Multi-pass microscopy

Microscopy of biological specimens often requires low light levels to avoid damage, which yields images impaired by shot-noise. An improved measurement accuracy at the Heisenberg limit can be achieved exploiting quantum correlations. If sample damage is the limiting resource, an equivalent limit can be reached by passing photons through a specimen multiple times sequentially. Using pulsed light, a self-imaging cavity, and temporal post-selection, we have demonstrated full field multi-pass microscopy.

Here’s a sketch of our microscope:


Our “8f” multi-pass microscope.

The four intracavity lenses, L1…4, form a 4f imaging system on the left (L1,2) and right (L3,4) sides of the sample plane, S. So, when a pulse of light enters the cavity and passes through the sample, it is imaged onto the outcoupling mirror, Mo. While some of the light leaks out (where it can be imaged), most of the light is reflected and re-imaged onto the sample. The process repeats, with an image again being formed on Mi, reflected, re-imaged onto the sample, and again imaged onto the outcoupling mirror. The light has now interacted with the sample three times, which can increase the signal — for example, a phase shift — by a factor of three.


A quartz wedge depolarizer, placed between crossed polarizers.

Let’s see what happens when we put in a sample. A nice, simple sample is a quartz wedge depolarizer. This is a simple device consisting of a birefringent quartz wedge mated to a non-birefringent piece of glass (to prevent beam deviation). Due to the wedge shape, the optical path length throughthe quartz varies as a function of position — but since it’s birefringent, this means the effective retardance of one polarization relative to another varies as well. In effect, it acts as a spatially-dependent waveplate. So, when viewed between crossed polarizers, we expect a sinusoidally varying intensity, as seen on the right.



But, if we pass through the sample multiple times, we build up the polarization rotation, and instead of seeing one oscillation we see many:


Of course, we achieve these multiple interactions not by going through multiple identical samples, but by reimaging the same sample multiple times. This yields the below data, where we pass through the sample m = 1, 3, 13, and 29 times.