The RP Photonics Software News > 2015-08-20

Reasons for Not Engaging in Computer Modeling of Laser Devices

Computer models can be extremely useful in laser development and scientific research. The key benefit is that they help you to develop a deep insight into what is going on in these devices, allowing you to build up a thorough understanding, including both qualitative and quantitative aspects. Of course, understanding is often the key to success: once you know how exactly your lasers work, you can quickly and efficiently optimize them, not wasting time and other resources. In scientific research, gaining a better understanding of things is even the central goal.

However, it is a fact that most people working in laser engineering or laser science are not dealing with computer models. This raises the question what is the reason for that. There may actually be many different reasons – some of them quite understandable and justified, while others deserve to be critically reviewed. In the following, I try to list and shortly discuss those reasons which came to my mind.

I do admit that I'm not neutral in this area, as I strive to sell simulation and design software in this area. However, you will see in the following that I'm giving honest advice, not pretending that modeling is the solution for everything.

I don't believe that it would work

With that opinion, you could be right or wrong, depending on the situation. Of course, you should engage in computer modeling only if you see a reasonable chance that it will work, and there are several conditions to be fulfilled:

• You need to identify the appropriate type of model for your purpose.
• You need appropriate simulation and design software, which you can trust.
• You will need to have all required input data.
• More conditions may have to be met in order to obtain results which really bring you forward. (See below for some examples.)

I am often asked whether computer modeling makes sense in certain situations. In many cases, I can confirm that it indeed promises substantial benefits, but it is also not rare that I have to tell them otherwise. For example, there are cases where I know that the essential physical processes are influenced by inputs which you can hardly get hold of.

To give you one example for difficult case, there are high energy pulsed fiber amplifiers based on few-mode fibers, and users are interested in predicting the beam quality which you can get from such a device. Unfortunately, the detailed beam evolution depends on many factors such as bending, fiber inhomogeneities, thermal effects and doping profiles; it is normally not realistic to hope that one would sufficiently precisely know about all these effects to make any reliable predictions. So at least for this particular aspect, one should not promise too much.

Another example is lamp pumping of solid-state lasers. At least the pumping process is so complicated that I would normally consider detailed modeling as impractical. This does not mean, however, that no modeling at all makes sense for such lasers. For example, one can still simulate the power evolution based on an assumed distribution of laser excitation, for example in order to optimize a laser resonator.

You may also like to read my article of 2014-10-20 on “Typical Limitations of Numerical Modeling”.

I can't trust the results

That is an issue involving different aspects. One of them is the question of model validation, which I have discussed in an earlier spotlight article. Particularly for home-made software, this can be a critical issue.

Of course, predictions from models may be wrong for many reasons – for example, inaccurate data of used components. However, you then at least get the chance to reveal that – which may actually be quite important.

One should be aware that real experiments may also not tell you the truth – depending, of course, on the type of question. The actual questions of interest are often not what is the performance of the particular implementation of a laser design, but rather how well it should work if the parts are fine, how sensitive it will react to certain imperfections, etc. Such questions can often be more reliably answered with a model. For example, a non-ideal laser crystal, exhibiting some quenching effects, will have a precisely known amount of such effects only in a model. And a laser built with components with off-spec behavior can also mislead you a lot.

I don't believe that it would be worthwhile

It is true, something which works is not necessarily worthwhile to do. You have to compare the possibly substantial efforts for getting into computer modeling with those of possible alternatives – for example, doing more experiments in the lab, hoping to get your results simply with trial and error. Unfortunately, it can be quite difficult to estimate the required time in advance.

At a first glance, one may often feel that the quickest solution is to simply try out in the lab whether something can work – for example, a certain laser design. At least, it may be simpler to imagine that you buy the required parts, put them together and see what comes out – compared with buying some simulation software, getting familiar with it and to the required calculations. Such a reasoning can be quite wrong, however:

• First of all, the investment of time and money into computer modeling is largely done just once and can then be utilized in many following experiments or development projects. Once you have the type of model, it can be by far faster to simulate the performance than first to build a real device and do actual measurements.
• If you put together your prototype in the lab and find that it does not work, you may actually not have learned very much. Usually, the laser does not tell you what is wrong, and what you would have to correct to get it working properly: is it a poor design, are some parts faulty or dirty, is it not properly aligned, etc. You may then feel that the laser could work better when exchanging certain parts, which may require you ordering those, waiting for them, realigning the laser, etc. Obviously, it can take a lot of time until you really get something working while not knowing what actually is going on inside.
• Even if it once works, it also does not tell you why it works, what you could do to improve it further, whether you have already achieved the optimum performance, whether you could achieve the same with fewer or cheaper parts, etc.

A quite reliable rule for experiments is that it always takes much longer than you initially thought!

Honestly, modeling may also take longer than you thought. At least, you will not have to wait for additional ordered parts to arrive, you will not put expensive parts at risk, and if anything unexpected happens, you can always “look inside” the model (i.e., access any internal data such as optical intensities or excitation densities) in order to find out what the problem is.

I don't need it

… because the performance of the real laser in the lab and not of an idealized laser is what is relevant for me: this may be, but don't you also want to know e.g. how the laser would perform if the laser crystal were ideal?

… because I already know enough about how my laser works: you may have some knowledge from textbooks of from scientific publications, you may have some equations to calculate certain things, but a computer model can probably give you for more comprehensive insight.

… because my laser works just fine: but don't you want to know whether it could work even better, whether it could be made with fewer parts, etc.?

… because a colleague of mine is taking care of that – that's fine, of course not everybody in the team has to do the modeling himself.

It's not my field – I am an engineer, not a scientist (or an experimentalist, not a theorist)

That is strictly speaking not an answer to the question why you are not engaging in computer modeling. There is no rule saying that you cannot do such things if you are an engineer or an experimentalist. Supposedly, you are interested in understanding your lasers so that you can e.g. develop and optimize them more efficiently. So you should consider any reasonable method of obtaining such an understanding.

It takes too much time to learn this

It is definitely wise always to consider the required time for getting into a new area, but one should not overestimate the effort when using high quality commercial simulation and design software.

As an example, our RP Fiber Power software can absolutely be used even by people knowing quite little on these devices. They can still collect the required input data like input powers, fiber length etc., and use fiber data as delivered with the software or possibly within the technical support. Based on that, they can calculate the performance of lasers and amplifiers for example, without knowing the physics details e.g. on light propagation in fibers or the excitation processes of laser-active ions. Of course, power users who really understand the physics can profit much more and use the same product for most sophisticated investigations. However, even quite simple laser and amplifier calculations can already be very valuable for a laser company. Furthermore, the latest version of the software supports so-called custom forms, which can be tailored to specific applications, so that one can even do very special simulations or calculations without understanding the details, as long as somebody else (e.g. me, within the technical support!) prepares such forms.

Of course, making models and the required software yourself is quite a different issue; this is really demanding and time-consuming, and may be worthwhile only if you cannot more efficiently use a commercial software, or if your primary interest is to learn a lot on physics and software engineering.

I do not know about it

Surely, many people do not make a conscious decision not to use computer modeling; they simply do not know that this option exists in their field. That is sad, of course, since they may lose important opportunities. Others may find it very difficult to obtain reliable and helpful materials on models, which leads us to the last section:

Learning About Modeling

I understand well that there is a widespread lack of knowledge on computer modeling in laser technology, whereas such techniques have become very common in other areas like electronics. One of the problems is that many people do not learn about modeling at universities, simply because the lecturers also don't know much about it. Therefore, I'm regularly working hard on presenting high quality educational materials on modeling:

• This newsletter treats the topic quite regularly.
• On our website, I have also published a comprehensive tutorial on modeling in the context of lasers and fiber optics.
• Our pages on simulation software do not only describe certain software products, but also provide interesting information on modeling.
• Particularly for our RP Fiber Power software, I have published numerous instructive case studies.
• Many articles of the RP Photonics Encyclopedia contain example cases where substantial insight is obtained from computer models.

I am also trying to convince organizers of scientific conferences to offer short courses on laser modeling, but so far with limited success.

Finally, working as a technical consultant I am often asked whether I could recommend some modeling in order to answer specific questions, or whether I could provide computer models for use by my customers. Similarly, people who are interested in simulation software ask me whether certain modeling exercises would be worthwhile. Indeed, it is a certain straightforward idea to ask somebody with a great experience in the field before forming an own opinion on whether one should engage in this area.

This article is a posting of the RP Photonics Software News, authored by Dr. Rüdiger Paschotta. You may link to this page, because its location is permanent.

Note that you can also receive the articles in the form of a newsletter or with an RSS feed.