THE USE OF NIOBIUM AS AN ANODE MATERIAL IN LIQUID FILLED ELECTROLYTIC CAPACITORS

Field crystallisation during the anodisation of niobium can be reduced to a minimum by employing a phosphoric acid electrolyte at low temperature. For certain other electrolyte systems, reduced crystalline growth rates are obtained by increasing the solute concentration or by adding either phosphate or ethylene glycol. Porous anodes were prepared from niobium powders of nominal particle size 8#m or 10m using sintering temperatures of 1400-1800C. The capacitance of these bodies was in the range 4,600-10,000 #F Vgm after anodising at 20C. Shelf tests on experimental capacitors filled with a concentrated sulphuric acid electrolyte indicated a maximum working temperature of 85C. Although chemical compatibility between the dielectric oxide and the sulphuric acid was maintained up to 125C, thermal effects on the oxide at this temperature resulted in unacceptably large increases in capacitance and leakage current. Endurance tests at 85C of capacitors rated at 35V/200gF or 63V/60F showed good stability of capacitance and dissipation factor over a period of 3000 hours. Final leakage currents were in the region of 10 #A#F -1V -1 at 85C and crystalline growth was greatly restricted.


INTRODUCTION
Niobium is in the same group (Vb) of the periodic table as tantalum but has roughly half the atomic weight. Both elements have very nearly the same atomic and ionic radii and their compounds are similar in properties. However, niobium has half the density of tantalum and since the price per kilogram of the two metals is about the same it presents an attractive substitute for tantalum in many applications, including electrolytic capacitors. Some of the relevant physical and chemical properties are summarised in Table 1. Although niobium oxide has a higher dielectric constant than tantalum oxide, a greater thickness of oxide per volt is formed when the former metal is anodised. In consequence, the capaci-tance values per unit area of oxide films on the two metals are similar for the same anodising voltage.
Niobium has two disadvantages with respect to tantalum in the context of capacitor applications. The first is the more marked tendency of anodic oxide films on niobium to undergo field crystallisation 1,2,3. For example, calculations by the present authors suggest that the radial growth rate of crystalline areas may be up to 1000 times greater on niobium than on tantalum, under the same conditions of anodisation. However, as indicated in earlier work3, the growth rate during anodisation can be much reduced by employing low temperatures. The second disadvantage concerns the greater sensitivity of anodised niobium to heat treatment. On heating at temperatures in excess of 125C (300C for tan-Atomic Metal weight talum) a semiconducting region is formed within the oxide4. This leads to substantial increases in capacitance and dissipation factor and a marked dependence of both of these parameters on voltage and temperature. In addition, there is evidence to suggest that in the case of niobium oxide some mechanical damage may also occur. The combined effects are so deleterious that it seems very unlikely that niobium could ever be used for solid electrolytic capacitors, the preparation of which requires the thermal decomposition of manganese nitrate. The present paper examines the possibility of using niobium for liquid filled capacitors. It was recognised that certain limitations in the voltage and temperature of operation would have to be imposed.
However, it was considered that this would be acceptable provided that the high reliability normally associated with tantalum capacitors was retained.

Foil
Initial anodisation studies were performed with a supply of 0.0025cm thick niobium foil of capacitor grade. This material was degreased in warm methyl ethyl ketone prior to anodisation. The anodising was carried out by applying a constant current of 2 mA cm -2, until the voltage reached a predetermined level at which it was held constant. The indicated times of anodisation given in this paper refer to the period at constant voltage.
As shown in earlier work with tantalum it is possible to obtain a guide to the radial growth rate of crystalline oxide by measuring the time to reach the first minimum value in the current-time characteristic at constant voltage. This value, which has been called tmin, increases as the growth rate decreases. Values of tmi n for niobium anodised in a number of different electrolytes are given in Table II. Unlike the results of similar experiments with tantalum, there was no apparent relationship between radial growth rate and the pH of the electrolyte. Of the electrolytes studied only phosphoric acid showed any marked ability to minimise the growth of crystalline oxide. Relative  Figure 1.
Increasing the concentration of the sulphuric acid electrolyte resulted in a substantial decrease in growth rate for concentrations of 20% or above (Table 111). This result is important, since the more concentrated solutions have physical properties which render them ideal for use as working electrolytes in liquid filled capacitors. Qualitatively similar results were observed after anodising in oxalic acid electrolytes of varying concentration.
In the light of the above results, experiments were performed to find out if the growth rate in other electrolytes could be reduced by making small additions of phosphoric acid (Table IV).
It was found that the phosphoric acid additions had a considerably greater effect in sulphuric than in oxalic acid. In a further experiment niobium foil was anodised at 50V for 360 min either in 40% sulphuric acid or the same electrolyte containing 1% phosphoric acid, the temperature of both solutions being held at 85C. The inhibiting effect of the phosphoric acid is illustrated in Figure 2. Phosphoric acid (b) X200 FIGURE 2 Degree of crystallisation after anodising niobium foil in H2SO electrolytes at 50V, 85C, 360 minutes: (a)40% H2SO4, (b)40% HSO4 + 1% H3PO additions were also found to be effective in reducing crystalline growth during anodisation in nitric acid.
The effect on the crystalline growth rate of adding ethylene glycol to an oxalic acid electrolyte is shown in Table V. For a 50% addition, the crystalline oxide covered less than 1% of the surface by comparison with approximately 100% coverage for the aqueous solution. It was also established that combined additions of glycol and phosphate to the oxalic acid solution had an additive effect.
On the basis of these experiments, it was concluded that phosphoric acid was a suitable anodising there was a considerable spread in particle size, in the range 2/am-25#m for powder A and 2#m-50/am for powder B Figure 3). The powders were coated with an organic binder and pressed into a cylindrical shape, 0.706 cm diameter and 1.188 cm long. The pressed bodies, which incorporated tantalum riser wires, had a density of 5.0 gm cm -3 After dewaxing, the bodies were sintered in vacuo at temperatures of 1400C 1600C and 1800C. The interior and exterior structures of bodies sintered at 1800C are shown in Figure 4 (powder A) and Figure 5(a)and (b) (powder B). The improved interior porosity of bodies prepared from powder A, a factor resulting from the lower average particle size, is apparent. Both exterior surfaces showed reduced porosity in consequence of 'smearing' on withdrawal from the die after pressing. The influence of reduced sintering temperature, illustrated in Figure 5(c) and (d), for powder B treated at 1400C was to increase porosity by reducing powder agglomeration. At the lower temperatures, the powder particles also showed less general smoothing and less rounding of the edges. Preliminary experiments quickly confirmed that low temperature anodisation would be essential if field crystallisation was to be minimised ( Figure 6).
In consequence, all the subsequent work was performed in 1% phosphoric acid at 20C. For anodising voltages up to 110V, a current density of 50 mA per anode ('-0.05 mA cm -z) was applied to 75V at which voltage the current was reduced to 20 mA per anode. For voltages above 110V, 50 mA per anode was applied to a voltage 50V below the forming voltage and the anodisation was then completed at  20 mA per anode. All the specimens were held at the forming voltage for 60 minutes.

Results
The influence of sintering temperature on capacitance and leakage current after anodising to l0V is shown in Table VI. Both parameters were measured at half the formation voltage in 40% sulphuric acid at 20C. The leakage current was recorded after 30 minutes polarisation and the capacitance immediately afterwards with a 50 Hz ripple voltage superimposed. The anodes showed an increase in capacitance of less than 2% in raising the applied bias voltage from 5V to 35V; in the same voltage range the dissipation factor decreased slightly.
As anticipated from the values of the particle size, powder A produced anodes with significantly greater surface areas than B. Reducing the sintering temperature from 1800 to 1400C caused the surface areas to increase by 50-60%. Not unexpectedly, this also resulted in higher leakage currents. The latter effect was much more marked for powder B, which was capable of producing anodes at 1600C and 1800C sintering with leakage currents comparable to tantalum. However, at 1400C sintering the difference in leakage current between anodes prepared from the two powders was much smaller.
The difference in purity between the anodes was reflected in the ability to increase the anodising voltage above 110V. Anodes from powder A prepared  Table VII and a selection of micrographs of the anodised bodies is displayed in Figure 7.
The measurements in Table VII Table VIII together with nominal ratings for the tubular types only. Leakage currents of the capacitors were measured at the rated working voltage and capacitance and dissipation factor at the same voltage with 50 Hz ripple applied.

Results
In order to determine the upper temperature of operation, the capacitors were initially subjected to shelf tests. So as to be able to differentiate between chemical and thermal effects the units under test were filled either with sulphuric acid or lithium chloride. The results for a period of storage of 100 hours at 125C are given in Table IX for capacitors from series 2 and 3. Both series showed substantial increases in capacitance and leakage current, the effect being more pronounced for series 2. However, the very similar changes for series 3 capacitors filled with the two different electrolytes suggested that the effects were related to thermal rather than chemical causes.
Further capacitors from series 3, having the button construction and filled with sulpuric acid, were tested on shelf at 85C for 4000 hours. Average capacitance increases were 6.4% and leakage currents, although increasing, remained below 0.3 x 10 -3pApF -1V -x at 20C. On the basis of   Table X. A similar test carried out on series 3 capacitors for 1000 hours showed an average capacitance change of 0.1%, stable dissipation factors and final leakage currents at 20C of 0.05x 10-3/.tA/.tF-1V -1 In general all the units tested displayed acceptable parametric stability, the general performance of capacitors prepared from powder B being superior in terms of leakage current and dissipation factor. However, field crystallisation during the endurance testing was minimal for anodes prepared from both powder supplies. Some additional electrical characteristics, namely leakage current and capacitance versus temperature, and capacitance and impedance versus frequency, are given in Figure 8 (1) series 2 (35V/135F), (2)  powders (nominally 4/.tm particle size) using a 90C anodisation temperature. However, in the present work the ratio of working voltage to forming voltage is about 0.6 of that conventionally used with tantalum and so a more realistic comparative value for niobium would be close to 5000 F V gm -1 In consequence, the price of these niobium anodes is similar to that of tantalum anodes prepared from the highest capacitance powders but for the same capacitance/voltage rating the niobium bodies are twice as large. Furthermore, the reduced purity of the 8/m niobium powder limits the anodisation of sintered bodies to about 120V and does not permit the preparation of capacitors capable of working much in excess of 35V. Anodes prepared from the 10/m powder sintered at 1800C can be safely anodised to 180V and operated at 63V. The decreased #F V gm -x value of these anodes is consistent with the larger particle size and higher sintering temperatures If powder of this purity can be obtained in a particle size approaching 4/.tm, it should be possible to produce anodes of the same physical dimensions as the tantalum bodies, having the same capacitance/ voltage rating, costing half as much and operating at least at 35V. A capacitor working at 50V might well be achievable.
In general, both the 35V and the 63V capacitors that have been subjected to endurance tests at 85C for 3000 hours have shown good stability of capacitance and dissipation factor. Final leakage currents at 85C were in the region of 10 -4 gA/.tF -Va value fully comparable with that of tantalum capacitors and field crystallisation was restrained to a minimum.

CONCLUSIONS
1) The radial growth rate of crystalline oxide during the anodisation of niobium can te reduced by employing a phosphoric acid electrolyte at low temperature.
2) Thermal effects limit the temperature of operation of niobium capacitors to 85C.
3) Capacitors rated at 35V/200/F and 63V/60/F have shown good stability on endurance tests at 85C. 4) Niobium powders of nominal 4/am particle size will be necessary before the anode cost can be reduced substantially below that of tantalum. at Towcester for preparing the niobium anodes, assembling the capacitors and for making some of the electrical measurements. Mr. G. Donachie is thanked for help with some of the experimental work and Mr. N. S. Griffin for the scanning electron micrographs. This work is published by permission of The Plessey Company Limited.