Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

Contents lists available at ScienceDirect

Int. Journal of Refractory Metals and Hard Materials

j ourna l homepage: www.e lsev ie r .com/ locate / IJRMHM

Hydrogen reduction of tungsten oxides: Alkali additions, their effect onthe metal nucleation process and potassium bronzes underequilibrium conditions

T. Zimmerl a,⁎, W.-D. Schubert a, A. Bicherl b, A. Bock b

a Vienna University of Technology, 1060 Vienna, Austriab Wolfram Bergbau und Hütten AG, 8543 St. Martin, Austria

⁎ Corresponding author at: Wolfram Bergbau und HüttE-mail address: t.zimmerl@wolfram.at (T. Zimmerl).

http://dx.doi.org/10.1016/j.ijrmhm.2016.06.0150263-4368/© 2016 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 March 2016Received in revised form 17 June 2016Accepted 24 June 2016Available online 25 June 2016

The addition of alkali compounds influencesW particle growth during the reduction of tungsten oxide using hy-drogen. Existing literature does not reveal themechanism and clearmode of interaction. To improve understand-ing of the reduction process, lithium, sodium and potassium compounds were added to a highly pure tungstenoxide prior to reduction, and interrupted reduction experiments were carried out at 750 °C. The resulting pow-ders were investigated by metallographic and chemical methods: X-ray diffraction, SEM and EDX analysis.The experiments confirmed that intermediately formed tungsten bronzes play an important role in the earlystages of reduction. Further experiments showed that bronzes are formed from potassium containing com-pounds in equilibrium with mixtures of W and WO2 or WO2 and WO2.72. Furthermore it is demonstrated thatthe kinetics of the reduction sequence differs significantly between doped and undoped tungsten oxide powders,as demonstrated by the on-line measurement of reaction water in the hydrogen process gas.The results help to understand how the presence of alkali compounds effects the reduction of tungsten oxides.Investigations in production scale with intentionally contaminated tungsten oxide showed that bronzes canplay a role in industrialised production.

© 2016 Elsevier Ltd. All rights reserved.

Keywords:ReductionTungsten oxideAlkali compoundsHydrogenNucleationMechanismEquilibrium

1. Introduction

E. Lassner and W.-D. Schubert [1] describe that hydrogen reductionof tungsten oxides is an important and well established industrial pro-cess for the manufacturing of high quality tungsten powder.

The overall reduction can be summarized by the chemical Eq. (1)given below.

WO3 þ 3H2↔Wþ 3H2O ð1Þ

However, as shown in earlier investigations by [2–7], a series of ox-ides is formed intermediately during reduction and a chemical vapourtransport process (CVT) takes place. The intermediate phase transfor-mations are WO3 → WO2.9 → WO2.72 → WO2 → W. The CVT process isbased on the formation of a volatile tungsten oxide, WO2(OH)2, which

en AG, 8543 St. Martin, Austria.

formswith thewater produced during reduction. Thus, a certain vapourpressure of this compound prevails during the reduction sequencewhich determines the morphology of the reaction products, in particu-lar of the metal powder particles formed, as described by R. Haubner,W.-D. Schubert, E. Lassner, M. Schreiner and B. Lux [2]. Each transitionstage during reduction has, therefore, to be considered from two view-points: phase nucleation, and phase growth.Whereas the first stage canbe considered to occur always on the surface of the reacting species (e.g.WO2) in the form of a vapour/solid reaction, the growth stage can takeplace via the volatile tungsten compound. These two stages are shownin the Eqs. (2)–(4) for the formation of tungsten metal:

WO2 þ 2H2↔Wþ 2H2O tungsten nucleationð Þ; ð2Þ

WO2 þ 2H2O↔WO2 OHð Þ2þ H2 formation of volatile tungsten compoundð Þ ð3Þ

WO2 OHð Þ2 þ 3H2↔Wþ 4H2O tungsten growth by CVTð Þ ð4Þ

It was demonstrated by W.-D. Schubert, B. Lux and B. Zeiler [6] thatin the case of pure tungsten oxide the nucleation of the metal phase

Fig. 1. Sketch of the (a) reduction furnace and (b) arrangement for reduction water measurements.

Table 1Production parameters of used potassium tungstates (molar ratios, TH … production tem-perature, tH … reaction duration) [13].

K2WO4 K2W2O7 K2W3O10 K2W6O19

KHCO3:WO3 2:1K2WO4:WO3 1:1 1:2K2CO3:WO3 1:6TH 600 °C 700 °C 800 °CtH 24 h 24 h 100 h

Fig. 2. Sketch of equilibrium ampoule with used compounds.

88 T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

becomes possible only at significantly lower hydrogen humidity, as ex-pected from thermodynamic data. This gap between the theoreticallyand kinetically possible areas of tungsten metal formation at a giventemperature is a result of the nucleation work which has to be donewhenWO2 is transformed intoWmetal. However, if the nucleation pro-cess is facilitated by a “nucleation aid”, formation of tungsten metal be-comes possible even at higher humidity because themoment a tungstennucleus is formed further growth of tungsten can readily occur via thevapour phase. The higher the humidity within the powder bed, themore WO2(OH)2 is available, and the faster is the particle growth.

Investigating the role of potassium on the reduction of so-called nonsagging tungsten, used in the lighting industries, J. Neugebauer [8] ob-served the formation of tungsten metal out of potassium tungstenbronzes (KxWO3) occurs earlier than out of WO2. In later investigationsfrom W.-D. Schubert, B. Lux and B. Zeiler [6] and R. Haubner, W.-D.Schubert and E. Lassner [9], several ternary tungsten compounds (tung-states, tungsten bronzes) and even metal powder particles (Mo, Co, Ni,Cu) were able to act as nuclei for tungsten deposition by CVT.

Alkali additions to tungsten oxides prior to their hydrogen reductionare known to influence significantly the properties of resulting metalpowders, in particular their particle size; e.g. additions are used

Fig. 3. Reduction boat after (a) 5 min and (b) 30 min reduction

industrially to produce tungsten metal powder with mean particlesizes well above 10 μm [1].

Two differentmechanisms have been proposed up to now to explainthe interaction of the additives on the reduction process:

• the catalytic action of liquid alkali compounds upon the reaction withwater vapour forming the volatile oxide hydrate [9]. Equimolar con-centrations of Li, Na and K have about the same effect on particlecoarsening, which supports this theory.

• the occurrence of tungsten bronzes, and their effect on the metal nu-cleation process, by J. Qvick [10,11].

It is the aim of the present work to find experimental evidence tosupport one or other of these mechanisms for alkali additions.

The term tungsten bronze was used for the first time by F. Wöhler[12] because of the compounds' metallic lustre as well as their interest-ing electrical properties. Their chemical structure is commonly writtenas MexWO3 (Me … metal, with 0 b x b 1). Depending on the x-valuethey may form hexagonal, triclinic, tetragonal or cubic structures. Theyare non-stoichiometric compounds and are known for their shinycolourations (purple, violet, blue, red or golden).

2. Material and methods

2.1. Starting materials

High purity tungsten yellow oxide was provided by WolframBergbau- und Hütten AG. Alkali compounds were added to tungstenyellow oxide (WO3) being potassium carbonate (K2CO3), potassium bi-carbonate (KHCO3), sodium carbonate (Na2CO3), and lithiumhydroxide(LiOH), all of analytical quality (Sigma-Aldrich).

Different experiments were made to study the effect of the alkalicompounds during the hydrogen reduction process. During equilibriumexperiments oxygen partial pressure was set by using mixtures of W/WO2 (low) and WO2/WO2.72 (high) within two-sectioned and

time at 750 °C (lithium added). Hydrogen stream from left.

Table 2XRD and SEM examination of powder bed after 5 min reduction time in dry hydrogen at 750 °C (lithium addition); APT psm. … APT pseudomorphs.

Pos. Colour Phases in XRD Morphology (SEM) Résumé (XRD + SEM)

1 Grey/blue WO3 − X-related Lumps of cuboids Li-bronze with monoclinic structure2 Grey W, WO2, Li-bronze APT psm. Tungsten metal already formed3 Steel-blue Li-bronze (cub), WO2 Cubes, APT psm. Li-bronze4 Brown WO2, Li-bronze (cub) APT psm., cubes Tungsten dioxide, Li-bronze

Fig. 4. Reduction boat after (a) 5 min and (b) 30 min reduction time at 750 °C (sodium added). Hydrogen stream from left.

89T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

evacuated quartz ampoules.WO2 andWO2.72 were produced via partialreduction of WO3.

2.2. Interrupted reduction experiments

Firstly a powder mixture was prepared by blending 1 kg WO3 withan equivalent of 2 g of alkali metal. In this particular case the additionof dopant was intended to be non-uniform resulting in alkali compound“islands” within the powder bed. This facilitated direct observation ofthe compounds formed (due to their characteristic colours as comparedto the surrounding non-doped matrix) and their interaction during re-duction process.

The reduction boat was filled with the mixture to a layer height of2 mm and pushed into the heating zone of a tube furnace (Fig. 1(a))at 750 °C under dry hydrogen (6 L/min). The reduction sequence wasinterrupted after 5 and 30min by pulling the boat into the cooling zone.

Further reduction experimentswere carried outwith a powder layerheight of 25mm. In this case the interruptionwasmade after 120min toinvestigate the reduction sequence under very humid conditions. Afterpartially infiltration with epoxide resin a vertical cross section of thepowder bed was prepared to investigate it from top to bottom, withoutany mixing of interlayers.

Table 3XRD and SEM examination of powder bed after 5 min reduction time in dry hydrogen at 750 °

Pos. Colour Phases in XRD

1 Brown WO2, Na-bronze (cub)2 Blue/ruby Na-bronze (tet), WO2

3 Blue/ruby Na-bronze (tet), WO2

4 Brown/violet WO2, WO2.72

Table 4XRD and SEM examination of powder bed after 30 min reduction time in dry hydrogen at 750

Pos. Colour Phases in XRD and SEM

1 Grey W, WO2, Na-bronze (cub)2 Golden Na-bronze (tet), W, β-W, W3 Orange Na-bronze (tet), WO2, W, β4 Brown WO2, Na-bronze (cub), W

2.3. Humidity measurement of the reaction water

The respective alkali compoundswere added in the form of an aque-ous solution in equimolar amounts (14mmol/kg) to the tungsten oxideto achieve a uniform dopant distribution. The powders were filled in re-duction boats and reduced in a tube furnace with dry hydrogen at750 °C and 900 °C.

Measurement of reaction water produced over timewas carried outby a thermal conductivity measurement unit (Fig. 1(b)) capable of fol-lowing the water formation over the whole reduction sequence. Itthus provides data on the kinetics of the reduction process.

2.4. Equilibrium experiments with potassium containing species

The starting materials were produced as described in [13] (seeTable 1).

One samplewas produced frompotassiumcontaining APT (2.5wt%Kbased on WO2.9 content) via spray drying and decomposition at 500 °C,1 L/min hydrogen flow and 2 h duration (designation: “2.5%K”).

Small volumes of the compounds from Table 1 and 2.5%K were indi-vidually sealed-in into sectioned evacuated quartz ampoules (Fig. 2)with a big volume of the oxygen partial pressure setting mixtures (W/WO2, WO2/WO2.72, weight ratios: 1:1). Temperatures during 24 h

C (sodium addition).

Morphology (SEM) Résumé (XRD + SEM)

WO2, lumps, APT-pseudos Still no metal formedLumps of needles, WO2 Na-bronze in tungsten dioxideLumps of needles, WO2 Na-bronze in tungsten dioxideAPT-pseudomorphs Tungsten brown and violet oxide

°C (sodium addition).

Résumé (XRD + SEM)

Tungsten metal formedO2 Lumps of blocky Na-bronzes, W, cubes of β-W and WO2

-W Lumps of Na-bronzes, WO2, W and cubes of β-WWO2, lumps of cubic Na-bronzes, W

Fig. 5. Reduction boat after (a) 5 min and (b) 30 min reduction time at 750 °C (potassium added). Hydrogen stream from left.

Table 5XRD and SEM examination of powder bed after 5 min reduction time in dry hydrogen at 750 °C (potassium addition).

Pos. Colour Phases in XRD Morphology (SEM) Résumé (XRD + SEM)

1 Brown/violet WO2.72 Tungsten violet and brown oxide WO2.72 and WO2

2 Blue K-bronzes (hex, tet), W Intergrown needle bronze structures W and K-bronzes

90 T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

dwell time were 700 °C for all compounds and 900 °C for K2W6O19 and2.5%K. Ampouleswere quenched after treatment to avoid reactions dur-ing slow cooling.

2.5. Analysis

The analysis of the powders included optical microscopy (LeicaDM4000), scanning electron microscopy (FEI Quanta 200), equippedwith an energy dispersive analysing system (EDAX Genesis), X-ray dif-fraction (Philips X'Pert PRO, Generator PW 3040/60, Goniometer PW3050/60, copper anode), chemical analysis (ICP-OES) and metallo-graphic investigations of the resin-embedded powders.

In case of lithium compounds no chemical analysis could be per-formed by X-ray emission analysis (EDAX).

3. Results

3.1. Interrupted reduction experiments

3.1.1. LithiumAfter 5 min reduction time the layer surface can visually be separat-

ed in two zones. The area closer to the hydrogen inlet shows a greyish

Fig. 6. Potassium bronze after H2O2-treatment. Replicas of tungsten grains are visible onthe surface of the crystals.

blue matrix with some grey speckles, while the rest of the surfacelayer is brown with several steel-blue “islands”. These steel-bluespeckles or craters formed exactly at the positions the dopant com-pound had been located before reduction. Fig. 3(a) shows a picture ofthe boat (dry hydrogen from left).

Results from examination of positions marked 1, 2, 3 and 4 inFig. 3(a) by SEM and XRD are given in Table 2.

After 30min reduction time the biggest part of the powder surface isgrey with some brown areas (increasing from left to right (Fig. 3(b)).XRD examination show that the amount of tungsten dioxide increased

Fig. 7. Cross section of the alkali-free reference after 120 min of reduction at 750 °C in dryhydrogen (height = 2.5 cm). Several individual images were set together to show thewhole powder bed.

Fig. 8. (a) Cross section of lithiumdoped powder bed after 120min of reduction at 750 °C in dry hydrogen (height= 2.5 cm). (b) Detail of position C in (a). Outer tungsten-rich shell (left:bright tungstenmetal particles, greyish lithium rich/oxide areas) and inner lithium-rich area (right: greyish lithium-rich/oxide areas). (c) Detail of position F in (a) (bright tungstenmetalgrains, greyish oxidic structures). SomeW grains are indicated with arrows.

91T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

from position 1 to position 4. Two additional phases are detectable: W-metal and lithium tungstate (Li2WO4).

3.1.2. SodiumAfter a reduction time of 5 min (Fig. 4(a)) the surface of the powder

bed shows a gradual transition from brown (left side) to dark brown/vi-olet colour. An array of bluish and ruby spots can be seen all over thesurface exactly where Na2CO3 had been located before reduction.

XRD-measurements and SEM-analyses supported the presence ofsodium bronzes (Table 3).

After 30 min reduction time the surface layer is mainly brown withgolden and orange spots along the whole layer length (Fig.4(b)). Onlyto the very left tungsten metal has formed. Table 4 shows the resultsof XRD and SEM imaging.

3.1.3. PotassiumThe surface of the powder bed after 5 min reduction in hydrogen is

shown in Fig. 5(a). The colour changes gradually from brown toviolet-brown, with some spots (craters) of dark blue formed whereK2CO3 had been located before the reduction. Details of XRD and SEMare listed in Table 5.

After 30 min reduction the powder was grey at the left side of theboat and increasingly brown to the right Fig. 5(b)). Contrastive greyareas (W)with dark violet or ruby spots (tetragonal potassiumbronzes)in the center are visible,whereK2CO3 had been located at the start of thereduction. SEM-images confirm needles of potassium bronzes in thecenter of the crater like structures.

Extraction experiments with 10 wt% H2O2-solution to remove tung-sten metal and oxides showed that nucleation and growth occurs

directly upon the bronzes (e.g. on potassium bronze after 5 min of re-duction, see Fig. 6).

3.2. Interrupted reduction experiments with high powder layers

The non-uniformly doped mixtures were used with increased pow-der layer height (25 mm, 750 °C). Pure WO3 was used as a reference.After 120min the partly reduced powder bedswere embeddedwith ep-oxide resin and cut vertically to show the cross section.

3.2.1. Undoped oxideFig. 7 shows the characteristic cross section of the powder bedwith-

out alkali addition. Three zones can be distinguished: a top layer (A –grey layer) of tungsten metal powder, a middle layer (B) composed ofa mixture of brown tungsten oxide and tungsten metal representingthe transitionWO2➔W, and a lower layer (C) comprising brown tung-sten oxide only.

3.2.2. Lithium additionThe cross section of the lithium-doped powder bed is shown in

Fig. 8(a). The top layer of the powder bed consists of tungsten metal(grey) and is sharply separated from the brown oxide areas. PositionsA and B show rings of tungsten metal surrounded the voids whereLiOHhad been located before reduction. These areas seemed to promotethe formation and growth of tungsten at higher humidity.

Below the top grey area tungsten brown oxide containing severalgrey rings (positions C, D, E, F, Fig. 8(a)) can be found. These rings arelithium-rich inside (left in Fig. 8(b)) and tungsten metal-rich outside(right in Fig. 8(b)).

92 T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

Even in the lower andmore humid regions of the powder bed tung-sten metal grains can be found close to the formerly lithium-containingvoids (Fig. 8(c)).

3.2.3. Sodium additionThe cross section of the powder bedwith sodium addition (Fig. 9(a))

contains orange sodium bronzes in the upper part (high-sodium) andorange-red-violet ones in the deeper sections (high-tungsten). Againthe top layer already consists of grey tungsten metal powder.

BSE-pictures of the positions A and B (Fig. 9(b) and (c)) show detailsof the rings surrounding the holes of formerly Na2CO3. Here metallictungsten was formed. Observations were similar for position C.

In position D in Fig. 9(a) no tungsten grains but two different sodiumbronzes (orange-red and violet) can be detected.

EDX analyses of the bronzes in the positions A and D show that thesodium content within bronzes increases from the bottom of the pow-der bed to the top.

3.2.4. Potassium additionFig. 10(a) depicts the cross section of the potassium-doped powder

bed after 120 min of reduction time at 750 °C. Big voids surroundedby tungsten metal can be detected where K2CO3 was located at the be-ginning of the reduction. The thickness of the tungsten ring is decreas-ing within the powder bed from top to bottom (see position C).

It appears that the reduction process uses potassium-rich areas toadvance into the powder bed where without addition only brownoxide was found (Fig. 10(a), position B).

Potassium bronzes can be found all over the powder bed. Theamount of potassium in the bronzes rises from lower to higher regionsof the powder bed as EDX analyses indicate.

Fig. 9.Cross section of (a) the sodiumdopedmixturewith details of position A andD after 120mposition B of (a). Both pictures show tungstenmetal inwhitish grey and the sodiumbronze in grwith small green arrows).

Back scattered electron images (Fig. 10(b)–(d)) of the indicatedareas (A, B, C) in Fig. 10(a) show that tungsten nucleated even at mosthumid conditions in the lower parts of the powder bed.

3.3. On-line humidity measurements during reduction

Fig. 11 shows the reaction water formation during reduction of puretungsten yellow oxide in comparison to alkali-doped tungsten oxides at750 °C and 900 °C, respectively.

The reaction curves consist of two separate regionswhich reflect thetwo main reduction stages; the first gives significantly higher water re-lease per time unit during formation of WO2 from WO3 (via the violetoxide WO2.72), the second, a lower but longer, during the formation oftungsten metal (WO2 → W).

In detail the humidity curves at 750 °C reveal a characteristic shoul-der in case of doped oxides at the end of the first reduction peak whichis due to the nucleation and growth of themetal phase even at this earlystage (still during the transition of WO2.72 to WO2). This shoulder isclear evidence of the presence of a nucleation aid facilitating metal for-mation at significantly higher humidity. In all these cases reduction timewas significantly decreased. At the end of reduction at 750 °C, water for-mation rate slows down in the case of Na- and Li-doped oxides. This canbe explained by the reduction of the respective tungstates (Na2WO4,Li2WO4) which are formed in the very late stage of reduction cycleand which can, at 750 °C, be reduced only with dry hydrogen (Eq. (5)).

Me2WO4 þ 3H2↔WþMe2Oþ 3H2O Me : Li;Nað Þ ð5Þ

At 900 °C, the reduction rate appears similar for all dopant variantsand the WO2 to W transition takes about 75% of the time compared to

in of reduction at 750 °C indry hydrogen. BSE image of (b) ring inposition A and (b) ring iney. In (c) twodifferent bronzeswere detected (light grey anddarker grey, border indicated

Fig. 10. (a) Cross section of the potassium doped powder bed after 120 min of reduction at 750 °C in dry hydrogen (height = 2.5 cm). BSE pictures of indicated positions in (a); (b) outerrim of position A (left lower corner: brown oxide with few tungsten grains, to the upper right corner: a lot of tungsten grains), (c) detail of position B with tungsten grains (big whitishgrey) and two different bronzes (lighter and darker grey) and (d) tungsten grains (whitish) embedded in fine oxidic structure in position C.

93T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

pure tungsten oxide. Again, doped transformation occurs at significantlyhigher humidity. No tail is formed at 900 °C because the reduction of thetungstates at this temperature is rapid.

In comparison to laboratory experiments some particular reductionexperiments with inhomogenous doped WO3 were carried out in pro-duction scale aggregates. Fig. 12 shows not fully reacted golden/orangesodium bronzes on an agglomerate of tungsten metal. To high amountsof sodium and inappropriate reduction conditions can lead to an incom-plete reduction.

Fig. 11. Reaction water curves (kinetic curves) of the reduction of tungsten oxide with hydrocompared to pure WO3 (undoped).

3.4. Equilibrium experiments with potassium containing species

Table 6 shows the results of phase analysis with XRD of reactionproducts after 24 h of heat treatment at 700 °C in equilibrium withWO2/WO2.72.

Except K2WO4 which stays unreacted, all other potassium tungstatesreact to form tetragonal potassium bronze when starting with higher po-tassium content and hexagonal potassium bronze startingwith lower po-tassium content. Violet and brown oxide can be detected in some cases.

gen at (a) 750 °C and (b) 900 °C with different alkali compounds (equimolar additions);

Fig. 12. After production scale reduction golden sodium bronzes (arrows) within the greymatrix of tungsten metal powder.

Table 7Phases after heat treatment: 700 °C at 24 h in equilibriumwithW/WO2. Starting materialin first line. ++… high, +…middle,−… low,−−… very low amount, and *… foundin SEM.

W/WO2 K2WO4 K2W2O7 K2W3O10 K2W6O19 2.5%K

K2WO4 ++ +K2W2O7 −Bronze tet ++ + − +Bronze hex ++ +WO2 − − − − − − ++W * * − − *

94 T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

Table 7 shows the results of phase analysis with XRD of reactionproducts after 24 h of heat treatment at 700 °C in equilibrium with W/WO2.

Again K2WO4 remains almost chemically unreacted, while all othercompounds react to bronzes (tetragonal starting with higher potassiumcontent and hexagonal with lower potassium content). Additionally de-pending on the starting material brown oxide and tungsten metal canbe detected in XRD or found in SEM.

SEM images of selected starting materials and reaction products areshown in Fig. 13 for 700 °C and lower oxygen partial pressure (WO2/W).

Results for treatments at 900 °C (high and low oxygen partial pres-sure)with K2W6O19 and 2.5%K as startingmaterial are shown in Table 8.

At 900 °C starting with K2W6O19 produced hexagonal potassiumbronze only (XRD, SEM). 2.5%K reacted to amixture of hexagonal potas-sium bronze and brown oxide or violet oxide depending on themixturein the other section of the ampoule (i.e. oxygen partial pressure).

Fig. 14 depicts SEM images of the compounds from Table 8 after heattreatment at lower oxygen partial pressure (WO2/W).

In Fig. 15 reaction products after treatment at 900 °C at higher oxy-gen partial pressure (WO2.72/WO2) are shown.

4. Discussion

All reduction experiments described in this work have confirmedthat ternary alkali compounds (alkali tungsten bronzes, alkali tung-states) are present throughout the reduction sequence. Tungstenbronzes were found to coexist with all other tungsten compoundsformed within the powder layer (WO2.9, WO2.72, WO2, W). At an earlyreduction stage (i.e. at high humidity within the layer) they formalkali-poor tungsten-rich bronzes whereas with on-going reduction(lower humidity) high-alkali bronzes are formed. Additionally the equi-libriumexperiments showed that potassiumbronzes form at 700 °C and900 °C at oxygen partial pressures comparable to the oxygen partial

Table 6Phases after heat treatment: 700 °C at 24 h in equilibriumwithWO2/WO2.72. Starting ma-terial in first line. ++… high, + … middle, − … low, and − − … very low amount.

WO2/WO2.72 K2WO4 K2W2O7 K2W3O10 K2W6O19 2.5%K

K4WO5 −K2WO4 ++ −Bronze tet ++ ++Bronze hex ++ ++WO2.72 +WO2 − +

pressures within the powder bed during reduction (mark: no hydro-gen). During these equilibrium experiments morphology of the startingmaterials changed, which indicates that there occurs chemical vapourtransport (CVT) reaction even without hydrogen as reaction gas.

J. Qvick [11] generalised his thermodynamic considerations inEqs. (6) and (7) below. Both equations support the results made withinthe current investigation.

MexWO3 þ 3 1−x=yð ÞH2↔x=yMeyWO3 þ 1−x=yð ÞWþ 3 1−x=yð ÞH2O withxbyð Þ ð6Þ

MeyWO3 þ 3−2yð ÞH2↔y=2Me2WO4 þ 1−y=2ð ÞWþ 3−2yð ÞH2O ð7Þ

Eqs. (6) and (7) demonstrate that tungstenmetal nucleates from thetungsten bronzes acting as seeds for rapid local growth of W particlesvia CVT. Nucleation of tungsten in alkali doped powders occurs at signif-icantly higher humidity than from WO2, as demonstrated by the deeppowder layer experiments. Thus, if the alkali-additions are uniformlydistributed within the tungsten oxide matrix, W nuclei are formed bythe gradual reduction of tungsten bronzes while no nucleation occurson the surface of the WO2 particles. Instead, WO2 reacts with the reac-tion water present forming the volatile (WO2)(OH)2 compound whichis transported to the growing tungsten nuclei.

Towards the end of reduction alkali tungstates with higher x-valuesare formed which are themselves reduced if reduced for long times, indry hydrogen and at high temperatures. Eventually the resulting oxidesor hydroxides would be vapourised in accordance with their thermalproperties.

The investigated alkali metals lithium, sodium and potassium be-have slightly differently. Lithium and potassium bronzes show a stron-ger effect as nucleation aid compared with sodium bronzes. Thus,tungstenmetal can be formedearlier (i.e. at a higher dynamic humidity)than with sodium additions. Consequently, with Li and K additions thetransition of WO2 to W can take place through larger powder beddepths thanwith Na. However, with Na additions the velocity of reduc-tion is still larger than in undoped oxide. This means that the influenceof humidity on theWnucleation process is less pronounced in the pres-ence of alkali additions and reduction can occur closer to the thermody-namic limit.

Humidity measurements indicate that early nucleation of tungstenmetal takes place in all alkali-doped powders at the end of the first re-duction step (i.e. shoulder at the end of the first peak in Fig. 11(a)).First nuclei form on the tungsten bronzes and small tungsten crystalsstart to grow during this period (see Fig. 6 for replicas of tungsten grainson tungsten bronzes after H2O2-treatment to remove tungsten metal).In contrast, a period of “drying-out” can be seen in case of the undopedoxide suggesting that for complete transition to themetal phase the hu-midity had to be reduced further still, see Fig. 11(b).

5. Conclusion

This investigation shows how alkali compounds act during the re-duction of doped tungsten oxides, making their effects “visible” in

Fig. 13.K2W2O7 a) before and b) after treatment; K2W6O19 c) before and d) after treatment; 2.5%Ke) before and f) after treatment; conditions: 700 °C, lower oxygenpartial pressure (WO2/W).

95T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

XRD and SEM investigations by using both heterogeneous and homoge-neous additions. Alkali metals react at the very beginning of reductionwith the tungsten oxide matrix to form tungsten-rich alkali bronzes.

Table 8Phases after heat treatment: 900 °C at 24 h in equilibriumwithW/WO2 andWO2/WO2.72.Starting material in second line. ++… big, +…middle,−… low, and−−… very lowamount.

W/WO2 WO2/WO2.72

2.5%K K2W6O19 2.5%K K2W6O19

Bronze hex ++ ++ ++ ++WO2.72 ++WO2 ++

These bronzes subsequently generate (“release”) tungsten metal nucleiunder comparatively humid reduction conditions. Such nuclei are ableto grow in the humidity gap between “thermodynamically possible”and “kinetically possible” and thereby form tungsten grains. This mech-anism enables reduction occurring at higher humidity within thereacting layers.

An experiment with intentionally contaminated tungsten oxide ex-hibits bronzes after reduction in production scale aggregates (Fig. 12).

During reduction, the bronzes follow a path from tungsten-rich toalkali-rich composition which leads to final evaporation of the alkalicompounds at a rate depending on reduction time and temperature.This agrees with available literature [10,11].

Both at 750 °C and 900 °C alkali-doped powders are reduced signif-icantly faster than undoped oxides during the WO2 → W transitionconfirming the fact that they are being reduced under more humidconditions.

Fig. 14. 2.5%K (a) after treatment; K2W6O19 (b) after treatment; conditions: 900 °C, lower oxygen partial pressure (WO2/W). Compounds before heat treatment, see Fig. 11.

Fig. 15. 2.5%K a) after treatment; K2W6O19 b) after treatment; conditions: 900 °C, higher oxygen partial pressure (WO2.72/WO2). Compounds before heat treatment, see Fig. 13.

96 T. Zimmerl et al. / Int. Journal of Refractory Metals and Hard Materials 62 (2017) 87–96

Acknowledgements

The authors appreciate thehelp ofW. Tomischko for electrotechnicalsupport concerning the on-line humidity measurements. Ch. Keller isgratefully acknowledged for fruitful discussions. We also thank A.Grearson for proof reading our manuscript.

References

[1] E. Lassner, W.-D. Schubert, Tungsten: Properties, Chemistry, Technology of the Ele-ment, Alloys, and Chemical Compounds, first ed. Kluwer Academic/Plenum Pub-lishers, New York, 1999.

[2] R. Haubner, W.-D. Schubert, E. Lassner, M. Schreiner, B. Lux, Mechanism of technicalreduction of tungsten: part 2, International Journal of Refractory Metals and HardMaterials 1983, pp. 156–163 (Nr. 4).

[3] O. Glemser, H. Ackermann, Gasous hydroxides. VI. Supplementary researchconcerning gaseous WO2(OH)2, Z. Anorg. Allg. Chem. (325) (1963) 281–286.

[4] H. Hellmer, W.-D. Schubert, E. Lassner, B. Lux, Kinetik der Wolframreduktion, 11thInternational Plansee Seminar '85, 1985.

[5] C. Choain, F. Marion, Sur les équilibres d'oxydoréduction des oxydes de tungstène,Comptes rendus hebdomadaires des séances de l'Académie des sciences 252 (3)(1961) 3258–3260.

[6] W.-D. Schubert, B. Lux, B. Zeiler, L. B, E. L, W.-D. S, B. L, The Chemistry of Non SagTungsten(Hrsg.) 1995.

[7] A. Bartl, Fundamentals of NS-Tungsten Powder Manufacture(Wien) 1997.[8] J. Neugebauer, Oxidations- und Reduktionsvorgänge im System K-W-O unterhalb

von 800 °C in H2O-H2 Gasgemischen, Planseeberichte für Pulvermetallurgie 23(1975) 77–85.

[9] R. Haubner, W.-D. Schubert, E. Lassner, Einfluss der Alkalidotierungen auf dieReduktion von WO3 zu Wolfram mit Wasserstoff, 11th International Plansee Sem-inar '85, 1985.

[10] J. Qvick, Influence of sodium on the reduction of tungsten (VI) oxide, Reactivity ofSolids 4 (1–2) (1987) 73–91.

[11] J. Qvick, Thermochemical calculations on the reduction of tungsten (IV) oxide withspecial emphasis on trace element behavior, Int. J. Refract. Met. Hard Mater. 3 (3)(1984) 121–131.

[12] F. Wöhler, Sur le Tungstene, Ann. Chim. Phys. 2 (1823) 43–53.[13] J. Gras, I. Hinz, G. Kirschstein, in: V. Haase, H. Katscher, G. Kirschstein, F. Schröder

(Eds.), Gmelin Handbuch der anorganischen Chemie - W - Wolfram - Sys. Nr. 54 -Erg.Bd.B3, Springer-Verlag, Berlin, Heidelberg, New York, 1979 (Hrsg.).