Sample weight is important for the accuracy of the results. Typically, automatic adsorption devices need to mesure at least 10 m2 for an sufficient accuracy. So this means that if your material has a surface area of 10 m2/g, you need to introduce at least 1 g of sample in your device. Thus, a surface area of 100 m2/g will require less material, 0.1 g, and 0.01 g will be enough if your material has a maximum surface area of 1000 m2/g. In conclusion, 0.05 g of sample can be enough only and only if the surface area you want to measure is high enough. Otherwise, you'll get a very bad isortherm from which you'll derive wrong values of porous texture. Just try and you will see.
When you weigh your sample, you also weigh the sample holder. Approximately 40 g. 0.05 g in 40 g is a very low proportion. If you have sample, 0.125-0.250 g is more suitable to avoid errors.
If you have a sample with little porosity (say, surface area less than 10 m2/g) and you use little weight (less than 50 mg, for example), the adsorbed amount will be very little and near the sensitivity of your equipment, so the error will be large.
If you have sample with high porosity (say, 1000 m2/g ) and use a lot of sample (say, above 200 mg), the adsorbed amount will be very large, and the measurement will take very very long time, and error will accumulate in each injection.
Alain gave a very good, general and simple rule of thumb: try to put around 10 m2 inside the sample holder.
The sample weight is a very important factor in determining accurately the BET surface area and porosity using N2 adsorption. Just like Alain mentioned, you'll need a sample weight that is large enough for adsorption of the gas. If the sample weight is too low, the result may not really represent the true porosity and surface area of the sample.
Want to support the discussion by asking a different question. Has anybody analyzed the sensitivity of BET SSA calculation results to the accuracy (e) of weighting results (empty tube and tube + sample)? Can you share the references or thoughts on this, please?
I couldn't find anything on this topic and think about making such sensitivity analysis.