1. Introduction
The increasingly miniaturization and integration require, for the modeling engineers and manufacturers of integrated circuits, to improve the behavioural models of the electronic components. Among the latter, semiconductor resistors are used in all components either as passive elements of an electronic circuit or as a sensor (pressure, temperature, chemical species, etc.). In all cases, and in order to preserve the integrity of the useful signal, their sensitivity to the influence parameters must be sufficiently well-known and controlled, either in an empirical way, or by analytical laws. Whatever the purpose of their use or their operating mode, the thermal resistance behaviour must be apprehended in the most accurate way. In a previous study [1], we showed that the thermal drift of a silicon resistor doped with acceptors atoms could be described by an equation of the second order and that the two temperature coefficients (TCRs) have a strong dependence with the doping concentration. Following the same idea, we modelled the influence of the doping concentration on the temperature coefficients of a resistor doped with donor atoms. Knowing that there are differences between the laws controlling the mobility of the electrons and the holes, one can expect that the resistivity of this material shows different behaviour according to the doping type.
2. Relation Resistivity-Mobility
The expression of a silicon resistance according to the temperature T is given by
(1)
where is a reference temperature; α and β are the first and second order TCRs.
Knowing that [1]:
(2)
where ρ and µ are respectively the silicon resistivity and the carriers’ mobility, it would be sufficient to model the carriers’ mobility behaviour according to temperature and doping to obtain the expressions of the two coefficients (TCRs). So far, no analytical model allows us to describe this dependence. On the other hand, there are several empirical models describing the influence of doping and the temperature on the carriers mobility, among which we used those of Arora [2], Klaassen [3] and Dorckel [4] as is shown in the Appendix. In order to evaluate the total carrier mobility (µi for electron and µj for hole), we use the Mathiessen’s rule which approximates it at low longitudinal field as the sum of four term, which are the four contributions to the carrier mobility: Lattice mobility (µi, L); donor mobility (µi, D); acceptor mobility (µi, A) and electron-hole scattering mobility (µi, j), and includes temperature dependence [5,6]:
(3)
Thus, using a limited development of the mobility’s expressions µ = f(ND, T) taken from these references, we used a method of parameters identification. This method is a term by term comparison of the mobility expression as a function of temperature, with Equation (2), for each concentration. Thus, we obtain the expression of α = f(ND) and β = f(ND).
3. Results
The obtained results allowed us to plot the curves in Figures 1 and 2. These figures show that for the three models, there is a strong similarity in the shape of the nonlinear curves α = f(ND) and β = f(ND). Moreover, the shapes of these curves are similar to those published in [1], as shown in Figures 3 and 4; moreover the minimum values appear for a given doping.
The comparison between the results previously published for a P-type resistor (Figures 3 and 4) and those studied in this work (Figures 1 and 2) should allow to quantify their respective thermal drifts as a function of the doping concentration in one hand, and on the other hand, to validate or not the non linear approach of their thermal coefficients (TCRs) in a given temperature range. Thus, by using the minimal values of the coefficients α and β taken from the curves in Figures 1-4, and by their substitution in the Equation (2), we plotted the curves of the relative resistance variation as a function of the temperature variation ΔT for the two types of doping
Figure 3. Variations of the first order thermal coefficient α as a function of doping concentration NA [1]. 0: Bullis [6] αmin = 200 ppm/˚C; 1: Dorckel [4] αmin = 400 ppm/˚C; 2: Arora [2] αmin = 980 ppm/˚C; 3: Klaassen [3] αmin = 1000 ppm/˚C
Figure 4. Variations of the second order thermal coefficient β as a function of doping concentration NA[1].
(Figure 5 for the type P; Figure 6 for the type N). Since the minimum value for different models do not occur at the same doping concentration for the α parameter than for the β one, then we have choose, for a same doping concentration, particular coefficients couples (αmin, β).
These values are used to evaluate the relative resistance change as a function of the temperature variation ΔT. The minimum values of the first and second order thermal coefficients were determined in order to predict the variation of the P-type and the N-type thermal drift resistance, for a given doping concentration.
Figure 5 shows clearly that for a couple αmin(NA) and βmin(NA) for the P-type resistor thermal variations can be considered as linear in a large interval of temperature. Figure 6 shows that the N-type resistor thermal behavior is highly nonlinear, whatever the values of αmin(ND) and βmin(ND). Therefore, these results may help the more complicated circuits to choose the resistors doping con