Dear Kulwinder, doing research on Hadoop is different from using it. Firstly, compiling Hadoop for Windows itself is a different topic for it depends on several other applications (like Windows SDK, maven etc.) I don't think doing research on a tool is a good idea or not, but you can certainly do research on data science related to big data. For instance, Hadoop falls under distributed computing systems (perhaps rather more for a level to state...). So, it might be better to concentrate either on parallel processing, grid computing, clusters, processing speed, job sequencing and aligning, memory management etc.
Dear Kulwinder, you can't install Hadoop directly on Windows. You have to use any flavor of Linux OS(its better to use Ubuntu 14.04 or 14.10). You can install Ubuntu under windows via VMware or Virtual box (available free on internet)and there you can install Hadoop under that linux OS. But my suggestion is to install Hadoop directly to ubuntu (i.e without using VMware). Because networking in Ubuntu is much easier than that of when it is installed under windows via VMware. Another thing, when you install hadoop under windows, the resources are being shared, so you will never get the actual flavor of Hadoop. Here are the links of Hadoop single node and multi node installation.