The h Bar

user54412

00:04

@FenderLesPaul Don't let profs push you around like that. Remember, they've seen hundreds of students come and go in the department -- it's not like they can only attain fulfillment in life if you in particular work with them.

0celo7

@ACuriousMind one of them involves a girl, and sadly she doesn't do diff geo

user54412

@ACuriousMind what if the differential geometry was index-based?

ACuriousMind

@ChrisWhite Ugh.

Tough choice, then ;)

user507974

00:21

it is a weird day for me... i might actually go bug people on the chemistry SE for a bit

@FenderLesPaul my vote is for ucsb

@BernardMeurer does @danielsank go to ucsb?

not sure why hed be sad otherwise

ACuriousMind

00:54

@user507974 He works at the Google lab in SB

and I think he went to UCSB before that

user507974

yea, Martinis group

seen presentations of theirs before

i was really annoyed because the nanosystems consortium reorganized itself this fiscal year and so they are not offering any internships for the summer due to the restructuring

and since i graduate ill be ineligible next year

DanielSank

01:22

@user507974 I was a grad student there. I still live in Santa Barbara working at the Google Quantum AI lab here.

user507974

@DanielSank when you say AI lab is this a seperate facility in downtown

or is it somewhere hosted in the CNSI

DanielSank

@user507974 It is neither down town nor at CNSI.

user507974

Goleta

@DanielSank im just thinking because I haven't heard about a google lab in the area

DanielSank

@user507974 We've only been here about a year.

user507974

@DanielSank oh ok,

hows it at google?

DanielSank

01:30

@user507974 Awesome.

Dream job.

user507974

@DanielSank what are they having you do there (that you are allowed to tell me about)

DanielSank

Surrounded by intelligent motivated people. Resources. Directed goals but lack of micromanagement so we can do what needs to be done to achieve those goals.

user507974

obviously its related to their qcomp systems

DanielSank

@user507974 Build a quantum computer.

Seriously that's the goal.

Although right now I'm reading about types of solder flux.

user507974

@DanielSank yea, what are you focusing on, fabrication, theory, noise?

DanielSank

01:32

@user507974 Me personally or the group?

user507974

if im not mistakened the google approach focuses alot on error correction systems to get coherent signals

personally

DanielSank

@user507974 Ehhh, we have several avenues of attack.

We're working on:
1) Gate model error corrected system
2) Analog quantum simulation
3) Quantum annealing

Those are the three main divisions of the hardware we work on.

#1 is the most advanced.

#2 is coming up real fast.

#3 is what DWave does and what we're learning to do.

Interestingly all three directions require the same hardware improvements, so the basic research anyone in the group does benefits all three.

user507974

im not familiar with analog qsim

DanielSank

@user507974 Basically, build a system whose Hamiltonian matches the one you want to simulate, and then just measure what happens :)

user507974

oh ok, so basically the goal would be to design a system thats like an fpga

where you can program it to create the desired hamiltonian

DanielSank

01:36

@user507974 Recently I worked a bunch on state measurement (also what I did in PhD). When I first started I did some theory/design. I also did a bunch of software. Now I'm getting back into some theory and a lot of noise theory/measurement.

@user507974 Uh, yeah sort of.

Any quantum computer is kind of like an FPGA though.

There's no op code processor. Everything works by direct interaction of the qubits.

In other words, there's no good separation between memory and processing.

Not so far, anyway.

user507974

what types of hamiltonians are these useful for

at least so far

DanielSank

Spin systems are the easy case.

We have a project working on chemistry problems, i.e. bond lengths in diatomic hydrogen. It's a toy problem to see if it works.

user507974

yea, thought so

what do the computational outputs end up looking like for analog qcomps

just a lot of eigenvalues?

DanielSank

Not my area really so I can't say for sure, but often people go for the ground state energy (whatever that actually means).

user507974

@DanielSank you mentioned noise though

DanielSank

01:41

@user507974 That's one of my specialties, yes.

user507974

which system do you work on currently that noise presents an issue for

DanielSank

@user507974 All of the qubits suffer from noise, dude.

user507974

@DanielSank well obviously, but i mean where is it worst

DanielSank

@user507974 Annealing.

@user507974 Actually, I have a very difficult theoretical problem related to qubit noise at the moment.

user507974

@DanielSank privy to share it?

DanielSank

01:45

The question comes down to how to measure analog noise using a qubit, which is hard because a qubit is a detector with 1 bit of resolution, and the result of each measurement is a random variable whose distribution depends on the noise at each measurement time!

So, I give you a qubit where the energy level splitting fluctuates in time and tell you to measure the spectral density of that noise.

Go.

There is some work on this and we're learning, but signal processing is always tricky and in this case it's extra tricky.

user507974

is the noise correlated more with teh qubits coupling with matter

or is it photons mnessing up the system

DanielSank

I don't know what that means.

user507974

like if you were to describe the noise source the answer is obviously everything

random photons

the enclosure vibrating

vacuum pressures

but is there anything that stands out

like a primary noise source

DanielSank

@user507974 Yes, it's fluctuating magnetic fields which seem to originate in the surface of the superconducting wires themselves.

It seems to have something to do with oxygen adsorbed in the metal surfaces.

This is a huge outstanding problem in our field.

It's been around for forty years and nobody has really decreased the noise level or even figured out exactly what causes it.

This is the biggest issue in our field.

user507974

@DanielSank is it just the fact that oxygen adsorption is occuring at random spots on the wire

DanielSank

01:50

@user507974 That's a pretty general statement which is almost surely trivially true :)

user507974

i mean is the random perturbations more the oxygens interaction (i.e. random results from the oxygen adsorbed) or the fact you dont know where that random perturbation is coming from

or maybe both equally play a role

because both can have very different solutions

ACuriousMind

@DanielSank At which temperature are the conductors?

user507974

probably ~100K right?

DanielSank

@ACuriousMind ~25 mK. Note that's considerably colder than outer space o_O

user54412

@DanielSank have you tried vacuums? :p

DanielSank

01:53

I like that the coldest places in the universe are in labs on Earth.

user54412

^ that is admittedly cool

DanielSank

@ChrisWhite ಠ_ಠ

user507974

@DanielSank unless there are alien labs that are colder

ACuriousMind

@DanielSank Only fluctuating magnetic fields, or also electric fields/radiation?

DanielSank

@user507974 Good heavens no.

That's waaaaay too hot.

@ACuriousMind Magnetism is the big problem. We don't radiate much.

user507974

01:55

@DanielSank meant to put < not ~, basically was making a safe guess since you mentioned superconductor

DanielSank

@user507974 The relevant temperature scale comes from the frequency of the collective mode we use as the qubit.

We're at ~5 GHz.

That's around 250 mK, so we have to be considerably under that.

user507974

so what is the biggest problem with the oxygen though

DanielSank

PSA: Apparently "no clean" solder flux really means "can't clean" solder flux. Don't use it unless you're operating a pcb manufacturing house.

@user507974 We're not 100% sure it's oxygen but the problem is that noisy magnetic fields make the frequency of the qubit fluctuate.

user507974

@DanielSank but is the noise it generates is poorly understood or the arbitrary spacial source of the noise is actually a bigger problem

DanielSank

So, if I have $|0\rangle + \exp(i \phi)|1\rangle$, the phase angle $\phi$ fluctuates in time.

@user507974 I don't understand the question.

The problem is that there's noise and we want to not have noise.

We'd be more happy to make it go away and not understand it at all as compared to understand it and not know how to make it go away.

user507974

02:00

@DanielSank its easier to make it go away though if you understand it

DanielSank

@user507974 Correct.

We're pretty sure it's oxygen. Now, ask yourself how the crap you're going to not have oxygen on your metals.

Amazingly, there are tricks and we're trying them.

user507974

@DanielSank crowd out with something that likes superconducting wires more than oxygen does?

DanielSank

@user507974 For example, yes.

user507974

@DanielSank that being said it also needs to make it thermodynamically expensive for the oxygen to adsorb

so it really has to adsorb ALOT

maybe just a coating of some kind would work wouldnt it?

DanielSank

Yes, if this were easy it wouldn't be a major outstanding problem for 40 years...

user507974

02:05

do the wires have to be exposed

DanielSank

@user507974 Meaning what?

You want to cover the wires in another material?

Ok, find me a dielectric material which has low loss at microwave frequencies at millikelvin temperatures.

Oh, right, there's barely any data on that because who the hell operates at millikelvin temperatures and cares about RF loss at the single photon level!?

user507974

@DanielSank people working on quantum computers do =)

DanielSank

@user507974 Nope.

Most of them said "screw that, we'll just use single layer devices" years ago.

No dielectric research there.

user507974

@DanielSank how thick are the wires?

DanielSank

To hard, and not enough potential for Nature papers.

@user507974 Say 100 nm to 200 nm thick.

Say 2 um to 10 um wide.

user507974

02:08

and what sides are exposed to a vacuum

the wire is etched on something id assume so at least 1 face is not exposed right

@DanielSank what type of superconductors

DanielSank

The top and sides are exposed, the bottom rests on a sapphire or silicon substrate.

We use aluminium.

Are you interested in this problem for real?

user507974

@DanielSank it definitely sounds interesting

DanielSank

@user507974 It is. Are you an experimentalist or theorist?

If you're interested in this, see papers by Robert McDermott.

user507974

@DanielSank my research background has been mostly computational but basically its because I've just never landed an experimental position for an internship =\

@DanielSank any good review articles in particular youd recommend

DanielSank

@user507974 Not off the top of my head. I'll email Robert and ask him.

user507974

02:13

@DanielSank thanks

DanielSank

@user507974 Ok sent.

I hope I remember to let you know what he says.

Well guys, I just learned how to buy solder.

Apparently there are two kinds: solder that works, and solder that doesn't.

2

user507974

@DanielSank im curious, how much noise does the substrate interface contribute

DanielSank

You want the kind that works.

Don't buy the other kind.

user507974

@DanielSank This

personal project?

DanielSank

@user507974 We don't actually know that yet, but we will soon.

@user507974 Nope. Lab.

user507974

02:15

@DanielSank so what type of solder is good for quantum computers

DanielSank

@user507974 There are two kinds of metal and two kinds of flux.

The metals are 60/40 and 63/37 tin/lead alloys.

This doesn't make much difference but the 63/37 is apparently a little easier to use.

The flux makes the difference.

There are four kinds.

There's rosin (R), rosin mildly activated (RMA), and rosin activated (RA). Those are listed in order of aggressiveness of the flux against oxides.

You basically always want RA for hand-done work.

user507974

@DanielSank right by the eutectic point i think right?

DanielSank

The only down side is that you have to clean the part when finished to get rid of residue because RA is corrosive.

Now, there's a fourth kind: no-clean (NC).

Apparently this stuff is terrible. It exists because it leaves less residue, so pcb houses like it, but in fact the residue that is there is harder to get off than RA flux.

user507974

@DanielSank that is probably the wrong type

DanielSank

So just use RA.

@user507974 Yeah I think so.

user507974

02:18

@DanielSank called it

DanielSank

@user507974 Yup.

user507974

even though you finished typing

i wasnt looking so it still counts

=)

@DanielSank is that what you ordered?

before

FenderLesPaul

@user507974 same

DanielSank

@user507974 I can't tell. I need to go get the stuff out of the workshop and look up the part number.

Right now I'm just ordering a bunch of sizes of the good stuff.

@FenderLesPaul Why's that?

user507974

@DanielSank there are some other good solders besides the lead tin though

user54412

02:21

@DanielSank so ligo had (and solved) a similar-ish problem

user54412

well maybe not too similar

user507974

oops tin

user54412

anyway, the vacuum in the ligo tubes would be measurably contaminated by oxygen diffusing out of the metal

user507974

@DanielSank are you guys using tin lead because you've determined the noise it introduces where you use it or is the solder you guys order some kind of special scientific grade

DanielSank

@ChrisWhite What was their pressure?

user54412

02:23

but apparently older metal was the answer -- the freshly-forged stuff was too gas rich

DanielSank

@user507974 What? Noise? We don't put solder close enough to the qubits for that to be an issue.

Also, solder is not magnetic.

user54412

@DanielSank good question

user54412

@DanielSank everything's magnetic -- you're just not trying hard enough ;)

FenderLesPaul

@DanielSank why's what?

user507974

@DanielSank well every part thats not etched ultimately is connected by solder so wouldnt it still perturb your input signals earlier on in the line

user54412

02:26

@DanielSank less than 10^-7 Pa apparently

DanielSank

@FenderLesPaul Ugh, sorry. Linked to wrong chat message.

@ChrisWhite Oh, that is actually high vacuum.

@user507974 Why would solder magically introduce noise?

Dammit I'm so bad at chat.

@FenderLesPaul I was asking why you said "Same" about @user507974's message about voting for UCSB.

FenderLesPaul

@DanielSank oh

Because UCSB is my top choice

user507974

@DanielSank well i kinda was asking about if it did make a difference

user54412

> At these low pressures the mean free path of a gas molecule is approximately 40 km, so gas molecules will collide with the chamber walls many times before colliding with each other. Almost all interactions therefore take place on various surfaces in the chamber.

user507974

the flux makes a differnce

user54412

02:29

that's really weird to think about

user507974

@ChrisWhite i like this statistic, very stat mech, much true

@FenderLesPaul what dicipline

FenderLesPaul

@user507974 quantum gravity

user507974

@FenderLesPaul so much math inbound

FenderLesPaul

@user507974 not string theory

DanielSank

@user507974 Doesn't.

FenderLesPaul

02:31

more along the lines of gauge/gravity duality (AdS/CFT), black hole thermodynamics, and quantum information

so I won't have much advanced math to deal with thankfully

DanielSank

@FenderLesPaul You'd be surprised where advanced math shows up.

user507974

@FenderLesPaul a and c scream ucsb

DanielSank

Here solve this:

Consider a cube.

You start on one vertex.

You make a random walk, with 1/3 probability to go to an adjacent vertex on each step.

What's the mean number of steps you take before arriving at the diametrically opposed vertex from the one you started on?

You can solve that using high school algebra, college level linear algebra, or techniques normally found only in QFT courses :)

FenderLesPaul

@user507974 a and c?

user54412

> gauge/gravity duality (AdS/CFT), black hole thermodynamics, and quantum information

> I won't have much advanced math to deal with

user507974

02:33

@DanielSank 1/3 chance to reach any adjacent vertex?

or for each vertex

DanielSank

@user507974 Any adjacent.

user507974

@DanielSank and by vertex you mean another corner of the cube (i know its kind of a stupid quesiton just need to be 100% sure)

DanielSank

@user507974 Yeah.

user507974

@DanielSank and the other 2/3 of the time you remain on the same vertex right

user54412

^ uhh... definitely a miscommunication here

DanielSank

02:37

@user507974 No. Each vertex is connected to 3 others. You have a 1/3 chance to go to any one of those three.

user507974

@DanielSank that was what i was making sure about, when you said 1/3 for any i thought you meant 1/3 total probability, but i had a feeling that was wrong

ok yea this one is pretty easy

DanielSank

@user507974 oh?

What's the answer?

user507974

@DanielSank pretty easy for a statistical physics question, id still need to write it out

DanielSank

@user507974 Tell me it's easy once you have an answer.

ACuriousMind

@FenderLesPaul How is AdS/CFT or gauge/gravity not "advanced math"?

DanielSank

02:40

Note that once you have a result I may or may not ask you for something like the mean square number of steps... :)

Then you will question you choices in life.

user507974

@DanielSank the tricky part comes from the fact that there are an infinite number of paths to reach the goal, so normalization is a bit tricky

DanielSank

@user507974 By "tricky" I do believe you mean "insanely maddeningly difficult".

user507974

@DanielSank i suspect it converges pretty quickly though so something closed form should pop out

or at least approximately good enough

DanielSank

@user507974 Like I said, give me a number.

user507974

@DanielSank opening up backpack

FenderLesPaul

02:44

@ACuriousMind I guess it depends on what you mean by advanced math

I meant that I never needed anything beyond basic undergrad physics math

user507974

@FenderLesPaul what techniques used for AdS/CFT that undergrads do?

@DanielSank so far interesting factoid for your problem is only odd numbered macrostates occur

ACuriousMind

@FenderLesPaul Haha, hear that @Danu?

I'm sure Danu will be pleased to hear he's doing basic undergrad math :D

DanielSank

@user507974 Sure.

0celo7

@ACuriousMind I'm convinced no one in this chat knows the difference between undergrad and grad

DanielSank

@0celo7 Correct.

Well, mostly you though because you've been subjected to this insanity without any real reference.

user507974

02:58

@DanielSank 27/4 steps?

DanielSank

@user507974 Nope.

user507974

@DanielSank the number of potential paths grows $2*3^{n-2}$ though right

DanielSank

@user507974 I don't know.

Never did it that way.

user507974

@DanielSank just to make sure, the answer isnt infinite is it

DanielSank

@user507974 Nope.

user507974

03:06

well i might take a jab at this later but have to leave atm

DanielSank

@user507974 Ok.

Have fun.

It's a cool problem.

user507974

wait

no that cant be right...

DanielSank

03:24

@user507974: Response from Prof. McDermott:

> I am not aware of any review articles. There are the recent experimental papers from Clarke (Steve Anton as first author), and some theory work recently from the group of Rogerio de Souza in UBC and Clare Yu at Irvine (who is working with us on thinking about oxygen).

user54412

@DanielSank Ok you sniped me

user54412

I didn't expect that number to be the answer

DanielSank

@ChrisWhite :)

Used a computer, didn't you?

user54412

you know me too well

DanielSank

@ChrisWhite Well, it's the smart thing to do. Never do a calculation if you don't already know the answer, eh?

user54412

03:32

hmm, looking at higher moments

user54412

am I supposed to see a nice pattern, or is it off to OEIS?

DanielSank

@ChrisWhite I have no idea what the moments look like.

It's just a question I use to needle people who solve this the easy way.

Ok gotta go.

Hungry.

Ciao, everybody.

0celo7

03:56

@ACuriousMind Actually, the opera was much better than differential geometry.

0celo7

04:07

@DanielSank I guess

what is "undergrad math" actually

Henry WH Hack

34528712987657122123100923322332232222

DanielSank

04:29

@0celo7 Ehhh, for physics majors it's typically multivariable calculus (although not always arbitrary dimensions), linear algebra, some amount of differential equations, basic complex analysis (contour integrals), some pseudo-understanding of orthogonal function expansion, a vague and totally screwed up knowledge of tensors, and maybe a passing familiarity with groups.

Did I miss anything?

0celo7

analysis?

topology?

> a vague and totally screwed up knowledge of tensors

Dude I'm scared about what they might teach us in an applied math class that covers tensors

by "groups" do you mean "representations of groups"?

DanielSank

@0celo7 More or less.

@0celo7 You should be scared.

@0celo7 No. Topology is useless for 99.9999% of physicists.

0celo7

@DanielSank a tensor is just a thing with indices

and the indices transform a certain way

>:(

DanielSank

You know I don't like that definition :)

But yeah, it's equivalent to the one I do like.

I prefer the definition that helps me derive all the important properties without having to remember anything.

0celo7

@DanielSank hence the ">:("

@DanielSank you need the chain rule

DanielSank

04:38

@0celo7 I can remember the chain rule because it's obvious if you remember what a derivative is.

0celo7

is it?

the proof requires a lot of care

DanielSank

@0celo7 So does proving that 1+1 = 2 but seriously who the hell cares?

A lot of extremely obvious things are hard to prove.

Go prove the intermediate value theorem. It's a total bitch but I don't care about proving it because it's completely obvious.

0celo7

@DanielSank we will in analysis

@DanielSank really?

what about all of HEP/GR/QM-th people

kevin Tah N.

05:18

Is there a free version to wolfram alpha step by step that shows all the symbols nicely too?

0celo7

@kevinTahN. I get it free through my school

kevin Tah N.

wow! That is so cool

@0celo7 It is a very good tool to practice and check calculations

Physics Meta

07:34

0

Q: Coloured text in questions and answers?

Using MathJax it is possible to make text of a different colour using e.g. $\color{red}{\text{this text would appear in red}}$ However, I rarely see this done and the times I have done it, it has always been edited out. Clearly, this feature could be misused (e.g. by changing the colour of larg...

discussion

DanielSank

@0celo7 Yes, really.

@0celo7 That depends on what you mean.

Most physicists who use GR don't give 2 craps about topology.

My fiancee uses it in astrophysics research. She needs to understand lensing. No topology there.

My Dad deals with GPS. You need GR there to understand the timing

No topology.

Slereah

08:11

Do they use

THAT book

amazon.com/Einstein-Matrix-Form-Derivation-Relativity/dp/…

Wait what the fuck is this article

arxiv.org/pdf/cond-mat/0607418.pdf

Danu

09:13

@ACuriousMind I get my mathematical self-respect from topology ;)

Not from AdS/CFT

Slereah

there aren't a lot of fields that actually use topology

Mostly just some GR and QFT people

I think some condensed matter people, too?

Danu

I'm going to be doing SCFT with topological defects, maybe I can use it a bit. Also Topological string theory, whatever that may mean

Secret

: 27873533 Why this analogy works, is it really just because both disciplines are nonlinear phenomenon...?

Slereah

09:40

Received (Day Month Year)
Revised (Day Month Year)
Accepted (Day Month Year)

Just because you're on arxiv doesn't mean you shouldn't fill your paper properly

user116211

09:59

@Danu: Is quantum superposition only valid for orthogonal basis set?

Slereah

what?

You can always superpose Hilbert space vectors

As long as they are normalized

user116211

@Slereah orthonormal?

Slereah

No

I mean

user116211

@Slereah mean?

Slereah

All vectors are gonna be a sum of orthonormal basis vectors

So even if they are not orthonormal

You can still write them as a sum of orthonormal vectors

So it's not really a problem at all?

user116211

10:06

@Slereah yeh!

Slereah

Plus this is a vector space so any sum of vector will still be part of the vector space

user116211

@Slereah you solved it.

user116211

@Slereah So, all vectors are superposition of orthonormal basis, right?

user116211

@Slereah: Another question

Slereah

Well that's what a basis is

user116211

10:09

@Slereah: Are all linear combinations quantum superposition?

Slereah

what do you mean

user116211

@Slereah Heard of MOT in quantum chemistry?

Slereah

No.

user116211

@Slereah v.v

user116211

@Slereah: Is it right then to say non-orthogonal basis?

Slereah

10:14

You can have a non-orthogonal basis, yes

Coherent states form a non-orthogonal basis

user116211

@Slereah We can ultimately convert the vector into a linear combination of eigenstates of orthonormal basis set, right?

Slereah

The point of a basis is that any vector of the vector space can be expressed as a linear combination of the basis vectors, yes

user116211

@Slereah I meant to say we can convert the non-orthogonal basis to orthogonal, can't we?

Danu

11:48

@Qmechanic I've got a request for you this time: Could you go through all resource-recommendation questions and make them into CW (or remove the tag?)? I did the first (earliest, that is) 5 pages of 50 questions already---maybe you saw my large waves of flags. There are about 13 pages left, IIRC. It doesn't actually take that long, but David Z told me to just stop the mass-flagging because it's too much to handle at a time.

Slereah

12:45

-1

Q: where to share the theory?

I have a verry logical theory, this theory is complex enough to make relativity look's simple in front of it. the probleme is that I am afraid about my right's, that any one can pretend to have it. what to do?

general-relativity

Sᴋᴜʟʟ ᴘᴇᴛʀᴏʟ