This effect is generally (but not always) related to the difference of thermal expansion coefficients between matrix and filler. Hence the PTC/NTC behaviour of the composite when cooled/heated.
I note that this effect was observed in the vicinity of the melting temperature (Tm), below the Tm the resistivity increases (PTC effect) and above the Tm the resistivity decreases (NTC effect), I would like to send me one or two articles references who treat these two effects and about the difference of thermal expansion coefficients between matrix and filler.
According to my experience NTC is observed when two conditions are met:
1. the conductive filler is applied in an amount which is close to the percolation threshold
2. there is a transition where the matrix shows an abrupt change in specific volume (reciprocal of density) - this is mostly a melting transition in a partially crystalline polymer.
NTC is due to the fact that the tunneling transition in conductvie clusters is very sensitive to inter-particle distances. At the melting the specific vlume increases abruptly, the inter-particle contacts (partly) break and the conductivity decereases. It can be used for self-regulatingg heating elements (e.g. de-icers in railway systems).
A similar reason can be ascribed to the observation that rubbers filled close to the percolation transition exhibit measurable change in conductivity either on extension or compression. (The first is used for extensional sensors, the second for traffic counting).
Incompatible blends where only one of the components contains the conductove filler are also very sensitive to processing conditions.
It is also well known that carbon balck filled melt which is a relatively good conductor if compression molded will exhibit much lower, and ansotropic conductivity when films are blown from it.
The sensitivity of tunneling to inter-particle distances also explains why repated heating-cooling cycles result in ill-reproducible conductivity curves. (See e.g. the early monograph of Gul', the Russian scientist). The same is true for the voltage dependence (I/V curves) upon cyclic loading.
Therefore I would say that the matrix effect is more important but the whole behavior of conductive/indulating composites is very much complicated. Especially producing semiconductive compounds close to the percolation thershold in a reproducible manner is art rather than science, it is very much know-how based technology.
In our measurements, the intensity of resistivity decreases with increasing of concentration of Carbon nanotube. So, In your opinion, the PTC/NTC effects are related to the composite (matrix/contenant). I would like to send me a bibliographic item that explain this behavior,