Last week I briefly discussed the [slit compression](/article/slit-compression/) effect that was affecting spectrometer design and performances. Today, I will discuss another effect in spectrographs which is *slit distortion*.

When we plot the chief rays traversing a spectrograph for a slit at various discrete wavelengths, we obtain a plot like Figure 1. Please note that this is the actual raytracing for the [spectroscope](/article/complete-400-800-nm-spectrometer-design-from-a-to-z/) I have designed earlier. Axis are given in millimetres.

![Figure 1 – Slit image by spectrograph at different wavelengths](/images/articles/slit-distortion/image1.png)

We can see that outer wavelengths have a moderate to strong distortion whereas wavelengths near the centre are almost straights. A close-up of the edge of the sensor is given in Figure 2.

![Figure 2 – Close-up on the edge of the sensor](/images/articles/slit-distortion/image2.png)

The effect clearly amplifies with the slit height as can be seen in the dot density. It is less than 2 µm for 1 mm slits, 5 µm for 2 mm slits, 12 µm for 3 mm slits and up to 20 µm for 4 mm slits.

If you process the spectra by summing vertically the columns to increase the SNR, you will suffer from a loss of resolution due to the distorted slit being imaged on several pixels.

I have computed this effect and displayed the result as an histogram in Figure 3 for a slit of 3 mm height (typical slit height at Thorlabs) and 3.45 µm pixels (pixel size of the camera I used in the spectrometer).

![Figure 3 – Spreading of the line across several pixels](/images/articles/slit-distortion/image3.png)

Most of the energy is spread on two pixels and you will have to reduce the slit to 2 mm to concentrate enough energy on the nominal pixel and reduce to 1 mm to completely remove the effect. If you have respected the golden rule of at-least 3 pixel for resolution of your spectrometer and that you are precisely working at the spectrometer limits, this effect can reduce your spectrometer performances by about 70% unless you either use a better algorithm for the sum or use a smaller slit size.

Please note that this effect is not present in our [Raman spectrometer](/article/diy-raman-spectroscopy/) since we do not use the full height of the slit due to the laser focalisation in one tight spot.