Carbide induced embrittlement on tungsten wires

Carbide induced embrittlement on tungsten wires

International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333 www.elsevier.com/locate/ijrmhm Carbide induced embrittlement on tungste...

202KB Sizes 5 Downloads 101 Views

International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333 www.elsevier.com/locate/ijrmhm

Carbide induced embrittlement on tungsten wires I. Gaal *, L. Uray Research Institute for Technical Physics and Materials Science, P.O. Box 49 Budapest 1525, Hungary Received 10 May 2002; accepted 17 June 2002

Abstract Heavily drawn and stress relieved non-sag tungsten wires are brittle at the ambient in torsion test in spite of their submicron sized fibre structure, when a part of their free surface is covered with tungsten carbide. The present study has shown that this kind of embrittlement may cease, if the stress relief anneal is performed at 1700 K for 15 min in an annealing atmosphere having an adjusted oxygen partial pressure of 104 mbar, while the samples remain brittle, when such annealing is carried out at an oxygen partial pressure of 106 mbar. Longer annealing periods at P ðO2 Þ ¼ 104 mbar exerts an adverse effect, as the strain at failure decreases gradually, when the annealing period is longer than 30 min at 1700 K. The oxygen induced decomposition of the carbide islands were monitored by means of their superconductivity. The carbon induced embrittlement upon stress relief of non-sag tungsten is also compared with the other types of embrittlement occurring in fibrous non-sag tungsten.  2002 Elsevier Science Ltd. All rights reserved. Keywords: Embrittlement; Non-sag tungsten; Decomposition of surface carbide; Effect of oxygen uptake

1. Introduction It is one of the salient features of the Coolidge technology that properly processed heavily drawn tungsten can be coiled onto a fine mandrel at the ambient. In addition, these coils remain ductile with respect to stretching upon their common stress relief anneal. This remarkable behaviour is closely connected with two features of the processing: (a) high drawing strains result in elongated fibrous grain structure with ultra-fine size [1,2] and (b) the fibre coarsening is very sluggish during stress relief, when a highly disperse phase suppresses grain coarsening [3]. In accordance with the general experience, one might, thus, expect that the ambient ductility of heavily drawn tungsten should persists upon annealing as long as the transversal mean linear intercept of the homogeneous fibre structure remains below 0.5 lm [1]. It has long been known that the technological exploitation of this kind of ductility is sometimes hindered by secondary effects. One may distinguish two sorts of them. The first group is closely connected to premature fracture that is initiated through localized heterogene-

*

Corresponding author. E-mail address: [email protected] (I. Gaal).

ities of the fibre structure [4]. The responsible microstructural features usually appear with a very low frequency and occupy only few, micrometer long parts of the wires (e.g. Fig. 1). The second group concerns localized splitting along some of the longitudinal grain boundaries upon deformation at the ambient. It is typical that the first extended splits do not yield to failure immediately; the samples fail only after the generation of a complicated system of spits at quite high strains [5,6]. This paper describes a spurious kind of embrittlement that is connected with the presence of carbide islands on the wire surface. Before the description of this type of embrittlement, it might be of interest to give a short description of the mentioned two sorts of secondary effects. It has been proven that the first group of secondary effects is closely connected to the presence of surface contamination. The main features of this effect are as follows: (a) Stress relieved coiled coils sometimes brake into two parts upon such a slight stretching that the plastic increase of the primary coil pitch is hardly detectable. This embrittlement has to evolve upon the heat treatment of the coils, since the wire material had previously withstood heavy deformation upon coiling at the ambient.

0263-4368/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 3 - 4 3 6 8 ( 0 2 ) 0 0 0 3 3 - 1

328

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333

Fig. 1. When a coiled coil fails upon slight stretching, the fracture paths often reveals a heterogeneous grain structure (upper SEM, the image width is 23.4 lm). The marked difference in the grain size in two suitable selected parts of the fracture path is visualized in the two lower micrographs. (The width of them amounts to 4 lm.) On the boundaries of the coarser grains bubble like features are revealed. (They are much larger than the upper limit of the size of a creep resistant bubble populations.)

(b) When every part of a broken coil is repeatedly restretched until the primary pitch angle becomes larger than 45, one usually gets only few (let say five) pieces of highly stretched coils parts. Therefore, one may conclude that this type of embrittlement concerns only few, very short parts of the coiled wire. (c) The fracture always starts in domains that underwent highly localized grain growth. (Contamination induced localized grain coarsening is often restricted to a fraction of the wire cross section (e.g.: Fig. 1).) (d) Model experiments has shown that this kind of grain growth may be closely connected with oxygen uptake [4,7]. (e) Similar embrittlement may be induced also through a great variety of metallic surface contamination [8,9] (Fe, Ni, etc.). (It is well established that in nonsag tungsten various solutes induce grain growth through bubble coarsening, if the impurity atoms diffuse into the bulk of the wire along grain boundaries [10,11].) One may conclude that the basic reason for the first group of the secondary effects is chemically induced grain coarsening [10,11], while the immediate cause of

the embrittlement is the locally coarse grain structure in the vicinity of the free surface (see Fig. 1). (In addition, it is evident that this type of embrittlement cannot be cured by any subsequent treatment, because the grain coarsening is evidently irreversible.) The second group of the secondary effects on the ductility of heavily drawn and stress relieved tungsten has been revealed by means of free-end torsion [4,5,12– 14]. Its salient features are as follows. (1) Wire samples in as drawn condition broke into two pieces only after quite extensive torsional strains, i.e. samples having a diameter of 170 lm and a length of 6 cm will brake after about 100 to 120 twisting turns. (Upon this deformation the torsional true strain at the wire surface, c, amounts to 0.9 to 1, where c is determined from the relation d c ¼ np : L Here d and L is the diameter and the length of the sample, respectively, while n denotes the number of the twisting turns.)

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333

(2) The extensive torsional deformation of as drawn samples was very heterogeneous along the wire length in a great variety of non-sag wires [5,6,12–14]. Upon torsion every type of the studied ‘‘ductile’’ as drawn samples split along the originally axial fibre boundaries with a quite marked frequency. The first splits already appeared at very low plastic strains, i.e. at c < 0:03. If the average surface strain was below c < 0:3, then the splits were constrained to a relatively small fraction of the sample length. In the sections with splits the true local surface strain, c amounted typically to 0.8, while the local strain on the non-split parts of the samples was less than 0.1. (3) At the ambient extensive split free twist could be achieved only by means of stress relief anneal [5,6,12– 14]. The majority of the studied wire types underwent split-free twist up to c ¼ 0:9 at the ambient, when the samples were pre-annealed at 1700 K for 15 min either in dry hydrogen [7], or in a vacuum chamber having a low adjusted oxygen partial pressure (P ðO2 Þ < 106 mbar) [5,6,12–14]. This paper was initiated through the experience according to which heavily drawn non-sag tungsten sometimes showed also an uncommon kind of severe embrittlement that does not fit into the frame of the described two types of embrittlement. The main features of this uncommon embrittlement are as follows (a) Certain pieces of some wire lots were very brittle in as drawn condition: they brake into parts after a torsional strain of c < 0:05 at the ambient. (In other words: embritteled samples having a diameter of 170 lm and a length of 6 cm will brake after six twisting turns.) (b) Such samples have strange features also after stress relief anneal (Table 1), because they show up excellent ductility only, when the oxygen partial pressure of the annealing atmosphere is not too low (i.e. P (O2 ) is about 104 mbar) and the annealing time is sufficiently short. In order to clarify the origin of this kind of embrittlement, one should bear in mind that oxygen uptake has generally an adverse effect on the ductility of tungsten and molybdenum [15,16], because oxygen segregation onto the grain boundaries cause severe grain boundary embrittlement. Therefore, one may suspect that the beneficial effect of oxygen upon stress relief is an indirect effect, that should be ascribed to the removal of Table 1 The number of twist turns at failure on two lots a non-sag tungsten after annealing at 1700 K at various adjusted P (O2 ) pressures Annealing time (min)

P (O2 ) mbar

Number of twist turns at failure

15 15 60 240

6  106 1  104 1  104 1  104

6 80 71 32

9 135 114 16

329

some sorts of carbide contamination. In order to substantiate this vision, one had to verify the presence of carbide islands on the (non-successfully) cleaned free tungsten surface, and one had to prove the close connection between the regained ductility and the removal of carbide by oxygen induced carbon vaporisation.

2. Experimental Since the fracture strain varied between broad limits from sample to sample in the lots to be studied, we prepared always four 70 mm long samples from a 280 mm long piece of wire. The diameter of the samples was 170 lm. The lubricant was removed by means of boiling NaOH solution, as usual. Thereafter, some samples were annealed in vacuum at 1700 K for 15 min by direct current heating. The annealing temperature was determined from the ratio of the hot to ambient resistance of the samples. The annealing was performed at an adjusted oxygen partial pressure of 104 or 107 mbar, respectively. (To this end the vacuum chamber was evacuated to a residual pressure of 108 mbar, and thereafter the oxygen pressure was adjusted by means of a micro-valve connected to an oxygen gas source.) After stress relief anneal the samples had a uniform fibrous grain structure. (The transversal diameter of the fibres was less than 0.5 lm, as it is usual at this wire size.) The as drawn and stress relieved samples were tested by torsion at the ambient in order to determine the failure strain. (The length of the samples between the grips amounted to 60 mm.) Although the presence of carbide contamination was tested in parallel unpublished studies by means of electron diffraction as well as by carbide sensitive etching, the present paper reports only on the monitoring of the surface carbide islands by making use of the superconductivity of tungsten carbides.

3. Effect of a superconducting surface layer on the resistivity In order to facilitate the discussion of the results, we determined also the resistive superconducting transition on the free surface of different tungsten wires that were coated with a carbide layer. The wire samples were cut from three different wire batches, the residual resistivity of which was appreciably different after an annealing at 1700 K for 1 h at P ðO2 Þ ¼ 106 mbar. (The resistivity of the coated and uncoated samples was virtually the same at 4.2, because the carbide layer was thin with respect to the wire diameter, and both the layer and the substrate were outside the superconductive region.)

330

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333

In order to grow the carbide layers, the cleaned samples were heated at 1700 K for 30 min in a vacuum chamber, the residual pressure of which was adjusted to 1 mbar through a C6 H6 source. (Before the application of the C6 H6 source, the residual pressure in the vacuum chamber was less than 106 mbar.) Since solid tungsten will be in equilibrium with W2 C at 1700 K according to the phase diagram of the W–C system, we may assume that the carbide layer consisted of W2 C, although the stoichiometry of the phase may change along the thickness of the layer [17–19]. The resistivity of the carbide coated samples (Fig. 2) was virtually independent of the temperature between 4 and 10 K, while the resistance ratio RðT Þ=Rð4:2 KÞ dropped to very low values ðRðT Þ=Rð4:2 KÞ < 103 Þ in a short temperature interval below 2.5 K, when the measuring current was less than 100 mA. (The details of the superconductive transition that takes place between 3 and 2 K are not shown in Fig. 2.) These results support the conclusion according to which the superconducting transition temperature of the studied layer was close to the tabulated value [17,19] of the transition temperature of W2 C. Let us emphasise that this result cannot be considered as a direct proof for the presence of W2 C, as the transition temperature of the refractory metal carbides is sensitive to the method of preparation [17–23]. (In this context let us note that the transition temperature of NbCx layers increases from 1 or 10 K, if x increases from 0.8 to 1. This experience may explain the discrepancy in the tabulated values of the transition temperature of the tungsten carbides [17–20].) The resistive superconducting transition of tungsten carbides (like that of the other type II superconductors) depends on the measuring current, on the substructure, as well as on the stoichiometry of the carbide [20–23]. It is well known that this current dependence becomes very marked, if non-continuous superconducting filaments are embedded into a normal conducting matrix [21,22]. This effect is closely connected with the fact that the

Fig. 2. Superconductive resistive transitions of the carbide layers grown on three different wire samples taken from three different batches.

superconducting current is concentrated into short superconducting filaments and consequently the magnetic field at the surface of the filaments will be high also at modest values of the total current. This experience suggests that the resistive transition will very markedly depend on the value of the measuring current, if superconducting islands (especially superconducting fibres) cover a normal conductor substrate (see Fig. 4).

4. Results It is a salient feature of the studied embrittlement that it is more or less spurious. This means that although generally low failure strains were observed on the as drawn samples of the lots selected for this study, a small fraction of the samples showed marked ductility. However, the brittleness persists, if the samples are stress relived at 1700 K for 15 min in a vacuum chamber with an adjusted P (O2 ) partial pressure of 106 mbar. Fig. 3 illustrates this feature by means of four simultaneously prepared samples taken from two different lots. (Let us emphasise that the failure strain was less than 0.1 on seven samples, while on one of the sample c amounted

Fig. 3. The strain at failure upon free end torsion on samples from two embrittlened lots of non-sag wires. The samples were annealed at 1700 K for 15 min at an adjusted oxygen partial pressure of 107 and 104 mbar, respectively.

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333

to 1 at failure.) In contrast to this, the failure strain of each stress relieved sample was higher than 1, when the annealing was performed at an adjusted oxygen partial pressure of 104 mbar (see Fig. 3). Since oxygen uptake can lead to severe grain boundary embrittlement in tungsten, it was of importance to determine also the effect of annealing time on the ductility. Table 1 shows that at a short stress relieve anneal (1700 K, for 15 min) the failure strain is high, when the annealing has been performed at an oxygen partial pressure of 104 mbar. However, the failure strain gradually decreased with increasing annealing time, when the oxygen partial pressure amounted to 104 mbar. The observed embrittlement was closely connected with the peculiarities of the surface regions of the samples, since the failure strain was always higher than 1, when a 10 or 20 lm thick surface layer was removed from the sample by means of electrolytic polishing. The temperature and current dependence of the resistive transition observed on the brittle samples (see Figs. 4 and 5) strongly supports the assumption that the embrittlement should be connected with tungsten carbide islands attached to the free surface. This assumption is supported through the following pieces of evidence. • The resistive transition in Fig. 4 takes place in the temperature region of the superconductive transition of W2 C (see Fig. 2). • The marked current dependence of the resistive transition proves [21,22] that the superconductor does not form a continuous layer (Figs. 4 and 5). • The resistive transition is absent on the ductile samples from which the a relatively thick surface layer was removed (Fig. 5).

Fig. 4. The unpolished samples (left) undergo resistive superconducting transition between 2 and 4 K, while this transition is absent on electropolished samples (right). (The samples were annealed at 1600 and 1700 K for 15 min at P ðO2 Þ ¼ 106 mbar.) The resistive transition depends markedly on the measuring current. (The value of the current in mA is shown on the left hand side of the curves.) When the annealing is performed at 1800 K, the resistive transition is absent on both kinds of samples.

331

Fig. 5. At 2 K the resistivity of the electropolished (ductile) samples is independent of the measuring current (curves N), while that of the brittle (unpolished) samples (curves S) depend markedly on the measuring current.

• The extent of the resistivity drop decreases with increasing annealing temperature, supporting the view that the superconductor-phase was gradually removed from the free surface through heterogeneous reactions like (Fig. 4). 2Wx C ðsolidÞ þ xO2 ðgasÞ ¼ CO ðgasÞ þ W ðsolidÞ

5. Discussion Scherrer [24] presented convincing pieces of evidence by means of AES and local chemical analysis that in certain lots the spurious local embrittlement of non-sag coils in the stretch test has to be ascribed to the presence of tungsten carbide along the fracture path. (He gave also good evidence for embrittlement that was closely connected to the presence of very fine Al2 O3 inclusions attached to the free surface of the non-sag tungsten wires.) The mechanism of this kind of embrittlement is well established [2], since a hard phase on the surface (in form of equiaxial particles, needles or layers) acts as local stress raiser. This effect is especially marked, if the phase is hard with respect to tungsten both in elastic and plastic sense of the word, as it is the case for W2 C (and WC). There are, however, also excellent evidences that support the view that spurious local embrittlement of the coils can be also closely connected with local, chemically induced grain growth driven by the uptake of various solute substances [10,11]. (The source of these substances is either the annealing atmosphere or a foreign layer attached to the free surface of the wire.) In case of the present lot this mechanism of embrittlement can be excluded, because the embrittlement ceases upon a certain type of annealing. Of course, the resistance of the grain boundaries against grain boundary fracture will depend on the

332

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333

actual chemical composition of the grain boundaries [15,16]. There are good pieces of evidence that segregated Fe [25] and C [10,11] increase the resistance of the grain boundaries against fracture, while oxygen segregation enhance the frequency of fracture along the grain boundaries [15]. Let us note that the present study actually revealed the competition between oxygen and carbon effects in the embrittlement. In this context two aspects are of importance. • The dissolution of surface carbides in tungsten should be too sluggish to enhance the fracture resistance of the grain boundaries, because surface carbides exert their adverse effects as long as they has not been removed from the free surface by gas–solid reactions. • The oxygen uptake of tungsten should be also quite sluggish at the temperatures of the stress relief, because oxygen uptake in a relevant period of annealing had not yield to embrittlement neither through grain boundary segregation nor through chemically induced grain coarsening at an oxygen partial pressure of 104 mbar. Of course, the development and composition of the surface carbides depend much on the metallurgical circumstances. One might consider two mechanisms for their formation. At first one can assume that certain traces of the graphite lubricant can be attached to the free surface, when the removal of the lubricant is not performed properly. These carbon traces react with tungsten upon the stress relief anneal and results in the growth of carbide islands, if the oxygen partial pressure of the annealing atmosphere is low enough. However, one may devise also an other mechanisms for the carbide formation, which should be active first of all on straight wires. At this mechanism, we have to bear in mind that the lubricant of the wire drawing consists of a tungsten oxide layer which is coated with graphite lubricant [26]. One may speculate that at any local failure the oxide layer will allow the interaction between metallic tungsten and graphite resulting in tungsten carbide grains or needles. (The simplest mechanisms of the oxide damage is definitely a scratch connected with the presence of a hard particle in the lubricant layer.) Similar effect is to be expected, if the thickness of the oxide layer is not uniform, because the oxygen content of this layer plays a salient role in the vaporisation of the graphite lubricant, if it is removed by a heat treatment in wet hydrogen.

6. Conclusions The surface carbide induced brittleness of ultra-fine grained tungsten can be cured upon stress relief annealing, when the annealing period is sufficiently short

and the oxygen partial pressure of the annealing atmosphere is not too low. This experience suggests that the dissociation of carbide into carbon dissolved in the grain boundaries (or in the bulk) is too sluggish to effect the brittleness, while the rate of the oxygen uptake along the grain boundaries should be also quite low, since oxygen segregation induced embrittlement appears merely after long annealing periods at the temperatures of stress relief.

Acknowledgement This work was supported by the Hungarian National Research Found (OTKA) contract no T 32730.

References [1] Seigle LL, Dickinson CD. Refractory metals and alloys II. New York: Interscience Publ; 1963. p. 65–111 (Metal Soc Conf, vol. 17). [2] Ashby MF, Gandhi CND, Taplin DMR. Acta Metall 1979;27:699. [3] Snow DB. In: Pink E, Bartha L, editors. The metallurgy of doped/ non-sag tungsten. London: Elsevier; 1989. p. 189–202. [4] Schilling W, Paschedag H. Techn Wiss Abh OSRAM Ges 1963;8:167. [5] Sz€ okefalvi-Nagy A. Scr Metall 1982;16:1009. [6] Uray L. Mater Sci Eng A 1989;112:89–92. [7] Neugebauer J. In: Rexer E, editor. Proceedings of Second International Symposium on Reinstoffe in Wissenschaft and Technik. Berlin: Akademie Verlag; 1967. p. 755–65. [8] Montelbano T, Brett J, Castleman L, Seigle L. Trans AIME 1968;242:1973. [9] Jansen HHR. High Temp––High Press 1986;18:173. [10] Gaal I. In: Pink E, Bartha L, editors. The metallurgy of doped/ non-sag tungsten. London: Elsevier; 1989. p. 162–5. [11] Gaal I, Uray L. In: Bildstein H, Ortner HM, editors. Proceedings of Thirteenth Plansee Seminar, vol. 1. Reutte: Metallwerk Plansee; 1993. p. 57–69. [12] Uray L, Gaal I. High Temp Mater Processes 1994;13:87–95. [13] Uray L. High Temp Mater Processes 1997;16:1–13. [14] Uray L. Z Metallkd 2001;92:386–90. [15] Kumar A, Eyre BL. Proc R Soc Lond A 1980;370:431–448r. [16] Smiti E, Jouffrey P, Kobylanski A. Scr Metall 1984;18:673–6. [17] Roberts WB. American institute of physics handbook. 3rd ed. New York: McGraw-Hill; 1972. p. 9-135, 9-147. [18] Roberts WB. In: Wearst RC, editor. CRC handbook of chemistry and physics, 67th ed. Boca Raton, Florida: CRC Press; 1986–1987. p. E-96. [19] Lassner E, Schubert W-D. Tungsten (properties, chemistry, technology of the element, alloys and chemical compounds). New York: Kluwer Academic/Plenum Publ; 1999. [20] Tinkham M. Introduction to superconductivity. New York: McGraw-Hill; 1975. [21] Tsuei CC. J Appl Phys 1974;45:1385. [22] Tinkham M. In: Garland JC, Tanner DB, editors. Electrical transport and optical properties of inhomogeneous media. New York: Am Inst Phys; 1978. p. 130–42. [23] Giorgi AL, Szklarz EG, Bowman AI, Matthias BT. Phys Rev 1962;125:837.

I. Gaal, L. Uray / International Journal of Refractory Metals & Hard Materials 20 (2002) 327–333 [24] Scherrer V. In: Ortner HM, editor. Proceedings of Tenth International Plansee Seminar, vol. II. Reutte: Plansee AG;1981. p 269–81. [25] Menyh ard M, Uray L. Scr Metall 1983;17:1195.

333

[26] Martens WAA, Brulez PJF. In: Bildstein H, Ortner HM, editors. Proceedings of Twelfth International Plansee Seminar, vol. 1. Reutte: Plansee AG; 1989. p. 523–38.