"A study in 2002 examined 50 Nobel prize winners from each of the three prizes for physics, chemistry and medicine. This study recorded the age of the scientists when they had done the work that was rewarded with the prize and found that the centre points for age were: physics, 34; chemistry, 37; medicine, 40 (Marchetti C, 2002.) A study published in 1993 investigated a similar data set and concluded that scientists tend to be the most productive in their mid-thirties (Stephan & Levan, 1993.) Finally, a 2008 study of 300 randomly selected bioscientists revealed that the most productive age was 36-40 (Falagas et al., 2008) based on the number of citations from their publications. These studies all indicate that a scientist's greatest potential for discovery is during their thirties"
If instead you measure productivity in number of articles published, this study (http://link.springer.com/article/10.1007%2FBF00142022) claims that productivity peaks at 45-49. However that number covers a wide variety of fields, not just science, and makes no attempt to account for the quality or significance of the work.
'On the ground' productivity, where the scientist is doing labwork ends before a scientist 'gets their name on papers'. That kind of work is different from the 'guiding the science' work. It's the boss who wins the Nobel Prize, not the student, though it's the student who is the 'productive' one in lab.
I'm sorry, I don't really have a source. My observation is based on the fact that during PhD/PostDoc years you have more time to spend on actual research. As you climb the academic ladder it is very probable that administrative/supervising/teaching duties become central to your job (unless you really don't want to).
Incidentally I remember also reading [1] that in certain fields, in this case mathematics, most of the groundbreaking research comes from younger mathematicians. Great contributions to the field from people over 40 are extremely rare.
[1] Simon Singh, Fermat's Last theorem. Ok, not great source but still!