With the change in temperature, the amount of free electron changes within the material, in turn, it changes the conductive part and hence the capacitive part. Thus composite of dielectric and conductive material leads to change the constant 

The attached paper can give a clear answer to this query. Moss relation is valid for both elemental and compound semiconductors. Of course, for oxide semiconductors some additional considerations are required.  In fact, dielectric constants of stoichiometric and non-stoichiometric oxides are not identical  and generally the electrical conductivity of the non-stoichiometric oxides is higher. Further study is suggested.

Thank you very much Sir

What we know for all semiconducting materials such as TiO2 is, that the static dielectric constant can be calculated from

- the high-frequency one (summarizing the contribution from the interband transitions in the visible and ultraviolett, see for example S. Shokhovets et al., J. Appl. Phys. 94, 307 (2003); http://dx.doi.org/10.1063/1.1582369 ) and

- from the the frequencies of the transversal-optical (TO) and langitudinal-optical (LO) phonon modes via the Lydanne-Sachs-Teller relation.

The fundamental band gap decreases with temperature causing an increase of the high-frequency constant.

In addition, the phonon frequencies shift to lower wve numbers with increasing temperature. For the multi-phonon matarials where one find TO ones es well at low wave numbers the increase is even more dramatic (have a look on the real part of the dielectric function for an Lorentzian-broedened oscillator on the low-energy side, epsilon 1 increases drastically close to the resonance frequency).

For a detailed analysis of IR data have a look on our recently published article Feneberg et al.:

"Many-electron effects on the dielectric function of cubic In2O3 : Effective electron mass, band nonparabolicity, band gap renormalization, and Burstein-Moss shift

DOI: 10.1103/PhysRevB.93.045203

R. Goldhahn has rightly mentioned the fundamental factors responsible for  the variation of the dielectric constant with temperature for the semiconductors in general and oxide semiconductors in particular. For theoretical calculations these factors have to be rigorously considered. However, for the experimental researchers working with oxide semiconductors more empirical calculations  may be adequate to explain the experimental data. However, the question of stoichiometry and non-stoichiometry for the oxide like TiO2 is an important consideration because  the oxides prepared in the laboratory using solution methods are generally non-stoichiometric like TiO2-x which affects the energy band gap and also the dielectric constant due to modification of  the current transport mechanism as a function of temperature.

Thank you very much Sir for your valuable suggestion.

Thank your very very much for such awesome explanations.