Hi Nima, if you are working in steels, you could review as option for phase studies conventional techniques as microscopy and additionally EBSD, and depending of sever factors you could obtain info using XRD. But take a look in papers related with your work and you can see the different methods for characterization of your material.
Depending on your material, you may look into atom probe tomography as well. EBSD on electron microscopy would usually be the first step. Just make sure you're university has the equipment.
Nima, if you go with "Map acquisition method" and perform phase analysis on that, the instrument software will populate the various phases present and will even present you the proportion of that phase. If you are after high accuracy you would require standards (i.e. sample with known composition). The major factors on which the accuracy of your phase analysis will depend are: (i) Use of standards relevant to your sample (ii) count rate you get from your EDS instrument (iii) how long you collect your spectra for (iv) spot size. Higher spot size gives high count rates but it compromises Secondary electron image but at the same time the Back scatter image is of better quality. With Map acquisition method the total time required for the full map to get good accuracy will be bit high as compared to the point and id option as there will be numerous number of points that will have to be scanned. Therefore, you will have to work out the total time you require for the map size you select. Based on my application area i.e. cement paste sample i require about 10-15 seconds on one point based on 10,000 counts per second, to get a good phase map and concentration. If you get higher count rates you can proportionately reduce the time of acquisition. But you have to work out the approximate number of points in the map size you chose.
Could you say anything about the number of phases you are expecting? Are they all known? Do they have a similar scattering behavior? Why do you exclude XRD, and why do you believe that techniques like SEM or TEM will give a statistically more relevant result? It is clear that higher resolving techniques enable a high precission but they are only rarely accurate for an extrapolation on your sample since the microstructure is inhomogeneous. Therefore, integrating techniques like powder XRD which cover a comparatively huge volume are more reliable.