Simultaneous estimation of heat transfer coefficient and thermal conductivity with application to microelectronic materials
Introduction
Recently, the development of wide range of simulation tools for microelectronic applications has been observed. They support thermal design, testing and optimization of microelectronic systems such as: sensors, integrated microsystems, microdevices, etc. Performance and reliability can also be investigated using several simulation methods [1].
In most cases of thermal calculations, numerical methods are used because of the complexity of geometry and boundary condition of modeled structures. The widely exploited approach involves Finite Element Method (FEM). Many commercially available software packages are commonly used in order to address thermal problems in microelectronics.
In order to solve real cases of thermal problem every simulation tool (either analytical or numerical) needs a description of materials used for structure fabrication, including the thermal properties of those materials. The thermal conductivity which affects level of temperature in the considered microstructure, and has an influence on level of thermal stresses and finally on the reliability of devices [2], is of crucial significance.
Test samples made in thick-film and LTCC (Low Temperature Cofired Ceramic) technology are considered in the presented paper. These technologies are widely applied for power modules fabrication, microelectronic devices packaging, sensors and actuators, and passive integrated components manufacturing [3], [4]. In many of the above mentioned applications the thermal management plays very important role, and thermal modeling is often used in order to investigate module performance or optimization. In all those studies knowledge of thermal properties of the materials as well as heat transfer coefficient is necessary in order to achieve reliable results.
A value of thermal conductivity can be found in data sheets of the material, delivered from the producer, but usually it may significantly vary with the time of storage, conditions of storage, parameters of the processing or manufacturing of the final device. Thus, in order to perform a reliable simulation or modeling the thermal conductivity of the material should be at least estimated based on some experimental data. There are many publications devoted to the thermal conductivity of electronic materials' measurements or estimations. Generally, one can notice that most of them are based on the construction of a dedicated device [5] or equipment [6], [7].
The second important factor for thermal modeling describes the heat exchange between considered device and environment. It is usually called heat transfer coefficient. This parameter is used to characterize the convective and radiative (in some temperature range) way of heat dissipation. It was previously reported that this factor shows strong dependence on the surface dimensions [8]. Although the problem of natural convection of horizontal plates was quite widely investigated there are only few papers devoted to that factor related to microelectronic devices. Thus an effective and simple method of heat transfer coefficient determination related to the microelectronic systems is required.
However, the problem of measurement of thermal conductivity and heat transfer factor has been previously exploited but there are only few papers related to this problem where both of those parameters are estimated in the one procedure [9]. In this paper we propose an alternative approach for simultaneous estimation of thermal conductivity and heat transfer coefficient.
Basically, we translate the above-described problem into a problem of parameter estimation of the heat transfer model. These kinds of inverse heat conduction problems [10] have been considered for a long time and are known to be rather difficult [11]. There exist many methods developed in order to solve such problems. Most of them are iterative algorithms that improve the initial guess of parameter values based on some goal function (e.g. square error function). Problem statement and conditions can significantly influence the convergence and thus the performance of such methods.
An application of repeated least square algorithm (RLSQA) [12] for parameters estimation is discussed in the presented paper. The algorithm is non-iterative and requires only the solution to two systems of linear equations. Convergence of this method is guaranteed, meaning the algorithm is robust and reliable. Generally, the eigenvalues of the heat operator are estimated basing on measurements in the first step of the algorithm. Then, unknown parameters are calculated based on the estimates of eigenvalues.
It is well-known that the location of measurements as well as placement of sources have a big impact on the estimation quality in case of distributed parameter systems. Thus experiment design methods should be applied in this kind of issues [13], [14]. This problem is also discussed in this paper in terms of sensitivity functions.
The whole procedure concerning thick-film and LTCC technology materials is shown in this paper, step by step. We begin with heat transfer descriptions in the test sample, then a basic introduction to the RLSQA method is presented. After that, the algorithm for thermal conductivity and heat transfer coefficient is shown including the exact form of all required data. The practical application of the presented method is given, as well. Measurement equipment used in the experiments is described in detail. The examples concern the alumina (96% Al2O3) and DP951 ceramics widely used in the thick-film and LTCC technologies. Test sample topology as well as the results of thermographic measurements of temperature field on the top surface test samples are shown.
It is known that both examined factors can depend on other physical quantities, especially on temperature. We assume that in the examined range of temperature this dependence is not so big, and the linear model is still valid. Moreover, all measurements are carried out by three different power levels of heat source, and three different mean temperatures of the sample.
Obtained results are fully comparable with previously published. It must be emphasized that application of the presented approach in not limited to the microelectronic materials characterization and can be successfully applied in other fields of the science. We would also like to stress that our approach is a typical estimation procedure, which cannot be treated as a typical measurement, but it can significantly help in preliminary investigations of the examined materials.
Section snippets
Temperature field in the test sample
The examined test sample can be modeled by a plate in thermal steady-state, made of homogenous material with thermal conductivity a1. Due to the small thickness of the sample a two-dimensional model can be applied. If there is heat exchange through the upper and the bottom surface with coefficients and , respectively, temperature field on the plate can be described as follows [9], [15]:where T(x, y)=Tr(x, y)−Ta is the temperature in
Estimation algorithm
One can notice that thermal conductivity as well as heat transfer coefficient measurement is in fact an estimation of unknown parameters of distributed parameter system DPS [14] described by the Eq. (1) based on the measurement of temperature distribution on the top surface of the sample. This kind of task is reported as an ill-conditioned inverse problem of DPS identification. There are many methods dedicated for such systems identification, mostly iterative search algorithms, which
Measurement location
A problem with design of optimal measurement places arises naturally when one considers parameter estimation of DPS. In our particular case the measurement location has impact on matrix V form, and simultaneously on the estimation quality of eigenvalues ξij using least square method. There is a number of methods dedicated to optimization, but one should notice that in our particular problem the measurement places are chosen before the estimation procedure. It means that usual thermographic
Experimental setup
In order to verify the performance and efficiency of the presented approach suitable measurement setup as well as test samples were prepared. The presented experiment concerns an estimation of thermal conductivity and heat transfer coefficient of the samples made of two ceramic materials with different thermal properties. The thick-film technology was used to manufacture conductive paths and heaters on the top surface of the samples. Details of the measurement setup as well as test samples
Numerical comparison of the RLSQA and Levenberg–Marquardt method
In order to compare the performance of described method with other well-known Levenberg–Marquardt method a numerical experiment was done based on simulated data.
Model Eq. (1) was used to simulate temperature field measurements with added normally distributed, zero mean random noise with dispersions equal to σ=1, σ=10 and σ=25 (only data for σ=25 are presented in this paper). These data were used in non-linear identification of parameters performed by RLSQA method (marked by RLSQA),
Summary and conclusions
In the presented paper a new approach for simultaneous estimation of thermal conductivity and heat transfer coefficient was presented. The method is based on the repeated least algorithm and its application to the distributed parameter system identification. In this article exact form of needed data are shown, as well as the details of the procedure.
The applied RLSQA method is non-iterative and its convergence is guaranteed. Most of iterative methods have the ability to get stuck in the local
Acknowledgements
The author wishes to express his thanks to Professor Ewaryst Rafajłowicz for his interest in this work, valuable comments and suggestions. This work was supported by the Polish State Committee for Scientific Research (KBN), Grant No. 4 T11B 010 23.
References (22)
Impact of fem simulation on reliability improvement of packaging
Microelectronics Realibility
(1999)Electrical and structural investigations in reliability characterisation of modern passives and passive integrated components
Microelectronics Realibility
(2002)A linear inverse model for the temperature-dependent thermal conductivity determination in one-dimensional problems
Applied Mathematical Modelling
(1998)- et al.
Practical issues in distributed parameter estimation: gradient computation and optimal experiment design
Control Engineering Practice
(1996) - et al.
Thermal modelling of hybrid circuits: simulation method comparision
Microelectronics Realibility
(2000) - et al.
Thermal characterization of microsystems by means of high-resolution thermography
Microelectronics Journal
(1997) - et al.
Data filtering applied to infrared thermographic measurements intended for the estimation of local heat transfer coefficient
Experimental Thermal and Fluid Science
(2002) - T. Zawada, A. Dziedzic, L.J. Golonka, Temperature and stress field in 3d microvolume resistors, in: Proceedings of the...
Application of thick films in ltcc technology
Informacije MIDEM
(1999)- et al.
Thermal conductivity measurements on thin films based on micromechanical devices
Journal of Micromechanics and Microengineering
(1996)
Cited by (16)
3D-transient identification of surface heat sources through infrared thermography measurements on the rear face
2020, International Journal of Thermal SciencesCitation Excerpt :In Ref. [10], the authors used an inverse matrix inversion method to reconstruct, spatially and in steady state, the heat flux density dissipated on a printed circuit. In Ref. [11], the objective of the study was to estimate the conductivity and the heat transfer coefficient of a body. In the study [12], the authors developed a method (Discrete Boundary Resistance) in order to determine temperature distribution in a semi-analytical way within tri-dimensional solid-state circuits and packaged devices in steady state conditions.
Study of the thermal insulation properties of the glass fiber board used for interior building envelope
2015, Energy and BuildingsCitation Excerpt :The minimum thermal resistance law is suitable for estimating the effective thermal conductivity of the composite material with low fractal dimension. According to the minimum thermal resistance law and the effective thermal conductivity principles, as long as a small unit of the thermal insulation material has the same equivalent thermal resistance with the entire material, their effective thermal conductivities are equal when only considering the thermal conduction [23]. Therefore, we assumed that the glass fiber insulation board consisted of a series of small rectangle units, and each unit contains one glass fiber surrounded by air.
Novel LTCC-potentiometric microfluidic device for biparametric analysis of organic compounds carrying plastic antibodies as ionophores: Application to sulfamethoxazole and trimethoprim
2011, Biosensors and BioelectronicsCitation Excerpt :The LTCC devices offer good electrical and mechanical properties, as well as high reliability and stability. They may integrate in a single unit all steps associated to an analytical procedure, being of special interest to micro-fluidic applications (Golonka et al., 2006; Lopes et al., 2007; Khanna et al., 2005; Patel et al., 2006; Bergstedt and Persson, 2002; Muller et al., 2005; Zawada, 2006) and lab-on-a-chip devices. The integration of μPOT sensors on LTCC platforms has been successfully proven in the past, although only a few works are reported.
The warm-up and offset stability of a low-pressure piezoresistive ceramic pressure sensor
2010, Sensors and Actuators, A: PhysicalLTCC microreactor for urea determination in biological fluids
2009, Sensors and Actuators, B: ChemicalCitation Excerpt :Special micromachining methods are used to produce spatial structures inside LTCC modules: laser cutting, computer numerical control (CNC) machining, using of sacrificial volume materials, acetone etching, and low pressure lamination [7–9]. The most important features of the LTCC technology in respect to microdevices fabrication are the following: 3D structuring, integration of various components in one module, possibility of combining with various materials [10,11,25]. The advantages of LTCC structure, in comparison with silicon one, are as follows: lower price, shorter design and development time and possibility of integration of fluidic channels, heater, sensors and package in one LTCC module [12–15].
Optimal Input Signals for Parameter Estimation: In Linear Systems with Spatio-Temporal Dynamics
2022, Optimal Input Signals for Parameter Estimation: In Linear Systems with Spatio-Temporal Dynamics