This work presents the solution:

http://pubs.rsc.org/en/content/articlehtml/2015/MH/C4MH00174E

The method is included.

In short, you calculate a reference state (bulk), a surface state (which level is the surface state of interest), and so by difference you get the working function.

Thanks Dr. Brivio. The paper is relevant and helpful. But correct me if i am wrong, core level alignment technique has been used in this paper by taking Pb 1s state as common reference and then band offset at MAPbI3/

MAPbBr3 & MAPbBr3/MAPbCl3 has been calculated and aligned w.r.t. already known MAPbI3 VB and CB edges. My query is how can I use this technique for band alignment for CH3NH3PBI3 and TIO2 interface since I no longer have a common reference state (like Pb 1s in the paper)? I will be very thankful if you could put some light on it.

a surface state (which level is the surface state of interest) : here, are you implying like MAI or PbI2-terminated (001) /(110) surface or something else?

One way, as mentioned, is referencing a core state on both sides of the interface, but that requires an all-electron non-frozen core method. My guess is that a decent interface using Wien2k FLAPW is pretty hard to do. Vacuum levels are messy and won;t lead you anywhere. Alternatively, the standard method, usable in pseudopotential codes, is explained e.g. in J. Phys. D: Appl. Phys. 31 (1998) 1273–1299. If you have different interfaces in the cell you may get different offsets and hence an apparent field. If you have actual interface charge, there are methods to deal with it (PRB 57, R9427-R9430 (1998)).

PS: I emphasize that the core levels, in case you use those, must not be frozen, otherwise they'll be no use. For example VASP PAWs have frozen cores, last I checked.

Dr. Fiorentini, Thank you so much for the papers as i now understand that macroscopic average technique is the answer to my above query. I think after making a superlattice of CH3NH3PBI3/ TIO2 (100) heterojunction over various atomic planes (containing the growth axis), I should use the LVTOT=.TRUE. in INCAR of VASP calculations to obtain the electrostatic potential along the growth axis. But since CH3NH3PBI3 and TiO2 are different in terms of stoichiometry and geometry , perhaps you can tell me, should i use any other tag in INCAR to avoid electric fields due to unbalanced charges or to avoid/compensate dipole formation at interface ??

P.S. I couldn't find this paper : PRB 57, R9427-R9430 (1998)

Paper attached. You'll get a field a) in periodic boundary conditions when the interface offsets are different (similar to two different surfaces with different workfunction); b) there are real interface monopoles as in the paper attached; c) like b) but worse, with different monopoles (rare); d) a superposition of a) and b). You can cope with a)-like cases using LDIPOL or some such in VASP, and with b) as explained in the paper. You may have a combination of these effects in low symmetry.

Thank you for the insightful answer Dr. Fiorentini. It indeed is helpful.

You might also want to look at the MacroDensity package on github published by Aron Walsh's group ...it will parse the LOCPOT file printed using LVTOT = .TRUE. in your INCAR and plot the local potential for you. It is super useful. From this you can determine the vacuum potential. Then, using ICORELEVEL = 1 in the INCAR, you can obtain the core eigenvalues in the OUTCAR file and use the formulas for valence band offset (VBO) and conduction band offset (CBO). You will also need to know the potential difference of the interface slab model as well as the bulk VBM/CBM eigenvalues for each material.

Interesting