Aqueous Corrosion Study of Melt-Spun NdFeB Ribbons with TiC Additions

ABSTRACT

INTRODUCTION

NdFeB alloys have assumed an important position due to their outstanding magnetic properties with energy products, (BH)_max, as high as 40 MGOe,^[1] leading to a substantial technical and economic impact on the permanent magnet industry. NdFeB magnets have found wide use in automobiles, personal computers, and some other commercial products. Processing of NdFeB magnets is usually accomplished by either conventional powder metal/sinter process, or rapid solidification (melt spinning) which is the focus of this paper. Melt-spun ribbons are pulverized and mixed with a polymer to produce "bonded" magnets. NdFeB magnets typically exhibit poor corrosion resistance to humid environments which severely limits their application.^[2-10] The poor corrosion resistance of NdFeB is linked to the microstructure which is characterized by the presence of multiple phases ^[11] including a phase rich in Nd, an active metal with a standard electrochemical potential of E°= -2.4 V. It has also been shown that both the Nd-rich phase as well as the Nd₂Fe₁₄B matrix phase absorb hydrogen in humid environments leading to decrepitation.^[12-18] Studies aimed at improving corrosion resistance of the NdFeB magnets have mostly emphasized either protective coatings or the addition of alloying elements. However, the effectiveness of coatings is not always assured and alloying usually reduces magnetic properties.

The production of NdFeB magnets made from rapid solidification circumvents many of the difficulties associated with powder metallurgy and allows a wider selection of ordinarily less soluble alloying elements. The microstructure of the rapidly solidified NdFeB is characterized by a nanocrystalline grain size.^[19,20] TEM investigation ^[21] has revealed an inhomogeneous grain size distribution from the free to the wheel surface side of the ribbon. It has also been reported ^[22] that melt-spun ribbons have lower overall rare earth content, i.e. less Nd-rich constituent than sintered NdFeB magnets. These features and the presence of a polymeric phase suggest a difference in corrosion resistance compared to conventional sintered magnets.

Recently, Branagan and McCallum ^[23-25] have proposed the addition of TiC to both melt-spun and gas atomized NdFeB in an effort to enhance processability without degrading magnetic properties. TiC refines the grain size of the hard magnetic phase by controlling nucleation, enhances glass forming ability, and reduces the cooling rate necessary to achieve optimum magnetic properties. Although TiC provides an interesting approach, its effect on the corrosion behavior of NdFeB alloy requires further study and is the objective of this investigation.

EXPERIMENTAL

Materials Preparation

A number of NdFeB materials were used in this study. Table 1 shows a listing of these materials, their method of preparation and the form in which they were tested in this investigation. Eutectic Nd₈₀Fe₂₀, NdFeB single crystal, TiC, and pure iron were also tested for comparison. The Magnequenchä samples consisted of Magnequench A (MQA) with an approximate composition of RE=30.6 wt% (RE= Nd and Dy), Fe=68.58% and B=0.82% alloyed with TiC. Powders for bonded magnets were made by milling melt-spun ribbons under low vacuum, yielding powders with particle size below 53 mm. Single crystals were prepared by a flux growth method. Some ribbons were used directly as electrodes for corrosion testing in order to observe possible differences in corrosion behavior between the free side and the wheel side of the ribbons.

Table 1--Samples used in electrochemical testing.
Sample	Alloy	Description	Wheel speed	Form tested
1	2-14-1	Nd2Fe14B	10 m/s	bonded
2	2-14-1	Nd2Fe14B	20 m/s	bonded
3	2-14-1	Nd2Fe14B	30 m/s	bonded
4	2-14-1+TiC(6at%)	(Nd2Fe14B)94+(TiC)6	12 m/s	bonded
5	MQA+TiC(6at%)	MQA+TiC	12 m/s	bonded
6	MQA+TiC(6at%)	MQA+TiC	20 m/s	bonded
7	2-14-1	Nd2Fe14B	21.25 m/s	ribbon
8	2-14-1+TiC(2at%)	(Nd2Fe14B)98+(TiC)2	16.25 m/s	ribbon
9	2-14-1+TiC(6at%)	(Nd2Fe14B)94+(TiC)6	10 m/s	ribbon
10	2-14-1	Nd2Fe14B	-	sintered
11	2-14-1+Co-V	Nd2Fe14B+5Co 3V	-	sintered
12	Nd-Fe	Nd80Fe14B	-	eutectic
13	2-14-1	Nd2Fe14B	-	crystal
15	TiC	TiC (99.5 wt%)	-	pellet
16	Fe	Fe (99.99 wt%)	-	plate

Electrode Preparation

Two types of electrodes were prepared for the electrochemical experiments, ribbon electrodes and bonded magnet electrodes. Ribbon electrodes were made by mounting small pieces in low viscosity epoxy resin and carefully measuring the exposed area. Before mounting, electrical contact was made to the back surface of the specimen by attaching a thin copper wire using silver paste. The surface of the mounted electrode was then very lightly polished (1200 grit SiC paper), ultrasonically cleaned in distilled water, then dried under a flowing stream of air. Bonded magnet electrodes were prepared from milled ribbon (< 53 mm dia.) mixed with 5 wt% epoxy powder and pressed into pellets (1 cm dia. and approx. 0.5 cm thick). The bonded magnet pellets were then mounted in low viscosity epoxy resin, and polished sequentially on 240, 320, 400, and 600 grit SiC papers, cleaned in distilled water in an ultrasonic cleaner and then dried by high pressure air. In order to determine accurate current densities (A/cm²), the active surface area of exposed NdFeB was determined by image analysis.

Electrochemical Testing Procedures

Figure 1 - Schematic diagram showing the experimental arrangement used in this study.

Three different electrochemical techniques were utilized in this investigation: potentiodynamic polarization, anodic dissolution at constant potential, and hydrogen charging/discharging experiments. All experiments were performed using an EG&G Model 273 Potentiostat. Potentiodynamic polarization and hydrogen charging tests were preceded by a determination of corrosion potential, E_corr, at open circuit for one hour. The corrosion current, i_corr, was determined by the polarization resistance method. The glass electrochemical cell consisted of a standard three electrode arrangement, a NdFeB working electrode, a saturated calomel (SCE) reference electrode, and a platinum mesh counter electrode, Figure 1. Potentiodynamic polarization curves were obtained in deaerated, stirred 0.9M Na₂SO₄ at room temperature at two different values of pH (4.4 and 6.5) using a voltage sweep rate of 20 mV/sec.

Constant potential experiments were performed by holding the applied potential at -600 mV vs. SCE for times of either 40 or 120 min. The rate of iron dissolution as a function of time was determined by atomic absorption spectrophotometry. Most samples were tested in 0.9M Na₂SO₄ at pH 4.4, but a few were tested at pH 6.5.

In the charging/discharging experiments the material was alternately charged (hydrogenated) by a cathodic current and discharged (dehydrogenated) by an anodic current. These experiments were carried out in room temperature 6M KOH by (a) applying a constant cathodic charging current (either 1 or 5 mA) for 30 minutes, (b) measuring E_corr for 5 min, and (c) applying a constant anodic current (1 or 5 mA) thus discharging hydrogen until a potential plateau was obtained corresponding to complete hydrogen removal. In order to examine the influence of hydrogen capacity on corrosion behavior, several experiments were performed where step (c) was omitted and a potentiodynamic polarization experiment was performed instead.

RESULTS AND DISCUSSION

Figure 2 - Typical anodic polarization curve for NdFeB alloys in 0.9 M Na2SO4 solution.

Potentiodynamic Experiments

Ribbon electrodes were prepared such that either the wheel side surface or the free surface could be exposed to solution.

Ribbons containing three different levels of TiC (0, 2, and 6 at. %) were tested. A schematic of a typical polarization curve obtained during the potentiodynamic experiments is presented in Figure 2. As observed, it shows four different regions: (1) cathodic region from 50 mV below the corrosion potential to Ecorr , (2) active region from ~ -0.85 to ~ -0.3 Volts, (3) maximum current region from ~ -0.3 to ~ 1.5 Volts, and (4) second activation region above ~ 1.5 Volts. Between the maximum current and the second activation region the decrease in current is probably due to formation of corrosion products. The polarization data for the set of experiments with ribbons is shown in Figure 3a for the wheel side samples and in Figure 3b for the free surface samples. From this data the corrosion potential, E_corr, and the corrosion current (i.e. corrosion rate), i_corr, were calculated using the polarization resistance method. The results obtained from data in Figures 3a and 3b are given in Table 2.

Figure 3 - Potentiodynamic anodic polarization curves (region near Ecorr) for the of NdFeB with and without TiC additions. (a) Wheel side, and (b) free side.

The most significant item of information in Table 2 is the noticeable difference in i_corr between the free side and the wheel side with the free side samples exhibiting larger corrosion rates. Furthermore, the difference in i_corr between wheel and free surfaces increases in the order 6 at% TiC < 2 at%TiC < no TiC. The addition of TiC minimizes the difference in corrosion resistance between the sides of the ribbon. Results also suggest that because of its finer microstructure, the wheel side surface has better corrosion resistance than the free side characterized by the presence of coarser grains. It is apparent that for both the free side and the wheel side, the addition of TiC causes an increase in E_corr and a decrease in i_corr. Increasing additions of TiC result in lower values of i_corr, which again is an indication of improved corrosion resistance.

The potentiodynamic polarization data for bonded magnet electrodes are shown in Figure 4 for NdFeB samples without TiC and in Figure 5 for samples with TiC (6 at%) additions. Calculated E_corr and i_corr values are given in Table 3. The effect of pH is expected, i.e., a more aggressive electrolyte (lower pH) results in significantly higher corrosion rates, i_corr. Wheel speed also has an effect on corrosion rate. For instance, i_corr decreases significantly for the first three 2-14-1 samples in Table 3, as wheel speed increases from 10 to 30 m/s. The addition of TiC to the bonded magnet electrodes causes an increase in E_corr and a decrease in i_corr for the same or similar wheel speed. For example, at a pH of 4.4 the i_corr value for the 2-14-1 with 6 at% TiC (12 m/s) is 105 mA/cm² whereas the 2-14-1 without TiC (10 m/s) is significantly higher 366 mA/cm². The same trend is repeated at pH 6.5. E_corr measurements also show that higher TiC additions result in more positive values of E_corr, an effect also observed for the ribbon electrodes.

Figure 4 - Anodic polarization curves for bonded NdFeB in 0.9M Na2SO4. (a) pH=4.4, and (b) pH=6.5.		Figure 5 - Anodic polarization curves for bonded NdFeB with TiC (6 at%) additions in 0.9M Na2SO4. (a) pH=4.4, and (b) pH=6.5.

Figure 6 - Prediction of the mixed potential theory for the improvement in corrosion resistance. Curve 1 represents the anodic curve for the alloy without TiC and curve 2, the anodic curve for the alloy with TiC.

The variation in E_corr and i_corr upon TiC additions, observed in Table 2 and Table 3, can be explained in terms of mixed potential

theory as shown in Figure 6. In any corrosion process, the corrosion current, i_corr, and the corrosion potential, E_corr, is given by the intersection of the anodic polarization curve, curve 1 for example, with the cathodic polarization curve. If the nature of the anode is changed, e.g. through a change in microstructure, such that the anodic polarization curve is shifted upward, as in curve 2, then a more noble corrosion potential, E'_corr, and a lower corrosion current, i'_corr, result. In all cases, experimental results shown in Table 2 and Table 3 are consistent with this mixed potential model. It is apparent from this analysis that the addition of TiC shifts the anodic polarization curves upward resulting in higher corrosion potentials and lower corrosion rates compared to samples without TiC.

Constant Potential Experiments

Constant potential experiments were performed on bonded magnet electrodes in order to determine the corrosion rate as measured by the rate of iron dissolution, Figures 7a and 7b. The rate of iron dissolution exhibits a roughly linear increase with time (Figure 7a) up to about 30-40 min., where the rate decreases as confirmed by experiments performed for 2 hours (Figure 7b). This decrease is probably due to formation of corrosion products on the surface of the alloy. These results suggest that the addition of TiC is particularly helpful in decreasing the rate of corrosion at potentials near E_corr.

Figure 7 - Dissolved iron as a function of time at -600 mV vs. SCE in 0.9M Na2SO4 pH=4.4 for melt-spun NdFeB alloys. (a) 40-minute tests, and (b) 2-hour tests.		Figure 8 - Schematic representation of hydrogen charging and discharging experiments.

Galvanostatic charge/discharge experiments

A schematic representation of the sequence of these experiments is shown in Figure 8. Corrosion potentials obtained before and after charging are compared in Table 4. It can be observed in this Table that the corrosion potentials for samples with TiC additions are at least 100 mV more noble than those for samples without TiC in 6M KOH. This difference indicates that the samples with TiC exhibit lower thermodynamic driving force for corrosion. The results presented in Table 4 also show a dramatic decrease in E_corr upon charging for all samples. This decrease in corrosion potential which ranges between 500-650 mV are indicative of hydrogen absorption and formation of a hydride.

Figure 9 - Potential vs. time for bonded NdFeB alloys during anodic discharge in 6M KOH. (a) 1 mA anodic discharging, and (b) 5 mA anodic discharging.

A measurement of the relative capacity of the material to absorb hydrogen can be obtained from the time required for anodic discharging. Complete removal or discharging of hydrogen corresponds to the time at which the potential reaches a plateau of about +400 mV as shown in Figure 9(a) and (b). The presence of a small plateau at about -900 mV is consistent with a two stage hydrogen desorption process for NdFeB alloys suggested by some authors, i.e. initial hydrogen absorption by the Nd-rich phase followed by absorption into the matrix Nd₂Fe₁₄B phase.^[26,27] The charge transferred, time of desorption, and measurement of the mass of the sample allows the calculation of hydrogen capacities in terms of mAh/g by the expression:

where D is the discharging current in mA, T is desorption time in h, and W is the weight of active material in g. The results calculated from equation 1 are presented in Table 5. It is clear that samples incorporate more hydrogen at 5 mA as evidenced in Table 5. Higher hydrogen absorption is observed for the samples without TiC additions as indicated by longer times of desorption and higher hydrogen capacities. These results show that TiC inhibits hydrogen uptake.

Immediately after cathodic charging, an anodic potentiodynamic polarization experiment was performed in order to determine the effect of hydrogen absorption on E_corr and i_corr. Results are shown in Table 6. The hydrogenated samples exhibit more negative corrosion potentials and higher corrosion currents compared to their counterparts with no hydrogenation. However, the change in corrosion resistance is not severe for the samples with TiC additions. This suggests that TiC additions mitigate the decrease in corrosion resistance due to hydrogenation. The relationship between hydrogenation and corrosion resistance could be explained by the formation of a Neodymium hydride, resulting in a volume expansion of the Nd-rich phase in grain boundaries and physical separation of individual Nd₂Fe₁₄B grains from the sample.

Corrosion potential comparison

Figure 10 - Ecorr for different samples in 0.9 M Na2SO4 pH=4.4.

Corrosion potential measurements are useful since they can provide information on the tendency of a material to corrode and to establish possible galvanic couples in a specific environment. Various E_corr measurements obtained in this investigation in 0.9 M Na₂SO₄ pH=4.4 are presented in Figure 10. Data in this Figure include for comparison special samples mentioned in Table 1. The results indicate the likehood of galvanic corrosion since there is a large difference between two phases in contact, i.e., the Nd-rich eutectic phase and the matrix represented by the single crystal. The difference of about 200 mV is sufficient to promote galvanic corrosion. In this case, the 2-14-1 matrix acts as a cathode, having the less negative potential, and the Nd-rich phase acts as the anode. Therefore, the Nd-rich intergranular phase will be preferentially attacked leading to intergranular corrosion. Furthermore, the quantity and exposed area of Nd-rich phase is much lower than the 2-14-1 producing an unfavorable area ratio between the anode (Nd-rich phase) and the cathode (2-14-1), thus accelerating corrosion.

Comparing the 2-14-1 ribbon with 2-14-1 bonded magnet, E_corr of the ribbon is more noble than the bonded magnets. Probably, the polymer binder decreases the corrosion resistance of the bonded magnet. Figure 10 also shows that E_corr for the optimum 2-14-1 (20 m/s) bonded magnet is more noble than that of the sintered 2-14-1 and therefore more corrosion resistant. If we compare the 2-14-1 ribbon with the 2-14-1 sintered magnet the difference is even larger. According to the E_corr data presented in Figure 10, the different 2-14-1 materials (with no alloying additions) can be listed in order of increasing corrosion resistance: ribbon > bonded > sintered.

CONCLUSIONS

The microstructure and corrosion behavior of NdFeB magnet alloys with and without TiC has been determined. All experiments for the bonded magnets show that the addition of TiC enhances the corrosion resistance as indicated by: (a) lower corrosion currents, (b) more noble corrosion potentials, (c) lower iron dissolution, (d) lower hydrogen capacity and (e) less increase of i_corr upon hydrogenation. The samples with TiC have _corrosion currents about a factor of 2-3 lower when compared to the samples without TiC. For ribbons, the effect of TiC is similar, as revealed by improved corrosion resistance in the potentiodynamic experiments. The potentiostatic experiments confirm that behavior as indicated by high rates of iron dissolution at constant potential. The deleterious effect of hydrogen on the corrosion resistance of NdFeB alloys was evidenced by the higher i_corr values compared to the non-hydrided samples. Two mechanisms possibly dominate the corrosion process in NdFeB alloys: galvanic corrosion between the different phases present in the microstructure and hydrogenation produced by the high affinity of the Nd-rich intergranular phase for hydrogen. The corrosion properties of the ribbons vary through the ribbon thickness. Higher corrosion rates were observed for the free side compared to the wheel side. However, upon addition of TiC, the ribbon becomes more homogeneous and this difference is minimized. The beneficial effect of TiC on the corrosion resistance of the NdFeB alloy can be attributed to refinement of grain size and enhancement of amorphous microstructures.

ACKNOWLEDGEMENT

The authors want to acknowledge Dr. R.W. McCallum and his research group at Iowa State University for providing the melt-spun materials used in this investigation.

References