The analog outputs on the 4461 (and the analog inputs, for that matter) are band-limited by digital lowpass filters. These low-pass filters have a brick-wall response in the frequency domain, which results in Gibbs-phenomenon ringing in the time domain. See this discussion, among others:
http://en.wikipedia.org/wiki/Gibbs_phenomenon
The ringing will be excited whenever there is significant signal power near the abrupt band edge, because the abrupt cutoff will result in signal power being located just below an absence of power in the spectrum. Whenever signal power appears in spectral isolation like that, you can expect ringing in the time domain response. Your square wave consists of a series of harmonics at frequencies below, near, and well beyond the digital filter's cutoff. When the digital filter cuts off the higher harmonics, the ringing is created.
To prevent the ringing, you must create a waveform mathematically that is bandlimited already. The result will be a square wave with slower rise and fall times, but no ringing.
Hope this helps,
Ed
