## Monday, September 22, 2014 ... //

### Why I remain a BICEP2 believer

Because I can see the non-dust pattern with naked eyes...

Just three days ago, I wrote a blog post about BICEP2. So I wasn't terribly excited to write down another blog post once the Planck Collaboration published a paper claiming that the BICEP2 could be due to dust, especially because I don't find the Planck paper to be terribly new, interesting, insightful, or game-changing.

A random image, taken from Perceiving randomness: egalitarian bias

They've been saying similar things since the spring and the arguments they presented today don't seem stronger than the previous ones.

Their fits don't seem to be too good without the error margins and as far as I can say, they are inflating the errors by inventing various kinds of "extra errors" (such as the "conversion error") in order to dilute and obfuscate the signal that they may have failed to discover, despite their superior gadgets and huge funding. This production of spurious errors sort of reminds me of Gerhard Schröder's invention of new taxes such as the environmental tax, the beverage tax, the bad weather tax, and others (Schröder wasn't a sufficiently arrogant hardcore thief to propose a carbon tax, however!).

Much of this tension is a clash of personalities. I think that what BICEP2 has shown is the experimental science of the best kind and unless some embarrassing error emerges (I really mean something like a loosened OPERA cable: it hasn't emerged so far), I will continue to think of them highly even if their discovery is ultimately reduced to dust (or another background). Like proper stereotypical experimenters, they didn't really believe a word that the theorists like to say (all proper experimenters think that gravity is actually caused by leprechauns and GR is just a theorists' fairy-tale for babies to sleep smoothly; but if a theorist needs the experimenters to empirically determine something, the good experimenters are as reliable as a vacuum cleaner). However, after they spent a very long time by efforts to show that their signal is due to something else, they published a paper with the discovery claim and it was undoubtedly right that they did so. Science couldn't operate if the publication of a discovery were viewed as a blasphemy.

On the other hand, I do feel that the Planck folks have the lack of audacity combined with prejudices and the plan to "just impose new limits" is their modus operandi, a dogma predetermined to direct their behavior for years to come. I am even afraid that they will abuse their stronger political power and their joint paper with BICEP2 (to be out in November 2014) already has a predetermined qualitative conclusion now – it will be just a new limit on $$r$$, the tensor-to-scalar ratio.

But let me stop with these ethical questions which are clearly dividing the people. There is a scientific substance. Everyone who is interested in cosmology would love to know whether the imprints of the primordial gravitational waves have been seen. I agree with those who say that this discovery, if true, is the greatest discovery in many years if not decades or a century. I would probably place it above the Higgs boson discovery because unlike the Higgs boson, it wasn't really guaranteed.

However, we must ask: is the discovery real?

Of course that I am not 100.00000% sure. But I still think it's significantly more likely than not that the BICEP2 discovery is genuine and the pattern they see is simply not dust. Why? Because it clearly doesn't look like dust. What do I mean?

Just look at the key "photograph" of the BICEP2 field with the polarized CMB signal. My question for you is the following: Can you distinguish this picture from the picture of the "colorful smoke" at the top of this blog post? I sincerely hope you can. But why are these two pictures so different?

The BICEP2 picture is very regular in some very specific way: you may talk about a preferred length scale, the typical distance of adjacent red blobs (or adjacent blue blobs). It pretty much looks as if the adjacent, nearby blue blobs (or adjacent red blobs) wanted to be a rather well-defined distance from each other – and the distance is something like 1/7 of the width of the picture (about 4° steps in the declination, the vertical axis – or about 8° steps in the right ascension on the horizontal axis; note that $$\Delta \phi$$ has to be multiplied by $$\sin\theta$$ to get a distance on the unit sphere). Do you agree? Needless to say, the step $$\Delta\theta=4^\circ$$ exactly corresponds to $$\ell=90$$ because one may squeeze $$90$$ waves per $$4°$$ on a circle that has $$360^\circ$$ in total.

On the other hand, the colorful smoke at the top doesn't seem to have a preferred length scale. To a large extent, it is self-similar. If you need to know, the colors in the "coloful smoke" picture were computed from functions defined by Fourier series, and the coefficients in the momentum representation were random numbers comparable to a (decreasing) power law of the momentum, to guarantee that the color is sort of continuous.

If you work on the sphere, the "momentum modes" become "spherical harmonics" and $$\ell$$, the orbital angular momentum, replaces $$|\vec k|$$. OK. What is the decomposition of the pictures to the spherical harmonics?

Open the new Planck paper on page 8. You find Figure 2 over there which has two similar (electric, magnetic spectral) parts and the top one-half of each part features three red almost straight lines (top) and three blue almost straight lines (bottom) which are the predictions for the dust. These monotonic predictions may leak to the polarization data and the resulting polarization, in the bottom part of the pictures, is less monotonic but it is still monotonic enough. The slopes are small and the local maxima are very mild.

It simply looks very different from the BICEP2 spectrum that seems to have a pretty clear local maximum near $$\ell=90$$ or so while the $$\ell=50$$ harmonics are weaker by a factor of two.

I don't want to say which particular increase or decrease in the BICEP2 spectrum most visibly contradicts the predictions from the "dust" hypothesis. If I wanted to choose the criterion that, in my opinion, discriminates between the BICEP2 observations and the predictions of dust more sharply, it's the presence of a preferred length scale (preferred value of $$\ell\approx 90$$, or preferred angular scale in the sky) that the BICEP2 seems to clearly and "repeatedly" see while, as far as I can say, all dust-like explanations would tend to predict a self-similar smoke-like pattern that has no finite preferred angular scale.

It's my feeling that if you quantified some "confidence level" telling you "how strongly BICEP2 sees a preferred finite angular scale smaller than the size of the window", you would get a pretty high confidence. And the fact that this preferred angular scale $$\ell\approx 90$$ agrees with the predictions of inflationary cosmology is one more consistency check for me to be even more certain.

Again, I am in no way 100.00000% certain that the gravitational waves are being seen over there. Of course that if there is a risk of something like a "loose cable" in the BICEP2 apparatuses, the confidence level may go to 50% or lower. But if there is no risk of an embarrassing error and the competing hypotheses are really dust-like, I feel that my confidence is still above 99% that what BICEP2 doesn't see a dust/smoke because that would have a much more self-similar, scaleless appearance.

For those reasons, it looks like Planck is a victim of a confirmation bias, a team looking for excuses why they haven't seen those things before their competition and trying to sell their own "absence of evidence" as "evidence of absence".

#### snail feedback (34) :

Dear Lubos,

I'm a bit puzzled by your comments. When you say that Planck invented extra errors, are you accusing the team of dishonesty? or merely speculating that the vested interests of the Planck team might have accidentally resulted in biases in their (bona fide) analysis? The type of biases that are often removed by with blinded analyses. Can you justify your claims that reported errors were inflated rather than accurate estimates of uncertainties?

Secondly, your invite us to compare of the dust spectrum in Fig. 2 with the BICEP spectrum with a peak at l~90. Isn't this comparison made for us in Fig. 9, a plot of the estimated dust spectrum and an r~0.2 CMB model? Can't we see in Fig. 9 that the two agree fairly well (although the agreement isn't quantified in the text)? i.e., that a dust spectrum could indeed have mimicked an r~0.2 signal including the peak at l~90?

Best, Andrew.

It also looks like a chessboard tilted 45 degrees. Dunno, might of course be an artifact of how the picture itself is done.

Dear Andrew,

in healthy circumstances, an experimenter is doing his own experiment instead of trying to undermine someone else's experiment by vaguely related and sometimes unrelated arguments. Because the task for an experimenter is to find something, the usual dishonesty of an experimenter usually manifests itself by his *underestimate* of the uncertainties - when uncertainties are understated, an experiment may look better than it is. But one may say that if someone is *overstating* some uncertainties, it is the mirror image of the same problem.

You used the word "dishonesty" so in this sense, you manipulated me into a particular wording, but my *honest* answer must be Yes, of course, that is what I meant.

At any rate, whether one wants to call it dishonesty or not, it's less important than the fact that if Planck's uncertainties end up being larger than BICEP2's uncertainties for the most relevant specific questions, then Planck isn't useful for deciding these questions, is it? Planck seems to want to claim that they have found a problem with the BICEP2 paper but without saying what the actual problem is. If Planck has big uncertainties, it doesn't mean that BICEP2 has too large uncertainties as well, does it?

Their figure 9 has nothing whatever to do with the "missing peak" problem I have discussed. This figure only shows that with those (it seems to me: inflated) values of D_ell(BB), the BICEP2 spectrum i.e. values of D seem to fit within the band for each "ell", except for "below ell=50" where BICEP2 is more suppressed. But the values of D_ell for individual values of ell, and their errors, isn't what my argument is all about. My argument is about the difference between D_90 and D_80, for example. BICEP sees this difference as larger than predicted by the dust model, I claim. Figure 9 shows no prediction for these differences such as D_90 - D_80.

So what I see that Planck is doing - and you seem to support them in doing so - is obfuscation of evidence. I am not questioning that if one doesn't look at the right places with the right apparatuses, he won't see any primordial gravitational waves. I am just questioning that Planck has found a real problem with the paper by BICEP2, a collaboration that apparently *did* look at the right things with the right tools.

Cheers
LM

I would caution: it is very hard to look at a map and conclude much about the origin of the signal. The reason is, any inference about the shapes, sizes etc of polarization hot and cold spots necessarily contains less information than the angular power spectrum. [In the absence of non-Gaussianity - an excellent assumption here.] And the power spectrum of dust *could* exactly mimic that observed by BICEP, as several detailed papers recently found.

Dear Orangutan, I think that your reasoning is circular and fundamentally wrong because it uses some wrong assumptions, too.

The structure of the information - local patterns vs power spectrum as a function of "L" only - goes in both ways. They are complementary and none of them is superior.

In particular, you may have some patterns that only depend on the celestial latitude yet they emulate any spectrum as a function of L you want. On the other hand, the particular picture of BICEP2 is "isotropic" (the blobs are equally separated in the vertical and horizontal direction) so one may exclude that the relevant patterns only depend on the latitude, among many other things.

It's very clear that if someone ignores this extra information - e.g. the uniformity of the contributions to the spectrum from all values of M (which is the Fourier-transformed translation of the isotropic character), he can't really reproduce or validate all the detailed data that are actually making up the BICEP2 discovery.

Fig. 9 shows binned data, with massive bins, unfortunately. A single bin spans l=40-120. So I agree that it's impossible to resolve differences between l=90 and l=80.

That said, the resolution of the BICEP data is also large, delta l ~ 50. I can't find any reason to say something interesting is happening around l~80-90.

I can't see any features in BICEP power spectrum that are definitely absent in the Planck extrapolated dust spectrum.

Why don't you just draw how the typical BICEP-sized window would look according to the model? Or, more easily, why don't you look at the colorful picture at the top of this blog post? I feel as if I were talking to a tree. You don't see because you don't want to see, do you?

I see that we've reached an impasse. I'll think about your points. Thanks for your time and best wishes, Andrew.

Actually in the full fig3 of the BICEP2 paper the noise is simulated at the lower right figure, no? The pattern is different, but a pattern exists due to the choice of l, I think.

Dear Lubos,

with respect to Gerhard Schröder's Steuersong it should be mentioned that this particular song and other influences seems to haved turned around his behaviour by 180 degree. So I think it is fair to say that history remembers him as a chancellor with a very good economic track record.

For those who haven't listened to it yet here is youtube link which still works:

Fair enough, one should also check for isotropy at least visually, and look for anomalies in the e.g. M-dependence.

From Sean Carroll's blog:

George Efstathiou says:
September 21, 2014 at 11:24 pm
As a member of the Planck Science Team, I would urge caution concerning the interpretation. What we are saying is that polarised dust emission in the BICEP2
field is high. But it may be that there is something left in the BICEP2 signal that
can be attributed to gravitational waves. We need to cross-correlated the Planck
maps with the BICEP2 maps. This analysis is underway.

lubos, don't be so nervous, because you were wrong. everything you claimed has failed: supersymmetry, strings, extra dimensions e.t.c. complete loser. Woit wins

I agree completely, Lubos. The Planck team’s behavior is really annoying. If they can produce their own evidence that the signal is absent they should publish it. If not they should remain quiet until they can.
Of course, anyone is free to find errors in BICEP 2 but such “errors" must concrete and not just a judgement call. The very fact that the Planck team is a competitor brings a strong requirement that their evidence of error is incontrovertible.

Dear Mikael, I totally realize that and agree, having watched Germany quite carefully during his tenure. He's been also reasonable in other matters I don't want to reopen.

This comment was about the cartoon Gerhard from the song only - which is at least equally ingenious.

For me, the crucial number to compare two different experiments would be the strongest statistically measurement of BICEP2, that r=0 is excluded with in 6.4 sigma for r=0.19+0.07-0.05

I cannot find an r in the Planck paper, though to read really through it would take me a long time, except in calculations of models. It is suspicious that they do not give the hard number that shows that there exists B polarization, no matter what the origin is. It seems to be obfuscating with bands calculated by models etc with r cuts, and not give their number if they are measuring B polarization regardless of origin.

6.4 sigma non zero is a very strong measure . I want to see their value for this.

Cloud Computing Online Training, ONLINE TRAINING – IT SUPPORT – CORPORATE TRAINING http://www.21cssindia.com/courses/cloud-computing-online-training-23.html The 21st Century Software Solutions of India offers one of the Largest conglomerations of Software Training, IT Support, Corporate Training institute in India - +919000444287 - +917386622889 - Visakhapatnam,Hyderabad Cloud Computing Online Training, Cloud Computing Training, Cloud Computing, Cloud Computing Online Training| Cloud Computing Training| Cloud Computing| Courses at 21st Century Software Solutions
Talend Online Training -Hyperion Online Training - IBM Unica Online Training - Siteminder Online Training - SharePoint Online Training - Informatica Online Training - SalesForce Online Training - Many more… | Call Us +917386622889 - +919000444287 - contact@21cssindia.com
Visit: http://www.21cssindia.com/courses.html

Dear Anna, a good comment, thanks.

The high-confidence BICEP2 claims more or less assumed the dust to be negligible which is supported by some previous papers and modeling on the dust.

If one decides that the dust is not negligible, it becomes hard to produce any real numbers because whatever one sees, if it is not instrumental etc. noise, may be emulated both by the gravitational waves and dust. To a large extent, they are indistinguishable. Well, they are indistinguishable if one only describes them by "one intensity figure" for the relevant L=90 range.

At the end, the dust and gravitational waves are still distinguishable and BICEP2 has done much more than Planck to actually distinguish them. But one must go beyond the "single figure" approach. It seems to me that Planck wants to avoid to even consider those things, and I think it's impossible to decide about the validity of the BICEP2 discovery without such a more detailed analysis.

Andrew is not me. Just sayin'

The data I saw from the Planck paper seems to confirm the BICEP2 and KECK results because, as you point out quite well in this post, the dust models can't explain the drop in polarization signal as small values of multipole.
All that the BICEP2/KECK team need to do is to expand their search area so that they get a data point at l=25. If they can show a drop in polarization between l=100, l=50, and l=25, then I think that they can convincingly argue that the B polarization signal is from gravitational waves and not from dust.

I would argue that the Planck "binned representation" was a little dishonest because the Planck team has the raw data on dust polarization and they can go to multipole values less than 25. So, why did they "hide the decline"?
(a) Isn't it in they interest to confirm gravitational waves?
(b) Why didn't the Planck team actually present their B mode polarization vs multipole just like BICEP2&KECK?

Dear Mike, thanks for this refreshing way to formulate what I (and probably others) also think.

Your reference to "hide the decline" was very witty. While laughing, I replayed the music video

and imagined the tiny adjustments of the lyrics that could make other verses relevant for Planck, too. ;-)

Thinking a little further on black hole evaporation, I guess that each particle type follows its own thermal distribution independently. If the black hole is cold (far infrared radiation), then it will emit mainly photons. Other particles are 'frozen out', as the probability of high energy emission is small. However, when the BH heats up so photon emission is on the order of 2 eV (a red hole!), then a significant number of electron neutrinos can be emitted http://en.wikipedia.org/wiki/Neutrino#Types, presumably with a Maxwell-Boltzmann velocity distribution corresponding to instantaneous temperature. I reckon this should speed up the evaporation, over and above that expected for photons alone, as now there are two significant channels of emission ('significant' in relation to the probability distribution). Similarly, when a significant chunk of the thermal spectrum crosses 170 keV for muon neutrinos and 500 keV for electrons/positrons, the evaporation rate should speed up further, as new particles contribute. Is that correct, or is it dominated by photons all the way up and we can ignore the other particles? If it is correct (that the evaporation rate speeds up at higher temperature, beyond that expected from photon emission), could this be a way to map new particles??? That is, if one had a pet mini black hole, and by measuring the photon emission rate over time, one could infer the masses of new particles by looking at the slope of the rate/temperature plot?

http://cdn.meme.am/instances/500x/54608853.jpg

Is the photograph some kind of a joke or satire? Could you please explain it to those of us who don't get it? Is that an attack on one of the great volunteers who work on the great experiment on the South Pole? If it is, what is happening in Hawaii? Why do you think that grad students and postdocs have "freedom" to decide when they may go to work to Hawaii?

Do you agree it's fair to conclude that the professional critics of BICEP2 suck in physics, in humor, as well as human beings?

And what about quark emission? While it is standard to say that quarks don't appear individually, is it possible for a black hole to emit individual quarks? That is, a virtual quark-antiquark pair are produced on the horizon, and the BH swallows one of the pair? Surely not even the strong force can overcome an event horizon so why will this not happen?

As the black hole temperature increases through each new particle rest mass threshold, that particle species should be Hawking radiated by the black hole. The Hawking formula is per degree of particle freedom, so the total evaporation rate increases as each new species can be emitted. The more it radiates, the smaller its mass gets and the black hole's gravitational temperature is inversely proportional to its mass, so its temperature climbs. Above 200 - 300 MeV, it should start emitting quarks. The size of the black hole and the time interval between emissions will be such that the quarks are asymptotically free on emission but then form colourless states as they move away from the black hole. At very high temperatures the black hole should be emitting every particle species. If we can find a tiny black hole, we can answer all those annoying questions about SUSY particles without having to build an accelerator!

Excellent! It is great to have people who know this stuff. That would make the calculation at http://en.wikipedia.org/wiki/Hawking_radiation#Black_hole_evaporation inaccurate for low mass black holes, but I guess nobody really knows how it will go at the end. As a point of interest, did you take all particle species into account to get E^-3? After all, the range up to 430 GeV contains all known particles, and the distribution of any one species will depend on the total evaporation rate. Not an easy calculation, I'm sure! Has anyone published it? Presumably yes, as it is so interesting, but I didn't hear of it (though I'm not in this field, as you might have guessed!!).

Yes, the E^-3 tail mathematically is set by how the black hole mass decreases with time. The E^-3 is only slightly modified by the inclusion of extra fundamental particle species: the way to see it is a small effect is to consider the energy range, say, from 1 GeV to 10 GeV. In that range there will be 3 new fundamental species (the tau lepton and two quark flavours) for the black hole to emit for the first time but, on the other hand the E^-3 falls off by a factor of 1000 over the same range. So those extra fundamental species can make little difference to the E^-3 fall-off.

You understand that Ukraine is a completely separate sovereign nation to Russia, right? That in the 21st Century, we've witnessed Russia invade a neighbouring sovereign nation? The equivalent to the United States invading Canada for religious / economic / historic gains in 2015.

No, Ukraine is not "completely" separate and sovereign relatively to Russia in your sense, and no, Canada doesn't enjoy this kind of "complete" separation and sovereignty relatively to the U.S., either. And no, Russia hasn't invaded Ukraine yet.

Actually not-today deployed megatonnage less than 2000 megatons-there ACTUAL data on nuclear war -http://www2.gwu.edu/~nsarchiv/nukevault/ebb480/.

and as matter of fact there still impossible to develop new thermonuclear designs without tests.

His opinion is flawed and so is yours. Do you think you know how many nukes Russia has? In any case, this will not happen who has the most as the whole world will perish and you with your flawed views will die too.

You also have poor understanding of Russian people. This means result will be one. Total mutual annihilation. Regarding American people . You are warmongering pussies.

Nobody trust Russia, its the U.S. nobody trust, have you been living under a rock? The insane government is the aggressor here, if you don't see it, you either don't have a clue about what is going on or your just a moron, plain and simple! Who is trying to take Yemen over again, who has gone into Iraq 3 times, and Ukraine, the U.S. is the aggressor, pull your head out of your airse!