Some Applications of Electrochemical Impedance Spectroscopy

Electrochemical impedance spectroscopy (EIS) has been widely used to characterize rates and mechanisms during electrochemical processes. In this study, EIS was employed to probe the behavior for simulated galvanic coupling of a carbon fiber/polymer matrix composite in the marine environment. In a later study the effect of cathodic 10n plating and ion implantation of 304 stainless steel on its corrosion behavior was examined. The electrochemical behavior of a carbon fiber/vinyl ester composite in 3.5% NaCl solution was investigated. Negative potentials were applied to the composite material to simulate galvanic coupling of metals. EIS enables time dependent data to be acquired non destructively from a single sample. The impedance of the material as a function of time and applied negative potential was measured. It was found that increased damage to the composite was induced by more applied negative potentials, as cathodic reactions were increased with decreasing potential. Experimental data from the measurement technique is also amenable to modelling by passive electrical circuit elements. The system was equivalently modelled by two interacting RCtype subcircuits representing the carbon fiber/moisture and vinyl ester/moisture interphases. The pore resistance Rpo determined from the model was found to offer a damage monitoring criteria for the composite material. In an examination of surface damage to composites, laboratory testings were found to simulate accurately long term surface damage from galvanic marine exposure m seawater. Surface examination after

plating and ion implantation of 304 stainless steel on its corrosion behavior was examined.
The electrochemical behavior of a carbon fiber/vinyl ester composite in 3.5% NaCl solution was investigated. Negative potentials were applied to the composite material to simulate galvanic coupling of metals. EIS enables time dependent data to be acquired non destructively from a single sample.
The impedance of the material as a function of time and applied negative potential was measured. It was found that increased damage to the composite was induced by more applied negative potentials, as cathodic reactions were increased with decreasing potential. Experimental data from the measurement technique is also amenable to modelling by passive electrical circuit elements. The system was equivalently modelled by two interacting RCtype subcircuits representing the carbon fiber/moisture and vinyl ester/moisture interphases. The pore resistance Rpo determined from the model was found to offer a damage monitoring criteria for the composite material.
In an examination of surface damage to composites, laboratory testings were found to simulate accurately long term surface damage from galvanic marine exposure m seawater. Surface examination after long term galvanic coupling in seawater indicated removal of the polymer matrix above carbon fibers in addition to previously found blisters. Therefore two types of damage, blistering and dissolution, occurred due to galvanic interactions seawater exposure. Blistering was found only 720 hours (30 days) at a potential of -0.65VscE, along with regions of polymer surface dissolution. Imposing a potential of -1.2V s CE resulted in exposing carbon fibers after the covering polymer layer was rapidly removed. Possible electrochemical mechanisms for the polymer dissolution process are discussed.
TiN and ZrN cathodic ion plated coatings were applied to a 304 stainless steel, and exposed to 0.5N NaCl solution. The TiN coatings were also ion implanted with Ti and Au to determine the effect of ion implantation on corrosion behavior. It was found that ion implantation did not enhance the corrosion resistance of the TiN on 304 stainless steel.
However the ZrN did protect the stainless steel from corrosion. It is suggested that the ZrN is inherently more protective by formation of a passive layer. TiN even with excess Ti ions from implantation did not form a protective passive layer. Corrosion interfaces of coated and/or implanted stainless steel were modelled by a simple parallel RC-type circuit due to strong interfacial bonding between substrate and coatings.
Modelled data indicated the charge transfer resistance of the ZrN was higher than TiN and was related to the enhanced protection of ZrN.

INTRODUCTION
Electrochemical impedance spectroscopy (EIS) has been proven to be a powerful non destructive technique in electrochemical studies especially for highly resistive materials such as painted metals where other electrochemical methods are inadequate (l-5). Several reviews (6,7) suggest two important areas of application, firstly rapid estimates of corrosion rates, and secondly identification of corrosion mechanisms, especially in the presence of an adsorbed film or an applied organic coating. Other investigators applied this technique to degradation of a porous electrode (8,9), corrosion behavior of inhomogeneous surfaces (10,11), characterization of passive films and localized corrosion (12)(13)(14)(15)(16). Electrochemical processes can also be examined by EIS in low conductivity media where meaningful de measurements are impossible if the ohmic drop is not eliminated (17,18).
Studies involving EIS to predict corros10n rates and behavior of coated and uncoated metals usually involve modelling of the experimental data to represent the electrochemical interface. The models consists of circuit design with passive elements such as resistors and capacitors. These passive circuits models are termed "equivalent circuit" models as they are analogs of the electrochemical nature of the corrosion process. Other circuit elements include for example the Warburg semi-infinite diffusion element. Various equivalent circuits for different electrode/electrolyte systems has been proposed as physical models from which corrosion rates are determined (8)(9)(10)(11)(12)(13)(14)(15)(16). However, 3 Theoretical Background of Impedance Spectroscopy When a purely sinusoidal voltage expressed as v(t) = V m sin (rot) where w = 27tf, is applied to an electrochemical cell, the resulting steady state current will be i(t) = Im sin (rot + 0) where e is the phase difference between the voltage and the current, measured in degrees. The phase difference is zero for resistive behavior, and 90° for capacitive behavior (see figure 1). Phase differences between o· and 90° are found for systems with both resistive and capacitive nature due to different properties of localized sites at the interface.
The impedance of the cell is defined as The imaginary and real impedance can be plotted on a complex Argand diagram as shown in figure 2.

5
The most commonly employed technique of impedance spectroscopy is the measurement of system impedance directly in the frequency domain. A single-frequency voltage is applied to the cell and the phase shift and impedance magnitude are measured, or alternatively the real and imaginary parts of the resulting current at that frequency are measured. When applied over a wide frequency range or bandwidth (ie. 1 mHz to 1 MHz), information on electrode processes involving a direct charge transfer · processes and/or diffusion mechanisms is revealed through modelling of the experimental data.

Simple Basis of Circuit Models
The corrosion interface in solution adjacent to the electrode surface consists of double layers of an inner layer of electrons and an outer layer of ions. Their inherent capacitive reactances are characterized by their relaxation time, defined as the time needed for any system to reach equilibrium upon application of a small amplitude of perturbation (ie. voltage step), or more realistically the distribution of their relaxation times due to inhomogeneous properties of local sites in the system(l 9).
The electrical responses of a cell of heterogeneous properties vary depending on the species of charge present, and the microstructure of the electrodes.  (23)(24). The Helmholtz model is used in this discussion.
In an activation controlled system, Rt is related to the corrosion current density, Icorr• by the Stem-Geary equation (25)  Other types of graphical representations are discussed in literature, such as the idealized Randles plot of Z' vs. ffi-1 /2 usually used for system of diffusion-controlled behavior, and plots of Z' vs. ffiZ" and Z' vs. Z"/ffi (26).
In real systems, complications of impedance behavior at lower frequency (0.1 -0.001 Hz) are often observed and explained as due to effect of concentration and diffusion of species across the interface (20).
The equivalent circuit for metal/solution system with this behavior is shown in figure 5 where the diffusion related element is in series with the resistance Rt. On a Nyquist plot this is observed as a straight line at low frequency usually mentioned as a diffusion tail, see figure 6. A special case of diffusion is the semi-infinite Warburg diffusion (19), with diffusion tail is at 45° to the real axis. Warburg diffusion on a Bode plot will give rise to a ffi -1 /2 behavior usually at low frequency limit, and is shown in the Bode plot in figure 6.
A very general equivalent circuit for an electrode/coating/electrolyte system is shown in figure 7 . The coating layer is represented by a parallel pore resistance, Rpo• and coating capacitance, Cc, and the metal/coating interface in general is represented by an interfacial impedance Zif proposed by Mansfeld, Kendig and coworkers (1,3). They proposed that the interfacial impedance Zif is equivalent to a parallel combination of Cdl and Rt for an activation controlled process, or series combination of Rt and Warburg diffusion element W for a diffusion controlled processes at the metal/coating interface. These are represented by circuits (a) and (b), respectively m figure 7. Walter (2,27 ,28)  interface. An equivalent circuit similar to that for a bare metal exposed to electrolyte (see figure 3) has been proposed for a perfect dielectric coating with no defects, and is referred to as a quasi-homogeneous 3-D model of metal/coating system (29).
A more complex equivalent circuit was proposed to described a coating with defects which is referred as inhomogeneous 3-D layer model (29). This was used to describe corrosion behavior in the presence of corrosion products and of pitting corrosion on aluminum alloys (9)(10)(11)14). The inhomogeneous 3-D model is shown in figure 9.
In this study, the two most important models which provided the basis of analysis are those shown in figures 3 and 7. More specific models were developed from these two to further explain mechanisms corresponding to the impedance behavior and corresponding physical changes observed.

Experimental Procedure
The general experimental arrangement for EIS testing 1s as shown in Figure 10.

Circuit Analysis of Experimental Data
Interpretation of impedance data can be carried out using vanous methods of data analysis (30)(31)(32). These are available as commercial software such as EQUIVCRT (33), and LEVM (34). In the analysis, equivalent circuit models of the system are proposed which usually consists of combinations of passive elements such as R, C, L, and diffusion related element such as W. Complex nonlinear least squares  figure 11 the center of the semicircle could be below the Z' axis. This behavior is suggested as due to inhomogeneities in the electrode surfaces (36), both laterally and within the oxide film (12).
The constant phase element can be represented in its admittance form as Y cpe( ro) = Y 0 (jro )n = Y 0 ron cos (nrt/2) + j Y 0 ron sin (nrt/2) where Y 0 is the adjustable parameter obtained from the least squares analysis, containing the diffusion coefficient (33 ). This is a very general dispersion formula, and reduced to a Warburg semi-infinite diffusion at n=0.5, given as Y w(ro) = Y 0 (jro)0.5 = y 0 [ (ro/2)0.5 + j (ro/2)0.5] For n=O it represents a resistance with R= Y 0 -1, for n=l it represents a capacitance with C= Y 0 , and for n=-1 it re pre sen ts an inductance with L= Y 0 -1.
A parallel combination of CPE and resistance R yields an arc m the Nyquist plot and the depression of the arc depends on the parameter n.
The total impedance of this parallel circuit is and the time constant 't = (Y 0 R)l/n. For a circuit which consists of parallel R and C, the time constant is given as 't = RC. Equating the time constants gives RC = (Y 0R)l/n 13 then, the capacitance of interface modelled by parallel R-Y cpe circuit is The constant phase element is found to best fit for an interface with deviation from ideal capacitive behavior, as shown in chapters 2 through 4.

Procedure for EIS Modelling
The modelling starts with an analysis o. f the experimental data where the dispersion of the impedance data will be decomposed into subcircuits which will lead to an indication of many possible subcircuits.
The tentative identification of these subcircuits will provide the starting values for the adjustable circuits. The selected sub-circuits will be optimized using partial NLLSF. The optimized sub-circuits will be subtracted from the total dispersion. This will cause distortion to the rest of the dispersion. Usually the distorted part is removed to ease further analysis. Subtraction will be carried out until no reasonable circuit elements are represented in the distorted dispersion file. Final optimization of the subtracted sub-circuits will be performed by total NLLSF. An example of experimental data the circuit modelled to the data and the prediction from the model is shown in figure 11.             Rn ,por = pore resistance due to the electrolyte penetration. Negative potentials were applied to the composite material to simulate galvanic coupling of metals to it. The impedance of the material as a function of time and applied negative potential was measured. The data was adequately modelled by passive circuit elements. It was found that increased damage to the composite was induced by more applied negative potentials, as cathodic reactions were increased with decreasing potential. At a potential of -1.2V (SCE) holes were found in the composite surface after 90 hours of exposure. The pore resistance determined from the model used to fit the impedance data was found to offer a damage monitoring criteria for the composite material.

INTRODUCTION
Non-corrosion related blistering of glass fiber/polymer matrix composites in the marine environment is well known, and is thought due to an osmotic process which is diffusion controlled [l]. Water molecules from the environment diffuse into the polymer matrix under the influence of a concentration gradient existing between the initial moisture content in the composite and its saturation value.
Combination of diffusing water molecules . with water soluble material in the polymer forms a new concentrated solution which resides in pores and in turn creates an osmotic pressure with moisture in the composite. This pressure draws more water into the pores, creating further increase in localized pressure. As the osmotic pressure exceeds the yield stress of the resin, blisters form. No electrochemical process is thought to be involved in blisters formation for glass fiber based/ polymer composites.
Previous studies on the corrosion behavior between a graphite/polymer composite and metals, observed that galvanic corrosion can also induce blisters [2]. The galvanic interaction between a carbon fiber vinyl ester composite and anodically active metals such as steel and aluminum 2014 rapidly initiated blisters [2,3] . Over the same time period of exposure, up to three months, no blistering was observed in the composite in the absence of a galvanic coupling with metal. The site of the blisters was over the location where a glass cross weave in the uni directional carbon fiber weave was closest to the exposed composite surface. A schematic diagram of the cross section of the composite indicating the fiber locations is shown in figure 1. The larger the separation on the galvanic series between the cathodic carbon composite and the active coupled metal, the more rapidly blisters form and more metallic corrosion observed.
One major difference between diffusion controlled blisters in glass fiber composites and the electrochemically induced blistering m carbon fibre hand-prepared composite was the pH of blister fluid. It The potentials chosen for this study simulated galvanic coupling of different anodically active metals, e.g. low alloy steels and aluminum alloys, to the composite and also when cathodic protection is applied to a system containing coupled metals and composites.
The variation in electrochemical behavior of the composite with time at fixed values of Eapp will be determined by measurements utilizing impedance spectroscopy. Many studies have been conducted to evaluate the corrosion behavior of organic coatings placed on aluminum and steel. Using the sophisticated technique of electrochemical impedance spectroscopy (EIS), models of corrosion behavior for coated aluminum and steel had been proposed by various authors [ 4,5]. Generally, coated steel and some aluminum alloys can be represented by Model 1 shown in Figure 2, and most coated aluminum by Model 2, also shown in Figure 2.
As an initial basis of modeling the equivalent circuit of a composite coupled with metal, the general model was used ( see Figure 2 ). A schematic diagram of the composite is shown in figure   1 for reference of the circuit elements to physical entities.The equivalent circuit consists of Rs, the solution resistance; Cc, the capacitance of the matrix polymer separating the conductive carbon fibers from the environment; Rpo• the pore resistance of the matrix polymer due to the penetration of electrolyte from the surface through to the conductive fibers ; and Zif, the general impedance behavior of the metal/coating interface. In this study, Zif represents the impedance behavior of the interphase region between the carbon fiber and moisture. Of particular interest was to determine whether a specific model can be used to generalize the composite coupled With metals. From a previously proposed blistering mechanism (2), 39 particular passive elements for the model were predicted. It was expected that the impedance spectrum will show diffusion behavior at low frequency. Therefore, model 1 is not expected to represent the experimental data in this case.
Finally, a measurable damage criteria characteristic representing composite damage was sought. Initial Rpo values for coatings was previously found not to show any correlation with their long-term behavior [5]. However, this parameter ,Rpo• varies with wave, but at 90° to each other. The total fiber volume fraction, both carbon and glass fibers, in the vinyl ester polymer matrix is 26%. The volume ratio of glass fibers to carbon fibers in the composite was 1 :4.
Prior to each experiment, the surface of specimens were degreased with acetone and rinsed with deionized water. This treatment was not detrimental to the composite.

Experimental Setup
The electrochemical cell is schematically shown in Figure 3 and consisted of a test specimen and spherical glass joint clamped together. The area of the working electrode exposed to the electrolyte (

Circuit Analysis I Equivalent Circuit
Two different models were proposed initially to simulate the impedance behavior of carbon fiber vinyl ester composite at applied potential, see Figure 4. Model A, similar to other models proposed previously and shown as model 1, figure 2, was found not to fit the experimental data. This was predicted initially due to the diffusion based degradation mechanism operating for the polymer matrix.
Model A is limited to an ideal capacitive behavior at both the vinyl ester/solution interface and at the carbon fiber/moisture interface.
In this case the model is not accurate because of the composite physical properties such as cracks and pores, and the diffusion of water and migration of ions through the defects inside the composite.
Better simulations of the experimental data were achieved using However, some ionic conductivity was observed in the Bode plots with increased exposure. A slow decrease from initial values at frequencies below 30 Hz was noted as exposure progressed, indicating ionic conductivity. This was believed due to the slow penetration of water through the polymer matrix. Some microscopic porosity is typically present in the polymer matrix in composites.
The diffusion of water is sluggish because the driving force to induce this action is solely due to the concentration difference between the initial moisture content in the composite matrix and its moisture saturation value . This is shown on the Bode plots as the deviation of the phase angle from w-1 behavior. Through visual and scanning electron microscopy observation, ·DO damage was observed on the exposed surface of the composite (7).  (2) were found in this study at the applied potential of -1.2 V for 90 hours.

Impedance
The high pH of the blisters found previously (2) suggests that the hydroxyl ion remaining after cathodic reaction involving decomposition of water to form hydrogen at carbon fiber sites reacted at the glass fibers location. Only metals with corrosion potentials below the water decomposition potential initiated blisters when coupled to the composite, indicating that the oxygen reduction cathodic reaction was not effective in blister formation. The hydroxyl ions may have permitted rapid blistering of the composite by osmotic processes near the glass fibers while at the same time an electrochemical cathodic process occured above the carbon fibers 46 located closest to the surface. The increased rate of moisture uptake when the composite was galvanically coupled in comparison to uncoupled data would support the blistering mechanism. In addition, examination of samples from galvanic coupling over exposure periods much longer up to six months did indeed show damage similar to that found in this study over the carbon fiber locations (7). The data suggest that the electrochemical damage of the polymer matrix is independent of sample geometry.
It is suspected that the degradation process of the vinyl ester matrix responsible for holes initiates in the interior of the sample at the carbon fiber/ vinyl ester interphase region and moves slowly outwards towards the electrolyte until the carbon fibers are exposed to the solution. The decreasing Rpo values followed by a constant low pore resistance in the -1. ester is added to the moisture differential driving force. In addition, the moisture differential driving force due to concentration gradients will remam large due to the constant removal of moisture in the cathodic reaction. As proposed, the decomposition of water in the vinyl ester to form hydrogen at the carbon fibers will not allow an increase in moisture content to saturation levels. An increase in moisture transport rate is expected as a function of more negative potentials increasing the cathodic reaction rate in the composite.
Evidence was previously found in support of electrochemical interaction increasing transport rate when galvanically coupled composite gained more weight than the uncoupled composite (8).
With increasing time of exposure at the different Eapp, the phase angle maxima at high frequency shifted to the lower frequencies. Previous studies indicated that polymer swelling due to water uptake was associated with maxima shifts to higher frequencies as capacitance increased (9). In a separate study of the cathodic delamination of polybutadiene coated steel (10) shifts of the 49 high frequency maximum to lower frequencies were suggested to result from an increase in the area of metal exposed to the electrolyte. For an epoxy coated steel under cathodic current control (11), phase angle maxima shifts to higher frequencies were associated with increased delamination of the coating . However, it was proposed that the effect of decreasing coating thickness was to produce a phase angle maxima shift to lower frequencies . The stability of the polymer coating separating the cathode from the environment appears to control the phase angle maxima shift. A stable coating will maintain its thickness but diffusion through it will decrease film and charge transfer resistance as the substrate becomes active. Phase angle maxima shifts to higher frequencies will result in this case as found in other studies (11). When the coating is electrochemically degraded and slowly removed, its thickness is effectively decreased. The phase angle maxima then shifts to lower frequencies which supports the calculated thickness effects reported previously (11 ). The phase angle maxima shift in this study appears to be associated with degradation of the matrix material. The same degradation i.e. polymer dissolution, was also responsible for the low pore resistance values found after 50 hours of exposure.
For the -0.65 V Eapp, the values of pore resistance dropped at a decreasing rate during the period of the experiment. Surface damage was found on the sample when examined by scanning electron microscopy in the form of surface separations in the vinyl ester matrix over the carbon fibers (7). No blisters were found for the 90 hours of exposure but were found after 30 days of exposure.
As pore resistance appears to characterize polymer degradation, the combination of decreasing pore resistance and surface damage would indicate that the decomposition reaction for the vinyl ester initiates at the carbon fiber interface and propagates to the external surface.
As the cathodic reaction favoured is decomposition of water to form hydrogen production either intermediate hydrogen ions or hydrogen atoms or alternatively hydroxyl ion byproduct of the reaction would appear responsible for the decomposition of the composite. Further work is currently being conducted on these degradation mechanisms.

CONCLUSIONS
I. Impedance spectroscopy is applicable to studies of the degradation of carbon fiber composites by electrochemical processes.       In a laboratory simulation of the process, after 90 hours exposure in 3.5% NaCl with the composite at a potential of -0.65V scE regions of polymer surface dissolution were found. After 720 hours blistering was found accompanying dissolution. Imposing a potential of -1.2V s CE resulted in exposing carbon fibers after the covering polymer layer was rapidly removed.
Laboratory testing techniques were found to simulate accurately long term surface damage from galvanic marine exposure in seawater.
Electrochemical mechanisms for the polymer dissolution process are discussed.

INTRODUCTION
Carbon fiber reinforced polymer matrix materials are light weight, high specific modulus materials aimed at replacement of heavier metallic materials. One possible application is in the marine environment as replacements for aluminum alloys and steels. The composite will likely not exist in isolation, but be combined with metals. In these situations, galvanic couples form between the metal and . conductive carbon fibers in the composite.
For a galvanically coupled vinyl ester carbon fiber composite blisters formed in the polymer where an undulating glass fiber cross weave m the carbon fiber repeatedly approached the exposed surface (1). The pH of the blisters was measured at 10 to 11. The proposed mechanism for blistering involved the cathodic reduction of dissolved oxygen to form hydroxyl ions on carbon fibers. These ions then reacted with either the glass or components in the polymer to form an osmotic cell.
In a later study investigating a method to measure composite degradation due to electrochemical reactions (2), after 90 hours exposure at -0.65 V SCE no blistering was found. The non destructive electrochemical impedance spectroscopy (EIS) technique was employed to study damage to the composite. Nonlinear least squares fitting of the experimental data was conducted using the equivalent circuit software EQUIVCRT by Boukamp (3). Several possible equivalent circuit models were used to simulate these data including those proposed by Mansfeld and Kendig ( 4)  Severe degradation of the composite was found after 90 hours when a potential of -1.2 V scE was applied to the composite material. The polymer matrix was locally removed in the region directly above carbon fibers which left them exposed to electrolyte, figures 5 and 6. After 30 hours, black debris was floating in the solution, accompanied by a rapid increase in the current flowing in the sample. The debris was carbon fibers released from the surface after the vinyl ester resin was completely removed. In the region surrounding exposed carbon fibers, figure 5, the damage appears as thin channels running parallel to the carbon fiber direction and is similar to the damage previously found on galvanically tested samples. Rotating disc electrode studies in electrolytes similar to the one employed in this study ( 11) indicated that at potentials -0.2 to -0.5V s c E formation of hydroxyl ions and hydrogen peroxide is favoured as shown in reaction 1. [1] At more negative potentials in the range of -0. 7 to -1.2V scE a second reaction involving reduction of hydrogen peroxide to hydroxyl ions occurs, shown in reaction 2. [2] For both these cathodic reactions the rate limiting step is mass transport of oxygen to the cathodic surface.
Other possible reactions include reduction of protons, but at an extremely low rate for the pH ranges of 6 to 7 for this study. Oxygen reduction is occurring at a much faster rate than proton reduction. A second reaction to form hydrogen is available at potentials more negative than -I .2V scE and is shown as reaction 3.

[3]
This reaction is not mass transport limited and only depends on the supply of water and electrons. Hydroxyl ions are also produced by this reaction.
The reduction of oxygen reaction 1s mass transport limited and at slow rotational velocity, constant current density was observed over the potential range from -0.65 tol.2VscE (11).
Under the static solution conditions in this study, increased polymer dissolution rates accompanied increasingly negative imposed potentials. Production of hydrogen is a reaction that varies considerably with potential in static solution. At -0.650V scE, a negligible reaction rate from proton reduction reaction is available and would normally be overshadowed by the oxygen reduction reaction. The reaction rate is also mass transport limited. At -1.2V scE, the production of hydrogen increases rapidly by reaction 3 and is not mass transport limited.
The reactivity of the polymer as a function of potential and to the species available should indicate the cathodic reaction · for dissolution.
Hydroxyl ions do not dissolve the polymer, indicated by the stability of the polymer surrounding blisters containing fluid of pH 10 (2). Hydrogen peroxide was shown to dissolve epoxy matrix for an epoxy graphite composite (5) and as such may have contributed to the limited amount of damage at potentials of -0.65 VscE imposed potential. At the -1.2 VscE potential hydroxyl ion production is favoured by reaction 2. However the polymer dissolution rate, initially low and decreasing with time at -0.65V scE was rapid at -l.2V scE. This rate follows closely the hydrogen production over the potential range used in this study. At -1.2 VscE, hydrogen is favoured at the potential by reaction 3 and may be present in low amounts by proton reduction at the -0.65V scE potential.
A useful analogy may be that for hydrogen embrittlement of high strength steels. The fracture toughness under cathodically controlled conditions is a function of imposed potential in sodium chloride solutions, with lower fracture toughness values at more negative potentials over the range -0.65 to -l.2VscE (6). The accepted reason is that hydrogen is produced by cathodic reactions and the high strength steels are very sensitive to small amounts of hydrogen. The combination results in markedly decreased fracture toughness values with increasingly negative potentials. One possibility to explain the dissolution of the polymer is that they too are very susceptible to attack by atomic hydrogen as well as hydrogen peroxide.

Blister Formation
Although hydrogen is proposed as contributing to polymer dissolution, the reduction of oxygen reaction will also be proceeding, but at a constant rate over the potential range of this study. The hydroxyl ion 78 from the reaction is free within the composite after cathodic reaction on the carbon fibers. The pH of blisters was reported to be 10 to 11. The elevated pH results from hydroxyl ion diffusion in the vinyl ester to a location near the glass fiber where it can initiate a blister by initiating attack on the glass or combining with glass wetting agents. It should be noted that blisters can form by moisture diffusion and combination with water soluble components from the polymer (7). Blisters formed by this mechanism are usually acidic with pH of 3-4.· However long term exposure of the vinyl ester composite without coupling indicated that it was highly resistant to this mechanism of blister formation (1).
Blister formation from osmotic pressure is dependent on the chemical , diffusional and mechanical properties (7 ,8,9,10) of the matrix material. The stress within a blister from osmotic pressure will continuously increase while the necessary chemical components are available. The osmotic pressure overcomes the elastic stiffness of the material to move the matrix and increase the surface height which then indicates a blister. One upper limit of the process is when the pressure increases to a point where the blister bursts and the osmotic pressure 1s relieved. The second is when hydroxyl ion production ceases when oxygen in solution is depleted.
It is suggested that the blistering reaction initiates simultaneously with hydrogen production in the composite. An incubation period is required for the pressure to increase sufficiently to raise the surface of         TiN and ZrN coatings were applied. The TiN coatings were also ion implanted with N, Ti and Au to determine the effect of ion implantation on corrosive behavior. To quantify corrosion behavior, both potentiodynamic scans and electrochemical · impedance spectroscopy was employed. It was found that ion implantation did not enhance the corrosion resistance of the TiN on either the 52100 steel or the 304 stainless. However the ZrN did protect the stainless steel from corrosion.
It is suggested that the ZrN is inherently more protective by formation of a passive layer than the TiN even with excess Ti ions from implantation.

INTRODUCTION
Titanium nitride is well established as a coating to improve wear resistance of steels(l). It can be applied by several techniques, including reactive sputtering (2), and cathodic ion plating (3). At present cathodic ion plating is the technique of choice as rapid deposition rates in comparison to sputtering are available and it is desirable to deposit thick coatings for wear resistance. In addition the coating is amenable to substitution for gold due to its color similarity. To further enhance its service applications, it would be desirable for the TiN coating to provide improved corrosion resistance for the substrate it is deposited upon.
When ARMCO iron was coated with TiN by sputtering the substrates were electrochemically active in the test environment of sulfuric acid and therefore relied on the coating for all its corrosion resistance, its corrosion resistance was questioned ( 4 ). Additional protection mechanisms are required to improve the corrosion resistance of the TiN layer when placed on steel. To this goal, an assumption was made that it is the grain boundary structure of TiN that permits diffusion to the surface of the steel that is the mechanism responsible for corrosion. A schematic diagram of this mechanism is shown in Figure 1. Defects in the coating may also play a part, but the generally poor protection offered by TiN using several different coating techniques with different defect level suggested that the inherent structure of the material may highly influence the corrosion process. One technique which modifies structures is ion implantation (5). The objective of this research was to investigate the effect on corrosion behavior of ion implantation of TiN, applied as a coating on steel, with a second element in an attempt to modify its structure and enhance corrosion resistance. It is the results of this study 93 that will be reported in this paper. The aim of implanting several different elements was to determine the mechanisms of protection either by compressive stress increase in the film from 10n implantation or by enhancing the passive layer from the film of TiN on both a 52100 steel and on stainless steel. As an auxiliary part of the study to determine the influence of different coatings, zirconium nitride (ZrN) was also deposited on stainless steel by cathodic 10n plating to determine its inherent corrosion resistance in comparison to the TiN.

Sample Preparation
Substrates used in this study were AISI 304 stainless steel and 52100 steel. Their composition is shown in table 1. Discs 1.6 cm diameter and 0.3 cm thick for 52100 steel substrates were employed while 304 steel substrates were 3.2 cm diameter and 0.2 cm thick. Precleaning was carried out by degreasing with non freon degrease (NFD), rinsing with tap water, spraying with de-ionized water, and drying. The titanium nitride coatings were applied using cathodic arc plasma deposition technique to both plain and stainless substrates while the zirconium nitride was applied only to the stainless substrate. Each coating was 5 .0 µm in thickness. Nitrogen, titanium, and gold were implanted into the substrates up to a concentration of lxl016 ions.cm-2 at an accelerating voltage of 80 ke V. 52100 steels were tested in the following conditions: (1) bare surface, (2)  h. Electrochemical Impedance Spectroscopy.

95
Electrochemical impedance spectroscopy was performed on the S2100 steel and the AISI 304 stainless steel samples. The experimental setup for the impedance measurement is represented in Figure 2. The electrolyte solution was 0.5N NaCl solution and exposure time varied from 30 to 60 days for stainless steels, and from 8 to 10 days for 52100 steels. The exposed area to the electrolyte for stainless steel was 3 .25 cm2, and for 52100 steel was 0.95 cm2. Impedance measurements were conducted at open circuit potential and over a frequency range between 0.01 Hz and 100 kHz. A data density of seven frequency points per decade was used. Impedance spectra were represented in both complex impedance diagram ( or Nyquist plot), and Bode-phase angle plots. Data analyses were performed using a non-linear least squares fit method to obtain the equivalent electrical model for different substrate/electrolyte interfaces. Optimized parameters of the passive elements in the electrical circuit were plotted versus time.

Surface Examination
Scanning electron microscopy was employed to examine the initial surface condition of samples to detail surface characteristics prior to corrosion. A post corrosion testing examination was also conducted to identify the specific mechanism of corrosion for all the differing surface  Figure 14. Implantation of gold into titanium nitride coated stainless steel did not improve the resistance to the pitting attack.
The initial impedance value at 0.01 Hz is lower than that for the bare stainless steel. Pits were observed on the surface after 2 days of exposure to the chloride solution. Correspondingly, the phase angle at the lower frequency region and the impedance at the lowest frequency decreased over that time of exposure. The phase angle and impedance increased again after 24 hours, and dropped again to a relatively lower values three days later. This shifting was observed until the 14th day of the test period (shown in Table 2), and thereafter the impedance remained at 7xl04 ohm and the peak of maxima at the lower frequency region show a rest with the phase angle value at 70 degree.

Modelling of Impedance Data
It was found that a perfect dielectric model similar to that proposed for a quasi-homogeneous coating (6) (7). In the present study the iron component in each material is corroding. It is proposed that only the amount of surface area exposed that varies either by a passive film which suffers local breakdown or a TiN surface which permits corrosion of the substrate or from incomplete protection by implantation.
Q Figure 15. Model for bare steel substrates and samples coated and/or 10n implanted.
In figs 8 through 12 the experimental data is plotted along with the simulated data for the Bode representation of the circuit described above and employing appropriate values for the circuit elements. The double layer capacitance was represented by a general diffusion related element, Q, which is defined as a constant phase element. This element is mathematically written in its admittance form as Y*(ro) yo ( jro )Il ( 1 ) where Yo is the adjustable parameter used in the non-linear least squares fitting, and n is defined as the phenomenological coefficient which could be obtained from the slope of IZI on the Bode plot (8). A pure capacitance behavior is represented by n equals 1.0 (and the capacitance could be obtained by interpolated the impedance line of slope -1 to the IZI axis), however throughout this study n was found to be 0.90. This is understood as a deviation from an ideal dielectric behavior, or a leaky capacitor. This deviation is suggested due to the heterogeneity of the surface both laterally and within the depth of the oxide film which reflects the properties of the double layer (9). Therefore, the constant phase element should be a better representation of the double layer at the substrate/electrolyte interface.
The simple model was found to give the best fitting of the experimental data, for both bare and modified surface of 52100 steel and 304 stainless steel, and in both cases where uniform corrosion of the former and pitting on the latter were observed.
The optimized charge transfer resistance values were obtained from the least squares analysis (10)  It appears that film formation during exposure increased the charge transfer resistance for both the plain steel and the ZrN coated stainless steel. However the film on the plain steel was non protective, but the film on the ZrN was protective. Therefore although impedance spectroscopy will indicate film formation, it will not indicate which is protective unless a specific value of impedance for a protective film is assumed. Below lxl06 ohms.cm2, corrosion of the bare and modified surfaces of both steels were observed; same value below which corrosion under paint occurs (12). Other samples showed a decrease in charge transfer resistance or a stable charge transfer resistance below the assumed protection value. These films were not protective. In the case of the TiN coating layer was broken due to the chemical attack of the chloride ions. Having no free titanium in the coating did not allow it to repair the defects on the surface film and excess titanium did not appear to help. It is suggested that these circular defects acted as the preferential sites for pit initiation on TiN at which the chloride ions were incorporated into the film. This prevented further formation of surface film on the stainless steel surface, and allowed the pit initiation. The ZrN would appear to be naturally passivating in the particular environment as no pits were found and the charge transfer resistance suggested a passive film formed. Another possible mechanism is the ability of the ZrN to repair the film breakdown due to a high residual compressive stress reported for this coating (13). However it is thought unlikely compressive stress was a factor. Ion implantation would be expected to increase the compressive stress in the both types of coated layers but little variation in charge transfer resistance indicating increased corrosion resistance was associated with implantation of individual elements. There is possibility that compressive stresses due to TiN and ion implantations exceed the critical breakdown stress of the passive film, in which will produce breakdown of the film (14). The inherently better ZrN corrosion resistance over TiN would appear to dominate coating behavior.
Implanting a more noble metal on TiN coating decreased the Ecorr potential, and did not provide a higher resistance to the chloride attack.
The TiN coated stainless steel, and that further implanted with Au have equal charge transfer resistance to that of a bare stainless steel. Gold does not have the ability to induce passivation in the pH 6 and potential regions employed in this study (15). It could not repair the defects on the oxide film which allowed the chloride attack to take place. The implanted Ti on the TiN coated stainless steel did not protect the substrate from chloride attack.
Increased in the charge transfer resistance after the formation of pits could be due to the inhibiting effect of the corrosion products. This was found to be indicated by the increased phase angle at the lower frequency region (below 100 mHz), ie. the rest of the phase angle on the Bode plot after 14 days for the TiN coated stainless steel further implanted with Ti .

1.
Nitrogen gold and titanium implanted into 52100 steel, and into ion plated 304 stainless steel did not enhance protection from corrosion in 0.5N NaCL 2.
The corrosion mechanism was changed by implantation from uniform corrosion for plain steel to localized pitd.ng by the ion implantation. 3 TiN coatings before and after ion implantation of nitrogen, gold and titanium did not provide corrosion protection for 52100 steel or 304 stainless steel. The corrosion mechanism was local pit formation which exposed the steel substrate.

4.
ZrN did provide protection for 304 stainless steel m 0.5 N NaCl by formation of a passive film.