Because the high frequencies cause aliasing (the highest frequency which can be correctly assessed is half of the sampling frequency). A low pass filter attenuates the high frequencies.
If you sample a high frequency signal slowly the result looks like a low frequency signal. Play about in excel or some other plotting software to see what happens. For real signals the highest frequency you can sample properly is half the sampling frequency. Everything else needs filtering out (with a low-pass filter) before sampling because after sampling there is nothing you can do about it. Look up Nyquist Frequency on the internet or in your textbook.
That is used for band limiting the signal. The not so significant higher frequencies raise the sampling frequency value(fs>= 2fm).More the Fm more is the Fs required without much enhancement in the final signal.Hence for band limiting it is used.
I agree with all the colleagues above and add: Sampling is used to convert the analog signal which is continuous in time into a discrete time signal. It is a signal quantization in time. As it is required to minimize sampling rate for reproducing the waveform again, the sampling theory tells us that one has to sample the analog signal with a rate equal or greater than the Nyquist rate fs greater equal 2fm where fm is the maximum frequency contained in the the analog signal.
The sampling process itself is a mixing process such that one multiplies the analog signal with the sampling pulses to obtain the samples signal. By frequency analyzing this time sampled signal one finds the base band and copies of it around the integral multiple of the sampling frequency with an upper and lower side band of the baseband around each one. So in the first frequency interval between the zero frequency and the frequency fs contains the base band and an inverted copy of it . Since each one has a bandwidth= fm, and since they must not overlap or folded over, then fs must be greater than 2fm.
Since the signals generated naturally have high frequency tails one has to filter out this tail and limit the maximum frequency of the signal such that one can lower the sampling frequency and avoid aliasing or fold over distortion.
The LPF attenuates the high frequency components (frequencies higher than your message signal maximum frequency) present in the output of source signal. These are generated may be due to the non linearity of your generator or any other.......