Phys.org's New Indian telescope identifies its first supernova links to the recent arXiv Follow-up strategy of ILMT discovered supernovae. The International Liquid Mirror Telescope is no ordinary spinning puddle of mercury, it's got 21st century improvements!

The introduction begins:

The ILMT is a 4-m diameter zenith-pointing telescope located at Devasthal Observatory (Nainital, India). The first light of the facility was achieved last year (2022 April 29) and presently, it is in the advanced stage of commissioning. Unlike conventional telescopes, the primary mirror of the ILMT is formed by pouring approximately 50 liters of mercury into a recipient, which acts as a reflecting mirror. The effective focal length of the optical system is 9.44 m. The ILMT images are obtained using the Time-Delay Integration (TDI) technique. Given the fixed pointing of the telescope, the stellar objects move in the focal plane along slightly curved trajectories. Therefore, a dedicated five-element optical corrector is being used altogether with the CCD reading the electronic charges in the TDI mode (Gibson and Hickson, 1992; Hickson and Richardson, 1998). A 4k × 4k CCD camera (Spectral Instruments) is mounted at the prime focus of the telescope, which can secure nightly images in g, r, and i spectral bands with a total integration time of approximately 102 sec (in single scan).

Time-delay integration and liquid mirror telescopes are discussed further in:

Question: Instead of a five-element corrector for arced drift trajectories, why didn't the ILMT just use CCDs fabricated with slightly non-cartesian pixel layout?

I've done CMOS layout in a previous life, written (PERL) scripts that generate multiple layers in GDS, and compared to random logic (ASIC) circuits the imaging area (pixels) of a CCD is quite simple with relatively few layers and not particularly small nor critical dimensions.

Instead of trying to map slightly arced drift trajectories on to a straight row of pixels with a five element corrector, why not just ask for a CCD with the rows of pixels slightly curved?

The periphery of the die with the readout logic and on-chip analog amplification and ADCs with it's higher complexity would be unaffected. And plenty of advanced telescopes special order their large format CCDs to meet their specific needs since they're by definition low-volume, high-end applications.


1 Answer 1


There does not seem to be much readily available on the specific design or design criteria of the ILMT (or I'm bad at finding it - quite possible). The best I can find is Brajesh Kumar et al. (Monthly Notices of the Royal Astronomical Society, Volume 476, Issue 2, May 2018, Pages 2075–2085, https://doi.org/10.1093/mnras/sty298).

Of the five listed reasons a liquid metal telescope is useful, I will cite two:

Inexpensive technology: The cost of constructing a moderate aperture telescope (4-m diameter) is roughly 1/50 that of a conventional telescope of the same class


Easy image pre-processing: Unlike conventional imaging, here image pre-processing is comparatively easier. For example, the image reduction is performed by dividing each column by a one-dimensional flat-field. That can be achieved directly from the scientific data.

First, they wanted it inexpensive. Given the partners are Belgium, India, Canada, and Poland this is a 'budget' facility (not a major NSF or EU funded telescope). This would argue against considering a custom large area CCD development - that costs real money. If the corrector works, using an off-the-shelf 4k by 4k imager is a real win.

Second, the corrector indeed seems to do its job resulting in 'easy image pre-processing'. Even pretty fancy optics for the corrector can be pretty cheap compared with multiple custom development CCD fab runs to finally get a working chip.

  • $\begingroup$ "Even pretty fancy optics for the corrector can be pretty cheap compared with multiple custom development CCD fab runs to finally get a working chip." This is an unsupported sentence - one might write exactly the opposite just as easily. SE Answers need to support assertions otherwise folks can just say anything. $\endgroup$
    – uhoh
    Nov 28, 2023 at 15:13
  • 1
    $\begingroup$ Well, looking at the multi-year project plan for a fast CCD chip, based on a current design, I'd say I have some experience for (admittedly that I can't point to a reference). One-off advanced CMOS development costs are huge. $\endgroup$
    – Jon Custer
    Nov 28, 2023 at 16:09
  • $\begingroup$ But that's not at al what this is. You start with an existing design, and apply changes only to the array area (where the pixels are). You don't change the circuit, just make tiny adjustments to the pixel positions so they change from being in a straight line to being on very slightly curved lines. That can be done with a few dozen lines of layout script at most. Layout data is hierarchical so you just move the pixel centers, everything else (pixel structures, interconnects) just moves with that. $\endgroup$
    – uhoh
    Nov 28, 2023 at 20:03
  • $\begingroup$ Granted you make a new mask set, but these are not small features (CCDs have giant design rules compared to high speed logic, maybe quarter or half microns vs a handful of nanometers for modern logic) So a mask set is way cheaper and you can use the old i-line or 248 nm lithograpy tools on an older technology manufacturing line. So once again I don't believe that prose alone is sufficient to back up the assertion. It is not a new design, just a gentle adjustment to pixel centers. $\endgroup$
    – uhoh
    Nov 28, 2023 at 20:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .