Effects of TiCl4 purity on the sinterability of Armstrong-processed Ti powder

Article abstract of DOI:10.1016/j.jallcom.2008.06.097

The sintering behavior of titanium powder produced via the Armstrong process from two different grades of TiCl₄ was investigated by a combination of thermal, chemical, and microstructural analysis techniques. It was found that the use of lower

Full text of DOI:10.1016/j.jallcom.2008.06.097

Journal of Alloys and Compounds 473 (2009) L39–L43
Contents lists available at ScienceDirect
Journal of Alloys and Compounds
journal homepage: www.elsevier.com/locate/jallcom
Letter
Effects of TiCl4 purity on the sinterability of Armstrong-processed Ti powder
K.S. Weil∗, Y. Hovanski, C.A. Lavender
Department of Materials Science, Pacific Northwest National Laboratory, PO Box 999 Richland, WA 99352 USA
a r t i c l e i n f o
a b s t r a c t
Article history:
The sintering behavior of titanium powder produced via the Armstrong process from two different grades
of TiCl4 was investigated by a combination of thermal, chemical, and microstructural analysis techniques.
It was found that the use of lower grade TiCl4 leads to higher part density due to a difference in sintering
behavior that affects the entrapment of volatiles within the sintered body.
Received 25 February 2008
Received in revised form 10 June 2008
Accepted 11 June 2008
Available online 9 August 2008
© 2008 Published by Elsevier B.V.
Keywords:
cp-Titanium
Sintering
Armstrong process
Thermal analysis
1. Introduction
producing titanium by this process is proportional to the costs of
the reactants; i.e. the use of low-cost TiCl4 will substantially reduce
Most titanium commercially produced today is synthesized via
the Kroll batch process, in which purified titanium tetrachloride
(TiCl4) is chemically reduced with molten magnesium to form
metal sponge [1]. This largely replaced an earlier method of tita-
nium production, known as the Hunter process, that employs liquid
sodium as the reductant [2]. An advancement in titanium produc-
tion at the time, the Kroll process consists of a series of batch
steps, a number of which remain labor intensive, inefficient, and
expensive [3]. As a result, a number of process improvements and
alternative production approaches have been subsequently con-
sidered [4,5]. One of these is a low-cost alternative known as the
Armstrong process, currently under development at International
Titanium Products, LLC (ITP). Utilizing the same chemistry as the
Hunter process, the Armstrong process produces titanium pow-
der via the reduction of TiCl4 vapor injected into a molten sodium
stream [6]. Since the rate of sodium flow exceeds that required for
stoichiometric reaction with the incoming TiCl4, the excess cools
the reaction products and transports them to a separation vessel
where both the sodium and the salt byproduct are removed. The
process yields a continuous stream of commercially pure titanium
(cp-Ti) powder and through modification, can also directly produce
alloy powders such as Ti-6Al, 4V. Independent economic studies
prepared for the U.S. Department of Energy and the U.S. Defense
Advanced Research Project Agency [7,8] indicate that the cost of
the cost of Armstrong-produced titanium powder.
TiCl4 is typically produced by chlorination of titanium-bearing
minerals such as high-grade ilmentite and rutile [9]. Conversion
takes place in a fluidized bed reactor at 850–950 ◦C using petroleum
coke as a reductant and forms volatile metal chlorides that are
collected and separated by fractional condensation, double distil-
lation, and chemical treatment [10]. Iron chloride (typically FeCl3)
is the chief byproduct of fractional condensation and is sold as
an industrial commodity. Additional trace metal chlorides are
removed through multiple distillations, and vanadium oxychloride
(VOCl3), which has a boiling point close to that of TiCl4 (136 ◦C),
is removed by a complexation step [11]. Because of our interest
in determining whether a potentially lower cost source of TiCl4
would still yield a useable grade of titanium, particularly for non-
aerospace applications, a study was initiated to examine the effects
of TiCl4 purity on the formation, consolidation behavior, and result-
ing properties of commercial purity titanium powder produced by
the Armstrong process. The results presented here focus primarily
on differences in sinterability that were observed.
2. Experimental methods
Two grades of TiCl4 were obtained from the DuPont Corporation (Wilmington,
DE), one which was the standard high-purity material produced for TiO2 pigment
manufacture (referred to as Lot I) and a second grade that was prepared using fewer
purification steps (Lot II). Aliquots of each were sent to ATI Wah Chang (Albany, OR)
for chemical analysis, while the bulk lots were shipped to ITP (Lockport, IL) and
converted into two separate batches of cp-Ti powder via the Armstrong process
described above. To minimize oxygen contamination, the resulting powder lots
were packaged, shipped, and stored in argon prior to consolidation and testing.
∗
Corresponding author. Tel.: +1 509 375 6796 fax: +1 509 375 4448.
E-mail address: scott.weil@pnl.gov (K.S. Weil).
0925-8388/$ – see front matter © 2008 Published by Elsevier B.V.
doi:10.1016/j.jallcom.2008.06.097

L40
K.S. Weil et al. / Journal of Alloys and Compounds 473 (2009) L39–L43
Table 1
Results from chemical analyses conducted on the Lots I and II TiCl4 precursors and cp-Ti powders.
Element
Lot I TiCl4 (ppm)
Lot II TiCl4 (ppm)
Lot I powder (ppm)
Lot II powder (ppm)
C
N
O
54
<10
110
<10
<10
–
29
<10
491
506.1
<10
–
93
77
2400
<10
330
19
55
47
1780
3220
110
Trace metala
Na
Cl
21
0.31
BET (m2/g)
–
–
0.31
a
The list of typical trace metal impurities include chlorides of: silicon, tin, aluminum, vanadium, niobium, tantalum, tungsten, tin, zirconium [12].
The particle size distribution of each was determined by sieve analysis using
ASTM B214 and E11. The powders were uniaxially cold pressed at 550 MPa into
MPIF-10 test bars and 4 mm × 8 mm × 25 mm long dilatometry bars. The bars were
heated under 1 × 10−6 Torr vacuum at 10 ◦C/min to 1250 ◦C and held there for a
period 20 min–24 h. Density measurements of the green and sintered specimens
were conducted by simple geometric and mass measurements, as well as by the
Archimedes technique using distilled water as the displacement medium.
3. Results
As shown in Table 1, the chemical assays of the two grades of
TiCl4 exhibit relatively minor differences in C, N, and O impurity
concentrations and more significant differences in trace metal
impurity content. Shown in the same table, the results from ICP/MS
analysis of the corresponding titanium powder lots indicate that
these impurities are transferred during sodium reduction. Note
that there is virtually no difference in the concentration of residual
salt (NaCl) impurity in the two powders, 19 ppm for Lot I and
21 ppm for Lot II. However, there is 3.5× more residual sodium in
the Lot I powder than in the Lot II powder, 311 ppm versus 89 ppm,
respectively. Engineers at ITP reported no differences in processing
the two lots of TiCl4 and little difference in particle morphology
was observed between the two corresponding powders during
microstructural analysis. As seen in Fig. 1(a) and (b), the two pow-
ders display a spongy, coral-like porous structure. The individual
particles can be plastically flattened under fingertip pressure. At
higher magnification, Fig. 1(c) and (d), it is apparent that both pow-
ders are actually agglomerates composed of fine scale crystallites
Dilatometry measurements were carried out in an Orton 1600 series dilatometer
(Orton Co.; Westerville, OH) a heating rate of 5 ◦C/min over the temperature range of
250–1100 ◦C in flowing ultra-high purity argon (UHP Ar; Matheson Tri•Gas, Mont-
gomeryville, PA). Specimen temperature was measured with an accuracy of 1 ◦
C
and linear dimensional changes were recorded with an accuracy of 0.5 ␮m. After
cooling, each bar was characterized by X-ray diffraction analysis to determine if oxi-
dation occurred; specimens exhibiting measurable oxidation were not included in
the final sintering analysis. Thermal analysis of the powders was conducted at West-
ern Kentucky University using a TA Instruments differential scanning calorimetry
system (DCS; New Castle, DE) equipped with a VG Thermolab mass spectrometer
(MS; Thermo Scientific, Waltham, MA). All of the thermal analysis experiments were
performed in flowing UHP Ar (10 ml/min) at a heating rate of 10 ◦C/min. Microstruc-
tural analysis of the powder and of cross-sectioned and polished sintered bars was
conducted using a JEOL JSM-5900 LV scanning electron microscopy (SEM; Peabody,
MA). Elemental analysis of the TiCl4, titanium powders, and sintered samples was
conducted via induction coupled plasma mass spectrometry (ICP–MS) and intersti-
tial gas analysis at ATI Wah Chang.
Fig. 1. Low magnification micrographs of Armstrong processed cp-Ti powder synthesized using: (a) high-purity TiCl4 (Lot I) and (b) low-purity TiCl4 (Lot II). High magnification
images of: (c) high-purity Armstrong powder, and (d) low-purity Armstrong powder.

K.S. Weil et al. / Journal of Alloys and Compounds 473 (2009) L39–L43
L41
Table 2
Lots I and II bar densities after sintering at 1250 ◦C for 20 min, 8 h, and 24 h.
Density (% of theoretical)
Lot
As-pressed
1250 ◦C, 20 min
1250 ◦C, 8 h
1250 ◦C, 24 h
I
II
86.0
84.1
94.4
96.3
87.1
90.6
66.2
71.2
It has been previously observed that Armstrong-processed cp-Ti
powder exhibits swelling when sintered under excessive conditions
(i.e. high temperature and/or long soak times) [14]. In comparison,
properly washed Hunter fines do not. Although the same reductant
(Na) is employed in both processes, there are significant differ-
ences in the characteristics of the resulting powders. Because the
Armstrong process is a continuous (not batch) method of produc-
ing titanium powder and employs a significant excess of molten
sodium, the powder product contains a measurable amount of
unreacted sodium that is typically removed via volatilization prior
to washing away the NaCl byproduct. However, due to the lacey,
coral-like morphology of the powder, it is speculated that some of
the sodium may become trapped in elemental or compound form
internally within closed or semi-closed voids that are generated
during the reduction reaction.
By comparison, during the Hunter process TiCl4 reduction takes
place within a closed reaction vessel at temperatures higher than
those employed in Armstrong processing. The resulting product
that forms is not a powder, but a sintered spongy mass that is
subsequently crushed and treated in a series of vaporization and
washing steps to eliminate excess sodium and the NaCl byprod-
uct. The sponge must then be attritted to form a powder product,
although some fines are produced during the sodium reduction of
TiCl4. The typical morphology of these fines is much less porous
than the Armstrong powder, as shown in Fig. 4. Because of its low
surface area morphology and the lack of internal voids, this pow-
der contains very little free sodium; a typical chemical assay of
Hunter fines prepared from high purity TiCl4 is provided in Table 3.
As mentioned above, products prepared from Hunter fines exhibit
internal porosity only when the powder is not properly washed to
completely remove NaCl.
Fig. 2. Particle size distributions of high- and low-purity powders (Lots I and II,
respectively) produced by the Armstrong process.
measuring ∼3 ␮m in size. The particle size distribution for each lot
is shown in Fig. 2. On average, the powder prepared from Lot I TiCl4
is larger in size (809 ␮m versus 438 ␮m) and displays a trimodal
distribution instead of a bimodal one due to the presence of a small
amount of particles greater than 1680 ␮m in size. However, the
tapped densities of both are nearly the same, 8 and 7% of theoretical
respectively for the corresponding Lots I and II powders.
In preparing the test bars, both powders compacted approxi-
mately to the same degree. The green densities of bars prepared
from the Lot I powder averaged 3.86 g/cc, while those cold pressed
from the Lot II powders averaged 3.78 g/cc. However, a significant
difference in sintered density was observed between the two, as
shown in Table 2. Both sets of bars exhibit their maximum den-
sity after sintering at 1250 ◦C for ∼20 min and display a reduction
in density (i.e. swelling) with additional time at temperature. Note
that at each sintering temperature, the Lot II bars are consistently
higher in sintered density even though they initially display slightly
lower green densities. A comparison of the internal microstructures
of both specimen types shown in Fig. 3 underscores this difference;
after sintering for 8 h, compacts prepared from the Lot I powder
exhibit a greater concentration of pores and larger average pore
size than those fabricated using the Lot II powder. A similar phe-
nomenon has been reported for arc-melted buttons prepared from
Kroll and Hunter produced titanium sponge materials in which the
salt byproduct was not fully removed. In both cases, porosity was
caused by volatile species that formed during the decomposition of
residual hydrated magnesium or sodium chloride inclusions during
melting [13].
DSC analysis of the Lot I material displayed no indication of
an anomalous phase change or chemical reaction in comparison
with a reference Kroll produced sample. However, MS analysis
showed that upon heating a volatile species with an atomic weight
of 44, presumed to be CO2, begins to evolve around 500 ◦C and
that the degree of evolution increases with increasing temperature
up to 1450 ◦C. Comparisons between chemical assays of this test
specimen with the original Lot I powder indicated no significant
difference in sodium content, suggesting that the residual sodium
Fig. 3. Cross-sectional SEM images of the internal microstructures of sintered bodies fabricated from: (a) high-purity (Lot I) and (b) low-purity (Lot II) Armstrong powders.
Both specimens were sintered under high vacuum at 1250 ◦C for 8 h.

L42
K.S. Weil et al. / Journal of Alloys and Compounds 473 (2009) L39–L43
Fig. 4. Micrograph of Hunter processed cp-Ti fines synthesized using high-purity
TiCl4.
Fig. 5. Dilatometry curves for the compacted Lots I and II powders.
from powder production remains even after heating to 1450 ◦C. The
fact that sodium persists in the powder up to high temperatures and
that CO2 is generated during heating suggest that sodium carbon-
ate, or a hydrated form of the compound, is the likely source of
swelling. Anhydrous NaCO3 is known to decompose into Na2O and
CO2 when heated above ∼400 ◦C and the kinetics of this reaction
increase dramatically above 1200 ◦C [15].
Results from dilatometry measurements, shown in Fig. 5,
highlight specific differences in densification behavior observed
between the two different powder lots. For example, the dilatome-
try curve for the Lot I material displays four primary breaks in slope;
i.e. rate of expansion/shrinkage. The initial linear increase in com-
pact length from room temperature to ∼470 ◦C can be attributed
to thermal expansion; i.e. there is no significant degree of sinter-
ing that appears to occur over this temperature range. At about
470 ◦C, point a1, the sample begins to shrink at a nearly linear rate
up to 880 ◦C (point a2), a temperature close to the ␣ → ␤ transi-
tion (␤ transus) in pure titanium. At 880 ◦C, a sharp increase in the
rate of shrinkage is observed that abruptly ends at 1010 ◦C (point
a3).
The Lot II material also displays a low-temperature region of
linear thermal expansion with the first significant measure of
shrinkage occurring at ∼590 ◦C (point b1), although it could be
argued that an initial deviation from linearity might begin at 520 ◦C.
Over the temperature range of 590–780 ◦C (point b2), the com-
pact shrinks at a nearly linearly rate comparable to that of the Lot
I material. It then begins to undergo a higher rate of non-linear
shrinkage from 780 to ∼880 ◦C (point b3). Beyond 880 ◦C, again at
approximately the ␤ transus, the rate of shrinkage or densifica-
tion increases up to 1100 ◦C (point b4), then subsequently shows a
gradual decline but does not abruptly halt. Panigrahi et al. reported
nearly the same shrinkage behavior in micron-size cp-Ti (from pow-
der that appears to have been formed via a hydride/de-hydride
process based on the reported angular particle morphology): ini-
tial compact shrinkage that occurs at 650 ◦C and increases in rate
at 880 ◦C [16,17]. They and others have suggested that low tem-
perature sintering (<880 ◦C) is controlled predominantly by lattice
diffusion within the ␣-Ti [15–18]. Above the ␤ transus, several
mechanisms of sintering have been proposed, including pipe dif-
fusion through dislocations, [19] lattice diffusion from surface and
grain boundary sources, [20,22] and creep (via grain boundary rota-
tion) [16].
Thus, there are two primary differences in shrinkage behav-
ior between the Lots I and II powder compacts: (1) a difference
in temperature at which sintering appears to initiate, ∼470 ◦C ver-
sus 590 ◦C, respectively, and (2) a sudden cessation of shrinkage
observed in the case of the Lot I material. Based on the prior pow-
der characterization results, the key physicochemical differences
between the two powder lots are: (1) a higher average particle size
in the Lot I material, or more specifically a fraction of powder that
is greater than 1680 ␮m in size; (2) a measurably higher concentra-
tion of trace metal impurities in the Lot II material; and (3) a higher
level of “free” sodium (uncombined with chlorine) in the Lot I pow-
der. To determine the importance of the first factor, an additional
set of sintering runs were conducted on the Lot I material after
classifying it to exclude powder particles ≥1680 ␮m. The average
green density of compacts pressed from this classified material was
found to be 3.82 g/cc and the resulting densities upon sintering at
1250 ◦C for 20 min, 8 h and 24 h are reported in Table 4. Note that
there is no significant improvement in sintered density relative to
the unclassified Lot I powder (Table 2), indicating that the differ-
ence in sintering behavior between the two powder lots is impurity
based.
Diffusion in ␣-Ti is typically enhanced by group VIII transi-
tion metal elements (e.g., Fe, Co, and Ni) [21] and diminished
by metal species found on either side of this region in the peri-
odic table, e.g., Nb, Sn, and Si [22,23]. This may be one reason
why sintering initiates at lower temperature in the higher purity
Table 3
Results from chemical analyses conducted on cp–Ti Hunter fines produced from a
high-purity TiCl4 precursor.
Element
(ppm)
C
N
105
<20
Table 4
Densities in bars pressed from classified Lot I powder after sintering at 1250 ◦C for
20 min, 8 h, and 24 h.
O
1890
<10
Trace metala
Na
Cl
1300
1400
Density (% of theoretical)
As-pressed
85.1
1250 ◦C, 20 min
94.2
1250 ◦C, 8 h
87.3
1250 ◦C, 24 h
65.9
a
The list of typical trace metal impurities include chlorides of: iron, silicon, tin,
aluminum, vanadium, chromium, niobium, tantalum, tungsten, tin, zirconium [12].
Classified lot I

K.S. Weil et al. / Journal of Alloys and Compounds 473 (2009) L39–L43
L43
Lot I powder. That is, the type of trace metal impurities found
Acknowledgements
in the Lot II powder delays activation of the primary mechanism
for low-temperature sintering, ␣-Ti lattice diffusion. Because the
Lot I powder begins to sinter at lower temperature, closed poros-
ity will subsequently form in the compacted material at lower
temperatures as the microstructure evolves during sintering. Com-
bined with the fact that a high-temperature, volatile compound
such as NaCO3 is retained on the Armstrong powders, it is this
latter phenomenon that is likely responsible for the difference in
sintered density observed between the two material lots. That is,
swelling only occurs if the CO2 becomes trapped during the sin-
tering process; i.e. if the gas is encapsulated within a set of closed
pores that form prior to significant sodium carbonate decomposi-
tion.
The authors would like to acknowledge Dynamet Technology
for help in conducting initial characterization of the Lot I pow-
der, Mike Hyzny at DuPont, Inc. for preparing the two grades of
TiCl4, Lance Jacobsen at ITP for synthesizing the two powders lots,
and John Hardy and Benjamin McCarthy at PNNL for conducting
dilatometry testing. This work was supported by the U.S. Depart-
ment of Energy, Office of Energy Efficiency and Renewable Energy.
The Pacific Northwest National Laboratory is operated by Battelle
Memorial Institute for the United States Department of Energy (U.S.
DOE) under Contract DE-AC06-76RLO 1830.
References
Because the flow stress of cp-Ti is quite low at 1250 ◦C (<20 ksi),
the material will readily creep (i.e. swell) when subjected to an
internal pressure over extended time at elevated temperature
(>20 min), such as that exerted by the expanding CO2 trapped
within closed pores. However if decomposition occurs predomi-
nantly when the pore structure of the sintering body is effectively
still open, then little to no swelling should be observed. Thus,
we suspect that the higher concentration of trace metal impuri-
ties found in the Type II powder cause the delay in sintering that
allows CO2 to escape the compact prior to entrapment. Alloyed
Armstrong titanium powder (Ti-6,4) exhibits nil swelling during
sintering [14] and the mechanism responsible for this finding
may be the same as that proposed here. The net effect is that
the use of the lower purity TiCl4 source is beneficial in two
ways: (1) it is potentially a lower cost titanium precursor in the
Armstrong process and (2) the trace impurities allow higher den-
sity components to be fabricated via a low-cost press and sinter
approach.
[1] W.J. Kroll, Trans. Am. Electrochem. Soc. 78 (1940) 35.
[2] M.A. Hunter, J. Am. Chem. Soc. 32 (1910) 330.
[3] M. Jackson, K. Dring, Mater. Sci. Technol. 22 (2006) 881.
[4] G.Z. Chen, D.J. Fray, T.W. Farthing, Nature 407 (2000) 361.
[5] S.J. Gerdemann, Adv. Mater. Proc. 159 (2001) 41.
[6] D.R. Armstrong, S.S. Borys, R.P. Anderson, U.S. Patent 5,779,761, issued July 14,
1998.
[7] P. Kirchain, The Role of Titanium in the Automobile: Understanding the
Economic Implication of Three Emerging Technologies, Camanoe Associates,
Cambridge, MA, 2002.
[8] E.H. Kraft, Cost-Affordable Titanium Symposium Proceedings, TMS, Charlotte,
NC, 2004, pp. 27–34.
[9] I.J. Krchma, U.S. Patent 2,701,180 February 1, 1955.
[10] U.S. Environmental Protection Agency, “Titanium Tetrachloride Production,”
from Report to Congress on Special Wastes from Mineral Processing, Vol.
II, Office of Solid Waste, July 1990, p. 13-3, available at: www.epa.gov/docs/
epaoswer/other/ mineral/chapter13.pdf.
[11] U.S. Environmental Protection Agency (Office of Solid Waste), “Titanium Tetra-
chloride Production by the Chloride Ilmenite Process,” December 1995, p. 4,
available at: www.epa.gov/epaoswer/other/mining/minedock/tio2/tio2.pdf.
[12] X. Gu, S.E. Lyke, S.E. Mirabella, J.M. Shekiro, Jr., U.S. Patent Application
20060275195, filed December 7, 2006.
[13] S.D. Hill, R.V. Mrazek, Metal. Trans. 5 (1974) 53.
[14] A.J. Benish, A. Sathaye, P. Nash, R. Anderson, and T. Zwitter, in Advances in
Powder Metallurgy and Particulate Materials – 2005, vol. III, part 10 (2005).
[15] Solvay Chemicals Publication, http://www/solvayminerals.com/pdf/brochures/
heat effects.pdf.
[16] B.B. Panigrahi, M.M. Godkhindi, K. Das, P.G. Mukunda, P. Ramakrishnan, Mater.
Sci. Eng. A 396 (2005) 255.
[17] B.B. Panigrahi, M.M. Godkhindi, K. Das, P.G. Mukunda, V.V. Dabhade, P. Ramakr-
ishnan, J. Mater. Res. 20 (2005) 827.
[18] C. Herzig, R. Willecke, K. Vieregge, Phil. Mag. A 63 (1991) 949.
[19] R. Sundaresan, A.C. Raghuram, R.M. Mallya, V.K.I. Asu, Powder Metall. 39 (1996)
138.
[20] R.M. German, Sintering Theory and Practice, John Wiley & Sons., New York.,
1996.
[21] G.E. Rhead, Scr. Metall. 6 (1972) 47.
[22] M. Köppers, C. Herzig, M. Freisel, Y. Mishin, Acta Mater. 45 (1997) 4181.
[23] G.M. Hood, Defect Diff. Data 95–98 (1993) 755.
4. Summary
The sintering behavior of titanium powder produced via the
Armstrong process from two different grades of TiCl4 was inves-
tigated by a combination of thermal, chemical, and microstructural
analysis techniques. It was found that the use of lower grade
TiCl4 leads to higher part density; a phenomenon attributed to a
delay in the onset of powder sintering. Because of this delay, the
internal microstructure of the compact remains open at higher
temperatures, allowing volatiles to more freely escape and avoid
becoming entrapped within the sintered body where they cause
swelling/bloating during the final stages of sintering.