The SI unit of thermal conductivity (k) is watts per meter-kelvin (W/(m·K)); that of k3/ρ is (W3/(kg·K3), where ρ stands for density (kg/m3). With the molecular weight (Mw) expressed in g/mol, we have for (Mw/1000)·k3/ρ the (W3/(mol·K^3) unit, being this quantity conveniently expressed in molar basis. We can approximately predict (estimate) the thermal conductivity (k) of AxBy by (tentatively) accepting the additivity of the molar contributions of A and B for the above mentioned ratio. This leads to: k ≈ [(ρ/Mw)·(x·MwA·kA3/ρA + y·MwB·kB3/ρB)/(x+y)]1/3. Here, ρ, ρA, and ρB, are the densities of AxBy, A, and B, respectively; while MwA, and MwB, are the molecular weights of A and B. Mw* is a mean molecular weight defined for AxBy after the stoichiometric coefficients as Mw* = (x·MwA + y·MwB)/(x+y). After its substitution at the prior equation, we obtain: k ≈ {[ρ/(x·MwA + y·MwB)]·(x·MwA·kA3/ρA + y·MwB·kB3/ρB)}1/3. This formula, thought to approximately predict the thermal conductivity (k) of AxBy after that of A and B, requires the density of AxBy to be previously known or estimated. Its generalization to three or more components is obvious.
The formula derived at the above answer requires the density of AxBy to be previously known or estimated. This conducts us to the question of how can the density of AxBy be predicted (estimated) from that of its components, if ignored. This is the subject now addressed. It can be noticed that the ratio (Mw/1000)/ρ expresses the molar volume (Vm) with the SI unit (m3/mol). By accepting the additivity of the molar contributions of A and B for the molar volume (Vm*) of AxBy, we will have Vm* ≈ (x·VmA + y·VmB)/(x+y), what can be rewritten as: Mw*/ρ ≈ (x·MwA/ρA + y·MwB/ρB)/(x+y). Here, again, Mw* = (x·MwA + y·MwB)/(x+y). We can now predict the density of AxBy as given by ρ ≈ (x·MwA + y·MwB) / (x·MwA/ρA + y·MwB/ρB), what can also be easily generalized to three or more components.
The equations derived at the above answers can be combined ─ by substitution of the later in the former ─ what may be found convenient if the density of AxBy is ignored. This leads to the ‘mixture rule’: k ≈ [(x·MwA·kA3/ρA + y·MwB·kB3/ρB) / (x·MwA/ρA + y·MwB/ρB)]1/3, or, given in terms of molar volumes: k ≈ [(x·VmA·kA3 + y·VmB·kB3) / (x·VmA + y·VmB)]1/3.
The relationships derived at previous answers allow us to predict the thermal conductivity of some substance of molar formula AxBy after that of its components (A, B). The obtained ‘mixture rule’ averages the contributions of each component for the cube of the thermal conductivity, weighted by the product of their molar volumes by their respective stoichiometric coefficient (x, y), that is to say, weighted by their partial molar volumes if the substance could be considered as ideal solution (possibly solid). Generalization for substances composed from more than two components (e.g. AxByCz) is obvious. Such a mixture rule may perform reasonably for a compact material, but may perform significantly worst for a porous material. I have addressed the subject of predicting the thermal conductivity of a porous material ─ although specifically considering a foam ─ while answering to another question at this forum: https://www.researchgate.net/post/Is_the_formula_for_thermal_conductivity_of_metal_foam_correct2