Umm, not sure I need any, it's just basic physics. Newton's laws of motion, and a bit of fluid dynamics.
Anything moving through air experiences a resistance force due to the air, and so if you have two similar objects, and one has "better" aerodynamics (e.g. by having a lower coefficient of drag and/or frontal area), then by its very nature, the one with the lower CdA requires less force to travel at any given speed through the air, and so clearly it will also use less energy to arrive at any given speed from a slower one.
Now if you are comparing objects that have a different masses (nothing suggests an aero wheel has to be heavier than a less aero wheel) or presents different moments of inertia, then one can also do the math, and you'll find that the difference then depends on how much the CdA of the system has been reduced compared to difference in mass, and the moments of inertia. You can pretty much discard the moment of inertia, as the difference in energy demand when accelerating bicycle wheels with a different moments of inertia is very small.
but let's consider an example, take for instance my rear Velocity Aerospoke 32 spoke standard bike track wheel and my Zipp disk. Not I don't know the weights precisely of each, but let's be generous and say the Zipp weighs 0.5kg more. It's nothing like that much heavier, but let's take worst case.
Let's also assume we have a bike + rider with the Aerospoke = 80kg.
And let's also assume they go from 0 to 50km/h in 10 seconds, which is a pretty solid acceleration.
Power required just to accelerate (not including all the other resistance forces - this is just power required to change kinetic energy so we can isolate the impact of different masses)
Now let's add the 0.5kg to the system. Power demand for acceleration component is now
= 776.4W, or an extra 4.8W due to the additional mass.
OK, I know (because I have done the testing) that my Zipp lowers my system CdA by 0.023m^2 over the Velocity Aerohead. I also know that similar scale reductions in CdA are possible with similar deep section rims, such as 808 or 404. And that's just for the rear, not the front.
If I do an integration across that 10 second acceleration (and make an assumption of an even rate of acceleration, although it will under estimate
the aero savings since the rate of acceleration will slow due to both neuromuscular fatigue kicking in after 5-6 seconds and the curvilinear rate of increase in air resistance during an acceleration), and compare the difference in power demand for overcoming air resistance, the more aero wheel saves me 11.2W for that 10 seconds, more than double the power "cost" of the over estimated
mass increase. So a net gain there. And keep in mind I'm over estimated the mass cost and under estimating the aero benefit.
Now the aero savings are very small at the slower speeds at the start, and build up significantly as speed increases. If I had done same for an acceleration from say 30-50km/h over 5 seconds, then the saving due to the better aerodynamics during that 5 seconds is 21W and the cost due to extra mass is 7W. Net gain of 14W.
Then consider that once up to speed, the steady state power demand difference between the wheels at 40 km/h is 19W and at 50km/h it's 37W.
So not only does the heavier aero wheel reduced the net power demand during an acceleration from a standing start, it's also now significantly easier to maintain speed once you've stopped accelerating (which is most of the time), or indeed enables one to go faster for the available power. And if the starting speed is higher, the benefit of aero over weight during acceleration is even greater.