Affordable High Pulse Energy Ultrafast Laser Design

Age of nanosecond:

If you look back on the history of lasers, you can clearly see that every stage is a huge step forward compared to available technology. The first CW lasers were a miracle compared to conventional light sources - the light they produced was perfectly directional. Then the micro and nanosecond lasers melted and evaporated material like metal on very short time scales. Picosecond lasers did the same thing but without heating the material much. Finally, femtosecond lasers with their huge electric field directly teared off electrons from the atoms. Even though femtosecond lasers for material processing are available from multiple manufacturers, the common applications are still stuck in the age of nanoseconds. Ultrafast lasers are simply too expensive to use in family workshops or DIY labs.

But they can do so much more:

• LIPSS (periodic surface structures induced by the standing wave of the femtosecond pulse in focus) (3) (PDF) High spatial frequency periodic structures induced on metal surface by femtosecond laser pulses (researchgate.net)
• laser filament generation (plasma column created by periodic self-focusing and de-focusing) Filament propagation - Wikipedia
• cutting and engraving all materials (glass, ceramic, semiconductor, plastic, wood, textiles ..) Laser cuts glass – at the cutting edge | TRUMPF
• particle acceleration and X-Ray generation without high voltages (the electrons and ions get accelerated by the electric field of the laser pulse) X-Ray Generation - LIGHT CONVERSION

 

To make these lasers more available, they should not cost 50`000 – 100`000 USD (Pika-Pika?) while nanosecond lasers with similar average power cost 1`500 – 30`000.

To understand what makes the price different so huge, we must dive in into details and come up with a design which will enable wide adaptation of the femtoseconds.

Why so expensive?

There are three main reasons why the price is so high.

• It's not easy to design a high pulse energy laser.
• The design is always complicated.
• Testing of the laser needs special equipment.

 

Design of Ultrafast Lasers

Lasers are made of a laser gain medium (laser crystal) and two mirrors. To achieve short pulses, one needs to add a component or method in the laser cavity which creates the short pulses (called mode-locking). During the design there is a lot of points on the design checklist which need to be figured out. Here is a list of things which covers the most important points:

Pump absorption, thermal lensing, Kerr lensing, consider three or four-level lasing scheme, pump saturation, small signal gain, gain saturation, reabsorption, laser damage to components by pump or signal, nonlinear index changes, nonlinear effect (second harmonic generation, four-wave-mixing, stimulated Raman scattering, stimulated Brillouin scattering), self-phase modulation, dispersion compensation (GTI mirrors, chirped mirrors, prism or grating pair compressors), second and third order dispersion of materials and coatings (linear and nonlinear), nonlinear absorption and two photon absorption, spectral bandwidth (time-bandwidth product), absorption and emission cross sections, cooling of the crystal and pumps, pump wavelength shift with temperature, quantum defect, linear and transversal modes in the cavity, crystal length and doping concentration, optical design, linear and nonlinear polarization dependent effects, cavity losses (imperfect mirror reflections), stray light (unwanted pulsing or depletion of population inversion), optical path length change due to thermal effect, thermal expansion of the optomechanics, vibrations, tolerances, amplified spontaneous emission (ASE), soliton propagation

The most of the above have physical limitations (for instance linear and nonlinear index are connected), others have simply technological limitations (for instance no pumps with high enough power available). The hard part is really finding the right set of parameters which works from the laser start (quasi-CW) till the pulses are available for the user.

 

Design types

1) MOPA

The most popular design direction is the one called MOPA (master oscillator power amplifier). Basically, there is a laser oscillator which produces extremely low pulse energy and average power femtosecond pulses. A low power oscillator is easier to design because a lot of effects are intensity dependent (nonlinear) and those can be neglected. There are not many technological limitations there either - femtosecond laser oscillators were already developed a good 20 years ago. Once the pulses are available from the oscillator, the laser system needs to select some of them using a fast electro-optic switch. Then the pulses will be stretched, amplified, and compressed. This technique is called chirped pulse amplification. Again, the goal is to avoid those nasty nonlinearities. There are usually 2 or more amplification stages. It's almost like you must build at least 3 lasers plus add a lot of switches and monitoring to make sure those lasers working together. The large number of components make these lasers expensive.

A broken 20W Coherent Monaco femtosecond laser from eBay. Price: USD 45`000

 

2) Thin-disk

High power thin-disk oscillators are dealing with the temperature and nonlinear problems spatially rather than temporally (stretching and compressing of pulses). There are no amplification stages required. The most important part of the design is that the laser crystal is very thin (<1mm, 0.1mm or so) and instead of cooling it from the sides, it's cooled more efficiently from the back. The laser beam diameter in the cavity is large so the intensity is low (those nasty nonlinearities!). Just add the two end mirrors to the laser cavity, some mode-locking (Kerr-Lens or Saturable Absorber) and you are good to go. Unfortunately, it's not that simple. The thin laser crystal needs the pump light and pulses to pass through it multiple times (10, 20, 60 ..) because the absorption is low and the gain is small due to the low thickness. This complicates the alignment because there are a lot of mirrors, light spreads out in the long cavity, tolerances add up. It also doesn't help to bring the cost down, that the idea is rather new, patented and the main manufacturer is a premium one. (Blue square if you know what I mean.)

Image Source: Ultrafast thin disk laser – Ultrafast Laser Physics | ETH Zurich (Martin Saraceno, martin@saraceno.info)

 

What I haven't mentioned but will add to the total material cost is the control electronics, pumps and pump drivers, power supplies, output monitoring, cooling of pumps and crystals. This will be always part of the cost.

 

Testing of Ultrafast Lasers

Background noise and quality of mode locking requires a fast photodetector and an expensive RF frequency analyser.

• Analysing the output laser spectra needs a spectrum analyser or spectrometer.
• The pulse length can be measured using an autocorrelator or FROG/SPIDER device.
• To characterize the beam profile, you need a beam profiler.
• To look at pulse trains you need a fast oscilloscope.

 

The total testing equipment will easily cost used/cheap 20-40k USD while using more quality devices will start 20-40k per piece. Renting some of these might be possible, but it's very rare that even the manufacturers don’t always have this option. Delicate equipment will get eventually damaged when shipping it every few weeks to a different location and there are only a couple of people who would require a rental unit anyway. Making the renting scheme economical is a rather impossible task.

 

DIY-Optics Design

1) Recently the ultrafast laser design went from "being afraid of nonlinearities" in the direction of using the nonlinearities. I mean the Kerr Lens Mode-locking technique replacing the saturable absorbers with finite lifetime. In my opinion, nonlinearities are an important part of the femtosecond regime and we should use the nonlinearities even more, instead of being afraid of it.

2) We must (simply) reduce the number of components and the size of the laser. Thin-disks made the first step, but they lost something on the way.

In my design the laser beam is a ring (the crystal is pumped in a ring shape), mode-locking (and output coupling) is done using an interferometer mirror. Let me explain.

Optical simulation model of the cavity (in FRED Optimum Optical Simulation Software)

Ring

The choice of ring beam is rather unusual - it adds divergence not just to the ring width but to the ring diameter too. However, it has the advantage that the pump and the signal power in distributed on a larger area while all parts of the beam are still very close to the cooling in a cylindrical shaped laser crystal. The crystal thickness can be still normal, so pump absorption is good.

The ring shape is beneficial for dealing with stray light too: if the large middle area of the disk (Gaussian beam) is pumped, it would amplify a lot of other modes which partially overlap with the Gaussian ground mode. Stray light will more likely hit the pumped area too. The ring shape is something very specific, hard to hit without a purpose.

A further obvious disadvantage of the ring beam is that it has a hole in the middle. If the user wants to process with the laser an area, it will be harder to get a good coverage. Ring beams are sometimes used with optical tweezers, Bessel beams for producing small spot with high depth of focus, and I've seen somewhere a particle acceleration application too. So, the ring is not used very often, but this issue can be solved by beam shaping. It's possible to convert a Gaussian beam into ring so it is also possible the other way around!

 

Simulated laser output (Irradiance not to scale)

 

Mode-Locking

For mode-locking I decided to use something from the past (1988!), called Nonlinear Michelson Interferometer (NLMI). These guys did nice experiments when CO2 lasers were in fashion.

https://www.researchgate.net/publication/237363903_Passive_mode_locking_with_a_nonlinear_Michelson_interferometer_transient_and_steady-state_characteristics

Before explaining the NLMI, let's take a step back. While designing the laser I faced two hard to solve issues.

1) To achieve high pulse energy with a relatively low average power (10-20W) I would need to make the cavity very long (hundreds of meters) to keep the repetition rate low. Traditional multi-pass amplifier is not an option because of the ring pumped crystal, and it adds a lot of components (see thin-disk).

2) The ring shape and the width both have a divergence, and this beam profile can only work in a limited cavity length. As the ring width expands, eventually it will get as large as the ring diameter.

 

These two issues kind of work against each other. The cavity must be short and long at the same time. The solution is a regenerative amplifier type of scheme. In a regenerative amplifier there are two mirrors and an optical switch. The switch lets the pulse from the oscillator into the cavity, it amplifies in a multi-pass way (always the same path) and then the switch lets the pulse again out. This needs a very-very fast switch which knows (!) when it should open, and it needs an oscillator where the pulses come from. 

The NLMI happens to be both the switch and the mode locking device creating an oscillator. Time to explain how it works. The NLMI is a conventional Michelson interferometer (beam splitter, two mirrors .. 1 per arm) with a modification that in one of the arms there is a nonlinear media. As an example, let's take the situation when the interferometer is aligned to zero output. In this case the incoming (low intensity) light interferes with itself in a way that at the output there is destructive interference. The incoming light has no other way to go, it is reflected towards the source – it works as a mirror.

The nonlinear media adds the convenience that as soon as the intensity of the pulse get's high enough, it will change the refractive index (nonlinear index) of the material in one arm. The increased refractive index means longer optical path in one of the arms and as a result the interferometer is misaligned - (part of) the pulse is coupled out. There are some other effects to account for like self-phase modulation and linear GVD from the nonlinear media. These can be compensated be replacing the mirror of the nonlinear arm with a GTI mirror.

 

NLMI schematic

 

The laser output of this design is expected to be 100uJ per pulse at 100fs pulse length and 1053+/-10nm. The beauty of the NLMI is that it’s intensity dependent. Most Yb:CaF2 lasers achieve pulse lengths of 80-200fs. Even if the resulting pulse length is somewhat longer, more pulse energy will be stored in the pulses before they can couple out. (Intensity= Pulse Energy / Pulse Length / Beam Area)  

Simulations of the laser output shows that the natural operational mode is “GHz Burst-mode”. The number of pulses in the burst and their peak power can be controlled via piezo translation of the interferometer mirror and by adjusting the pump power.

 

Output pulse energy

 

The burst-behaviour comes from the intensity dependent reflectance of the interferometer. First the reflectance is low at low intensity and increasing with intensity. This promotes the creation of ever shorter pulses (mode-locking). Once the reflectance gets to one and the pulse is further amplified, the further phase shift makes the NLMI to couple out the pulses. Once the population inversion created by the pump pulse is completely drained, the remaining pulse energy couples out or fades inside the cavity. The distance between the pulses inside the burst comes from the cavity roundtrip.

Burst mode is not a bad thing. In the recent two years more and more companies come up with their Burst mode attachment module. Laser processing dynamics by GHz bursts are not well understood yet, but this feature seems to further improve ablation rate and quality.

Mechanisms of ultrafast GHz burst fs laser ablation | Science Advances

 

Limitations

Achieving smaller pulse energies for microscopy would need a different initial setup of the interferometer, but the burst-mode-like behaviour might be a disturbing factor in these applications. Increasing the pulse energy needs redesign (larger crystal and ring diameter) but using the design tool developed at DIY-Optics reduces the required effort. However, the pulse energy can’t be increased without limit. The main limitation is the laser induced damage threshold of the crystal, (likely to happen at about 200uJ pulse energies) and the thermal lensing which get’s stronger if the ring get’s further away from the cooled surface. Further amplification (like the MOPA solution) scheme can be done with external modules. Amplitude Systems (Amplitude - Laser manufacturing company) is the expert with million dollars cost per system.

As the setup is using an interferometer, it will be sensitive for length changes. Couple of minutes heat-up time will be required (like with other femtosecond solid-state lasers) so the thermal index change and the thermal expansion of the components is settled. Vibrations can cause length changes in the setup – this is a potential issue which can be hopefully overcome with vibration-free design or smart optics tricks.  

Measuring each point of a beam of light : News Center (rochester.edu)

 

Price and Ultrafast Laser Market

The proposed simple laser design can be built in Switzerland from ca. 15`000 USD total material cost and ca.  10`000 USD labour. If the usual 20% profit margin is applied, the laser will sell for ca. 30`000 USD. That’s with off-the shelf components. With process optimization and volume orders the price can be easily brought below 20`000 USD.

Compared to the current market situation this is extraordinary. The company Litilit sells the same average power module (Industrial grade femtosecond laser (litilit.com)) with the same pulse energy and somewhat longer pulses for 70`000 EUR. Somewhat higher power (30W) lasers are available from EKSPLA FemtoLux 30. Industrial Femtosecond Laser — Ekspla for about 100`000 EUR, the Jasper 60W from Fluence Jasper X0: High Power Femtosecond Fiber Laser - Fluence.technology is 110`000 EUR, the Chinese Huaray 50W costs 60`000 EUR HR-Femto-50 Industrial Femtosecond Fiber Laser - Femtosecond Lasers - Huaray Laser, NKT  aeroPULSE FS50 - NKT Photonics or IPG 20-30W IPG Photonics Corporation lasers sell for 40`000-50`000 EUR and finally, premium Trumpf and MKS SpectraPhysics lasers might go for about double the price per Watt average power. These prices are never published on the websites so take these data with a pinch of salt.

Of course, this design is not easily comparable with all those high-end lasers. They can do more because they have a lot more components – so they can invent features, use components differently which are built-in already. But does everyone need that? No. If I just wanted to have 100uJ femtosecond pulses at 5-50kHz for benchtop laser structuring, x-ray generation – I would have to buy a >60k machine with 100-1000kHz repetition rate. This is not cool.

 

P.S.

Discussion with the scientific community (ReseachGate, Meetoptics, Reddit) did not reveal any showstoppers yet, but we have to admit that femtosecond laser design is difficult and there is never guarantee that this will work as expected.. so, this is only a design proposal at the moment.

I would be more than happy to build it and enable the implementation of high energy femtosecond lasers for new, real-world applications (X-ray machines, particle accelerators, maybe even fusion using laser filamentation). As outlined above, such a development project needs expensive parts and test equipment, and DIY-Optics GmbH has only part of the equipment available at the moment. If you are looking to invest into innovative initiatives in photonics and laser technology or support this project in any way, don't hesitate to contact me at daniel@diy-optics.com or LinkedIn (Daniel Csati | LinkedIn). Cooperative development with other photonics/laser companies is also welcome. 

Please feel free to comment.