 |
Each year on the tenth of April, I reminisce about the public defense, decades ago, of my doctorate thesis at the University of Twente. I was already incorporated in the army at the time, and changed uniform for dress suit for one day. It was the closure of four years of painstaking work. However, a research project resembles a book or a movie, in the sense that it is never truly finished. Several years ago I made a digital version of my doctorate thesis. Why would I republish my dissertation many years after graduation, without an actual need for revision of its content? There are two reasons for this.
First, I found while moving to a new home, forgotten diskettes with the sources of simulation software as well as most of the sources of my thesis. Luckily I still had a drive that could read such media. It proved straightforward to compile on Linux the simulation software that I wrote for UNIX. Since my research concerned a subject in computational physics I had actually resurrected my former laboratory, at home. You will understand that I could not help myself revisiting the work I did twenty years earlier.
Second, I wanted to convert the thesis to a modern format for distribution through digital media, but it proved not possible to produce an exact digital copy. Original typesetting and plotting was done with decommissioned software, so text and data had to be converted. Also, graphics had to be made anew, as original postscript files were lost.
Performance of workstations has increased so much, that what took a year around 1990, can now be done in a few days. I leisurely spent free time to find answers to some questions that were left to rest after the graduation party was over. This gave me the opportunity to resolve some loose ends in the thesis the way I had in mind when writing the original version. Finishing my dissertation before I was incorporated in the army was a race against time - I was new to rheology and the rheology department was new to simulations, computer capacity was but a small fraction of what it is today, software had to be written from scratch, and Brownian dynamics was a relatively new technique in microrheological modeling.
I included some of the newly found results in the revised text with the sole intention to complete the work where it was left open in the original version. Of course I preserved the essence of the original thesis and kept it consistent with state of the art when the original manuscript was written, otherwise the material would have become anachronistic. For example, at the time it was not yet feasible to incorporate long-ranged hydrodynamic interactions in a non-trivial nonequilibrium three-dimensional simulation, so studying the neglect of hydrodynamics, a central theme in my dissertation, was a legitimate research subject at the time.
Allow me to digress to explain an interesting development in the late 1980's that saw the onset of decentralisation in academic computing. Up to that time, number-crunching jobs were typically run on central mainframes. When I did my research, workstations became sufficiently powerful as to start competing with mainframes for many applications. Researchers were quick to adopt these workstations as budgets for computer centre facilities generally were limited as those were shared by many. That is one of the reasons Fortran was a dominant language - it gave researchers most results for the allotted resources.
The photo in this paragraph shows an example of such workstation. I acquired this particular one since it was a cost effective alternative for computing budget on the VAX 8650 at the university's computer centre. This desktop could run many of my simulation jobs, leaving it processing round the clock, all days of the week. Colleagues were intrigued to see simulations running in real time. My largest jobs ran on supercomputers - a CONVEX 220 at the university and a CRAY Y‑MP at SARA, Amsterdam. For these large batch jobs, the workstation functioned as a terminal to dispatch jobs and retrieve output, and for data processing afterwards.
The new digital publication would be what I wanted my thesis to have looked like. Of course, this new version is not the one that my committee approved, though I like to think that the committee would have agreed to this new edition since it arrives at the same conclusions based on the same arguments plus extra results demonstrating that the original conclusions would not have been different if we would at the time have had these extra results. This can be confirmed by comparing the digital version to the original printed thesis. So finally, after all these years, I could consider this research project as satisfactorily complete.
Recovering the old diskettes did not just yield a digital version of my thesis. Later on I used the simulation software as a platform to experiment with simulation techniques. After that I built a Beowulf cluster and wrote from scratch multi-processor simulation software to run on it. This setup dwarves the supercomputers mentioned earlier in this post. If only I would have had such facilities at my disposal when I did my doctorate research …
Studying rheological behaviour through computer simulation - a scientific activity that should be called computational rheometry - in particular shear induced ordering in colloids, continues up to this day. I found that my thesis was not the only one that is concerned with the implementation of a simulation method for dispersions, exploring how to operate it such that reliable results are obtained, and then connect results to theory and experimental results.
Even though my career turned away from this research subject after military service, it is gratifying to see that we were on the right track, and did what we could do with the hardware at hand, circa three decades ago. Making a digital edition of the thesis brought back vivid memories of those hectic years of youthful eagerness, working long days and dealing with serious setbacks, to have in the end a learned committee approve our academic rite of passage.
|
|
© 2002-2024 J.M. van der Veer (jmvdveer@xs4all.nl)
|
