DETERMINING CONDUCTIVITY AND THICKNESS OF CONTINUOUSLY VARYING LAYERS ON METALS USING EDDY CURRENTS Erol Uzal, John C. Moulder, Sreeparna Mitra and James H. Rose Center for NDE Iowa State University Ames, Iowa 50011 INTRODUCTION Modifications to metal surfaces are important for many products; they can improve the interaction of the product with its environment, while retaining the structural properties of the bulk metal. Surface modifications provide properties such as good electrical contact as well as resistance to wear, corrosion and high temperatures. Consequently, it is desirable to develop nondestructive methods for characterizing near-surface properties, such as the electrical conductivity and magnetic permeability. In this paper we present an eddy current method to determine the structure of continuously changing surface layers Recently, several groups [1-7] have studied the use of eddy-current testing to characterize samples produced by coating an otherwise uniform plate of metal with a single metal layer (e.g. cladding or a metallic paint). Accurate estimates of the thickness and conductivity of the layer were obtained from measurements of the impedance as a function of the temporal or spatial frequency of the probe. These estimates depend on the ability to accurately model the coil's impedance as a function of the conductivity and permeability of the layer and base material. The work of D. H. S. Cheng [8], of Dodd and Deeds [9] and of C. C. Cheng, Dodd and Deeds [10] provides the relevant analytical models for plate geometries. These authors give simple closed-form formulas for the impedance of an air-core coil over a layered metal plate that has discontinuous piece-wise constant changes in the conductivity and magnetic permeability. Much less is known about the eddy-current impedance if the conductivity and permeability vary smoothly in the near-surface region. In this paper, we present an inversion method for characterizing samples that have smoothly varying near-surface conductivity profiles. Such profiles might be produced, for example, by case hardening, heat treatment, ion bombardment or by chemical processing. The structure of this paper is as follows. First, we review an analytic solution for the impedance of an air-core eddy-current probe over a layered metal plate whose conductivity varies as a hyperbolic tangent. Second we report measurements of the impedance of a variety of layered samples as a function of frequency. Third, we show that the impedance measurements can be inverted to determine the variation of the conductivity with depth. Finally, the paper is concluded with a brief summary. Review of Progress in Quantitative Nondestructive Evaluation, Vol. 12 Edited by D.O. Thompson and D.E. Chimenti, Plenum Press, New York. 1993 251
FORWARD PROBLEM Figure 1 shows the geometry of the problem. Consider a cylindrical, n-turn air-core coil next to a metallic half-space (z < 0). The coil's axis is perpendicular to the half-space's surface. The magnetic permeability is assumed to be everywhere that of free-space, Ilo. The conductivity o(z) is assumed to assumed to be zero outside the metal (z > 0), to depend only on the depth, z, and to become constant for sufficiently large (negative) z. We will consider conductivity profiles that can be parameterized in terms of a constant plus a hyperbolic tangent (1) This conductivity profile exhibits a smooth, monotonic change of 0. The parameter a controls the steepness of change. The hyperbolic tangent profile was chosen because an analytic solution of the problem is possible, and because it can represent a fairly large class of monotonic, smoothly varying profiles (see Figs. 2,4-6). There are four parameters in Eq. (1). 02 is the conductivity of the substrate, 01 is related to the surface conductivity, z = - c is the inflection point in the profile and a measures the degree of grading of the profile. We will assume that the substrate conductivity 02 is known. The forward problem is to determine the impedance of the coil from the given coil geometry and material data. For a conductivity profile in the form of Eq.(l) this problem was solved by the authors [11]. The experimentally determined quantity is the difference in impedance for two measurements: (1) the layered half-space and (2) a half space of the base material (no layers). We subtract the impedance for case 1 from case 2 and report the difference, /{Z. The subtraction reduces errors due to imperfect modeling of the coil, and facilitates comparison to experiment. The impedance difference for an n-tum coil is (2) z Fig. 1. Geometry of an n-turn air-core coil over a half-space. Conductivity of the halfspace is given by (1). 252
where (3a) G;:: F(Il+v, ll+v+1, 211+1; Yo) (3b) (3c) (4a) (4b) (4c) (4d) (4e) and Yo;:: 1/( l+e--i:/a>. (4f) Finally, F denotes the hypergeometric function. The equation for the impedance, (2), can be numerically evaluated in a quick and straightforward fashion. EXPERIMENT All impedance measurements were taken with an HP 4194A impedance analxzer, which is capable of measuring complex impedances at frequencies between 102 and 1 ()O Hz. For the measurements reported here, we confined our measurements to 399 points lying between I khz and I MHz. The coil and its associated cable (10 cm long) were connected to the impedance analyzer and the coil was mounted in a fixture over the specimen to permit placing the coil on the surface in a reproducible manner. Measurements of the coil impedance were obtained both on the layered material, Zl and on a part of the substrate not covered by the layer, ~. The difference of the two impedances,!:j.z;:: ~ - Zl' was recorded at each frequency. 253
The construction of samples was one of the major difficulties addressed in this work:. The basic problem relates directly to the purpose of this paper. Up to the present time, no good non-destructive method has existed for determining the variations in the near-surface conductivity and permeability of a metal. For example, measurements were made on a titanium plate that had been heated in air to create a case-hardened surface region ("alphacase"). However, we were unable to nondestructively measure the conductivity as a function of depth for this type of sample, and consequently cannot use this type of sample to stringently test the theoretical models. A second series of samples were created by stacking metallic foils (typically 20 foils with a thickness of 25 f.ull each) to create a piece-wise continuous approximation to a continuously varying conductivity profile. In this way, we were able to obtain precise information on the conductivity as a function of depth at the cost of giving up the smoothly varying nature of the profile. However, for the frequencies that we are using, the penetration depth (wavelength) of the critically damped eddy-currents is much greater than the 25 f.ull thickness of the individual foils. Consequently, the discrete nature of the foils will not be resolved in the impedance measurement. By stacking a sequence of Cu, Ti and other foils on a Cu substrate, we can simulate a system whose conductivity gradually goes from that of Cu (at the substrate) to that of Ti (at the top of the layered structure) as illustrated in Fig. 2. Various surface profiles were modeled by combining thin foils of copper, aluminum, zinc, nickel, molybdenum and titanium. The substrate material was made of either copper, titanium or aluminum 7005. Typically 20 thin foils were used to approximate a continuous proftle 0.5 mm thick. All the measurements were carried out by placing the stack of foils in contact with a given substrate and the probe then placed upon the foil under a small spring load. Measurements of!lz were found to be sensitive to small variations in lift-off between measurements on and off the layers and spring loading on the probe helped to achieve reproducible results. Since eddy currents flow parallel to the surface, we expected no effects due to lack of bonding between the various metallic layers. This assumption has been verified by comparing otherwise identical bonded and unbonded samples [I]. The averaged value of several identical measurements on each layer sequence was used for inversion. For the stacking sequence in Fig.2 the impedance was calculated in two ways: (1) using the solution of Cheng et al., which is valid for an arbitrary number of discrete layers, and (2) using the tanh solution for the approximate continuous profile. A comparison of these two calculations and the measured impedance is shown in Fig.3. INVERSION AND RESULTS The inversion method that we used is probably the simplest one possible. Namely, we used Eq.(2) to compute llz for a variety of layer parameters (al,e,a). We then found that set of parameters for which the theory curve was as close as possible to the experimental data. The least squares norm was our measure of closeness. Explicitly, we defined a cost function N Q = L ( IllZ theoiy I - IllZ exp I )2 (5) i=l Here, the sum is over a set ofn frequencies (typically N = 20). Q was minimized by using a simplex direct-search procedure. Figures 4, 5 and 6 show three examples of inversion results. Figure 4 shows a conductivity profile that increases rapidly from the surface, whereas in Fig. 5 the conductivity increases deeper inside the material. Figure 6 shows a profile that increases gradually over a large distance. In all three cases the inversion (continuous curve) gives a good approximation to the average change in conductivity. 254
:g ti) 4 '-"?;> :g f 2. r--j /11 0-0.0 I- 1 0.2 0.4 0.6 Depth (mm) 0.8 1.0 Fig. 2. Example of a piece-wise continuous conductivity profile constructed with metal foils and a hyperbolic tangent approximation to it.. 0.5 -r----------------, 0.4 i :, 0.3 8 10.2. 0.1 --0-- -6-- tanh Chen~etal. expenment 0.0... --... ---,---... --... ---1 o 100 200 Frequency (khz) Fig. 3. Impedance difference for the example in Fig. 2. Experiment, numerical computation for the layered profile, and theoretical calculation using the tanh profile. 255
6 : ~ 4.q fi u 3 ::l "0!::: 0 u 5 V 2 V 0 II 0.0 0.2 0.4 0.6 0.8 1.0 Depth (mm) Fig. 4. Conductivity profile of a layered sample that varies rapidly near the surface and the result of the inversion (smooth curve). 256
6 5 i en '-' 4 ~ :~ () 3.g 6 u 2 0.0 0.2 0.4 0.6 0.8 1.0 Depth (mm) Fig. 5. Conductivity profile of the layered sample that varies relatively rapidly further inside the solid and the result of the inversion (smooth curve). 6 5 i en 4 '-'. ~.s: E 3.g 8 2 V vy o 0.0 l- 0.2 I ~ II,... - /~ 0.4 0.6 Depth (mm) 0.8 1.0 Fig. 6. Conductivity profile of layered sample that varies relatively gradually and the result of the inversion (smooth curve). SUMMARY We have presented an inversion method for layers of metals with a smoothly varying conductivity profile. Assuming that the conductivity profile is a monotonic curve, the inversion gives its important parameters. We have tested the inversion method on model conductivity profiles constructed by stacking a large number of thin foils of differing conductivities. In summary, we have demonstrated that the surface conductivity, approximate thickness and the degree of grading of continuously changing conductive layers can be determined from frequency dependent eddy current measurements. 257
ACKNOWLEDGEMENT This work was supported by the Center for Nondestructive Evaluation at Iowa State University. The authors gratefully acknowledge the careful measurements contributed by U. Hafeez. REFERENCES 1. 1. C. Moulder, E. Uzal and 1. H. Rose, "Thickness and conductivity of layers from eddy current measurements," Review of Scientific Instruments., vol. 63, No.6, pp. 3455-3465, 1992. 2. S. 1. Norton, A. H. Kahn and M. L. Mester, "Reconstructing electrical conductivity profiles from variable-frequency eddy current measurements," Research in Nondestructive Evaluation, vol. 1, pp. 167-179, 1989. 3. S. 1. Norton and 1. R. Bowler, "Theory of eddy current inversion," Journal of Applied Physics in press. 4. 1. R. Bowler and S. 1. Norton, "Eddy current inversion for layered conductors," Research in Nondestructive Evaluation, in press. 5. N. J. Goldfine, "Magnetometers for improved materials characterization in aerospace applications," submitted to Materials Evaluation. 6. S. M. Nair and 1. H. Rose, "Reconstruction of three dimensional conductivity variations from eddy current (electromagnetic induction) data," Inverse Problems, vol. 6, No.6, pp. 1007-1030, 1990. 7. S. M. Nair and J. H. Rose, "Exact recovery of the DC electrical conductivity of a layered solid," Inverse Problems, vol. 7, No.1, pp. L31-L36, 1991. 8. D. H. S. Cheng, "The reflected impedance of a circular coil in the proximity of a semiinfmite medium," IEEE Transactions on Instrumentation and Measurement, vol. 14, pp. 107-116, 1965. 9. C. V. Dodd and W. E. Deeds, "Analytical solutions to eddy-current probe-coil problems," Journal of Applied Physics, vol. 39, pp. 2829-2838, 1968. 10. C. C. Cheng, C. V. Dodd and W. E. Deeds, "General analysis of probe coils near stratified conductors," International Journal of Nondestructive Testing, vol. 3, pp. 109-130, 1971. 11. E. Uzal, I. C. Moulder, I. H. Rose and S. Mitra, "Impedance of coils over layered metals with continuously variable conductivity and permeability: theory and experiment," unpublished. 258