The DFTB+ software (http://www.dftbplus.org/) is a good choice. It has many options (including Extended Lagrangian Born-Oppenheimer apart from normal MD runs). the latest version is publicly available without registering, so it's easy to try.
The only problem you may encounter is lack of parameters for some atoms, hence first you may want to check for it (http://www.dftb.org/parameters/download/).
Thank you so much dear Karol Strutyński. I have another question, recently I am working on copper clusters but I have one problem I am unable to make structures of multi atoms clusters like Cun (n=20, 21.......).Sir, kindly tell me the process of making structures of these types large clusters, I will be thankful to you.
I cannot help you there, as I have no experience with such stuff.
When you have no leads as to the structure of new molecule/cluster you may try old route: make "random" cluster by hand and run some MD then pick some representative structures from the run (by energy or how they look) and minimize them all.
I know that it is not much of an insight, because It's time and attention consuming.