From working with Richard Feynman at Caltech to recently winning the Dirac Medal, theoretical physicist Deepak Dhar in conversation with TNIE’s Tamreen Sultana shares about his journey, experiences in research and what it takes to pursue science in India.
What sparked your interest in science as a child?
When I was young, my father encouraged me to follow science. He would bring a British popular science magazine called ‘Understanding Science’, which was a bit expensive by Indian standards. Of course, it was in English. I had to first learn the language, and then to understand the concepts. Eventually, I became better in English, and also developed some interest in science.
You studied at Allahabad University, IIT Kanpur and then Caltech. How did these places shape your approach to physics and research?
All of these were very important in developing my understanding and knowledge of basic science. Caltech is certainly one of the world's best places for this kind of work, so I am very happy to have had the opportunity to learn there.
You worked with Richard Feynman at Caltech. What did you learn from him?
There were lots of things to learn from him, even just by watching him. The first thing I recall about him was that he was not very pompous and was very kind to students (unlike some other people we meet!).
He was also able to explain things in very simple terms without using too much jargon. I learnt the importance of making sure that students understand something well. And the importance of constructive criticism, which helps them do better.
Why did you return to India after your PhD?
When I finished my PhD, I looked for post-doc positions at a few places in the US but didn't get any. Then there was an offer of a post-doc position from TIFR, that seemed like a good choice, so I came back.
Is there a fundamental difference between doing a PhD in the US and India?
I don't think making a very general comparison between a PhD in the US and a PhD in India is very useful. There are some very good places in the US and there are lots of not-so-great places there as well. Similarly, there is considerable variation in the quality of institutions in India.
Certainly, the best universities in the US are better than the best in India. I do not think many Indians will disagree with this statement. But several institutes in India are very comparable to institutions among the top 50 in the US, though perhaps not the top five.
For a general reader, what is statistical physics trying to explain?
We know that bulk matter is made up of lots of atoms. Being able to predict the properties of bulk matter is part of understanding the world around us. We should be able to predict properties such as what happens at a particular temperature, or the behaviour of water, oil and things like that.
If we know the behaviour of water molecules, in principle, we should be able to predict what will happen if you have a large amount of water. But it turns out that this is very hard to do using the exact equations of motion.
The great advantage of statistical mechanics is that you don't need to know the exact state of every individual particle. Most of the time, even if you don't know the exact state, you can predict some properties quite reliably. For example, pressure can be predicted quite well, as can temperature and other quantities.
That is what statistical mechanics does: it says that many properties of a large number of particles can be predicted even if the individual behaviour of each particle cannot be predicted.
Could you explain the Abelian sandpile model in simple terms? What were you trying to understand?
A property of the sand that is familiar to everybody this: you take a pile of dry sand and add grains to it very slowly. When you add a grain, you cannot control its position very well. Sometimes the grain will simply sit on the pile. Sometimes it will move a little and settle there. Sometimes it causes a big avalanche. You may drop one grain, but 100 or 1,000 grains can come down.
What is happening is that you have a system which is being driven by adding grains, but its response is sporadic, irregular and burst-like. Sometimes nothing happens and sometimes a large event occurs. This is not very periodic. The interval between events is random and not easily predictable.
These kinds of avalanche-like relaxations are what people were trying to understand. They made a model called the Abelian sandpile model. I will not describe the model in great detail here, but what they could show was that the model had this basic behaviour: you add one grain, sometimes very little happens and sometimes a much bigger avalanche happens.
But this had only been shown using computer simulations. There was no theoretical way of predicting this behaviour. What I did was to show that you can define some quantities called addition operators. In this model, these addition operators have the property that, if you take a configuration of sand and add a grain at one place A and then another, at place B, or add the grain at B first and then at A, the final configuration is the same after the avalanches are over, and sandpile has a stable configuration.
This property — that the final answer does not depend on the order in which you add the grains— is called the Abelian property.
Using these properties, along with other properties of the model, various quantities could be theoretically calculated. This was the prototypical model of these kinds of driven systems, which are called self-organised critical systems. It was useful in helping us understand the general behaviour of such systems.
There are many other examples of such driven systems, such as earthquakes and solar flares and rain.
Is the randomness in the sandpile model fundamental, or does it emerge from rules?
Of course, randomness in the model is put in the rules. There are two types of randomness in this model. One is that the configuration on which you add one more grain is not fully specified, so it is random. Where you add the grain is also random.In addition, there may be randomness in the toppling rules. So there are different types of random elements, but the net result of all of these is the same: avalanche-like relaxations. The detailed behaviour of these avalanche-like relaxations may vary depending on the details of the system.
You are only the second Indian to receive the Dirac Medal. What does the recognition mean to you, and could it help theoretical physics in India?
I am very happy to realise that some work which I did around 35 years ago is still remembered and recognised as important. That is a matter of some satisfaction.
Whether it will help other physicists depends on what future brings. If the award helps convince the government to increase funding for science, then it helps. If it doesn't, then it doesn't help. By itself, an award cannot improve research, but if it makes decision-makers change their priorities, then maybe it will help. Another possible positive effect could be that the career preferences of some promising young people may be influenced by such awards.
What were you doing when you learnt you had won the medal? Did you expect it?
When we do our work, we are not looking for awards. I knew I was nominated for the award, but did not think that it was likely to happen. There are always a large number of very good nominees for such awards, and the probability of success is not high.
When I heard the news, I was sitting in my office and checking my email. One of the emails said that I had been selected for the award; a few days later, the director of ICTP would call me, but I should keep the news confidential till the official announcement.
Your medal citation spans equilibrium and non-equilibrium statistical mechanics, optimisation, neuroscience and AI. Is there a common thread?
I think this part of the text from the citation that you mentioned mentions the work of all the winners. My own cited work comes from different periods. In general, I work in statistical physics and the properties of systems that vary in time. These work evolve, but depends on earlier work and ideas. The general theme in my work is to study how non-equilibrium systems behave and organise themselves.
Is there an unsolved problem you would still like to pursue?
There are a large number of unsolved problems. There are several of them which I have tried and failed to solve.
So I don't think this is the right place to make a list of these problems. But yes, there are many problems, not just one, and I would be very happy to solve any one of them.
So far, I have not been able to do it. Maybe I won't be able to, but I can keep trying. That is the way we work in general. Once in a while, if you try a problem, maybe you succeed.
AI is increasingly being used to solve complex mathematical problems. Will it change how researchers work?
I don't really know too much about AI, but my knowledge is based on what I read and hear. I think AI gives you the description, but it doesn't give you any understanding as to how that description and the predictions arise. I was brought up to think that a scientist should look for understanding somethings about the world around us better, and that is what scientists strive for in their lives, successfully or not. I don't see AI providing much of this understanding.
What would you tell young researchers who feel pessimistic about doing research in India?
I would say that there is a glass which is either half full or half empty. Of course, things are not great here in India. But I think they are certainly better now, than they were in the time of S N Bose, Raman or Saha, who managed to do great work in spite of the conditions, because they overcame the hurdles in their paths.
So I would say that things are certainly not as good as they could be, and they would be better if they improved. But for one individual person, it is not good to put the entire blame on other people and the social environment for one’s failures. Some `zamane ko badal dete hain’ spirit is needed.
