The h Bar

DanielSank

04:10

@EmilioPisanty you around?

Emilio Pisanty

05:56

@DanielSank vaguely

Why?

Yashas

07:16

Is there a way to modify the close vote reason? I accidentally clicked on X in an attempt to mark Y.

John Rennie

@Yashas no there isn't. I've found myself in exactly the same position in the past.

John Rennie

07:54

If you have the free time I strongly recommend Graham Farmelo's new book The Universe Speaks in Numbers. My review is here.

I'd guess for most of us only the last chapter will be new, but the book is well written and entertaining even when it's covering ancient history like Dirac and Einstein :-)

pss 1

@JohnRennie sir may I know in which field of physics you worked..

John Rennie

@pss1 colloid science

Interface and colloid science

Interface and colloid science is an interdisciplinary intersection of branches of chemistry, physics, nanoscience and other fields dealing with colloids, heterogeneous systems consisting of a mechanical mixture of particles between 1 nm and 1000 nm dispersed in a continuous medium. A colloidal solution is a heterogeneous mixture in which the particle size of the substance is intermediate between a true solution and a suspension, i.e. between 1–1000 nm. Smoke from a fire is an example of a colloidal system in which tiny particles of solid float in air. Just like true solutions, colloidal particles...

Slereah

morning

The East Wind

@JohnRennie Hey

@Slereah Morning

John Rennie

@TheEastWind hi :-)

The East Wind

08:01

Just a quick doubt. What would happen if we perform the double slit expermient but instead of the screen used to obtain the interference patter we used a metal instead

John Rennie

@TheEastWind Do you mean a metal plate with the two slits cut into it?

The East Wind

I mean would we observe the photoelectric effect (provided the light is of sufficient energy) or interference patterns?

No I'm talking about the final screen where we obtain the pattern

John Rennie

I guess there would be photoelectrons, but they wouldn't affect the diffraction pattern. Light would still go through the slits and interfere to form the pattern on the screen.

The East Wind

But I read that at any given instant in time light can only show one type of nature?Either wave or particle.

Wouldn't what you said imply that light behaves both as a wave and as particles at the point where it strikes the screen?

John Rennie

Each individual photon can either pass through the slits or it can strike the metal and produce a photoelectron.

The photons that pass through the slits behave like waves, while the photons that eject photoelectrons behave like particles.

The East Wind

08:06

I am talking about the screen where we obtain the pattern. After the light has gone through the slits.

John Rennie

As a general rule light behaves as a wave when it's propagating through space and it behaves like a particle when it is exchanging energy with something.

The East Wind

So that would mean no interference pattern?

John Rennie

A photon that passes through the slits behaves like a wave when it is travelling between the slits ad the screen. Then when it hits the screen it transfers its energy to the screen and behaves like a particle.

The interference pattern is created in the region between the slits and screen where the photon is behaving like a wave.

Slereah

why must we spend all our life discussing the double slit experiment

7

Death to QM I say

Let's all discuss metrology instead

The East Wind

In the region b/w the slits and the screen? In thin air?

Isn't it created on the screen which is placed past the slits?

John Rennie

08:12

Suppose the screen is a photographic plate. Photons interact with the plate like a particle, so one photon will create one spot on the plate i.e. it behaves like a tiny particle hitting the point at the plate and leaving a mark at the impact point.

But when you send lots of photons you find the dots build up to form the diffraction pattern.

There is a nice illustration of this here:

50

A: Is it wrong to say that an electron can be a wave?

What is a wave? From sound and water waves we come to an association with sine and cosine variational behavior. Wave equations are differential equations whose elementary solutions are sinusoidal . In water waves and sound waves and even electromagnetic waves what is "waving", i.e. has a sinuso...

That's for electrons not photons, but the principle is the same.

The East Wind

Now those are interesting images..

"There is obviously interference, yet each particle also collapses to a single point when it hits the film. Which means that when the interference occurs each particle is travelling alone thus not being interfered with by anything else. Each particle, in essence, interferes with itself.

Every so often people would come up with new interpretations of this phenomena. It used to be said that each electron's probability distribution act as a wave thus interfere with itself. Now it is said that electrons are just excitations in the electric field "

That's something definitely weird..

John Rennie

@TheEastWind it's less weird than you think.

The East Wind

Dont we sometimes stretch models/assumptions too far, at the point wehre they seem(at least to me) intuituvely wrong?

John Rennie

Particles are described by quantum field theory, where they are excitations in a field. In some circumstances it is a good approximation to describe as particles and in other circumstances it is a good approximation to describe them as waves. But both of these are just approximations.

It is important to understand that an electron or photon isn't sometimes a particle and sometimes a wave. It is a quantum field that sometimes behaves like a particle and sometimes behaves like a wave.

The East Wind

Dont we abuse the meaning of the word "behave"?

John Rennie

08:25

The theory that describes electrons and photons is quantum electrodynamics and this is around 80 years old so it's a well tried and tested theory.

The East Wind

I mean a LOT of chemistry is explained by thinking of electrons as particles(basicity of compounds,electronegativity and so on). How would we able to explain these effects by thinking of electrons as waves(even approximately)

John Rennie

@TheEastWind a lot of molecules have delocalised electrons. How do you explain that if the electron is a particle?

The East Wind

I mean that the lone pair of electrons are particles, which can form a pi bond between two atoms.

John Rennie

@TheEastWind particles have a well defined position. Waves don't. So if an electron is delocalised isn't it more like a wave than a particle?

The East Wind

dont particles also only have a probability of being somewhere at any point in time. So that randomness in position can be explained via this?

Mikhail

08:48

I'd say the novelty was that these objects were discrete. Your detector could catch/eat/consume a discrete quantum of the field, it would also seemingly "disappear" from the other detector, something that classical fields don't do. Hence the "mechanics" of "quanta".

Also the vacuum noise, which leads to a lot of fun things like the Casmir effect, or the less well known tendency for light in a vacuum to experience chromatic dispersion because it hits stuff.

0

A: Polarization in high energy vacuum non-linearity?

You need Quantum Electrodynamics (QED) to understand vacuum polarization. In a nut shell, imagine that the vacuum is filled with virtual electron-positron pairs, and as the high-energy light propagates it interacts with these pairs and they separate creating a polarization-like effect. In some ...

Physics Meta

11:41

0

Q: How do I retrieve the starred questions?

I hope this is the right place to ask. If not I will delete and ask somexhere else (where though?). I noticed there is a star I can click next to vote arrows close to questions title. So I thought I could bookmark them this way, but I cannot see any effect, nor can I find a list where those marke...

support

Secret

12:16

I really like to think quantum fields are neither waves nor particles. It just when given a correct circumstances, it interacts in a way that is similar to how particle or waves interact with something, but it is always a quanutm field, never suddenly becoming something else

The thing that makes QFT much less straightforward is there isn't really much place one can build an ontology from the mathematical formulism without risk overinterpreting with classical intuition

e.g. as Acuriousmind once reminded me, that the quantum field is really an infinite dimensional generalisation of e.g. momentum operator, thus nothing is actually physically produced when we operate a quantum field onto a quantum state in the mathematics. It is only the expectation value that has physical meaning

JMac

I mean, a computer program which to the question of the form "Does x exist?" or "Does x not exist," returns the answer "x both exists and does not exist." is computable. That "phrase" or answer can be computed. Even though the answer itself is computable, it expresses something that is not computable (we cannot simulate a universe where x both is and is not). @Wolphramjonny — nielsbohrden 13 mins ago

This reminds me a lot of the person trying to push the idea that if they could get a computer program to print X = X and X =/= X it proved we were a simulation or something

Secret

The rarest thing in our universe is not antimatter

It's dialetheia

In fact, I will bet that given unbounded amount of time, the probability of having a physical infinity is much much closer to certainty than the probability of obtaining a dialetheia object

Secret

12:39

List of human thinking habits:

1. Love hierarchies

2. Love categories

3. Love ontology

4. Love asking why

5. Fear of the unknown and the uncontrollable

etc.

List of nature thinking habits:

1. Really, really, really have issues with letting diletheia to exist
2. Really love ensemble averages

FIN

Captain Bohemian

13:05

I like to explore the unknown but I really hate the uncontrollable.

the latter may feel like a quagmire where you are making Sisyphean effort constantly.

Captain Bohemian

13:25

the exploration of the unknown which leads to unveil something new to me usually delights me very much.

user351417

15:38

Any 10k rep users about?

user351417

Can someone look at physics.stackexchange.com/questions/478964?

user351417

Is it similar to physics.stackexchange.com/q/478991?

user351417

Ah, spectacular. Qmechanic posts a comment saying that the existing one is a duplicate of the deleted one just as I cast that flag.

user351417

Ah never mind. The post was undeleted : P

Kyle Kanos

16:02

@JMac not sure how a device built under the principle that the LNC is absolute could violate said LNC...not sure how a potential violation would also indicate anything more than a broke computer...

JMac

@KyleKanos That was what all the comments on the question had tried to explain, but the OP kept insisting that it "represented something physical" and therefore for some convoluted reason had to be describing something physical because he could make a program output seemingly contradictory statements

the worst part IMO was he basically just made it print out "A = True; A = False" and conclude that it necessarily meant the program considered it both true and false. From my limited understanding of code, it seemed more like they were changing A from true to false and printing both results

Semiclassical

16:17

@JMac seems like that's the equivalent of treating whether a coin is heads/tails as an immutable fact of the universe, and then acting shocked when it changes

Kyle Kanos

Yeah, print("True" if A else "False") can't print both answers, unless the underlying language is broken.

But you can probably make a computer language that doesn't care about LNC & output both. But it'd take special handling of the syntax

Semiclassical

LNC?

Kyle Kanos

And even then, I quite imagine the language bring basically useless

@Semiclassical law of non-contradiction

Semiclassical

ah

ACuriousMind

@JMac I can make a program print that by just telling it to print the string "A = True; A = False" :P

JMac

16:30

@KyleKanos It was programmed in C or C++ IIRC, maybe Python. I'm pretty sure it was at least a common language

Kyle Kanos

Okay, then it'd never work.

Though maybe one could do it in C with some clever, evil macros....

JMac

@ACuriousMind Yeah, they basically did that, but added in a step where it printed based off the status of a variable, and then just changed that variable between calls so that it was true, then false. I might be remembering wrong, but I'm pretty sure it was that trivial

Kyle Kanos

395

A: Loopholes that are forbidden by default

Interpreting the challenge too literally That is, if the challenge says "write a function that, given a number n, returns the n-th prime", posting something equivalent to: function f($n) { return "the $n-th prime"; }

ACuriousMind

@KyleKanos You didn't say "Standard loopholes apply", so I'm fine!

Kyle Kanos

But they're forbidden by default!

ACuriousMind

16:38

If true can be false, "by default" can be "not by default" :P

Kyle Kanos

My brain hurts reading that

ACuriousMind

That means it's working!

Stan Shunpike

16:55

hello

I am working with a 4-bar linkage system that i'm modeling in sympy. I've got the equations worked out for how the system behaves, but i have some questions about how dynamical systems work

Specifically, suppose I have like a pendulum with a spring.

I can plot a graph of the position of the mass at the end through time.

But let's suppose I change one of the parameters, like the length of the rod or the size of the mass or the spring constant.

Then the paths it would sweep out would be different.

JMac

Did you have a question based on that, or are you still typing?

Stan Shunpike

still typing

Now let's suppose I ran some experiments

and I generated some kind of distribution P(trajectory | parameters)

Now let's suppose the length, mass, and spring constant were all unknown. And I was given the path that a spring system swept out

Could I do the reverse and use the path to infer the parameters?

I just dont know what a "path" space would look like

and I'm a bit confused about how to think about that

cuz what I want is P(parameters | trajectory)

im done

JMac

What do you mean by P(Parameters|Trajectory)?

Stan Shunpike

17:17

um so like

let me see if i can give an example

Kyle Kanos

I wouldn't think there's a nice invertible feature/function to go from the two distributions

Your initial conditions also matter, especially in coupled systems

Stan Shunpike

so like if i was given this

or let's suppose several paths

there's no way to do some kind of learning that would allow me to estimate the parameters?

JMac

@KyleKanos When I studied mechanisms, we definitely didn't learn one. The process basically worked out to "play around with parameters until you get the desired motion", except for a couple specific mechanisms that were a bit more studied. Designing a 6-bar mechanism for example was just a mess of excel spreadsheets that related how each input would change the output, and we just adjusted the lengths and stuff

Kyle Kanos

Right

Stan Shunpike

for.rest/articles/The-Spring-Pendulum-with-Sympy

this where i got the plot

Kyle Kanos

17:25

You could try a minimization technique to get the parameter space down, but it's basically going to be getting the trajectory for each possible input, comparing the computed to the given & adjusting

And you'd have $n$ dimensions to scan through

JMac

@StanShunpike You mention you're modelling a 4-bar mechanism, are you modelling them as rigid members? If so, I'm not sure how related this spring pendulum example will be (unless you only intended that to show what you wanted to plot)

Stan Shunpike

@JMac Yeah, it was just to illustrate the general idea

i'm going to model several different systems

and i just wanted to know if it was possible to infer the parameters from a trajectory plot but it sounds like not

thanks for the info! saved me a bunch of time

JMac

Yeah, I would just work it the other way around with your code, make a code that can plot the trajectory based on your inputs, then just play around until you get the desired trajectory

Stan Shunpike

Ok, that makes a lot of sense. Do you know what kind of algorithms are used to like try to make a system like a pendulum match a particular trajectory?

Kyle Kanos

Generically, you're minimizing the difference of the two trajectories

So any minimization algorithm ought to work

(one trajectory is the input, the other is what you computed with the given parameters)

Stan Shunpike

17:33

yup, ok cool. i'll see what i can find. thanks!

JMac

@KyleKanos like make a program that generates the output based on input, and then use a numerical method that adjusts the parameters until the desired output is approximately right? Even if that's not specifically what you meant, I've never thought of it like that before but that makes a lot of sense

Kyle Kanos

That is what I meant. In 1D, we normally call these "root finders"

Finding the x s.t. f(x)=0

JMac

Yeah, that method makes a ton of sense. I don't know if I'll ever be in a position to apply it, but I need to try to remember that one. Almost tempted to bust out some old textbooks and try it out

Kyle Kanos

It can get hairy for multi-parameter searches, as it's possible to find local minima instead of global.

JMac

@KyleKanos I could see it being complicated to actually get a good implementation... but compared to trying to figure out an analytical method that can start with desired trajectory and output the mechanism requirements, that method seems pretty simple

Stan Shunpike

17:48

Ok so like let me lay out a bit more clearly what I want to do

because I went and thought about it a bit

and there could be some other issues

So basically here's what I'm going to do

I'm going to create some simple linkage system. For instance, like the 4 bar one, or a pendulum or whatever.

But I'm going to add "motors", which I will represent as torques at the joints.

And i'm going to initiate the system such that, one of the motors is broken

but that which one is stochastically chosen

and I don't know it

don't know which one it is

What I want to do is basically detective work

So suppose I had a double pendulum system

with a motor at each revolute joint

and suppose the motors have some program normally

but one of them fails or deviates some how

I want to basically try to deduce, by playing around with adding forces to the system or constraints to the angles

which motor is broken

vzn

@Stan hi havent seen you in ages. did analyze something vaguely similar in college mechatronics class, we used fourier analysis to find 1 out of 2 motors with messed up bearings.

Stan Shunpike

@vzn nice to see you guys too! well, i completed my undergrad in econ and decided i didn't like econ, and am going back to doing physics related things. currently taking a robotics class.

so here i am! lol

hmmm

vzn

@StanShunpike amazing that was going to be my next question! congrats on your new degree + return! robotics is such a great field too. for your problem my question is, its not clear what the motors are doing on the pendulums...?

Stan Shunpike

Well, I haven't worked that out specifically yet the mechanics. I'm envisioning a robot where a motor fails but a situation where we have no sensors and need to make inference.

Currently, this is my first robotics course and the professor emphasizes probabilistic graphical models, with little consideration for use in actual mechanical settings

His emphasis is more on topics like Kalman filters and SLAM

So I don't have a specific example off hand of what the motors are doing yet.

vzn

@Slereah because its basic yet still not fully understood + PS feynman loved it :P

Stan Shunpike

18:00

Maybe I need to figure that out first

@vzn also, Fourier analysis is widely used from what i understand in different areas. Is this application u used from a particular discipline?

vzn

@StanShunpike if the motors are doing something rhythmic and not random, then fourier analysis probably would do a lot. my "application" was maybe from control system theory or monitoring, mechanical engineering, mechatronics etc, not sure the exact subfield... also it sounds a lot like machine learning might be applicable to your general idea. its making a lot of new inroads into physics & many science areas these days...

Stan Shunpike

I'm starting my masters this fall and that will be my specialization

masters in statistics

how would machine learning be relevant though?

In my particular case, i will only be able to get a low sample size from the robot in question

so any large samples would have to be generated by simulation

vzn

@StanShunpike statistics, great choice! and ML is very strongly connected to statistics these days, aka "data science". yes "low sample sizes" can be a difficult aspect of robotics + ML... is this a masters project? simulations are increasingly acceptable/ applied in robotics... eg google is going a lot in that direction... its basically an ML classification problem...

Stan Shunpike

ohhhhhhhhhhh i see what you mean

that's smart!

interesting

didn't think of it like that

vzn

:) the idea is you have a lot of physical inputs into the ML system and it classifies as ok vs broken etc... via training... the time series aspect of this makes it harder for ML which is maybe better suited to "instantaneous" physical classifications...

Stan Shunpike

18:11

why would time series make it harder?

like so in my pendulum example, the pendulum swept out a single path. presumably that's like a static thing

so that seems like something you could feed into an ML alog

algo*

if you have trained the algo on a bunch of simulations

of different paths

enumaris

looks like data science is happening here...

vzn

@StanShunpike the ML has to monitor a timerange of data before it can make a decision, a single sample in time probably isnt enough to differentiate. yes the way to feed the time series data would be part of the analysis/ decomposition. you could maybe feed a fixed set of points/ samples over n seconds as the training vector etc.

Stan Shunpike

i could have like a multiclassifer type function with softmax

vzn

note that pendulum(s) are classic chaotic behavior/ dynamics...

Stan Shunpike

correct

enumaris

18:13

huzzah! data science for physics!

Stan Shunpike

@vzn so a lot of my intro ML classes were like centered on things like image processing or text analysis

yeah exactly @enumaris

so like

vzn

@enumaris right down your alley right? :)

Stan Shunpike

do you have any recommended resources for physics specific applicaitons

enumaris

I saw at talk at GTC about using computer vision to classify galaxies

Stan Shunpike

because my physics i think is fine for this but ive never used ML and physics together

enumaris

18:14

that's more of an astronomy application

vzn

@StanShunpike image processing is not so far away from this app because its basically training on sets of input vectors etc

Stan Shunpike

ohhhhhhh

true

enumaris

@StanShunpike yes, in Reinforcement Learning the Bellman equation is a Fredholm equation of the second kind. So you can do, e.g. light transport using reinforcement learning algorithms

:)

Stan Shunpike

brilliant! wow this is so cool

ML and physics, like my two favorite things

vzn

@enumaris yes ML is really taking off in cosmology! almost a killer (science) app for physics ML so to speak. but think it will have wideranging impact on many areas of physics & we're seeing early days.

Stan Shunpike

18:15

@enumaris wow that's amazing

enumaris

arxiv.org/abs/1712.06115

here's a paper about it from Nvidia

3-D rendering is all about light-transport after all :)

go, let your imagination go wild!

I have not had a chance to read that paper yet...but it's definitely on my to-do list

vzn

@enumaris maybe heard some about that. ML is going to be Very Big in scene generation in near future. reminds me, did you ever see this site? extraordinary, it was announced earlier this year thispersondoesnotexist.com

Stan Shunpike

@enumaris thats wild. it's amazing. because the examples in ML classes tend to ignore physics, which is actually a shame now that im thinking about it because its so useful for robotics systems as well as other things in physics

@vzn scene generation?

vzn

@StanShunpike ML is finally really starting to take off Bigtime in robotics, paradigm shift in play, have seen a few refs on that recently... am expecting really big breakthroughs in "motion planning" etc... you guys have seen boston dynamics spotmini havent you? very smooth/ lifelike/ biological-looking motions, basically all ML afaik... (would like to learn more about their algorithms)

enumaris

:D glad I could be of help

vzn

18:22

@StanShunpike billions of dollars are riding on "artificial scene generation" in video games and movies. ML is starting to make inroads as in the nvidia research. it can create more than just scenes, it can create artificial worlds...

enumaris

@vzn I did not see that site, but I went to the seminar where the Nvidia guy presented how they generated those images

they had to make some modifications to GANs to allow for high resolution image generation and style transfer

It's definitely a space I'm personally interested in

vzn

@enumaris am very excited by it too lately its a huge "gamechanger" o_O think its way cool that youre doing it on a pro level, the field is very vibrant/ dynamic these days, very long dreaming of working in it, but itd be a tricky "lateral" move for me at this point...

Stan Shunpike

18:38

@vzn @enumaris so what do you guys focus on?

like your areas of research / interest / expertise?

enumaris

@StanShunpike my current area of research is kinda like conversational AI...natural language understanding, human dialogue understanding...that kind of stuff

if I had to put my "expertise" into topics, I would say Physics, Deep Learning, and NLU

the 3-D rendering interest comes from my interest in deep learning

vzn

@StanShunpike my day job is java + database + ui, but like to/ enjoy tracking cutting edge stuff from CS for yrs etc... recently researched/ blogged quite a bit on AGI...

Physics Meta

19:34

0

Q: How does the reason for holding a question correspond to the actual question?

My most recent question (Why do we assume fundamental properties like charge or mass do not consume anything?) was put on hold given the following reason: "We deal with mainstream physics here. Questions about the general correctness of unpublished personal theories are off topic, although specif...

support

Emilio Pisanty

19:52

9

Q: What does "finite but unbounded universe" mean?

In Einstein's book about relativity, he says that his theory predicts that the shape of the universe would be finite but unbounded. But how is this possible? What's the difference between an infinite and an unbounded universe? Therefore, what does it mean for a universe to be "finite but unbo...

general-relativity cosmology differential-geometry universe topology

Good job by @tpg2114 protecting that one

JMac

Is that the one that just had a nonsense answer in the LQP queue? I remember recently reading something from that in my queue

Kyle Kanos

20:50

Sigh... Ought to stop & finish reading before laughing at the way it's written thus far

Though kinda curious that the three deleted answers on that thread are all from 1 rep users.... wonder if they tried answering other questions...

enumaris

I can't view the deleted answers D:

Physics Meta

21:32

0

Q: Is the « peer pressure » badge something to be proud of?

This may sound like a silly question but I believe it is not. I recently deleted a question in order to rewrite it, as there were (a lot of) concerns regarding the definitions used and more generally its tone and clarity. But the system has no way to tell if I intend to undelete at some point or ...

discussion

PM 2Ring

23:04

There's also a current question about the boundary of the universe in Astronomy: astronomy.stackexchange.com/questions/31807/…

danielunderwood

@enumaris I first read that as "controversial AI"...

heather

@StanShunpike I would just like to say that your username's great =)

Transcript for