Did you mean "center frequency"? If so, even for a 20 dB attenuation at 1000 times the bandwidth in the stopband, I run into numerical issues and can't get such a high order, and thi sis for a Cauer/elliptic filter. You could try a 2nd order, but that would mean a Q of f0/BW=5000. You'll likely need an xtal for that. A 32768 Hz for watches is in the order of tens of thousands. But, before that, can you say what it is for? It sounds like an XY problem.

No I don't mean center frequency I mean bandwidth of the pass band filter,the range of frequencies at which the attenuation is higher than - 3dB

MissMulan‭ I meant about the 10 kHz part because, as Olin says in his answer, that's not a characteristic of a bandpass filter. But you can define it in terms of center frequency and bandwidth (and attenuations in both pass- and stop-band). You haven't answered, though: what purpose will that filter serve? There may be different, less absurd ways to do it -- because a 2 Hz bandwidth at 10 kHz is absurd. For a 2nd order, the settling time will be 4.6 s for 1%.

Ya I agree Q will be really high.