Assuming that the quantum efficiency of the two sensors is equal, the amount of light collected by the two sensors would be equal. The question is how do you compare the results? If you look at the two images at 100% in Photoshop, the 45.7 mp image will be much larger and noise will be more apparent with the larger image. To make things equal, you have to view the images at the same size. You could resize the 20.9 mp image to 45.7 so that the two images would be viewed at the same size. Alternatively, you could downsize the 45.7 mp image to 20.9 mp.
The 20.9 mp sensor collects more light per pixel than the 45.7 pixel. If you downsize the 45.7 mp image to 20.9 mp you have 4 read noises which add in quadrature, while with the 20.9 mp image there is only one read noise. If read noise is relatively high with both sensors, the 45.7 mp image will have a worse SNR, while if read noise is small, the differences will be smaller. In general, the larger pixel sensor will have an advantage. With the very low read noise of current sensors, the differences will be smaller.
For an excellent technical analysis, see Emil Martinec's essay
here.
Bill