Vladimir V. Kulish Mem. ASME e-mail: mvvkulish@ntu.edu.sg School of Mechanical & Production Engineering, Nanyang Technological University, 50 Nanyang Ave., Singapore 639798 Vasily B. Novozhilov Assoc. Mem. ASME Faculty of Engineering, University of Ulster, Shore Rd., Newtownabbey, Co. Antrim, BT37 0QB, United Kingdom An Integral Equation for the Dual-Lag Model of Heat Transfer An integral relation is obtained between local temperature and local temperature gradient for the dual-lag model of heat transport, which substitutes classical Fourier law at short time scales. Both the heat flux lag and the temperature gradient lag are considered, however, it is shown that only difference between the two affects temperature profile. Being applied at exposed surface of material, the integral equation predicts surface temperature variation for any form of imposed heat flux. The solution is tested considering solid heating by a picosecond laser impulse. Results are compared with the classical solution of the parabolic heat transfer equation and available experimental data. 关DOI: 10.1115/1.1797034兴 Keywords: Analytical, Heat Transfer, Laser, Thin Films Introduction Heat transport at small scales receives increasing attention due to application in modern electronics 关1兴, 关2兴, and 关3兴. Many commonly used devices, such as personal computers or cellular phones, operate already on nanosecond time scale, and the spatial scales for the energy transport are of the order of a single atom. It has become established that the heat transport equation is different from the classical one at such time scales 关4兴. In particular, one has to modify the Fourier law, which relates heat flux and the temperature gradient. According to Fourier law, heat flux adjusts immediately to the imposed temperature gradient, i.e., there is no relaxation time for the heat flux. At those scales where the Fourier hypothesis fails, one has to take into account the lag between the heat flux and the temperature gradient. In the most consistent way, such model for micro-scale heat transport has been proposed by Tzou 关5兴, who introduced the two lags, namely the heat flux time lag and the temperature gradient time lag. Such a model represents a new type of constitutive relation between the heat flux and the temperature gradient, which supercedes the Fourier law at small scales. Being combined with the energy conservation law, such relationship leads to the energy transport equation, which is different from the classical parabolic one. Numerical solution of the modified heat transport equation has been attempted in 关5兴, using phase lags as small parameters and obtaining Taylor expansion of the constitutive relationship. The present study generalizes analysis of the one-dimensional homogeneous energy transport equation with the dual phase lag. The integral form of the solution is obtained for the most general form of the constitutive relationship. It is shown that the general formulation, which does not involve approximate Taylor series, in fact leads to quite simple heat transfer equation and integral relationship for the solution. Analysis is restricted to a one-dimensional semi-infinite domain with constant material thermal properties. Note that materials behave as thermally thick in most microheating applications. Further, in many typical cases 共such as the one considered below兲 the temperature rise is very small; therefore the constant material properties assumption is an excellent approximation. The energy transport equation is reduced to the form, closely reminiscent of the classical heat transfer equation, but with the time lag. Integral equation, which relates local values of the temperature and its gradient at any point inside the domain, is Contributed by the Heat Transfer Division for publication in the JOURNAL OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division October 22, 2002; revision received June 3, 2004. Associate Editor: G. S. Dulikravich. Journal of Heat Transfer obtained. This equation is then applied for prediction of the surface temperature under heating conditions, and the results are compared with the available experimental data. Mathematical Model and Preliminary Analysis Consider an infinite solid, ⫺⬁⬍x⬍⫹⬁. According to the dual-phase-lag model proposed by Tzou 关5兴, for any ⫺⬁⬍x ⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ there is a lag between the local temperature gradient, T/ x, and the local heat flux, q ⬙ , q ⬙ 共 x,t⫹ q 兲 ⫽⫺k• T 共 x,t⫹ T 兲 ; x ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (1) Here q is the phase lag of the heat flux, T is the temperature gradient phase lag, and k is the thermal conductivity of the material. This is in contrast to Fourier’s law, where heat flux adjusts immediately to the imposed temperature gradient q ⬙ 共 x,t 兲 ⫽⫺k• T 共 x,t 兲 ; x ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (2) Constitutive relation 共1兲 results in the new form of energy conservation equation, which in general form is written as T 1 q ⬙ 共 x,t 兲 ; • 共 x,t 兲 ⫽⫺ t cp x ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (3) The nature of the modified equation can be established if one expands the left and right hand sides of Eq. 共1兲 to get q ⬙⫹ q 冋 册 q⬙ T 2T ⫹o 共 q 兲 ⫽⫺k ⫹T ⫹o 共 T 兲 ; t x xt ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (4) Upon substitution into the conservation law 共3兲, the following heat transport equation is obtained 关2兴 q 2T 1 T 2T 3T ⫹o 共 q 兲 ⫹o 共 T 兲 ; ⫽ ⫹ ⫹ T ␣ t2 ␣ t x2 x 2 t ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (5) for the temperature field, T(x,t), in the domain. It is clear from Eq. 共5兲 that the lagging behavior should be taken into account for the processes whose characteristic time scales are comparable to q , T . Copyright © 2004 by ASME OCTOBER 2004, Vol. 126 Õ 805 Different approaches may be taken to solve the heat transport equation with the lags numerically. Tzou 关5兴, for instance, considers the truncated Eq. 共5兲 and then obtains a solution of this equation in the Laplace space. The solution thus obtained cannot be inverted. Therefore, approximate methods, such as partial expansions and the Riemann–Sum approximation, have to be applied. Although the numerical results agree quite satisfactory with most of the experimental results, the exact solution is believed hard—if at all possible—to obtain. In fact, truncation of the constitutive relation 共1兲 is not necessary. Even involving conventional finite volume methods, one can easily solve the conservation equation 共3兲 with the general constitutive relation 共1兲. Moreover, in the next section we demonstrate that a simple integral equation can be obtained involving the constitutive relation 共1兲. Therefore, the problem is considered here in the most general form, involving the both lags and untruncated constitutive relation. Note that Eq. 共1兲 can be written in equivalent form as q ⬙ 共 x,t 兲 ⫽⫺k• It is obvious now that the temperature and its gradient are related locally. Taking the derivative of Eq. 共12兲 d⌰ 共 x,s 兲 ⫽⫺C 1 共 s 兲 • dx ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ (13) and comparing to Eq. 共12兲 itself, ⌰ 共 x,s 兲 ⫽⫺ L ⫺1 ⫺ T 2T 共 x,t 兲 ⫽ ␣ 2 共 x,t⫺ 共 q ⫺ T 兲兲 ; t x 冉冑 冉 ⌬ ␣ •exp ⫺ •s s 2 冑 冕 ␣ • t 0 冉 u t *⫺ 冑 T 共 x,t 兲 ⫽T 0 ⫺ T 共 x,t 兲 ⫽T 0 ⫹ Consider now the problem for semi-infinite solid t苸 关 ⌬ ;⬁ 关 ;x苸 关 0;⬁ 关 (8) This domain is initially in thermal equilibrium. Therefore, the initial and boundary conditions are written as 共 ⬁,t 兲 ⫽0 (9) The condition at x⫽0 depends on particular specified rate of solid heating given by the flux q ⬙ (0,t), and is a direct consequence of constitutive relation 共6兲. Here the excess temperature is introduced as ⫽T⫺T 0 , and ⌬ ⫽ q ⫺ T is the difference between the lags. Differential equations with lag are readily handled by Laplace transform. Taking transform of Eq. 共8兲 with respect to t, L( (x,t)) ⫽⌰(x,s), and involving initial condition, one gets the ODE dx s •exp共 ⌬ •s 兲 •⌰⫽0 ␣ (10) with the general solution ⌰ 共 x,s 兲 ⫽C 1 共 s 兲 e ⫺ 冑s/ ␣ exp共 ⌬ •s 兲 •x ⫹C 2 共 s 兲 e 冑s/ ␣ exp共 ⌬ •s 兲 •x The latter must be bounded at x→⬁, therefore, C 2 (s)⬅0, and the solution is written as 806 Õ Vol. 126, OCTOBER 2004 冊 • 冉 冊 冊冊 冑 冑 ␣ • ⫽ u t⫺ ⌬ 2 t⫺ ⌬ 2 共 x,t⫺t * 兲 dt * ⫽ 共 x,t 兲 x 冑 冕 ␣ • (12) T 共 x, 兲 x t⫺⌬ /2 冑* 0 t ⫺ 1 冑 k c p • 冕 t⫺⌬ /2 ⌬ ⫺ 2 •d q ⬙ 共 x, ⫹⌬ 兲 冑 0 (15) (16) (17) ⌬ t *⫺ ⫺ 2 •d (18) Being applied at the surface (x⫽0), Eq. 共18兲 provides the means to predict behavior of the surface temperature of the material, T s (t)⫽T(0,t), based on the imposed flux q s⬙ (t)⫽q ⬙ (t,0), T s 共 t 兲 ⫽T 0 ⫹ 1 冑 k c p • 冕 t⫺⌬ /2 0 q s⬙ 共 ⫹⌬ 兲 冑 t *⫺ ⌬ ⫺ 2 •d (19) Surface temperature is the most important parameter in microscale heat transfer applications. For the case q ⫽ T ⫽0, the above equation reduces to integral relation obtained for the classical heat transfer equation by Kulish and Lage 关7兴. Model Validation A test solution is obtained by integrating Eq. 共19兲, in order to predict surface temperature of the material, subjected to prescribed heat flux. The heat flux applied at the boundary imitates laser impulse according to Gaussian distribution in time, namely 冋 冉 冊册 ⬙ exp ⫺ q s⬙ 共 t 兲 ⫽q s,max (11) ⌰ 共 x,s 兲 ⫽C 1 共 s 兲 e ⫺ 冑s/ ␣ exp共 ⌬ •s 兲 •x (14) Integral equation 共17兲 relates values of temperature and its gradient at the same point, and holds everywhere inside the domain. In terms of the heat flux, Eq. 共17兲 can be written as Integral Equation ⫺ 2 ⌬ 2 ⌬ t *⫺ 2 (7) Equation 共7兲 has the same form as the classical heat transfer equation, but with the lag q ⫺ T in the diffusion term. An immediate and important conclusion from Eq. 共7兲 is that solution in the dual-lag model does not depend on q and T separately, but on the difference q ⫺ T only. d 2⌰ 冊 ⌬ d⌰ 共 x,s 兲 ␣ •exp ⫺ •s • s 2 dx or, after simple re-arrangement, t⫺t * ⫽ , ⫺⬁⬍x⬍⫹⬁, ⫺⬁⬍t⬍⫹⬁ 1 共 0,t⫺⌬ 兲 ⫽⫺ q ⬙ 共 0,t 兲 , x k 冉 冑 The inverse Laplace transform of Eq. 共14兲 can be found involving the table transform 关6兴 (6) Substitution of 共6兲 into the conservation equation 共3兲 yields 共 x,0兲 ⫽0, s exp共 ⌬ •s 兲 •e ⫺ 冑s/ ␣ exp共 ⌬ •s 兲 •x ␣ where u(t) is the unit step-function, and applying the convolution theorem 关6兴. This results in the following integral equation: T 共 x,t⫺ 共 q ⫺ T 兲兲 ; x 2 共 x,t 兲 ⫽ ␣ 2 共 x,t⫺⌬ 兲 ; t x 冑 t⫺b 2 ; t⭓0 (20) with b⫽10 ps and ⫽5.0 ps 共Fig. 1兲. Characteristic time of the considered heating process is, therefore, of the order of picoseconds. In the course of computations, the maximum value of the ⬙ , was 103 W/m2 . surface heat flux, q s,max Transactions of the ASME Fig. 1 Laser impulse for the test case Material properties are taken as those of gold 共Au兲, i.e., k ⫽315 W/共m K), ␣ ⫽1.2495⫻10⫺4 m2 /s, q ⫽0.7438 ps and T ⫽89.286 ps 关5兴. The time lags cited in 关5兴 have been determined experimentally. For a given heat flux history, integration of Eq. 共19兲 is straightforward, and is carried out by explicit time advancement using trapezoidal rule for integration. In order to handle singularity, analytical integration is performed in the vicinity of the upper limit. The time step is chosen in such a way that the relative error of the solution does not exceed ⫽10⫺6 . Upon achieving this criterion, the solution also becomes independent of further reduction in time step. The evolution of the normalized surface temperature, ⫽(T s ⫺T 0 )/(T m ⫺T 0 , is shown in Fig. 2. Also shown in this plot is the profile for the classical heat transfer equation ( q ⫽ T ⫽0). It is obvious from Fig. 2 that the classical solution deviates significantly from solution of Eq. 共19兲 at time scales comparable with q , T , and, therefore, cannot be used at such scales. In particular, the maximum surface temperature is predicted at later time by the classical equation, and the shape of the temperature profile is quite different from the solution of Eq. 共19兲. An interesting mathematical exception is the case q ⫽ T , where the solution of Eq. 共19兲 coincides with that of the classical equation. However, such a case is unlikely for real materials, since the lags seem to differ significantly 共see above example of gold兲. Note also that for real materials T ⬎ q so that the upper limit in the integral 共Eqs. 共17兲–共19兲兲 is always positive. The lags q , T are material properties, which are known, however, with limited accuracy. It is worthwhile, therefore, to investigate sensitivity of the solution to these parameters. Plotted in Fig. 3 is a family of solutions that are obtained for different values of ⌬, which deviate ⫾10% from the base values given in 关7兴. The sensitivity is small, compared with difference between the classical and dual-lag solutions. From Fig. 3, one can see that the less value of the lag is, the closer the solution is to the classical model 共no lag兲. Fig. 2 Normalized surface temperature obtained for a bulk sample of gold „Au… in the case of the dual „B… phase lag, and classical energy equation „D… Journal of Heat Transfer Fig. 3 Time evolution of the surface temperature for different values of the phase lags: „B… ⌬ÄÀ88.5422 ps „ q Ä0.7438 ps, T Ä89.286 ps…—base case; „F… ⌬ÄÀ97.3964 ps; „D… ⌬ ÄÀ79.6880 ps Fig. 4 Comparison between the numerical solution and experimental data Finally, Fig. 4 shows the numerical solution of Eq. 共19兲 compared with the available experimental data by Brorson et al. 关8兴 and Qiu et al. 关9兴. Direct comparison of the maximum temperature with the measurements was not possible, since experimental data 关8兴 and 关9兴 are presented in normalized way. The maximum excess temperature obtained in the present calculations was ⌬T max⬇0.06 K. Conclusions The heat transport equation with the dual-phase lag constitutive relation is considered. In contrast to previous studies, general formulation of the constitutive relationship is retained in the analysis. It is shown that such a general formulation results, in fact, in much simpler solution of the problem. Heat transfer equation with the dual lag is reduced to the form, similar to the classical parabolic equation. The solution of such equation depends only on the difference between the two lags. Therefore, the lagging behavior of the particular material is, in fact, described by the single parameter. The method of Laplace transform has been used to obtain relationship between the local temperature and the local heat flux for the considered model. Such a relationship is written in the form of a simple integral equation. Being applied at the surface, the integral equation provides a convenient way to predict maximum 共surface兲 temperature during heating. The integral equation has been tested for the case of material heating by a picosecond laser impulse. The result has been compared with the solution of the classical 共parabolic兲 heat transfer equation and with the experimental data available in literature. In case of zero lags 共or for time scales much longer than time lags兲, the proposed integral equation reduces to the previously established integral relation 关3兴 and 关7兴, applicable to problems, described by classical Fourier law. OCTOBER 2004, Vol. 126 Õ 807 Nomenclature b cp k L q⬙ q s⬙ s t t* T T0 Ts Tm u(t) x ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ mean of normal distribution, s specific heat of solid material thermal conductivity, W m⫺1 K⫺1 Laplace transform operator local heat flux, W m⫺2 surface heat flux 共at x⫽0), W m⫺2 Laplace transform variable time, s dummy time variable, s absolute temperature, K initial temperature 共at t⭐0), K surface temperature 共at x⫽0), K maximum temperature during heating, K unit step-function co-ordinate normal to material surface, m Greek Symbols thermal diffusivity, m2 s⫺1 excess temperature, ⫽T⫺T 0 , K Laplace transform of normalized non-dimensional temperature, ⫽(T s ⫺T 0 )/(T m ⫺T 0 ) ⫽ variance of normal distribution, s ⫽ density, kg m⫺3 q ⫽ phase lag of heat flux, s ␣ ⌰ ⫽ ⫽ ⫽ ⫽ 808 Õ Vol. 126, OCTOBER 2004 T ⫽ temperature gradient phase lag, s ⌬ ⫽ difference between lags, ⫽ q ⫺ T , s References 关1兴 Brorson, S. D., Kazeroonian, A., Moodera, J. S., Face, D. W., Cheng, T. K., Ippen, E. P., Dresselhaus, M. S., and Dresselhaus, G., 1990, ‘‘Femtosecond Room Temperature Measurement of the Electron-Proton Coupling Constant in Metallic Superconductors,’’ Phys. Rev. Lett., 64, pp. 2172–2175. 关2兴 Qiu, T. Q., and Tien, C. L., 1994, ‘‘Femtosecond Laser Heating of Multilayered Metals—I. Analysis,’’ ASME J. Heat Transfer, 37, pp. 2789–2797. 关3兴 Kulish, V. V., Lage, J. L., Komarov, P. L., and Raad, P. E., 2001, ‘‘FractionalDiffusion Theory for Calculating Thermal Properties of Thin Films From Surface Transient Thermoreflectance Measurements,’’ ASME J. Heat Transfer, 123共6兲, pp. 1133–1138. 关4兴 Cattaneo, C., 1958, ‘‘A Form of Heat Conduction Equation Which Eliminates the Paradox of Instantaneous Propagation,’’ Compte Rendus, 247, pp. 431– 433. 关5兴 Tzou, D. Y., 1997, Macro to Microscale Heat Transfer: The Lagging Behavior, Taylor & Francis. 关6兴 Abramowicz, M., and Stegun, I. A., 1964, Handbook of Mathematical Functions: With Formulas, Graphs, and Mathematical Tables, Dover, New York. 关7兴 Kulish, V. V., and Lage, J. L., 2000, ‘‘Fractional-Diffusion Solutions for Transient Local Temperature and Heat Flux,’’ ASME J. Heat Transfer, 122, pp. 372–376. 关8兴 Brorson, S. D., Fujimoto, J. G., and Ippen, E. P., 1987, ‘‘Femtosecond Electron Heat-Transport Dynamics in Thin Gold Film,’’ Phys. Rev. Lett., 59, pp. 1962– 1965. 关9兴 Qiu, T. Q., Juhasz, T., Suarez, C., Bron, W. E., and Tien, C. L., 1994, ‘‘Femtosecond Laser Heating of Multilayered Metals—II. Experiments,’’ Int. J. Heat Mass Transfer, 37, pp. 2799–2808. Transactions of the ASME