It's the peaks in the spectral domain. The cochlea is lined with little hairs that are tuned to different frequency bands. When they're stimulated by sound waves, the nerves they're attached to create tiny electrical impulses which propagate into the larger networks of neurons that make up the auditory cortex.

When a neurophysiologist uses the term "entrainment" it usually means that they either observe neuronal firing rates or spectral content of energy over populations that is somehow directly influenced by properties of the auditory or visual stimulus. In cases like this, it's like "we played an x hertz tone and shortly after stimulus onset we saw energy at x hz appear in a population of neurons"