materials science and engineering- a volume 352 issue 1-2 2003 [doi 10.1016/s0921-5093(02)00864-x]...

Upload: jasbir-s-ryait

Post on 14-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    1/7

    Mechanical alloying of MoSi2 with ternary alloying elements.Part 1: Experimental

    A.J. Heron, G.B. Schaffer *

    Division of Materials, School of Engineering, The University of Queensland, Brisbane, Qld 4072, Australia

    Received 22 July 2002

    Abstract

    Phase evolution during the mechanical alloying of Mo and Si elemental powders with a ternary addition of Al, Mg, Ti or Zr was

    monitored using X-ray diffraction. Rietveld analysis was used to quantify the phase proportions. When Mo and Si are mechanically

    alloyed in the absence of a ternary element, the tetragonal C11b polymorph of MoSi2 (t -MoSi2) forms by a self-propagating

    combustion reaction. With additional milling, the tetragonal phase transforms to the hexagonal C40 structure (h -MoSi2). The

    mechanical alloying of Al, Mg and Ti additions with Mo and Si tend to promote a more rapid transformation of t -MoSi2 to h -

    MoSi2. In high concentrations, the addition of these ternary elements inhibits the initial combustion reaction, instead promoting the

    direct formation of h -MoSi2. The addition of Zr tends to stabilise the tetragonal phase.

    # 2002 Elsevier Science B.V. All rights reserved.

    Keywords: Mechanical alloying; Molybdenum disilicide; Rietveld

    1. Introduction

    Mechanical alloying (MA) is well known as a non-

    equilibrium powder processing technology and has been

    applied to many systems, including MoSi2 [1/8]. The

    MA of molybdenum and silicon has been studied

    extensively with most researchers reporting a sponta-

    neous high temperature synthesis reaction after only a

    few hours of milling to form the tetragonal C11bstructure (t -MoSi2). Continued high energy deformation

    induced by mechanical milling has also been shown tofavour the formation of the hexagonal C40 polymorph

    (h-MoSi2).

    Alloying of MoSi2 with a third element to remove

    oxygen is one strategy used to reduce the deleterious

    effects of SiO2 at elevated temperatures [9]. The alloying

    behaviour of MoSi2 has been widely studied, particu-

    larly with Al to form Mo(Si,Al)2 [9/12]. As shown in the

    ternary phase isotherms proposed by Brukl et al. at

    1600 8C [13] and by Yanigahara et al. [14] at 1550 8C,

    Al has some solubility in MoSi2 and in concentrations in

    excess of 3 at.%, promotes the formation of the C40

    structure. The Al substitutes for the Si atoms on the

    {110} planes of the C11b structure, which are equivalent

    to the close-packed {0001} planes in C40. This coincides

    with a minor charge transfer away from the Al sites,

    which bond covalently with the Mo [15]. The MA of

    MoSi2 with Al has recently been investigated by Costa e

    Silva and Kaufman [9]. The h-MoSi2 phase is alsostabilised by both Ti and Mg, although the Ti subsitutes

    for the Mo and forms a stable disilicide with the C54

    structure. The C40 structure is stable at intermediate Ti

    levels [16,17].

    The aim of this work was to study the MA of Mo with

    Si to form MoSi2 and its alloying with Al, Mg, Ti or Zr.

    These elements were chosen for their strong chemical

    affinity for oxygen. This paper shows the progress of the

    reactions using X-ray diffraction, analysed using Riet-

    veld analysis. A companion paper presents a computer

    simulation of the reactions as they occur during MA

    [18].

    * Corresponding author. Tel.: '/61-7-3365-4500; fax: '/61-7-3365-

    3888.

    E-mail address: [email protected] (G.B. Schaffer).

    Materials Science and Engineering A352 (2003) 105/111

    www.elsevier.com/locate/msea

    0921-5093/02/$ - see front matter# 2002 Elsevier Science B.V. All rights reserved.

    PII: S 0 9 2 1 - 5 0 9 3 ( 0 2 ) 0 0 8 6 4 - X

    mailto:[email protected]:[email protected]
  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    2/7

    2. Experimental

    The powders were obtained from Cerac Inc., the

    powder details are presented in Table 1. The mechanical

    alloying was performed in a Spex 8000 mixer/mill. The

    powders were placed in a high strength steel vial with

    20)/8.4 g hardened steel ball bearings. The powderswere mixed in proportion such that the total powder

    mass of each mill was 8.4 g, giving a constant charge

    ratio (ball mass to powder mass, Cr) of 20. The vial was

    sealed in an inert argon atmosphere with an oxygen

    concentration B/1000 ppm. Prior to each mill, the vial

    was cleaned by milling with alcohol for 1 h and then

    sand blasted and the loose powder was removed by

    compressed air. Molybdenum and silicon powder were

    milled for times between 2.5 and 100 h with up to 16.7%

    of Al, Mg, Ti or Zr (all concentrations in at.% unless

    otherwise specified). After each mill, the loose powder

    was removed for analysis.

    3. Rietveld analysis

    Diffraction patterns were collected using a Phillips X-

    ray diffractometer with Cu Ka

    radiation and a nickel

    filter. Scans were performed using a low angle back-

    ground holder with diffraction angles 15/908, a step size

    of 0.058 and a scan speed of 18 min(1. The phases

    present were identified using the mPDSM search/match

    program. Due to the large discrepancy in the initial

    powder sizes, it was difficult to obtain X-ray diffraction

    patterns for the unmilled powder mixes. The X-ray

    diffraction patterns from the powders after 2 h of

    milling were, therefore, used as a reference for the

    powders milled for longer times.

    The X-ray patterns were analysed using the LHPM9

    Rietveld program [19] to give the phase proportions.

    Rietveld analysis refers to any method that uses a

    mathematical description of the crystal structure to

    calculate a theoretical X-ray diffraction pattern. The

    description of the crystal structure includes information

    such as the space group, atomic positions and lattice site

    occupancy, preferred orientation and unit cell size. In

    order to determine a theoretical diffraction pattern, the

    X-ray diffraction process, including source of experi-

    mental error, is also modelled using parameters for the

    radiation wavelength, background noise and adjustment

    to peak shape due to grain size, peak asymmetry and

    adsorption effects. The theoretical pattern is then

    compared to the experimentally observed pattern and

    recursively refined to minimise the difference between

    them.

    The computer generated diffraction patterns are

    governed by several parameters, including unit cell

    parameters, atomic positions, scale factors and terms

    to define the peak shape. The refined scale factor from

    Rietveld analysis gives an indication of the integrated

    peak intensities. The broad, low intensity of the peaks

    are accounted for by increasing the Lorentzian para-

    meter. Hence, if two diffraction patterns with the same

    phase proportions are analysed, one with tall sharp

    peaks and the other with low broad peaks, Rietveld

    would keep the scale factor constant and vary the

    Lorentzian parameter to account for the peak shape.

    The Rietveld method can only match phases in the

    experimental patterns that are included in the input file.

    Thus, a traditional search/match process must first be

    undertaken to determine which phases are observable

    before the Rietveld analysis can be initiated. Due to the

    broad peaks in several of the diffraction patterns,

    especially in the mills with high proportions of Ti and

    Zr, it is difficult to determine all the phases present with

    complete certainty. This is particularly problematical for

    amorphous phases. The converse of this problem must

    also be considered. If a phase is selected in the input file

    but is only just resolvable above the background in the

    experimental pattern, or not observable at all, the

    Rietveld analysis tends to overestimate the phase

    proportion by keeping the scale factor constant and

    increasing the Lorentzian parameter. Hence the experi-

    mental pattern is matched by a broad, low intensity

    peak with a large integrated area. When this was

    observed during the Rietveld refinement process, the

    phase in question was considered not to be present in

    significant proportions and was removed from the input

    file. It is for this reason that Rietveld occasionally does

    not report the existence of a particular phase when it

    may be argued that its peaks can indeed be obser ved.

    To account for amorphous phases and those that fall

    below the detectable limit, the phase proportions were

    corrected by assuming that the Mo never becomes

    unobservable and is always accounted for as either

    elemental Mo or in the MoSi2 phase. This assumption is

    based on the high atomic scattering factor of Mo,

    particularly compared to Si. The corrected molar phase

    proportions and the associated error analysis are derived

    in Appendix A.

    Table 1

    The size and purity of the starting powders

    Powder Size Purity (%)

    Mo (/325 mesh 99.9

    Si 3/6 mm 99.999

    Al (/325 mesh 99.5

    Mg (/325 mesh 99.6

    Ti (/150 to '/325 mesh 99.5

    Zr 1/3 mm 99.8

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111106

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    3/7

    Fig. 1. X-ray diffraction patterns for the mechanical alloying of 33.3 Mo/66.7 Si.

    Fig. 2. The Rietveld refinement for the sample containing Mo and Si powder, without a ternary element, milled for 2.6 h.

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111 107

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    4/7

    4. Results and discussion

    Fig. 1 shows the X-ray diffraction patterns for themechanical alloying of Mo and Si in the absence of

    ternary additions. After 2.5 h of milling, the diffraction

    pattern shows strong Mo peaks with relatively weak and

    broad Si peaks. The reduction in the intensity of the Si

    peaks during the early stages of mechanical alloying is

    commonly observed and is attributed to the intimate

    mixing of the two elements and the strong absorption of

    Mo [2,8]. At 2.6 h, the elemental precursor powders havetransformed to t -MoSi2 and the peaks are sharp and

    well defined. This is indicative of a combustion reaction

    between 2.5 and 2.6 h. A large portion of Mo is still

    evident after combustion. It is possible that some

    regions of the powder mixture were physically isolated

    from the combustion front, or were not involved in

    sufficient deformation events for the combustion reac-

    tion to proceed. The diffraction peaks after 10 h have

    broadened considerably compared to the peaks for the

    pattern after 2.6 h of milling, indicating deformation

    and grain size refinement. The tetragonal structure

    transforms to the hexagonal structure as milling con-tinues up to 100 h. The Rietveld analysis immediately

    Fig. 3. The change in the relative phase proportions during the

    mechanical alloying of 33.3 Mo/66.7 Si, as determined by Rietveld

    analysis.

    Fig. 4. X-ray diffraction patterns for the mechanical alloying of 27.8 Mo/55.6 Si /16.7 Al after (a) 2, (b) 5, (c) 10, (d) 20 and (e) 50 h.

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111108

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    5/7

    following combustion is shown in Fig. 2 and the relative

    phase proportions over the entire milling period are

    shown in Fig. 3. After 100 h, t -MoSi2 and h -MoSi2reach steady state proportions of 31.7 and 68.3 mol%,

    respectively.

    After 2 h of milling, the powders containing 16.7% Al,

    Mg and Ti have not reacted and still show strong

    elemental peaks. As milling continues for up to 50 h, Al,

    Mg and Ti additions promote the formation of h-MoSi2,and other than the powder containing Ti, exhibit sharp

    diffraction peaks. The series containing 16.7% Ti

    exhibits very broad, low intensity peaks after 50 h,

    indicating a fine nanocrystalline structure. The X-ray

    diffraction patterns of the system with 16.7% Al are

    shown in Fig. 4. In contrast to the systems containing

    Al, Mg and Ti, the system containing Zr largely inhibits

    the formation of the hexagonal polymorph. The X-ray

    diffraction patterns of the system with 16.7% Zr is

    shown in Fig. 5. The ZrSi2 phase has the C49 structure

    and the C40 phase does not form in the quasi-binary

    ZrSi2/MoSi2 system [17].

    The effect that each element has on the relative phase

    proportions after 50 h of milling are shown as a function

    of alloy content in Fig. 6. As shown above, the addition

    Fig. 5. X-ray diffraction patterns for the mechanical alloying of 27.8 Mo/55.6 Si /16.7 Zr after (a) 2, (b) 5, (c) 10, (d) 20 and (e) 50 h.

    Fig. 6. The concentration of the tetragonal C11b polymorph as a

    proportion of the total MoSi2 content after 50 h of milling showing

    that Zr stabilises the tetragonal phase whereas Al, Mg and Ti stabilise

    the hexagonal C40 phase.

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111 109

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    6/7

    of Al, Mg or Ti to MoSi2 favours the formation of the

    hexagonal C40 polymorph over the tetragonal C11bphase and complete transformation to the hexagonal

    variant occurs at some critical level of alloying element.

    In contrast, the tetragonal phase persists for all Zr

    concentrations tested.

    5. Conclusions

    The crystallographic structure of mechanically al-

    loyed molybdenum and silicon with various ternary

    additions was monitored using X-ray diffraction and

    Rietveld analysis. In the absence of a third alloying

    element, the reactants combust during milling to form

    MoSi2. Initially, this has the tetragonal C11b structure,

    but this decomposes during further milling to the

    hexagonal C40 phase. The addition of titanium pro-motes the transformation of the tetragonal MoSi2polymorph to the hexagonal one. Magnesium and

    aluminium additions also sustain the transformation,

    but to a lesser extent. Zirconium stabilises the tetragonal

    phase.

    Acknowledgements

    This work was funded by the Australian Research

    Council.

    Appendix A

    Rietveld analysis may be used to estimate the relative

    phase proportions of the observable phases within a

    sample. However, if a phase becomes amorphous or the

    quantity falls below the detectable limit for Rietveld

    analysis then the phase proportion cannot be accounted

    for in the determination of the observable phase

    proportions. To account for the unobservable phases,

    it is assumed that the Mo never becomes unobservableand is always present as either elemental Mo or in the

    MoSi2 phase. The total number of atoms within the

    sample may be determined from the experimental molar

    phase proportions

    Atotal0NA(pMo'pSi'pA'pMoSi2'pU) (A1)

    where pU is the number of unobservable moles of (Si'/

    additive), NA is Avagadros number and the relative

    molar phase proportions, determined experimentally

    from Rietveld analysis, are given by pMo, pSi, pA and

    pMoSi2 for the Mo, Si, addition and MoSi2 phases,

    respectively.

    The atomic phase fractions are, therefore, given by

    AMo01

    Atotal

    NA

    pMo'

    1

    3pMoSi2

    ASi0

    1

    Atotal

    NA

    pSi'

    2

    3pMoSi2

    AA01

    Atotal(NApA)

    AU01

    Atotal(NApU) (A2)

    Since AMo is constant and equal to the initial atomic

    phase fraction, given by

    AMo01

    3

    (1(AI) (A3)

    where AI is the initial atomic fraction of the additive, the

    first line in Eq. (A2) may be rewritten as

    1

    3(1(AI)0

    NA

    pMo '

    1

    3pMoSi2

    NA(pMo 'pSi 'pA 'pMoSi2 'pU)(A4)

    Rearranging to solve for pU gives

    pU0

    pMo '1

    3pMoSi2

    1

    3 (1( AI)

    ((pMo'pSi'pA'pMoSi2 ) (A5)

    The total number of moles in the sample is given by

    Mtotal0pMo'pSi'pA'pMoSi2'pU (A6)

    Therefore, the normalised phase proportion, pU? , of

    unobservable Si and additive phase relative to the total

    number of moles is given by

    p?U0pU

    Mtotal(A7)

    It follows that if pU? of the phases were unobservable

    then the initially determined experimental molar phaseproportions represented only (1(/pU? ) of the moles

    present in the sample.

    Therefore, the corrected molar phase proportions are

    given by

    p?Mo0(1(p?U)pMo

    p?Si0(1(p?U)pSi

    p?A0(1(p?U)pA

    p?t-MoSi2 0(1(p?U)pt-MoSi2

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111110

  • 7/30/2019 Materials Science and Engineering- A Volume 352 Issue 1-2 2003 [Doi 10.1016/s0921-5093(02)00864-x] a.J Heron;

    7/7

    p?h-MoSi2 0(1(p?U)ph-MoSi2 (A8)

    The errors in the relative phase proportions may be

    calculated from the Rp counting statistic returned by the

    Rietveld analysis program, LHPM9 [19,20]. Rp is

    defined as the absolute error between the calculated

    and observed diffraction intensity, summed over all datapoints in the specified diffraction angle range.

    Rp0

    Pjyi;obs (yi;calcjP

    yi;obs(A9)

    If Sp ,obs and Sp ,calc are the observed and calculated

    scale factors, and yi1 is the true intensity for data point i,

    then

    Rp0

    PjSp;obsyi1(Sp;calcyi1jP

    Sp;obsyi1(A10)

    Rp0jSp;obs ( Sp;calcjP

    yi1

    Sp;obsP

    yi1

    (A11)

    Rp0jSp;obs ( Sp;calcj

    Sp;obs(A12)

    The error in the absolute phase proportion is

    jpp;obs (pp;calcj

    pp;obs0

    jSp;obsZpVp ( Sp;calcZpVpj

    Sp;obsZpVp0Rp (A13)

    The upper, pp', and lower pp

    (, limits of the corrected

    phase proportions for phase p , may be determined by

    p?p;obs0p?p;calc9Rpp?p;obs (A14)

    p'p 0p?

    p;

    calc

    (1 (Rp)(A15)

    p(p 0p?p;calc

    (1 'Rp)(A16)

    Considering the unobservable phase in Eq. (A7) the

    upper limit for the relative phase proportions may be

    calculated as

    f'p 0p?p;calc

    p?p;calc 'p?U '(1( Rp)

    (1' Rp)

    Xni"p

    p?i;calc

    (A17)

    Similarly, the lower bound may be found

    f(p 0p?p;calc

    p?p;calc 'p?U '(1 'Rp)

    (1 (Rp)

    Xni"p

    p?i;calc

    (A19)

    For the data here, the residual error, Rp , was typicallyless than 0.15, and the proportion of unobservable phase

    was typically less the 30%. Using these values it may be

    shown that the corrected phase proportions are

    bounded by an error of 10 mol%.

    References

    [1] R.B. Schwarz, S.R. Srinivasan, J.J. Petrovic, C.J. Maggiore,

    Mater. Sci. Eng. A155 (1992) 75.

    [2] A.J. Heron, G.B. Schaffer, J. Mater. Synth. Proc. 2 (1994) 335.

    [3] B.K. Yen, T. Aizawa, J. Kihara, Mater. Sci. Eng. A220 (1996) 8.

    [4] B.K. Yen, T. Aizawa, J. Kihara, J. Am. Ceram. Soc. 79 (1996)2221.

    [5] E. Ma, J. Pagan, G. Cranford, M. Atzmon, J. Mater. Res. 8

    (1993) 1836.

    [6] S.N. Patankar, S.-Q. Xiao, J.J. Lewandowski, A.H. Heuer, J.

    Mater. Res. 8 (1993) 1311.

    [7] D.L. Zhang, J. Mater. Sci. Lett. 14 (1995) 1508.

    [8] L. Liu, F. Padella, W. Guo, M. Magini, Acta Metall. Mater. 43

    (1995) 3755.

    [9] A. Costa e Silva, M.J. Kaufman, Mater. Sci. Eng. A195 (1995) 75.

    [10] A. Costa e Silva, M.J. Kaufman, Scripta Metall. Mater. 29 (1993)

    1141.

    [11] J.S. Jayashankar, E.N. Ross, P.D. Eason, M.J. Kaufman, Mater.

    Sci. Eng. A239/240 (1997) 485.

    [12] M. Fu, J.A. Sekhar, J. Am. Ceram. Soc. 81 (1998) 3205.

    [13] C. Brukl, H. Nowotny, F. Benesovsky, Monatsh. Chem. 92 (1961)967.

    [14] K. Yanigahara, T. Maruyama, K. Nagata, Metall. Mater. Trans.

    12 (1993) 1200.

    [15] A. Arvantis, S. Diplas, P. Tsakiropoulos, J.F. Watts, M.J.

    Whiting, S.A. Morton, J.A.D. Matthew, Acta Mater. 49 (2001)

    1063.

    [16] W.J. Boettinger, J.H. Perepezko, P.S. Frankwicz, Mater. Sci. Eng.

    A155 (1992) 33.

    [17] H. Nowotny, H. Kudielka, E. Parthe, Plansee Proceedings, 1955,

    p. 166.

    [18] A.J. Heron, G.B. Schaffer, Mater. Sci. Eng. A, in press.

    [19] C.J. Howard, R.J. Hill, AAEC Report no. M112, 1986.

    [20] R.A. Young (Ed.), The Rietveld Method, Oxford University

    Press, 1993.

    A.J. Heron, G.B. Schaffer / Materials Science and Engineering A352 (2003) 105 /111 111