Digital photonics and laser lightning rods from AKL22

I was at AKL 22 a couple of weeks ago: the international laser technology congress at Aachen, Germany. I was present for the interesting technology conference and the live presentations at Fraunhofer ILT.

I have to say that after 2 years of Covid and only being able to assist to virtual conferences if available was a breath of fresh air. Virtual events are nice and you can still learn a few things, but physical events are something else. You are invested in what you hear and there is less distraction than in the office. The best part is that you can network on the side, which is almost impossible with virtual events. Not to forget visiting traveling and visiting new places…

This was my first time at AKL and in Aachen, so I can’t tell much about how it changed from the last time. I can only talk about what I noticed this year, and there was a lot of interesting points.

Digital photonics galore

First of all, Digital Photonics was everywhere. This was actually the first time I heard the term. I know of digital optics: making computer generated holograms and other discrete optical elements. But digital photonic in today’s definition is something else.

This seems to be a big new trend in products containing optical devices and in laser machinery. From my understanding, it is an extension of the whole “software eats the world” zeitgast. To put it in other words, companies are shifting their effort toward software integration, even in optics and photonics.

To illustrate this: it is no longer a question of using a clever idea to improve the cutting rate of a laser against glass, by using for instance Bessel beams or another trick based on some new understanding of the underlying physics. No. Digital Photonics is instead about the “bruteforce” optimization of the processes through software. It is about finding the right parameters using computers, especially machine learning and artificial intelligence. It is about having an automatic process go through scores of trials, use machine learning to create a model and use that model to find the best parameters.

To be fair, I am not entirely convinced by it. It does sounds great on paper, and I am sure there are a few applications where it can give good results. But from my experience and from what I used to read a few years ago, I know that a lot of people in the industry over estimate the capacities of artificial intelligence and machine learning. This is a good idea for processes that are already very well documented, for instance glass cutting or hole drilling. But as it was said during the talks, more and more processes, especially those requiring the use of femtosecond lasers, are very specific to the clients. Each machine is unique. So if the client decides to tackle a new product, a new type of engraving or laser processing then I don’t think a simple machine learning algorithm can find the right parameters that easily when it doesn’t know what to look for.

I am not against machine learning at all. I am actually quite interested by it, but from my understanding, machine learning is only interesting for some carefully chosen applications. Most notably, I can cite the use in Optical Character Recognition, and olso to correct aberrations in photos taken by smartphones. Of course, it doesn’t necessarily mean that the photos are good [2]. Still, these algorithm are absolutely amazing to me. They are able to make pictures taken through 5 mm of diameter optics with a tiny sensor and “cheap” lenses almost look like they were taken by a reflex. This is part of a new trend called computational photography. Kind of reminding me of digital photonics, but for photography…

One other thing I liked during the presentation of some machine manufacturer was the description of their idea for the future of the industry. Thinking of some company that makes laser pieces, typically for the watch industry or for medical applications, they were describing some sort of website or portal where the client can pass an order for, say, 1000 pieces. During the order, the client import a CAD file or some other file describing what he wants the company to deliver. The website sends the data to the closest factory able to make those pieces. According to the clients wish, the machine test the parameters and use machine learning to tweak them to reach the desired cutting angles, depth, surface finish and whatnot. Finally, the pieces are made and sent automatically. No human interaction, all automatised.

Of course, this is a dream: there are always the need for engineers behind those fully automatised equipment to fix them and to take care of unscripted cases. But I liked the idea: to be able to control all parameters and limit the risk of a mistake that people are most likely to make.

Additionally, I think digital photonics might hinder the discovery of new techniques and effects that a person skilled could identify, but that a computer could not even see. If a company wants to be really innovative, the only way is to have people do the work. Why? Because people make mistake and because from those mistakes, some great finding and new innovative processes can be found. But it is not only about making mistakes, it is also about keeping an eye out, it is about a deep understanding of the physics, it is about making connections that would not be obvious.

To sum up: if you are in a company doing run-of-the-mill processing with photonics tooling, then digital photonics might be of interest for you. If you are a company that is trying to innovate, trying new things, then you should pass this trend and go for the good old photonic engineer to do the parameter searching.

Laser lightning rods

Aside from this topic of machine learning and artificial intelligence in photonic application, there were a lot of demonstration from laser companies. One that caught my attention was the talk from Trumpf. It caught my attention because of 3 words: “Laser Lightning Rods”.

This was a total surprise for me to see those words. I knew that the lasers were getting more powerful, especially in the ultrashort range. New models are reaching 200 W at 250 fs [3]. It was surprising for me to see laser manufacturers using the idea of the laser lightning rod as a viable, commercial reason to sell a laser.

It was also the occasion to learn that a project is underway here in Switzerland! This project happens right now in the Eastern part of Switzerland, at Mount Säntis [45]. It is the product of a collaboration between several research institutes from Europe (and Switzerland). This project also involves industrial partners especially Trumpf with the hard task to create a laser source powerful enough to trigger the filamentation in the air. Started in 2017, they should have done the first experimentations in 2020, and they should have given the result in 2021. I suppose that Covid made sure that those deadlines were missed, and I have no idea where the state of the project is right now. The project website seems a little lagging since last december. All I know is that they have been looking for a PhD student for a 4 year study at EPFL who will most likely be starting soon this year, if not already.

Exciting to see that this is going somewhere. I’ll keep an eye on any advances they make and I will write some more details about how they are doing those experiments from their latest papers.