshin,Scanning Electron Microscopy and X-Ray Micro-analysis, 2nd ed., Plenum, New York (1992).

13. D. T. Schwartz, J. Electrochem. Soc., 136, 53C (1989).14. D. T. Schwartz, J. Elect rochem. Soc., 140, 452 (1993).15. B. Tribollet and J. Newman, J. Electrochem. Soc., 130,

2016 (1983).16. D. T Schwartz, P Stroeve, and B. G. Higgins, AIChE J.

35, 1315 (1989).17. B. C. Baker and A. C. West, This Journal, 144 164

(1997).18. G. M. Whitney, Personal communication (1996).19. P C. Andricacos, C. Arana, J. Tabib, J. Dukovic and

L. T. Romankiw, J. Electrochem. Soc., 136, 1336(1989).

10'

10°

101

l02

-----5-----Eq,(5) — —

S

n = 1

n = 2

n=3

0.2 0.3 0.4 0.5 0.6