Site web ESPCI
ARNE HESSENBRUCH (AH) : Did you study physics ?
Jan Friis JORGENSEN (JFJ) : No, not at all, I am not a physicist. Many people think that I am, but actually I was educated in electrical engineering, got a master degree with a special within biomedical engineering. And I think I’ve worked around seven years within different fields of medical engineering, oral visuality and also all the sound diagnostics that we’ve took care. When I started in a small Danish company called DME, which is still existing, and producing STM, SPM, microscopes.
AH : When was this ?
JFJ : I think it was back in 89. I worked there for almost two years and in the beginning I was not meant to work on the STM, but they were behind schedule, so I was assigned to the project and I got stuck. I stayed there for almost two years.
AH : And you had never heard the STM before joining the company ?
JFJ : Not really.
AH : After joining and learning about the STM, what did it mean to you ? Was it an exciting instrument ?
JFJ : Sure, very exciting, it was new. It was very new that people could visualize atoms, achieve atomic resolution, so it was of course attractive to work with. But as a software engineer, you are a little bit of an outsider because all the good people were physicists. It was physicists who had invented the microscope ; nevertheless I recognised several problems in the microscope : even with very fine resolution there was much distortion in the images. Already at that time many people had tried to correct it - of course in hardware which is also the best way to do it if you can - but there are still many problems that could not be solved this way. So I thought why not ? But there was no time for such work in that company, there was not enough resources, and there were many other things to do. I got a chance to work on it only upon leaving the company in November 1990, when I started on an industrial Ph.D being hired by IBM Denmark and collaborating with the Danish Institute of Fundamental Metrology and the Danish Technical University, the Image Processing Department. I finished in 1993. I could almost define the project as I wanted. I had already had seven years of experience working in private companies. But now I had to study again, and of course that was hard but it was also good to know about all the problems which I wanted to solve. It was a kind of a niche, because, as I have said before, most people at that time were physicist working on it, and so they had different approaches for solving the problems, and I used my small capabilities to solve that.
AH : Could you elaborate on the problems ? The tip and sample overlap, is that the kind distortion we are talking about ?
JFJ : Well, that was the one distortion people had been very focused on : how well the tunneling process worked when scanning over the atoms and several people produced very good theories on how to understand these mechanisms, but when you were scanning over a hundred nanometers or more it didn’t really matter. There were other sources of error in the equipment itself : the hysteresis of the scanner and of different kinds of noise in the operations. These were actually the main problems and I guess within science it was given too little attention. Of course it was more exciting to look at atoms.
AH : Was it a case of low and high status research ?
JFJ : I don’t know, it might have been. Anyway I think I found a niche ; I solved some hysteresis problems in new ways that nobody had considered before. And it is a part of our living today.
AH : Are there more distortions ?
JFJ : You mentioned the tip-sample which is of course a problem. It is understandable that to measure something very small, the probe should be at least as small to get a good image.
AH : That’s not an issue you can address with your tools, is it ?
JFJ : I can not address it today. At that time, when I made my Ph.D, I addressed the problem but I did not solve it, because I didn’t have the time for it, but other people solved it in meanwhile - I mean as best as you can. With software you can do a little, but you can’t make everything perfect. Nonetheless software helps you a lot. First of all, you need an understanding of the shape of the tip. Some tips are simply too poor to use for imaging but others are acceptable. With these I can use software for correction and reconstruction, and this can lead to more accurate measurements.
AH : Were you alone in your field ?
JFJ : There have been more papers on tip characterization and on understanding the tunnelling mechanism, while there have been few on hysteresis.
AH : But in your niche were you completely alone, or were there other people working on it, say around 1990, when you were doing your Ph.D.?
JFJ : There were a few working on metrology, but not with these automated image processing tools. I think at that time it was quite unique. There was nothing out there to copy, so I had to invent the tools myself.
AH : How did the opportunity to do an industrial Ph.D. come about ?
JFJ : There was an ad in Ingeniøren, a Danish periodical for engineers.
AH : Whose advert ?
JFJ : IBM Denmark and the Danish Institute of Fundamental Metrology.
AH : Was Kim Carneiro behind this ?
JFJ : Yes. I found it natural to also connect the Technical University to the project. So I ended up with three partners.
AH : The Ph.D topic was to write software to deal with the distortions in the machine itself, the noise coming from within ?
JFJ : Yes. Maybe there was also another approach, because there had been people before doing some software by making certain algorithms and demonstrating that it could work from the very beginning. I tried to enable others to use this and I built the house bigger and bigger. It’s the still the same building but we have taken it from Unix to normal PCs to really reach many people. I also stayed also a couple of months at IBM in Rueschlikon where the STM and AFM was invented.
AH : When did you go there ?
JFJ : It must have been 91 or 92 - I don’t remember exactly.
AH : Were they interested in your project ?
JFJ : Sure, sure.
AH : Did they also recognize it as a niche ?
JFJ : Yes, but there was also skepticism. Some said : "well, people have been making software before, but when they leave the software gets lost because nobody knows how to use it and continue to work on it."
AH : This is a general problem of software, right ? That it has to be made user friendily and to become independent of its maker.
JFJ : Yes. It’s hard to get enough attention within a scientific institute I think, because their focus is on something else ; there are barriers. At the beginning only physicists worked with the STM and nanotechnology. Now there are physicists, chemists, biochemists, biologists - a lot of people who previously did not communicate. They need to learn from each other now. And you can not build an STM without using a lot of different sciences. One of which is of course software. But at that time, people were happy simply to see an image on the screen after pushing a few buttons.
AH : So the demand for your software has developed with the increasing expectations of what you could do with a STM ?
JFJ : Maybe, I think at the time nobody believed you could start a company based on image processing for scanning probe microscopy.
AH : Because the market for SPM was very small ?
JFJ : Very small, but when I finished in 93 I considered commercializing and I discussed it. Nobody believed in it.
AH : Now it is feasible to run this as a business because the market is large enough ?
JFJ : SPM is of course a niche but from a software engineer’s point of view it’s simply image processing. There’s no shortage of images requiring treatment such as scanning electron microscope images and optical microscope images. We now have enough expertise to address the other markets too, so we kind of expand from the nanometer range of SPM to many other things around us.
AH : I would’ve expected the noise and distortion in other instruments to have been completely different so that you would not be able to draw on your expertise with SPMs.
JFJ : Many of our tools can be applied to other images. Some distortions are indeed peculiar to SPM images, but there are also some generalities. For instance, satellite images are scanned line by line the same way an STM image is, and this line by line scanning can give the same kind of distortions and artifacts in each type of image.
AH : I see ; that is indeed very general. As a physics student I scanned photographs of the solar surface using an optical photometer and of course it scanned in just this way. Interesting ! When you started your Ph.D., did you already think in terms of such generality ?
JFJ : Sure ! You have to. I mean I didn’t go into the field to just to be an SPM person. I wanted to learn something which I could use widely. Of course the SPM is interesting in itself, but image processing is also interesting in itself.
AH : All right ; in 1993 you finished your Ph.D. What were your options then ? Could you have gone to IBM ?
JFJ : Probably yes ; I didn’t really try. I didn’t because I really wanted to stay in the field of STM. I think there was still a lot of work to do so I continued at DFM.
AH : What about the big STM companies like Digital Instruments and Park, could you have gone to them, were they interested ?
JFJ : At that time, I didn’t have much contact with them, and they do not pay much attention to single individuals. You really have to scream loudly to get their attention. Of course they know me now ; I have visiting several such companies. I visited Digital Instruments back in the autumn of 99, and I gave my talk twice. The second one was in an R&D meeting and that brought me a lot of attention. There is potential for some cooperation with them. Actually we have a non-disclosure agreement with them to solve their instruments’ hysteresis problem. But they act slowly.
AH : What was attractive about DFM for you after the Ph.D ?
JFJ : That I could continue the work on image processing within the same field, SPM. Also, while working on SPM, I looked for a post-doc position. I finally found one at the [U.S.] National Institute of Science and Technology, which had relations with DFM, because both are metrological institutions. At NIST they also worked with SPM and I could do some image processing there. So, it was a very attractive position for me.
AH : Did they have a big outfit for STM problems at NIST ?
JFJ : Not that big, but they were working on what I would call a high risk project : a "molecular measure machine". The idea was to measure several milimeters across while keeping atomic resolution. At least atomic resolution it is very hard to measure even 100 nanometers across. In addition, they not only wanted atomic resolution but also to use interferometers and things like that in order to make all the accurate measurements. Of course, the more equipment you add, the heavier the construction and much can go wrong. So, it was a high risk project. I actually think it is still running. Obviously they have learned a lot by working on these complicated projects and I contributed a little software.
AH : How long were you at NIST ?
JFJ : One year. I came back and worked at DFM for almost two years.
AH : And you started your own company. Thinking about intellectual property, how have you managed this ? For example, write software at NIST, they presumably get the rights ? How does this work ?
JFJ : Well, I carried most of my software with me, so it was mostly a matter of porting it to run on their Sun machines. The difficulties centered on being able to read and handle their special file formats. So, I wouldn’t claim that I made an invention while at NIST.
AH : The experience at NIST rather taught you something about generalizing your software to be used on various other systems.
JFJ : Yes.
AH : Presumably that is a continuous story now that you put it on Windows, as you mentioned.
JFJ : Yes, we jumped from Unix to Windows. The market for Unix was not that big even then and it is not really growing.
AH : So what did you learn from the NIST experience ? What did you take home ?
JFJ : Well, a lot of connections. Very often, the best you get is getting to know people and to discuss problems with them. I recognized some new problems within SPM I hadn’t known about before, which of course leads to new ideas for solutions.
AH : For example ?
JFJ : We are getting deep into the way you scan. Scanning with a tip, you might have some friction making the tip bend. When scanning from left to right it bends one way, and when scanning from right to left it bends the other way. At the time most people thought that it was only in the AFM that you have these problems. Probably the most important thing I learnt was from their instrument that was more accurate than any I had ever seen before which enabled us to track very small residual errors. This showed me that the problem just mentioned is more general. We needed algorithms to improve the accuracy in order to measure at the sub-pixel level. Some of the algorithms which we have developed more recently for metrological systems are based on my experience at NIST.
AH : Let me make sure I understand. How would you know about the friction ? From a systematic difference in the scanning this way and scanning that way.
JFJ : Yes, you can see systematic dissimilarities between left to right and right to left.
AH : But that would only work if you stay in one line and just go to and fro, right ?
JFJ : You could do that but you can also take every second line. Usually, if you make an image, it’s only from left to right. They were scanning one line left to right and then the second line back the other way, and you can address all kinds of different hysteresis problems. And if the lines are not aligned then you need to analyze how much the odd lines shifted compare to the even lines. That’s very technical.
AH : But this would work only for a homogeneous surface, wouldn’t it ? You know, if you have a very inhomogeneous surface where every line is different, you can’t really tell what is the error and what is a sample, right ?
JFJ : Yes and no. The best thing is of course if you have something like a test structure on a homogeneous surface, but the closer the scan lines are together, the higher is the correlation, so by image processing and correlation techniques, you can actually correct for them. For instance, you can take every second line and make a cross relation and from that see that the line in between had been shifted, maybe just by 0.5 pixel.
AH : I see - and this shift is likely due to the bending of the tip scanning the sample surface ?
JFJ : You can hysteresis in more ways. You can have it in the piezo itself and indeed in other mechanical parts.
AH : And so what you have to do is to identify all the various kinds of hystereses and adjust ?
JFJ : All the non-linearities.
AH : Okay, let’s get back to the chronology. The STM project continued at DFM after your post-doc ?
JFJ : Yes, I got a permanent position which included other tasks. I became head of the consultancy section. So, I was able to continue my image processing work, but only part time.
AH : Consultancy for whom ?
JFJ : Everybody who wanted to buy our services. So, of course, one service I wanted to sell was image processing.
AH : Any image processing ?
JFJ : Yes, but we particularly wanted to sell our SPM expertise. People who come to DFM to get some images of a surface and a report. We had a few jobs, and the number of these jobs have increased.
AH : I am curious about the volume of this demand over time, if you have a sense of it. You started in 95 ; presumably there were very few few companies in 95 asking for such services ? Actually, was it companies that came for your services or government bodies ?
JFJ : No, it was university research institutes. They are still the majority but more and more high-tech companies are now using SPMs.
AH : Do you have a sense of how many SPMs there were in Denmark in 1995 ?
JFJ : I think we arranged our first user meeting in 93 or 94. We were only 20 or so ; in 96 maybe 50 people attended our user meeting.
AH : And each person corresponds to one instrument ?
JFJ : No, no, it’s hard to say, because each one of those people may have had five, six, seven, eight instruments, or even more. But still there weren’t that many in Denmark and there aren’t even today.
AH : Can you put a number on it ?
JFJ : No, it’s hard for me to put a number on it. I know for certain that DFM has four which I use. One is shared with DTI, the Microelectronic Center and the physical department here have one. I am not actually sure whether it is in working order, but they will eventually build a new one.
AH : They will build a new one ? They don’t buy off the shelf ?
JFJ : They build it or buy from the Aarhus group.
AH : The DFM services are for people who come with their data right ? They don’t come with an STM, so that, say, you would build the software into their software package.
JFJ : No, we never did that. Of course there are many ways you can use USB expertise. But the kind of job I had was like the one where I simply received some images from a company in America and had to give some feedback on distortions and such things. So, they can bring images that I analyze. Other customers came with surfaces to be measured and analyzed. We SPM-recalled and analyzed.
AH : Working here at the DFM, did you use your software in real time, adjusting while the SPM scanned ? In other words, did you build the software into the SPM ? Or did you measure the data first, and then run the software on the data set ?
JFJ : I never got to build the software into SPM. Before I left for NIST, we had started a project at DME with the intention of integrating our hysteresis algorithms into their software but we never finished. I forget why, but it also had to do with my going to America.
AH : How much of the starting your own company was your own desire to do it and how much was the current pressures in academia ?
JFJ : There was no pressure
AH : None at all ?
JFJ : No, no. You have to fight for what you want.
AH : You wanted to start your own company ?
JFJ : Sure ! And nobody pushed, not at all.
AH : In that case, the market issue we broached briefly before must have been very important. The number of SPMs in existence must be a crucial one for you. But we talked only about the number of SPMs in Denmark, and obviously you want to sell globally.
JFJ : Oh yes. Denmark is very small market. We could almost exist without it.
AH : What precisely are you selling ? The service of unscrambling data or the software ? I have noticed that you do offer free downloads on your website. What’s the business rationale for doing that ?
JFJ : Well, we are not doing any conventional marketing, so this is a actually crucial part of the marketing. People can go by themselves and find the product and try it and if they are happy, they might want to buy the full package.
AH : Okay, it’s a test, it’s a demo.
JFJ : It’s a demo. It’s very easy to give them the full version, we can just email them a key and they can implement it in ten minutes.
AH : Do you have any sense of how many people download this program ?
JFJ : Yes, around 2000 different people.
AH : 2000 downloaded demos ?
JFJ : Yes. We are preparing a release and once we have it, we will send an email out to to everyone who has downloaded in the past and then I will be able to update on our numbers. It has been a long time since I checked but I think it will be around two thousands.
AH : Presumably the idea is to sell new versions and upgrades ?
JFJ : We are working very gently with this. When they buy, they get one year of free upgrades. It’s part of the deal that whenever we have improvements for the modules they bought we send them without further cost for the next year. This is big selling point because we are upgrading very fast. We don’t want our customers to have an obsolete product.
AH : So this a long term business plan ? I mean, small businesses don’t usually make a profit in the first year, and one couldn’t expect that of you either ?
JFJ : That’s true
AH : Are you in profit yet ?
JFJ : Yes, right now we are.
AH : Wow, that’s great !
JFJ : We had some positive surprises in February [2001]. So, it’s quite a good development ; it keeps us busy.
AH : How many people are you ?
JFJ : Only three. I have had a permanent employee for one year now and we have also been using some students who have since left. No we have three permanent people.
AH : And you are all software people ?
JFJ : Yes.
AH : No marketing people ?
JFJ : No, not yet.
AH : So that will happen ?
JFJ : Yes. It has always been part of our business plan. But the timing is not right now.
AH : How do you work out the business structure with the not-for profit institutions that you collaborate with ?
JFJ : It’s not a problem. That’s why we are here. It’s definitely an advantage. It’s most important for us to have somebody around us using the software heavily in their work.
AH : They can tell you about the troubleshooting ?
JFJ : Yes, but not only errors, also new ideas and feedback on what they need. That is very important to us and we encourage all of our customers to give us feedback. New ideas are often built into free upgrades.
AH : What’s in it for the Danish Technical University to have you here ?
JFJ : I do not know. Of course, we are paying rent here and we can contribute to their visibility, showing that they are helping new companies. Actually, we are talking about the Danish Institute for Fundamental Metrology which is not a part of DTU. It’s like DTI, a self-owned technological service institute ; most of their income stems from project money from the government. We have a license agreement with them. They get 5% of everything we sell and they are a part owner.
AH : Are they under pressure to show their relevance for Danish industry ? Is that what you meant by visibility ?
JFJ : Sure, if you can show that a successful company was spun off an academic or government institute the latter would have an easier time getting money later on.
AH : What competition have you got in this field of unscrambling SPM images ?
JFJ : I think we are probably the only company making software exclusively. There are some other companies selling software to go with hardware, the latter being their main business.
AH : Such as Thermo Microscopes, and Digital Instruments ?
JFJ : Thermo Microscopes are only selling software to their own customers, I think. There is a small Spanish company coming up with some software right now. So, I have to consider whether they are a competitor or not. When I started I expected my major competitor to come from the United States. Don Chernov has a company called Advanced. He had something and it was very expensive - it was still in the DOS world. I think it has changed since then and in this respect I don’t really regard them as competitors. So right now we are sitting pretty and have only a few competitors. And for those people who really want accurate and serious measurements, the competition is very small.
AH : So you are not worried about people downloading and doing some reverse engineering and so on ?
JFJ : One should of course always pay attention to such issues.
AH : Thank you very much !
Fin de l’enregistrement
