Hacker News Reader

Comment by mathattack | Hacker News Reader

I'll be trying to explain the history of renormalization in quantum field theory

I'm hard pressed to think of a situation where this would come up in a conversation with non-scientists.

As a PhD candidate in high energy nuclear theory, I've observed a lot of 'math wacking', i.e. reading theory papers which include pages and pages of detailed equations without a single experimental data point. Just take a look at arXiv/nucl-th if you want to see what I'm talking about. These same guys go to conferences, poo-poo less first-principle calculations which at least attempt to describe experiment, and espouse their untested pet theories as if it were gospel.

To be honest, I've been extremely disappointed by the scientific process in high-energy particle physics. People are so tethered to their theory convictions, that they accept derivations as physical realities and dismiss the importance of experimental validation. The old farts in the field are also completely averse to computational/numerical methods and dismiss the importance of production quality code development.

So while I embrace the sentiment that 'math is the essential language of physics', if you can't embed the mathematics into a simulation which describes reality, then what are you doing as a scientist?

The space of possible experiments is enormous. If someone takes time to find a conclusion that the theory predicts from the first principles but for which experimental data are lacking, may it be somehow valuable? Especially if the theoretically predicted effect is somehow interesting.

The goal is to disprove theory's not to validate them. Remember any valid counter example can disprove a trillion pieces of supporting evidence.

http://www.nytimes.com/interactive/2015/07/03/upshot/a-quick...

You can always over fit for existing data. Useful theory's should suggest a novel experiments to test them.

The experimental data may prove or disprove a theory. But when it's lacking, there's no way to tell. So finding areas where experiments were not made, but can be made, and where the theory predicts some clear results, can serve as a means of falsification.

I agree that they hate production quality code and numerical methods, but in some of these DFT codes for example (HF is only somewhat better), you fix some numerical bug and a bunch of theoretical results agreeing with experiment go straight out the window and with it your confidence in the computational model. These 'old farts' have probably seen this happen way too many times and on top of it from people who lack the basic theory knowledge, let along experimental, to point out a crap result from the code.

I can see it happening. As a non-physicist programmer who knows a little bit about a little bit through popular science (like the people being discussed in the article), I know "just enough to be dangerous" about renormalization. i.e., that it was for a while a controversial practice derided by opponents as "handwaving problems away with division by infinity".

I also know that much smarter people than me don't consider it a problem anymore, but it's exactly the kind of gap in the door that self-taught "experts" would use to say that the whole edifice of science needs to be torn down as incorrect and replaced with their personal theory.

"handwaving problems away with division by infinity"

I'm being very pedantic here but it's a bit better to think of it as subtracting infinity from another infinity.

eh, raise the terms as an exponent and subtraction becomes division. :)

In log scale yes.

Renormalization is not the same thing as regularization. Regularization is a mathematical trick while renormalization has a very physical interpretation, for example, its use in statistical mechanics. In QFT, they are related and I think sometimes conflated, which is unfortunate.

You can easily describe renormalization without math because it's a physical concept. Regularization is more difficult. If the OP decided divergent integral regularization was an interesting conversation topic then I am on your side.

Please do then. Explain it without math, I mean.

The entire content of renormalization is summarized as this, given a calculation, you make it match the interaction at some energy. Assuming that the fix is essentially choosing the "scale" of quantities (say, voltage of the electric field, or how strongly electrons tug on other electrons) you are calculating, you fix that scale by making sure that after including it in your theory now, that your result matches the measured interaction at your chosen energy scale exactly. That is renormalization.

If you do this, you find that you need to choose a scale that "absorbs" the infinities of your calculations (yes, an "infinite" scale). But, the stuff that is not absorbed is a function of energy, and this actually agrees very well with experiment beyond the scale you normalized to.

Pretty sure he means people like you when he talks about his non-physicist friends.

If it's a part of Xcelerate's work, and he knows people who are not physicists, but are genuinely interested in what he does all day, I can see it coming up.

You'd be surprised, a mutual friend brings up your occupation (I'm a computational astrophysics grad), and then they inevitably ask you about black holes, and then you have to start delving into pretty complex topics.

Reply on news.ycombinator.com