You are on page 1of 4

Scripta Materialia 55 (2006) 11471150 www.actamat-journals.

com

Activity coecients for dilute solid solutions of Al in Ni


Yong Jiang,a John R. Smitha,* and Anthony G. Evansb
a b

Delphi Research Labs, Shelby Township, MI 48315, United States University of California, Santa Barbara, CA 93106, United States

Received 2 August 2006; revised 9 August 2006; accepted 10 August 2006 Available online 20 September 2006

We present a method for predicting activity coecients in dilute binary solid solutions from rst principles and apply it to Al in Ni. Accurate results require knowledge of the temperature dependencies of both enthalpy and entropy: with the latter being dominant. The activity coecient of Al varies by 15 orders of magnitude over 400 K 6 T 6 1700 K: whereas that for Ni only varies by <4%. Available experimental data are consistent with these results. 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Nickel alloys; Density functional; Thermodynamics; First-principle electron theory; Activity

For the prediction and interpretation of hightemperature diusion and related phenomena in solids, knowledge of thermodynamic activities is essential [1]. For example, the properties of Ni (Al)-based alloys used in hot section components (for power generation, aircraft propulsion and automotive catalytic converter supports) depend sensitively on elemental activities [2,3]. For dilute solid solutions, however, activity measurements are dicult and therefore sparse [48]. Here, we introduce a rst-principles methodology applicable to dilute solid solutions, and apply it to Al in a Ni host. The only similar computation known to the authors, that for various solutes in Si [9], did not consider the vibrational and electronic contributions to the solution energies. A principal thermodynamic variable is the chemical potential l, related to the activity aAl (expressed for Al in Ni) by ln aAl T ; x lAl T ; x l0 T =kT : 1 Al Note that aAl depends on the dierence between the chemical potential lAl of Al in the solution Ni1xAlx (x is the atomic per cent of Al) and that in pure, crystalline Al, l0 . An alternative form, written in terms of the Al ambient partial pressure p(T, x) for Al in solution and p0(T) for pure, crystalline Al: aAl T ; x pT ; x=p0 T ; 2

provides a means to measure the activity. The Al activity coecient cAl is dened as cAl T ; x aAl T ; x=x: 3 When the host metal provides stronger atomic bonds than pure Al, aAl < 1. This is the situation for Al in Ni, since Ni has a higher melting temperature and cohesive energy. Note that, as T ! 0, aAl ! 0 and cAl ! 0. Also, as x ! 0, cAl(T, x) ! cAl(T) (Henrys law), aNi(T, x) ! 1, and cNi(T, x) ! 1 (Raoults law). To compute aAl in dilute Ni solution we must nd lAl T ; x l0 T DH T T DST ; Al 4 where DH and DS are, respectively, the dierences in enthalpy and entropy per Al atom in solution and in pure Al. For one Al atom in a Ni host, DS conf k ln x: Combining Eqs. (1) and (3)(5), cAl exp DH T T DS n-c T =kT ; 6 where DSn-c is the non-congurational entropy dierence (or excess entropy), consisting of vibrational and thermal electronic contributions. In order to determine the temperature dependences of DH and DSn-c, we next determine the temperature dependence of the solid equilibrium volume V*(T) by minimizing the Gibbs free energy [11,12] with respect to V at each T: GT ; V H 0K V Gvib T ; V Gconf T ; V Gel T ; V ; 5

* Corresponding author. E-mail: john.r1.smith@delphi.com 1359-6462/$ - see front matter 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.scriptamat.2006.08.037

1148

Y. Jiang et al. / Scripta Materialia 55 (2006) 11471150

where G(T, V) is computed from rst principles. To be more specic, 0 K total energy and relaxation calculations are conducted for a series of volumes V. Then, the vibrational free energy and corresponding entropy and enthalpy are computed at each V by (   1 ) Y htq;a 2 sinh Gvib T k B T ln 2K B T q;a X 1X 8a htq;a k B T ln 1 ehtq;a =K B T 2 q;a q;a S vib T and H vib T Gvib T TS vib T : Next the congurational Gibbs energy, X Gconf T kT xi ln xi ;
i

oGvib T jV ; oT

8b

8c

can be computed for the solution (zero for the pure metals), as well as the thermal electronic component to the Gibbs free energy Gel(T, V) due to thermal excitation [11]. Finally, the equilibrium volume V*(T) is obtained by minimizing G(T, V) with respect to V at each T. Repeating the above calculations with V*(T), we obtain the enthalpy and entropy of components and the solution at each T, and further the temperature dependences of DH and DSn-c in Eq. (6). First principles solutions of the KohnSham equations [13] are obtained using the density-functional theory (DFT) code VASP [14] within the generalized gradient approximation (GGA) [15] for exchange-correlation, using ultra-soft pseudo-potentials [16]. To apply our method to the dilute solution Ni1xAlx, we create x $ 0.03 by inserting one substitutional Al atom in a (2 2 2) face-centered cubic (fcc) Ni lattice. This solution is suciently dilute that cAl(T, x) ! cAl(T). The ground-state structure of the alloy lattice is optimized until the total force on each ion converges within 0.02 eV/A. To ensure the maximum cancellation of numerical errors, all relevant energetics for the pure Ni and the Ni.97Al.03, solution are calculated using the same (2 2 2) fcc supercell and computational parameters, within the quasiharmonic approximation [17] (as described above). The plane-wave cuto energy is set as 400 eV, and Brillouin zone (BZ) integrations are performed using 10 irreducible k-points. Phonon frequencies are calculated using the direct method [18] implemented in VASP at wave vector, q = 0 (C point). For simplication, the electronphonon interactions and magnetic contributions are neglected [19]. Such a treatment has been successfully used to predict the bulk thermal properties of various metals and alloys including Al, Li, Na, Cu, Ag, Ni, NiAl, and Ni3Al [11,12,2022,3]. To obtain equilibrium lattices of Ni.97Al.03 at elevated temperatures, the supercell is relaxed statically at 0 K, and assumed to have the same thermal expansion as pure Ni. To demonstrate delity, the equilibrium lattice constant a0 and bulk modulus B0 of Ni at 300 K are obtained and compared with experiment. The calcula-

tions yield a lattice constant, 3.540 A, which compares with measured values of 3.52 A [23] and 3.524 A [24]. The calculated bulk modulus, 186.2 GPa, compares with an experimental value of 186 GPa [23]. Over the entire temperature range, our computed values of the Ni lattice constant agree with measurements [25,26] within $1%. Additionally, we have compared computed Ni enthalpies and entropies with experiment [27] (Fig. 1). (As the absolute enthalpy has no real meaning, we assign the zero enthalpy reference at 0 K.) Evidently we must include quasiharmonic vibrations and thermal electronic excitations in order to obtain good agreement for Ni at all temperatures. The harmonic and the simple quasiharmonic calculations without thermal electronic excitations both underestimate the enthalpies and entropies, because the relatively high electronic density of states across the Fermi level causes the temperature-dependent contributions to become signicant at high temperatures. The isolated Al substitutional site in Ni is much more energy favorable than the interstitial site (by 4.40 eV at 0 K), and thus assumed for the overall temperature range throughout this study. We note that Al has a relatively low melting temperature, Tm = 933 K. Due to the diculty of simulating a molten phase using rst-principles DFT, we choose the enthalpy and entropy data for pure Al from the JANAF tables [27] for all temperatures. The resultant energies of solution, DH(T) DH(0) and TDSn-c (=TDSvib TDSel), (Fig. 2), indicate that both vibrational and thermal electronic excitations contribute importantly to the enthalpy and entropy. DH decreases by $0.36 eV over 0 6 T 6 1700 K, while TDSn-c increases by $0.93 eV. For T < 933 K, the slope of TDSn-c is about twice that of DH (consistent with that found for Si in Al [28]). But, it is about 56 times larger when T > 933 K, mainly due to the rapidly increasing entropies of liquid Al. The discontinuity at T = 933 K corresponds with the experimentally-determined latent energy of melting for Al. The dierent eects of DH(T) and TDSn-c(T) on the Al activity coecient are identied and compared in Figure 3. Note that due to partial cancellation between

Enthalpy, H
0.4

0.0

Energies (eV)

-0.4

-0.8

Q-harm+el
-1.2

Q-harmonic Harmonic Experiment

Entropy, -TS

-1.6 0 400 800 1200 1600

Temperature (K)

Figure 1. Calculated enthalpies and entropies for Ni using the harmonic approximation, and the quasiharmonic approximation with and without thermal electronic excitations, in comparison with the experimental values [27].

Y. Jiang et al. / Scripta Materialia 55 (2006) 11471150


1.0

1149

Q-harm + el
0.8

-T S n- c

Q-harmonic

0.6 0.4 0.2 0.0 -0.2 -0.4


933 K

H(T)- H(0)

-0.6

400

800

1200

1600

Temperature (K)

Figure 2. Values for the enthalpy of solution, DH(T) DH(0), and the non-congurational entropies of solution (TDSvib TDSel) calculated using the quasiharmonic approximation, with and without thermal electronic excitations. The steps at 933 K indicate that the reference state of Al is changed at its melting point from fcc to liquid.

the temperature dependencies of DH and TDSn-c, the enthalpy change at 0 K, DH(T = 0), provides reasonable values of cAl up to $600 K (the dashed curve). The Al activity coecient cAl varies by 15 orders of magnitude over 400 K 6 T 6 1700 K. This strong temperature dependence can be understood by inspecting Eq. (6). Namely, it depends exponentially on (DH DSn-c)kT: which, in turn, is large and negative, due to the fact that AlNi bonds are signicantly stronger than AlAl bonds. Comparison with available (albeit sparse) experimental data [48] shows an acceptable agreement (Fig. 4a), except for the data point at 980 K [5]. The implication is that the experimentally-ascertained Al activity at 980 K [5] is erroneously low, as also surmised in another assessment [4]. The activity of the host Ni, aNi, is quite dierent from aAl, because the Ni neighbors are predominantly Ni atoms, similar to the case for pure Ni. Thus one would expect cNi ! 1 and aNi ! xNi (Raoults law). Through the GibbsDuhem [10] relation, we can simply correlate the two component activities as xd ln aAl 1 xd ln aNi 0: which for x ! 0, can be approximated by 10a

Energy Difference (eV)

Ni 0.97Al 0.03
10-2

ln cNi x=1 x ln cAl :

10b

10-6

10-10

10-14

10-18 400

by H(T=0) only by H(T=0) & Sn-c(T) by H(T) by H(T) & Sn-c(T)


800 1200 1600

Temperature (K)

Figure 3. Calculated activity coecient of Al in Ni.97Al.03 using various approaches to Eq. (6). Here DH(T) and DSn-c(T) are calculated in the quasiharmonic approximation with thermal electronic excitations.

Figure 4b shows that as expected the host Ni activity coecients are close to 1 and in fact they only vary by less than 4% over the same temperature range. We have reported a rst-principles method for predicting activity coecients in dilute binary solid solutions and applied it to Al in a Ni host. Calculations show that the 0-K enthalpy provides reasonable Al activity values up to $600 K, due to partial cancellation between temperature dependencies of enthalpies and non-congurational entropies of solution. However, at higher temperatures, the temperature-dependent entropy term dominates. That is, for T > 600 K, temperature-dependent entropies and enthalpies must be included in the activity computations. Over the range 400 K 6 T 6 1700 K, the Al activity coecient varies by 15 orders of magnitude while the Ni activity coecient

Al Activity Coefficient, Al

Ni0.97 Al0.03
10
-2

a
Ni Activity, Activity Coefficient

1.00

Ni0.97 Al0.03
0.98

Al Activity Coefficient, Al

Calculated
10-6

Activity Coefficient Ni
0.96

10-10

Activity aNi
0.94

Experiments
3% Al, 980 K
10-14

15% Al, 1300 K 10% Al, 1423 K 9 & 6% Al, 1600 K

0.92

Experiments

10-18 400

1% Al, 1650 K
800 1200 1600

0.90 400

b
800 1200

Ni aNi
1600

Temperature (K)

Temperature (K)

Figure 4. (a) Calculated activity coecient of Al in Ni.97Al.03 in comparison with the available experimental measurements: r [5], [6], h [7], n [4], and s [8]. (b) Calculated values of the activity and activity coecient for Ni in Ni.97Al.03 in comparison with the experimental results (by interpolating using experimental 5% Al at 1600 K [4]).

1150

Y. Jiang et al. / Scripta Materialia 55 (2006) 11471150

varies by only $4%. This strong temperature dependence of the former arises mainly because it depends exponentially on the (relatively large) enthalpy of solution of Al in Ni. The authors thank Dr. Xiao-gang Wang for many fruitful discussions and gratefully acknowledge AFOSR support from grant FA9550-05-C-0039.
[1] R. Arroyave, T. Eagar, Acta Mater. 51 (2003) 4871. [2] W. Zhang, J.R. Smith, A.G. Evans, Acta Mater. 50 (2002) 3803. [3] A.Y. Lozovoi, Y. Mishin, Phys. Rev. B 68 (2003) 184113. [4] K. Hilpert, M. Miller, H. Gerads, H. Nickel, Ber. Bunsen. Phys. Chem. 94 (1990) 40. [5] V.M. Eskov, V.V. Samokhval, A.A. Vecher, Russ. Metall. 2 (1974) 118. [6] A. Steiner, K.L. Kormarek, Trans. Metall. Soc. AIME 230 (1964) 786. [7] N.C. Oforka, Indian J. Chem. 25A (1986) 1027. [8] V. Merlin, N. Eustathopoulos, J. Mater. Sci. 30 (1995) 3619. [9] K. Iwata, T. Matumiya, H. Sawada, K. Kawakami, Acta Mater. 51 (2003) 551. [10] D. Gaskell, Introduction to Metallurgical Thermodynamics, 2nd ed., Hemisphere, New York, NY, 1981. [11] Y. Wang, Z.-K. Liu, L.-Q. Chen, Acta Mater. 52 (2004) 2665. [12] R. Arroyave, D. Shin, Z.-K. Liu, Acta Mater. 53 (2005) 1809. [13] W. Kohn, L. Sham, Phys. Rev. A 140 (1965) 1133. [14] G. Kresse, J. Furthmuller. Available from: <http:// cms.mpi.univie.ac.at/vasp/vasp/vasp.html>.

[15] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244. [16] D. Vanderbilt, Phys. Rev. B 41 (1990) 7892. [17] A.A. Maradudin, E.W. Montroll, G.H. Weiss, I.P. Ipatova, Theory of Lattice Dynamics in the Harmonic Approximation, 2nd ed., Academic Press, New York, NY, 1971, pp. 68188. [18] K. Parlinski, in: M.R. Johnsn, G.J. Kearley, H.G. Buttner (Eds.), Neutrons and Numerical MethodsN2M, Proceedings of the Workshop on Neutrons and Numerical Methods (AIP Conf. Proc. 479, Grenoble, France, 1998), AIP, Woodbury, NY, 1999, p. 121. [19] The ferromagneticparamagnetic (Curie) point transition of Ni is at Tc = 631 K, located fairly low in the temperature range of interest: 400 K 6 T 6 1700 K. [20] A.A. Quong, A.Y. Liu, Phys. Rev. B 56 (1997) 7767. [21] S. Narasimhan, S. de Gironcoli, Phys. Rev. B 65 (2002) 064302. [22] J. Xie, S. de Gironcoli, S. Baroni, M. Scheer, Phys. Rev. B 59 (1999) 965. [23] C. Kittel, Introduction to Solid State Physics, 6th ed., Wiley, New York, NY, 1986. [24] M. Ellner, K. Kolatschek, B. Predel, J. Less-Common Met. 170 (1991) 171. [25] J. Bttiger, N. Karpe, J.P. Krog, A.V. Ruban, J. Mater. Res. 13 (1998) 1717. [26] Y.S. Touloukian, C.Y. Ho, Properties of Selected Ferrous Alloying Elements, McGraw-Hill, New York, NY, 1981. [27] M.W. Chase Jr., C.A. Davies, J.R. Downey Jr., D.J. Frurip, R.A. McDonal, A.N. Syverud, JANAF Thermochemical Tables (J. Phys. Chem. Ref. Data 14 (Suppl. 1) (1985)), American Chemical Society and American Institute of Physics, New York, NY, 1985 (Part I and II). [28] V. Ozolins, B. Sadigh, M. Asta, J. Phys.: Condens. Matter 17 (2005) 2197.

You might also like