JAVIER: Okay. Hello, everyone. Welcome to the BOOST 2021 session on boosted Higgs plus heavy flavor. So the format will be -- you know, the speakers will have sort of five minutes to present their lightning version of their talks. Afterward, we'll take questions and discuss for ten minutes, and we'll have to stay on time. So we'll kind of move quickly after those ten minutes. There is a Google Doc for collecting some questions. But if you haven't written it there, that's okay. Just use the hand raise function in Zoom if you also want to ask a new question live. So let's move straight into our first speaker. Yajun.
YAJUN: Yes.
JAVIER: Thank you.
>> Recording in progress.
YAJUN: Can you confirm that you can see my screen?
JAVIER: Yes, perfectly.
YAJUN: Okay. Great. So hello, everyone. It's my pleasure to present the boosted Higgs b b-tagging in ATLAS. I'm Yajun He, LPNHE of Paris. If you happened to watch the video for the long talk, it will show you all key info, and you can check with your lecture for more details. First, a brief introduction. As resolved region, we're using two small R jets tagged as b-jet individually. However, at high energies, decay products of b hadrons are highly collimated, and we're limited by statistics. With resolved H to b b-tagging. As you can see from this plot, this is the result from -- one of the results from the resolved semi-leptonic analysis. And 250 GeV -- the statistical uncertainties are dominant. Therefore, at high energy region we're using a large R jet and its associated subjet to identify Higgs to bb, as you can see from this cartoon. Two methods presented in the long talk, as well as the applications for the double b-tagging method and the neural network-based tagging techniques.
In double b-tagging method, we're using variable radius track jets, and the most popular algorithms. MV2 with a single discriminant as output, and DL1r with three flavor probabilities. The Higgs to bb is the large R jet with 2 b-tagged ghost associated VR track jets. The performance and calibrations of this method based on the isolated VR track jets and jet flavors. The isolated means there is no association requirement to large R jet. And the efficiency and scale factors are measured based on the jet flavors with different events. Here I want to highlight the scale factor here is data to simulation scale factor. This scale factor is to correct the tagger, the efficiency self-tagger in Monte Carlo to match it to the missing data. And the right plot shows the performance of the different b-tagging algorithms. On the light flavor jet rejections. You can see the blue one is the L1R and the red one is the MV2. The DL1R has bad performance. This method is adopted in a lot of analyses in ATLAS. For standard model studies with double b-tagging method, more granularities are added at high pT region compared to the resolved tagging method.
As you can see in the left plot, it is one of the results from the boosted semi-leptonic VH. So we have more granularity with pT higher than 450 GeV, compared to the resolved in the intro slides. And in the right plot is signal strength measured in differential fiducial region with inclusive all hadronic Higgs bb, more granularity, as well as higher pT. Also, for new physics, the double b-tagging method performs good in high resonance mass range of 0.5 to 1.5 TeV, but degraded at very high masses. In the left plot, it shows you the new heavy resonance searches, using WH signature. You can see the green one is using the resolved topology and the blue one -- the acceptance with boosted topologies. And with the two b-tags, the acceptance increases first, and then decreases.
Because we are -- it's highly overlapped, at very high energy. And we have a similar observation with the searches of new heavy resonance Di-Higgs. As you can see in the red plot. Like in the full b signal region, the acceptance increases first and then decreases. And besides the double b-tagging method, we have more advanced tagging techniques. For example, in the ttH production analysis, the neural network algorithm is used to choose Higgs candidates in analysis. And also we have an X bbtagger. It's a neural network using large R jet and subjet as input, defining log likelihood discriminant with output from this tagger, and in the middle of the plot, you can see it's a performance of X-bbtagger, compared to the double b-tagging method. The blue point is with the X-bb, and the green is with the double b-tagging with the L1R and the gray is double b-tagging with MV2. And the X-bb has better performance than double b-tagging method.
And it is flavor information-based tagger, and as you can see in the right plot, the multijet rejection is -- depends a lot on the flavor composition of the large R jet. Here. And for this tagger, they dedicated the topology requirement in the calibrations.
JAVIER: Sorry, the five minutes are up.
YAJUN: Okay. It's finished. And the calibration based on the large R jet topologies with the two VR track jets -- we're using the bb topologies, and as you can see in the left plot, and for the top mistagging efficiency -- the lepton ttbar, the result you can see in the middle of the plot, and also we checked the modeling with the gluon bb events. And you can see using different flavor templates, we have a good agreement in that kind of predictions. And in summary, the boosted Higgs bb-tagging is very active development, the double b-tagging method -- we have very interesting results with this method and advanced techniques in both Higgs bbtagging, and also the calibration and we're looking forward to this application in physics analysis. That's all from my side. Thank you.
JAVIER: Thank you so much for that very... Lots of topics covered in that lightning talk. Does anyone have questions from the audience? Any questions? I don't see any hands. So I might go to the Google Doc. I see Lukas was connected. Do you want to ask your question?
LOUKAS: Thanks for the talk. The main question I had was the ones I write here. I don't know if you want me to read it or if you have the chance to read them and have some answers.
JAVIER: You can go ahead and ask in case someone hasn't read through them already.
LOUKAS: So the slides refer to the long version of the presentation. So slide 5... It's about the efficiencies of the mistag grade, the scale factors, at least in the plot, they go up to something like 200 GeV, 300 GeV, which is probably the relevant pT region... Is much larger. So it's just what the region showed? Or do you have also scale factors to higher pT?
YAJUN: Yes. We have dedicated procedure to extrapolate the scale factors. To high pT. Since... Yes. Since we are limited... The data events are limited at very high pT. So we have some dedicated extrapolated procedure. And in the extrapolation, the central value -- we keep the central value stays the same as the highest pT bins, however, we are using the simulation-based event to extrapolate the uncertainties. Uncertainties to the higher pT jet.
LOUKAS: So if I have a little bit more time, do you compare, for example, different generators? How do you use exactly the simulation to extrapolate...
YAJUN: Basically we have... We computed new scale factors. For example, different generators, and we calculate difference of the new scale factors to the central value, and tag it as the uncertainties, and add all uncertainties in quadrature -- the total uncertainties, in the extrapolated pT bins.
LOUKAS: Thank you. And then my second question is: If you compare the scale factors from the gluon to bb... To the Z to bb ones... How...
YAJUN: Unfortunately, now for the gluon to bb, we only did the modeling checks. We don't have the scale factors yet for gluon bb. Sorry that I don't have the direct comparison here. Hopefully we will have the comparison soon.
LOUKAS: Thank you.
JAVIER: Thanks. So I don't know if Santeri is connected...
SANTERI: Hi, yes, I'm here. Thanks for the nice talk also, from my side. So this is about slide 14 in the long slides. I think it's probably slide 7 in the short version. If you can put it up.
YAJUN: I can switch to the long one. 14. So for the uncertainties... They're basically different for the gamma and the plus jets. For the jet γ, smaller than 450 GeV, the uncertainties are mainly from the statistics. Here that γ modeling and the spurious signal -- and also for the plus jets PTPs, higher than 450 GeV, the leading uncertainty is a very small difference, globaling? The dominant uncertainties, a fit model... The jet mass resolution. And the Z modeling.
SANTERI: All right. Thanks a lot. And then a small follow-up question. I was just wondering about the binning for those scale factors. So looking at the plot here on the right... It looks like... I mean, even surprisingly, it looks like your statistical uncertainty is quite small, even above 600 GeV. So... I mean, maybe... I mean, my first idea would be that maybe you could actually allow to have a finer binning also in that tail, if you have lots of events there. But then there seems to be this jump. I don't know if it's really at 600 or if it's an effect from the binning. This jump in the scale factor. From 1.2 down to something like 0.6. So can you comment a bit on the binning? And do you know what's going on around 600 GeV?
YAJUN: Yeah. This jump is not exactly happening at 600 GeV. It's somehow from the binning. So one of the reasons is that... So we have... The multijet distribution, the slope of the multijet, changes a bit in functions of... In different jet pT bins. And also here... The things -- we cut at 600 GeV. So some slope, a bit smaller than 600 GeV, are smeared with a lower pT bins, where there's more events. So there's not exactly at 600 GeV. And there's one, after binning, it appears here. And for the statistics here, yes. I agree with you. It's a bit small. However... We do have enough statistics.
But all of the statistics -- the contribution of the signals is a bit... Is small. Especially at high pT bins. So the statistic uncertainties -- since we see the multijets directly to the data -- this statistic is mainly because a lot of statistics of the multijets. I don't know if it's clear like this.
LOUKAS: I think this was a really helpful answer. Thanks a lot.
JAVIER: Thanks a lot again for your talk. And for the good discussion. We should move on now to Daniel's talk.
DANIEL: Can you hear me?
JAVIER: Yeah, great.
DANIEL: Okay. My screen should be sharing now.
>> Recording stopped. Recording in progress.
JAVIER: Okay. Please go ahead.
DANIEL: Okay. So due to the short time, I'm going to focus entirely on the new intrinsic charm result. So... That's the title of the intrinsic charm at the LHCb experiment... But I'm also happy to discuss other results from my longer talk in the discussion session. So when I talk about intrinsic charm, I'm really talking about valence-like intrinsic charm. So specifically some intrinsic charm component of the proton that looks like a valence quark PDF. So we're searching for this by studying c jets produced in association with a Z boson. And if you look at the diagrams on the right, you can see essentially you're using a gluon for one of the protons to probe the charm content of the other proton.
And having the Z in there means you're high Q². So the theory is a bit more certain. And then if we're in the forward region, we're probing very high x, which is obviously where we expect a valence-like PDF to differ from the normal extrinsic charm. And indeed, you can see that if you look at this plot at the bottom. So this just shows three different theory predictions as a function of the Z rapidity. So going from very central over here to very forward over here. The blue is if you have no intrinsic charm. This I think is just a collection of different theory predictions. And then the green is what you get if you take the PDF prediction and essentially say: Be consistent with the data. But you can put in intrinsic charm. Where the existing data allow. And the red is just a straight prediction from light front QCD. If you put in a 1% intrinsic charm content.
So the real thing to notice here: There are two things to notice. First off, if you go to the most forward region, you can really see there's quite a difference between the no intrinsic charm and the intrinsic charm predictions. And secondly, this is really unconstrained by our current data. This IC allowed band is very broad in this uppermost region. This is really something that needs to be addressed. So how do we tag our charm jets? We use a 2D displaced vertex tagger. So essentially we're reconstructing displaced vertices, which come about because we have charm hadrons with a finite lifetime.
And then we're performing a 2D fit to the corrected mass. And also to the number of tracks. And this allows us to distinguish the charm from the beauty and also from mistagged light jets. These templates come from flavor enhanced calibration samples, which we can talk a bit more about, if anyone is interested. And essentially, we're fitting this in bins of jet pT and Z rapidity. You can then unfold in the jet pT. And you take the ratio to the total number of jets to get the quantity on the previous slide.
So the main question here is: Once we've got the number of tagged jets, how do we determine the efficiency? This is also done in a data driven manner, using dijet events. So I'll quickly walk you through it. So we take dijet events where we've got low asymmetry, and they're fairly back to back, in the transverse plane. And then we require one of our jets, which we'll call the tagged jet, to have a displaced vertex in it. The other one, which we're calling the probe, is the jet we're interested in.
And because the tagged jet has a displaced vertex, and these are fairly back to back, this is very enriched in heavy flavor. If this is likely to be heavy flavor, then you're looking at di-heavy flavor production. We then -- looking at our probe jet -- use a less efficient but easier to determine tagger, specifically exclusive charm decay, so D0 decay π or D+ to K2π. And determine the number of prompt decays. This is very simple to correct for efficiency coming from simulation and the fragmentation fraction and branching fraction of the given decay, to give you the total number of charm jets in the full sample. This is our denominator. On the same sample, we tag for displaced vertex candidates, this is our numerator, and you get efficiency by dividing them, which you can use in your DV tagged Z plus charm.
We determine the efficiency as a function of jet pT. However, within our fiducial region, above 20 GeV, this is fairly flat, about 20%. We still bin in this. Before we look at the results, a quick word on the systematics. So as you can see, the dominant systematic is from the charm tagging itself. This is actually still smaller than our statistical uncertainty at the moment. But for the future, it's worth noting that almost all of these systematics cancel when you take the ratio between two bins. So for the future, you can have very precise measurements with double ratios. So here are the results. And as you can see, the gray bands are our results. And then the color points are the same theory predictions from before.
You can see this is very clear enhancement in the uppermost bin, compared to what you would expect from the blue. Which is no intrinsic charm. And we're pretty consistent with the intrinsic charm points. This inconsistency here is at about 3 standard deviations, but just point out that this is before there's been any global PDF analysis to really determine the significance. So this can be quite non-Gaussian. Shouldn't be considered 3σ deviation just yet.
So in summary, this is the first really direct probe of intrinsic charm, looking at Z plus charm in the forward direction. Very clear enhancement in the highest rapidity bin. And if global PDF analysis gives the same level of significance, then this is really the first unambiguous evidence for intrinsic charm in the proton. It was statistically limited, but the run 3 dataset should give a definitive answer. So stay tuned!
JAVIER: Thanks so much for that lightning talk. Does anyone have questions from the audience? Great comic. Raghav?
RAGHAV: Nice set of slides and measurement. I'm just gonna read this comic. It's not really a question. So you have this fragmentation function measurement. Right? In Z jets, that came out last year. Or the year before. So now you have Z to charm. So is there any plans to do, like, kind of charm fragmentation in that scale?
DANIEL: So the real point for this analysis was to look at intrinsic charm. This was the most interesting thing we could do here. So now we have the charm tagging set up, I imagine other Z plus charm analyses are going to follow behind this one. Not personally aware of the charm fragmentation one. But it's certainly doable.
RAGHAV: I think that would be definitely interesting. Because you saw... I mean, you guys have the very interesting Jψ and jet measurement. That shows completely non-monotonic behavior, as opposed to Pythia. And the forward kinematics, you have the fragmentation. It would be cool to combine those two together.
JAVIER: Great. Thanks. Any other questions? If not, I see... One in the Google Doc that is actually kind of related to one I had. So I'll kind of combine them. Which is: For the BDT, the charm tagger, so... Can jet substructure techniques or similarly kind of Deep Learning techniques help with tagging jets in LHCb? And if so, kind of what techniques do you think might be pursued in the future?
DANIEL: So I think for charm jets, really... So these two quantities give you almost everything. So I think the efficiency of this 2d tagger is almost the same as the run 1 tagger we had. Which has a lot more features with the displaced vertex. It was still all based on the displaced vertex. But I think for charm, this is sort of pushing the limit. So I'm not sure you could gain much. For beauty, including more features definitely helps. I'm not entirely sure whether you could gain more by adding anything beyond what was included in the BDTs, in the run 1 measurements. The run 1 measurements -- we have BDTs that were based on various properties of the displaced vertex. And also... A few kind of higher level properties of the jet.
I mean, in principle, maybe you could get something more by including all the features of the jet structure. But I think really... LHCb sensitivity to these displaced vertices is such that they drive our ability to identify heavy charm, heavy flavor.
JAVIER: Thanks. I also had another question about... So I guess what would be the implications from this measurement? I guess perhaps other than the Higgs production rates that you mentioned earlier. And are there any models that kind of predict this rapidity depend result? Any models that predict that... Okay. Go ahead.
DANIEL: In terms of what the underlying changes are, I mean, this charm gluon interaction between two protons, I think, is enhanced quite a lot by a 1% intrinsic charm. I saw a paper... It doubles this contribution kind of when you're colliding two protons. It also increases the rate of charm charm. So essentially everything comes down from that. One of the higher level effects this can also have is what's happening in air showers when you're looking at -- if you're trying to look at atmospheric neutrinos and you've got air showers from cosmic rays hitting the top of your atmosphere, a big background is the charm produced from those. If you have more charm than you expect, you're gonna produce more charm hadrons than you expect. So really nailing this down will help with the background determinations there.
That was the only other specific example I have currently to mind. But it's... Yeah. Essentially... Anything where you're colliding protons and interacting with the constituents. If there's more charm in the proton, then your production rates are all gonna be slightly different.
JAVIER: Thanks. Phil, we have time for one quick question.
PHILIP: Quick question. How do you control the extrinsic charm production?
DANIEL: Extrinsic charm is entirely included within these predictions. So if we had only extrinsic charm, then you would have something that looks like the blue line. And if you had an intrinsic charm plus extrinsic charm, you might have something that looks like the red line. So it's really part of the measurement. You don't need to distinguish between them.
PHILIP: Did you try to look for Z plus Jψ production, to try to factorize the two?
DANIEL: That hasn't been done in this study, no. Really, the key point of this is just... If we look in the forward region, then there's such a large difference between intrinsic charm plus extrinsic versus just extrinsic that you don't really need to factorize them out. It's almost a counting exercise. I agree if you're looking for, say, something that's more C-like, or you're looking for things in the central region where the enhancement is much more subtle, then this would really be needed to kind of probe the intrinsic charm. It's really just intrinsic plus extrinsic is expected to be... Double the value of extrinsic load, once you're in this highest rapidity bin.
PHILIP: Okay. Thanks.
JAVIER: Okay. Thanks, everyone. Thanks, Daniel, for the talk, and everyone for the discussion. So let's move to the final talk. From Javier.
JAVIER: Yes. Hello. Let me share my screen. Okay. Can you see this? Very good.
>> Recording stopped. Recording in progress.
JAVIER: So the measurement of the transverse energy energy correlations and the second is measurement of heavy flavor jet fragmentation using b plus measurements. So for the TEEC... To remind you that this is simply a measurement of azimuth between jets -- for the events, you have the product of those fractions. We can also define the asymmetry as the forward part of this distribution, minus the backward part. And this is done in order to reduce contributions which are isotropic. (inaudible). So these are shown on the plots, on the number two. The TEEC function is the one on the left, and the asymmetry is the one on the right. This is compared to several Monte Carlo predictions.
We can see that the first of all -- Herwig, which is next to -- that's not describing the TEEC. For the endpoints, -1 and +1, while the rest do their job more or less decently. These are used in order to extract the strong coupling constant. By comparing with next to leading order predictions in pQCD, these are normalized, the uncertainties are taken into account using nuisance parameters. So this is the highest Q determination of αs to date, given that the last point is at 4.2 TeV. And what we see is that it is highly compatible with both the world average and with other measurements. So what we do here is essentially measure the TEEC and its asymmetry in ten bins of the scale, and then extract for these bins. Now moving quickly to the second measurement, it's a measurement of b fragmentation using the decays of b measurements (inaudible). And this is done because important uncertainty in final states with b quarks is the fragmentation of b quarks into hadrons. So for instance, such measurements include Higgs to b bar, ttbar production, top mass, et cetera. The inputs that are currently being used to tune the Monte Carlo on HF fragmentation are still using LEP and SLD data. So it's good to have new data from the LHC. Especially at the high energies that are available now.
In order to (inaudible) the Monte Carlos using this eta. So for the measurement, the reconstruction of the b hadron is possible fully by using this decay. So the invariant mass of the B measurements, and what we do is to measure the longitudinal and transverse components of the momentum with respect to the momentum of the jet. So essentially these two variables. This product and the momentum of the B meson and the cross product. We can see in the plot... The Monte Carlos which are currently being used, Herwig, Pythia, et cetera, do not agree within themselves and also do not agree with the data out of the box. So this is a useful measurement to change this.
So in order to reconstruct the B mesons and determine the purity of the B sample, what is done is to use invariant mass fits to the mass of the plots -- the invariant mass of the two muons and decayon. The signal model is double Gaussian, this is 1 minus the (inaudible). Then we have resonant background, B+ to Jψπ, and then the combinatorial background, which is a combination of Jψ and some other track. Which is parameterized as a straight line. So each of the bins in the pT of the jet, the pT of the foundation variable, either (inaudible), I would say a fit to invariant mass is done in each of the bins and the purity is determined. Essentially the fraction of each of these backgrounds and the fraction of signal is shown on the plot below. As a function of C for one bin in the jet pT.
So finally, after unfolding, after determining the systematic uncertainties, one can compare to the predictions by (inaudible) Monte Carlos, and this is what is done. Here we are comparing to four Pythia models, two Herwig models, two Sherpa models, changing parameters in the parton shower, the fragmentation function, and also internal parameters within the LB fragmentation function. This parameter, which is the power of X denominator in the LB fragmentation function, is tuned to the LEPβ, and this is what is called the A14rb tuning. This is the best. While it is true that all Pythia models are doing a decent job, and we also see that the Sherpa sample with the cluster fragmentation model -- does not describe the data well, especially at low pT, which means high C, as you can see in the plots, and also the Herwig 7 sample with the parton shower, that does not describe the data either. We have seen that this is due to the fact that the pT is not correctly set in this sample.
So in conclusion, we have two tests of QCD. One tests the parton shower and the global event geometry using transverse energy energy correlations. And the other tests both the fragmentation properties and the parton shower using the fragmentation of b quark jets to (inaudible). These are compared to the current Monte Carlo expectations, and we also compare the TEEC to the NLO signal regions to determine pQCD. This is the high Q² determination of αs and shows good agreement with the current world average. We have tested the fragmentation models for b quarks, including the Lund-Bowler versus Peterson and other models such as the string model versus the cluster model. And have also explored the sensitivity of the B production to the gluon bb splitting. This is all from me. Thank you.
JAVIER: Thank you so much. Do we have any questions from the audience? If not, I can ask some from the Google Docs. The first question is about if there's any hope of measuring the TEEC in particles rather than jets. Even just the charged particles, the endpoints of the cross section, especially narrow angle, contained interesting information from the QCD.
JAVIER: Yes. This is not only possible, but desirable to do. For determining αs, I think this is possibly not the best method. You will have larger uncertainty due to the hadronization model and the underlying event and so on. But of course, this is also interesting... Implementation of testing... This is definitely possible. It's something that could nicely be done. Yes.
IAN: Could I just ask a quick follow-up to that? For that, how much do the charged particles help you? So are they necessary to move to a fully inclusive or from jets to particles? Can it be done without...
JAVIER: I'm not sure of the question.
IAN: If you want to move from jets to doing it on all particles, is it the tracks that really allows you to do this? Or...
JAVIER: Well, in principle, yes. Because the neutral particles for doing it with neutral particles... We would need to use the clusters. But of course there is a lot of uncertainty on doing that. And the efficiency is not... Is probably not the best. So I would say that measuring it with charged particles is the best way to test these kinds of effects in the endpoints of the distribution.
IAN: Okay. Perfect. Thank you very much. That's very helpful. Thank you.
JAVIER: If there are no other imminent questions, I had another question about the... I guess it was on slide 15 of the long talk. Trying to remember now. The plots showing the gluon fraction of Gbb. And how it was different in the... In data, versus in the shower modeling. And I guess my question is... Yeah. It seems like there's... I guess it's hard for me to tell from this. Because my first question was... Is this... Showing the gluon splitting and the hard scattering, it's kind of normalized. It's just showing what those two shapes look like independently?
JAVIER: Exactly. All of them are normalized to unity. So... This plot is just to see that the fraction of gluon splittings can be tuned from this data.
JAVIER: I guess that was my question. Can you tune separately this increased presence of gluon splitting to BB, which as you know, I'm sure is very important for a lot of other measurements and analyses that have this as a major background or have this as something that is used, especially with these taggers for Higgs to bb and things like this. Is it possible to basically create an additional handle that you can tune the shower generator for this? And if so, would that potentially impact any other observables? So yeah. How would that interplay with maybe other observables that are currently well modeled, maybe?
JAVIER: Yeah, of course, for doing that, tuned to the Monte Carlo... This is always difficult. Because as you say... You can tune one measurement and spoil another measurement. There are measurements in ATLAS currently which are measuring explicitly the (inaudible). So the idea would be to use a global fit to make this tuned. Not only this measurement, but also in order to exploit all the measurements that we have, use them jointly. Not only one.
JAVIER: I see. Thanks. And Raghav?
RAGHAV: Can I just quickly follow up on that point? I agree with you that we should not tune ourselves away from something that might... We might measure later. But at the same point... What exactly are we tuning here? Right? I mean, you're not really touching... Are you touching some of the string parameters inside these models, for example?
JAVIER: Well, what would be tuned is the amount of gluons that get split into (inaudible), essentially.
RAGHAV: But that's still perturbative, then. You're not tuning any non-perturbative parts of the model?
JAVIER: Well, in the shower, the shower is... Includes a number of splittings. Because the (inaudible) is 2 to 2. For the gluon splitting, you tune internal parameters of the shower.
RAGHAV: I see. Okay. That's all. Thanks.
JAVIER: Any other last minute questions? If not, then let's thank all the speakers. Especially for keeping the short time. Which was difficult. And thanks, everyone, for attending, and the lively discussion.
MATTHEW: Thanks, everyone. We'll reconvene in 13 minutes. On the hour.
This event will be live captioned.
ANDREAS: Okay. Let's get back to this session. The topic is spin physics and QCD measurements. So let's get started with the talk from Oleg on jet angularities in Z+jet production.
OLEG: Can you hear me? Let me just share my screen. Can you see my screen now?
ANDREAS: No.
>> Recording in progress.
OLEG: Good day. I'm Oleg Fedkevych, and I'll show you my work, called jet angularities in Z+jet production. So we decided to provide some predictions at a prime accuracy level for the specific substructure of the observables called jet angularities. Defined as follows. Oops. Can you hear me, guys?
>> We hear you, but the screen is gone.
ANDREAS: I have no clue what's going on. Sorry.
>> It's strange. Because it looks like you're drawing something.
ANDREAS: Yeah, but I don't...
OLEG: Okay. It looks like... I lost the control over my screen. Sorry. I don't see anything.
CLEMENS: I can stop sharing for you.
OLEG: Okay.
CLEMENS: And Andreas, you can try to share?
ANDREAS: Yeah. There you go. It's not full screen.
OLEG: Okay. Sorry. It's the first time something like that happens to me. Previous talks were okay. So okay. How do you go? Just going to show my slides and I'm gonna...
ANDREAS: Yeah. Just tell me when to flip.
OLEG: Okay. You're showing the wrong version of the talk.
ANDREAS: Oh, okay. That's not great. Do you see it?
OLEG: That was the introduction, basically. Z+jet production. Yeah. So the right version of my talk... Yeah. We decided to work with angularities, which I define as follows. The formula from above. And the sum there, over i, runs over all jet constituents. And Δi is the distance in the azimuth rapidity plane, where i is a given parton or hadron and these are combinations of jet axis in the azimuth rapidity plane. So we decided to provide predictions for the observables. Not all of them are good for theoretical studies, because it requires α bigger than zero. So we decided to consider three commonly used cases. (inaudible) and also work with (inaudible). We use a algorithm with β=0 and Z cut 0.1. Can you switch to the next slide? We have some of our results. The Monte Carlo predictions... Made with leading order and next to leading order Monte Carlo. For the leading order, we use Pythia, Herwig, and Sherpa. So these are given by the dashed pink line, dotted blue line, and also the green envelope. And in the case of the Sherpa, in the case of the green envelope, we also use merging to higher jet multiplicities, up to two jets.
And the red envelope is our main prediction, our main Monte Carlo prediction. We use Sherpa to high multiplicities. So one can draw two conclusions. If we compare leading order predictions at large values, we see that Monte Carlo -- different Monte Carlo models agree. But if we go to the small values of λ, we see that predictions between different Monte Carlos... We get basically disagreement. We see the Pythia line goes down, for example. And this can be cured to some extent if we switch on grooming. So we see that on the right for the groomed observable Monte Carlo, different generators are more or less in agreement. But we see that next to leading order predictions differ quite a lot from leading order predictions. So we have different uncertainty bounds and different shape. So our conclusion is we have to use the state of the art Monte Carlo while talking about jet angularities. Then one can also take bin ratios between these different predictions and extract non-perturbative corrections, and these corrections can be used to account for non-perturbative physics in our sum predictions. Like underlying event and hadronization. And please go to the next slide.
And these are basically our final predictions. So... Okay. It's a lightning talk, so I skip a lot of details, but basically we implemented resum predictions at NLO+NLO' accuracy level. These results are automated, and therefore very flexible. One can also apply our predictions to basically any set of selection cuts. And here I have comparison between our final predictions given in gray, so gray shaded areas are the summation. And the red envelopes are given by the NLO Sherpa predictions. And we see that for example for the Les Houches singularity, both approaches disagree basically in the whole range of binning. And also grooming doesn't fix the situation. However, if we go to the next observable... Can you please... Yeah. So for example, if we consider the jet thrust, we see in the whole range of binning... Except one bin... On the left... We get very nice agreement between both approaches. Nevertheless, we see resum predictions still have somewhat larger uncertainties compared to Sherpa. And therefore for this particular setup, it also would be nice to get some further improvement in the accuracy. And the same situation is for the groomed observables. On the right, but here we see that... We have some disagreement around this soft drop transition point given by 0.1. Can you go to the final slide, please? Before finishing, I would like to show you some preliminary comparison. Again, the data recently collected by the CMS collaboration. On the left we have Les Houches angularity, in comparison against... Compare data against predictions of Sherpa and resum predictions, now given by the red shaded area, the color coded is somewhat different. Pay attention. And Sherpa predictions are given by the blue envelope. So we see that if we talk about the picture on the left, we see that actually our resum predictions fail to describe the data, whereas Sherpa does a very nice job. So we have agreement in the whole range of binning. And on the right, if we consider the jet thrust, we see that both approaches do a nice job, so we see basically both resummation and Monte Carlo can describe data in the whole range of binning. So we see that uncertainties from the data overlap with uncertainties from the theory. And that's very nice. So these are preliminary predictions, and more predictions are coming soon, as more data will become publicly available. For example, those who are interested to compare predictions against the data at high pT values. Because here we have 100 GeV, which is not the best option for the resummation.
But we see that for the case of the jet thrust, we still do pretty good job. And that probably brings me to the end of my talk. I don't know. Am I... I probably ran out of time already. So... Yeah. Sorry.
ANDREAS: Okay. Anyway, thank you for the quick summary. That was good. So let's start with questions from the audience first. Any raised hands? There's one question in the Google Doc maybe we can clarify that first. So... How do you find NLO accuracy? What logarithms do you resum to NLO? Do you recapture all the algorithms of angularities that depend on jet radius or Z cut or non-global logarithms?
OLEG: Thank you for the question. The precision is defined in the same way as in the paper -- if we talk about NLO plus NLO, concerning the logarithm... Basically if I got the question properly... We basically... We don't provide predictions for Z cut or some other observables. We just concentrate on Z jet angularities. So I guess we don't have all the logarithms we need to describe Z cut. But we also didn't try to do that. So basically when we talk about NLO plus NLO' accuracy, we talk about only three angularity cases, with and without grooming.
ANDREAS: I actually have one more question. Let me find the slide.
OLEG: This one?
ANDREAS: Maybe let's take this one. So the Sherpa uncertainties are much lower than the theory prediction uncertainty. So I wonder... Is this mainly due to the scale or the non-perturbative uncertainties? These larger uncertainties for the resum calculation? And if there's any ways on improving on those in the future?
OLEG: Okay. Uncertainties are actually coming from both. From resummation, because we have some additional parameter, summation scale, which we can also vary, and also from the three generators. Because they also give some quite big different... As you have seen, for example, in the previous slide. Leading order predictions also differ somewhat from the next to leading order. So I guess one can improve it like... By improving the accuracy of the computation. That would be one way. And the second way would be to... Consider non-perturbative corrections like... To resum slightly differently.
So here... Well... We're trying to be very conservative. We try to take all different models which are on the market, but in principle... Yeah. The question is if it is necessary or not. So that would be another option. Also, I mean, here for this particular set, our study has shown that if you talk about k factors, they're also quite large. And that is also what somehow can affect -- jeopardize our resummation, if you talk about NLO corrections.
So maybe one could also try to work in a slightly different set up, where the NLO corrections are not so big. The k factor can be up to roughly 2, for the setup, which is a bit too large. Oops, sorry if you hear some noise. From outside. Yeah. Just to summarize three ways. One, to improve (inaudible) resummation. Which is quite difficult to do from this point of view. One to reduce the NLO corrections, and finally to... Not to consider three different models of non-perturbative corrections, but take only one from Sherpa, let's say. Because as we have seen, Sherpa does a good job here.
ANDREAS: Okay. Thank you. That was a very complete answer. Thanks. Yeah. I don't see any raised hands right now. There's one. Felix, please go ahead.
FELIX: Thanks, hi. I was wondering... So if I understand correctly, in these Z jet events, there are a couple of phase space constraints on the total event topology. How are you taking those into account, especially in the analytic resummation?
OLEG: Can you open my long version of the talk, just to make clear what constraints are... Scroll down, please. Stop. The previous one please... Just one level up. Okay. Yeah. So you're talking about those two. Right?
FELIX: Right. Are you worried... I guess they could also somehow introduce some large logs or so. Right? I don't know how you would resum those. Or if you take those into account in some way.
OLEG: No, we don't take it into account. Basically we just keep Z boson offshell. So in our plugin, and then we just impose the selection cuts. But we don't really... Kind of... Take it in our resummation. Into account.
FELIX: That comes in leading order? Is that how one should think of it?
OLEG: I guess so. To be honest, I'm not the person who performed the actual resummation. So maybe if we have... My coworkers, maybe they can help me. But as far as I know, we don't really do any... We don't get any additional logs due to these constraints.
ANDREAS: I see Gregory raised his hand.
GREGORY: One of the problems, if you don't put any cuts, is that the overall cross section for Z+jets, leave aside any substructure measurements at the moment, the overall cross section for Z+jets has not get access when you go from leading order to --
OLEG: Can you go down? Sorry, just to illustrate that, here is --
GREGORY: The asymmetric cut tends to improve that slightly. And in a way, that cut in the region -- in principle, I guess, if you put a very tight cut, on the asymmetry, then indeed you should get resummation of that parameter. But the thing -- you're still in a region where you're okay and fixed order should do a good job. However, and it slightly reduces the k factor. However, even in that case, the NLO k factor, just for the cross sections, the normalization that goes into these plots, is already quite big, which means it comes with fairly big uncertainties as well. And in a way, that might be something that one should think of, in the sense that even if you measure... Substructure measurement, you're still sensitive to the overall cross sections for the process. In that sense, it would be interesting to compare what's in Z+jets versus dijets, where in principle, the cross section is better known or has less uncertainties.
FELIX: I was wondering if that could somehow lead to some of the differences you guys find, comparing to Sherpa or something like that.
GREGORY: For the resummation, I think there's no... Assuming Z cut small. So there's no specific resummation. We're doing the standard resummations assuming Z cut small. There's no finite Z cut correction. In the context of the Les Houches thing, it's interesting to look into. Whether some of these agreements can be lifted if you insert a finite Z cut correction, the same way they're present for any... Say, for example, (inaudible). We know they're small in the case of the threshold or the mass. There's the question of whether they're still negligible in the question of the (inaudible), but we haven't looked into that.
FELIX: Okay. Very cool. Thank you.
ANDREAS: Okay. I think it's time to move to the next talk. So thank you again, Oleg. Alejandro, can you please share your slides?
ALEJANDRO: Yes, give me a second. Hi, everyone. This is the short talk for the jet substructure measurements at CMS. In the long talk you saw the previous results. That's why I'm just gonna focus on the latest and greatest that we have. For this audience I think I don't have to convince you that the understanding of quark and gluon jets is a vital part of standard model and BSM measurements and searches. The latest studies that we have, these are motivated by the paper that is here.
And the λ variable that actually was introduced in the previous talk. So we studied all of these variables in a wide range of phase space. These help us to understand it better or simulations of those. The simulation you see here in these two cartoons -- the one, the dijet region, you can see the kinematic selection and so on. And the Z+jets region. Just as an aside, information in the dijet region -- we have the two jets, and the more central one, meaning the one with the small eta is the central one, and the one with the largest eta is the forward. Here it's nothing to do with the forward part of the detector. It's just the one with the highest eta. This is important, because as you can see in the plot on the right, for different pTs, this is the quark gluon composition of these jets that we're studying.
So here the black is the central dijet, the blue is the forward, and the Z... The red is the Z. So it's clear that we have a different quark gluon composition in these three jets. Now, as I mentioned, we studied bigger phase space, I mentioned this in the long talk. We need different parameters. And I'm showing you the Les Houches angularities. On the plot on the left, it's Z+jets and on the right it's central jets. Here you can see the detector label, the data, comparing to the two Monte Carlos that are widely used in CMS. MG5+ Pythia and Herwig++. And you can see the comparison between the data and the two generators. Here you can see clearly in the case of the Z+jets, the MadGraph is the one that is -- has a better agreement with the data, while for the dijet, I would say both have some disagreements in the entire region.
Now, more information was given in the long talk, but here let me show you the summary plots that we have, and again, I'm just focusing on the Les Houches angularities. But all of the rest are part of the paper. So in this case, first the two plots on the left -- you can see the data, compared to the two simulations. That I just mentioned. And for each of these bins, represent different types of jets. You can see here. What each bin represents. So focusing on the first plot, here is the ratio between the gluon enriched and the quark enriched region.
The simulation data. And the second plot is the split of the ratios. So gluon and quark enriched region. So from here we can clearly see that in the case of the gluon enriched region, the data, compared to what we call the older simulators or the ones we're using now, the data is basically in the middle of these two. Maybe I forgot to mention here: Each of these points is the mean value of the unfolded or of the simulators in each of these bins.
So yeah. I was saying here the gluon, the data is in between the two generators. For the quark enriched region, we can clearly see that the MadGraph 5 has a better agreement with respect to the data. Now, another interesting thing is that we check what we call new generators, basically state of the art generators, with the data that we unfold, and this is the plot that we saw. We were seeing here on the right. So we compared with Pythia 8 to different tunes. Herwig 7, Sherpa, and so on. And now this is the comparison of the data that we unfold before.
With these two... With this state of the art simulators. So here it is interesting, because we clearly see that if we focus on these plots, for the gluon enriched region, we see that the newer generators are modeling better the data. Compared to the ones that we have for the quark enriched regions. That is kind of the opposite of what we see in what we call the old generators.
So this is an interesting, very wide study of all of these variables. It's explained a bit more in the long talk. With that, let me finish saying that at CMS we have at least two more substructure measurements in the pipeline. So hopefully it will be ready by this year. So stay tuned for that. With that, I think... Yeah. We can go to the questions.
ANDREAS: Thank you, Alejandro. Let's start with the question from Matt.
MATTHEW: Hi. Thanks for this nice talk about this nice measurement. I have a question that might be very simple. But if you go back to the fourth slide... You're showing the new generators over here. And I guess that the only difference between the two Sherpas that you're showing is that one has an extra leg in the matrix element?
ALEJANDRO: I think so.
MATTHEW: So why is the substructure so different if the parton shower is the same? It looks like they differ quite a bit in the plot on the middle right.
ALEJANDRO: That's a good question.
ANDREAS: I think I can weigh in here. It's an additional jet simulated. Which actually changes the quark gluon fraction of the first one. So there's a change of quark gluon composition if you simulate more jets.
MATTHEW: So would you see big differences between Powheg and Pythia 8?
ANDREAS: We don't have Powheg in the comparisons yet, so we don't know yet.
MATTHEW: And you only have Herwig 7NLO, I guess? Okay. Cool.
ANDREAS: There's another hand raised from Raghav.
RAGHAV: Nice talk. I had a question. These angularity distributions have this large shoulder on the right side. Right? So when you compare your means, and you say that... Okay. You have this kind of sandwich. And you go to this newer tuned Monte Carlo and you say that it looks better for gluons, is that because you get the peak correctly? Or are they tuning better to the tails of the distribution? Because that has a better handle on... It might be useful to look at both mean and... I don't know. Some kind of width of these distributions, simultaneously?
ALEJANDRO: I think this is one of the questions that was -- that is in the Google Doc. For the summary plots, we concentrate in the mean, but you are correct that these will translate the mean difference between all of these. At least in the paper, that is coming in the next weeks, I think... We see all of these plots, and you can't really compare...
RAGHAV: Then you can see.
ALEJANDRO: You can see everything. This is a summary to have an overall idea what is going on with the generators and the data.
RAGHAV: That's fair. I also didn't see that in the Google Doc. Sorry if that was already asked.
ALEJANDRO: You're asking a question that another person did.
RAGHAV: Okay. Good.
ANDREAS: Okay. Max.
MAX: Thanks a lot. I think this is a bit related, now that we have the plot here. You had a very nice sort of comparison... You know, to many different generators, quantitatively with the mean. Did you make any sort of efforts to quantify the agreement more generally for the full distribution? Things like the widths, maybe even, because often those are very important for the quark gluon tagging and so on... Applications.
ALEJANDRO: I think this is a good point. We didn't include this in this paper. But I think for the people that want to see the distribution, you can see it clearly in the paper. All of these variables. Or also of course in the HEPData, where you can see the entire phase space. But we didn't include these as part of the summary paper.
MAX: The results are there because the raw distributions are all there. So that's important. Okay. Thank you. Have you had one more...
ANDREAS: One more can go and we switch.
MAX: Do you have the PTD plots floating around? Any of these. Yeah. Maybe the one that has the data-Monte Carlo ratio... Or data simulation ratio for PTD? Yeah. Yes, these. This is great. Should I understand anything significant from the fact that if I look at the left sort of four columns, there's this pretty substantial disagreement in all these variables and the means, and it's always that simulation is greater than β, and for PTD2, even though I think of it as something quite similar to a width, it ends up having the reverse the simulation data sign? Is there anything meaningful to that? Or is it just... Sort of a construction of the variable -- not too interesting.
ALEJANDRO: When I read this in the Google Doc, I didn't get it. Just so understand: You're saying why this is kind of flipped with what we have in the rest?
MAX: Exactly. Is there anything interesting that we learn about the disagreement, the nature of the disagreement, and so on, when these things are to the left or right of each other? Essentially?
ALEJANDRO: As far as I know, it could be that the PTD has kind of a reverse in behavior. In the sense that the mean of the PTD for gluons... For quarks... Is the opposite of the others. In average, these values are opposite with respect to the rest. So I think this is a known feature when we are comparing gluons and quarks. But...
MAX: That's all. Thank you.
ANDREAS: Okay. Thank you again, Alejandro. And let's move on. Alexander.
ALEXANDER: Yeah. Just share my screen...
ANDREAS: Switching to spin correlations.
>> Recording in progress.
ALEXANDER: Yes. So... I'm trying to give a short review of the talk, which is already online. And this is the work we did on spin sensitive jet observables and their resummation. Originally within the PanScales collaboration, with parton showers, in this talk, it focuses more on the observables themselves. So the idea was to come up with a set of observables which, on the one hand, we could use to test our shower framework, but also which could probe spin correlations in jet substructure. And to be clear, if you've seen -- watched the talk -- is that we haven't done any real Pheno studies yet. So at this point, the application is not entirely clear. We're thinking about several things, but I don't have a very clear application in mind. But we've already tested that things are going to work. But let me just recap how the two observables are defined.
The first one is an observable which is defined inside a jet. So you imagine you found a jet already with some jet finding algorithm, and you reclass with Cambridge/Aachen in the usual way, and you're interested in finding two branchings inside your jet, which are going to be sensitive to spin correlations, and at fixed order, the pictures that you have... Your hard probes, quark gluon, amidst soft gluon, which defines a plane with respect to the original hard parton, and then that gluon splits into gg or qq and that defiance the second plane and gives you the angle at fixed order.
And the way to define this in a jet is to do the reclustering, and then finding the highest kt branching with some Z cuts, some sort of soft drop, defines the first plane, you follow the softest branch, find the highest kt branching, and define the second plane. That gives you an observable between those two planes.
And then the other observable is -- we call ψ11', which is an observable which -- rather than correlations in ψ1 yet looks at correlations in partons in the two jets, the splittings across two jets. And this particular observable we like to think of as being some sort of EPR-like observable, because it's a classical way of thinking about EPR. You have two distinct sides of your event. Two things which classically should not be able to talk together, the two jets or two spins inside a jet, but you find correlations, which are not there, classically.
And the definition is more or less the same. Heavy jets, reclassify with Cambridge/Aachen, and you find the highest k1 branching with some Z cuts and that defines two planes. You can compute between them. And I won't tell you what it looks like at fixed order, but just to say in addition to the shower implementation that we have of these -- of spin correlations, we also extended the MicroJets code to include spin correlations at the single algorithmic level.
And this particular framework we've validated against a recent calculation of a triple energy correlator, just to say that these effects -- we have perfect agreement with the analytic results. The theory predictions. So one of the interesting things we found is that these particular observables are not just sensitive to the spin correlations of those two hard splittings. Sorry, not hard splittings. The two first splittings. But you get an all order effect for the correlations at all order, basically. So if you look at the modulation as a function of the cut, the Z cut, that you define in ψ12, you find that as you require the Z cut to become higher and higher, the effect of resummation becomes larger and larger, up to 15%, 20%.
On top of the fixed order prediction. And just to give you a sense of the sizes here, if you're looking at the Δψ12, with the symmetric z1 and z2 cut, if you look into the jet and don't care about whether it's (inaudible). You find 0.025. But with these numbers that we picked not completely arbitrarily, but without tuning them, you find them up to 4%. Which requires more study, to see if this is actually feasible. And I wanted to flash this last one. To say this has all been implemented in PanScales and has also been validated perfectly in the collinear limit. We get exactly what we expect from the MicroJets code and this triple energy correlator resummation. Yeah. I think that's the overview.
ANDREAS: Okay. Thank you very much. Any questions?
ALEXANDER: I think there were some on the sheet.
ANDREAS: Yeah. I think... To clarify better why this Z parameter... Yeah. Not sensitive to all orders, but what you define is...
ALEXANDER: It's a little bit of a technical... What I mean by it... If you move it at these ψ12 observables, even the EEEC observable has the same property -- if you compare the fixed curve with the all order resummation, it's not just a scaling of the curve, but actually a modulation in the resummation itself. When you include all the spin correlations to all orders, you get an effect there. And what I was trying to say with the D parameter -- you might ask -- spin correlations have been studied in the past. What is the novelty here? And it is that all the observables that are -- have been studied in the past, and even those which are sensitive to splittings inside jets, actually have the property that they don't really... They don't depend on the all order structure of the spin correlations. So for the D parameter, at next to leading log, you have the factorization where the differential information in the D parameter -- you only need to know as fixed order, whereas the resummation is actually contained in the C parameter. But I have to admit that this is a bit of a technical thing. But it's more to say... These observables that we have defined actually sensitive... Not just to have the spin information in the hardest emissions, but actually the spin information at all order.
ANDREAS: Other questions? I have one more myself. That is... Well, how... In what kind of final state could... Measurement of that be useful or even be sensitive to these small percent effects?
ALEXANDER: So I think there are two applications you can think of sort of immediately. One is that you could think of measuring this ψ12 directly in, say, Z+jet events. At the LHC already -- so you find your event and you tag a Z+jet event and look inside the jet to see if you can find these correlations. And we have done a few studies. So we think it's feasible. I don't have the plots to show that that's actually the case. But the other thing that you might want to think about is... If you, for instance, do quark and gluon tagging inside a jet, and you're just feeding some Monte Carlo information into your machine learning algorithm, if you don't get these effects correct, you might be looking at... You might be making errors on the order of a few percent. Which is something that you're not interested in doing. Because that ruins your quark gluon discrimination.
So I think there's also a possibility of using it as an input, in addition to other things in machine learning. Quark gluon combination. Because what you find is that quarks and gluons are very different. So they're basically... Either in plane or out of plane, in terms of these splittings. So they have very different behavior.
ANDREAS: That's very interesting. Okay.
ALEXANDER: But again, this needs study. I'm not saying this for sure is going to help your machine learning algorithms.
ANDREAS: Another question from Matt.
MATTHEW: Hey. This is maybe just... I think I could benefit from a little clarification, I guess, about this point. So my understanding was that these variables are sensitive to effects that are not in the Monte Carlo that we're using. The spin correlations are treated properly for pen scales, but maybe not in the Pythia parton shower. It would be difficult to do a measurement if it's not in the Monte Carlo we're using right now to try to unfold something like this? But maybe it would be okay?
ALEXANDER: So I should mention that in principle, the spin correlations are also contained in the Herwig shower family. So we haven't... In PanScales, it's been fully validated. The Herwig people can tell you exactly what's inside their shower. But in principle it should be the same as in the PanScales family.
They have some spin correlations in Pythia. But not all of them. But what's important to say is that it's a little bit unclear even in Herwig whether or not... So... Even if you want a shower that does not contain spin correlations, you'll see movement correlations induced essentially by the recoil of the shower, which are to some extent spurious spin correlation effects. And I think what I was also trying to say... Let's say you're trying to measure this and your shower does not contain spin correlations correctly and you're not using the right observables, you might find effects in your Monte Carlo that are definitely not there in the data. So no, you cannot just take Pythia and go out and measure this.
MATTHEW: So I guess I'm interested in... I guess we have several PanScales collaborators here. When do you think we might be able to start getting predictions with the PanScales parton shower to start looking at in the collaborations? Is this still quite a ways in the future?
ALEXANDER: I'll leave that to one of the more senior members that's connected. I saw Gavin and Gregory.
GREGORY: I should have raised my hand later. So let me postpone that, Matt. Actually I had a comment on what you just said about including spin effects. If you think about what you did already, measuring the Lund plane, for example, there are some effects, measuring the primary Lund plane, there are some effects there that depend on clustering, and clustering depends among other things on the site of one angle. In that context, we know that Herwig doesn't get non-global logs necessarily right, at that accuracy, at least. This is not something you've worried about when you've done the unfolding. And the same question there: Should you have worried about this in that unfolding? Knowing that you already have effects which are not well reproduced, not supposedly well reproduced by Herwig?
So there's a bit of the same kind of comment there. I'm not sure exactly how to take it. But at least to point out that the comments you just made... Already existed in previous measurements. Now, in terms of future PanScales predictions, I think there's still quite a bit of a way to go. First the moment is just... We just have an E+/- shower. So going to a shower that would be relevant for LHC. And then a few other things we want before being able to have genuine phenomenology predictions. I guess one aspect is pp. Another aspect is for E+/- matching with hard matrix elements -- these aspects, I think, are important, before going into starting to dig into phenomenology. Hopefully within the next couple of years we'll have something. But you know how making predictions is difficult about the future. So yeah. That's hopefully within next year... We'll be able to do something. So that's... Gavin, correct me if you want to be more optimistic or pessimistic. That's at least what I would say.
GAVIN: I think I agree with you, Gregory. We're progressing, but I don't think we're ready to tell you in a few months whether we're gonna have something out. I think the other thing, the other small comment, is that... These effects... The spin correlations aren't just a few percent. Because they're very different between the gluons with qq bar, they come with opposite sign, but gluons related to qq bar... Let me say that again. Gluons split into gg and the gluons split into qq bar, they come with opposite signs. And especially for the gluons splitting to qq bar, there's a very large effect. Alexander has shown this on the left hand side here. The top left plot. So there are in principle sort of 80% effects in G space spin correlations that you should be able to see. If you can tag somehow a gluon splitting to flavor/antiflavor. And I think the question of jet substructure with flavor tagging for individual particles at the end is something that is interesting, not just in this context, but is emerging -- like, it might be interesting also for future colliders. For example, E+/E- in understanding certain Higgs decays. There's a lot of open questions that become interesting once you ask questions about tagging the flavors of individual prongs. For example, (inaudible).
ANDREAS: I think that's a good time to switch. So thank you again. For this interesting talk and discussion. Now we have the last talk in the session, from Ian. Maybe you can...
>> Recording stopped.
IAN: So hopefully you can see my screen now. Is that good?
ANDREAS: That works. Go ahead.
IAN: Perfect. Thank you very much to the organizers for putting on this nice conference. Today I'm gonna talk about some work I've been doing primarily with Wa Jing, an exceptional student, Hao Chan, and later on work with another excellent student, Joshua Zandor. Since I had 21 slides, I decided it would not be useful to flash through them in a five minute period. So I decided to try to condense the whole story onto one slide, and I'll just kind of step through it. So this talk was based on essentially two different points, which I think are each interesting in their own right. But we kind of combine them for this particular application.
And so the first is understanding these quantum spin effects. In jets. Which was just very nicely talked about by Alexander Karlberg. And so everyone is familiar with the fact that if you have some decaying Z boson, it imprints its energy in the kind of macroscopic energy flow that you see in the detector. But as you start going to these higher point correlators, you can have more interesting effects, so if you produce some microscopic gluon, it has some spin structure, and so if you have two of the correlators out here, you can rotate them with respect to the third by some angle, and what you'll see is that the gluon spin structure is imprinted into a cos-2φ modulation. The fact that there's this 2, even though the gluon was spin one, is because it's coming from an interference effect between two gluons. So one plus one is two. And you get the nice cos 2φ modulation that was discussed in the previous talk.
So what we wanted to do was to start developing some theoretical tools to understand these higher point correlators more generally. So tools that would scale to understand, for example, the spin correlations in essentially arbitrary point correlators. And so the way we did this was to combine techniques from what's called the light ray operator product expansion, and the use of symmetries. And so for this particular three point correlator, shown here, the way we analyzed it was to start on this right hand side here, by looking at its symmetry structure. So these three points live in the transverse plane to the jet, so you can think of this as some 2 dimensional complex plane. And you can study the symmetry structure of the correlator in this plane.
And so, for example, if you boost the jet -- or the impact of the Lorentz symmetries, on this plane -- as you boost the jet, this dilates in and out, or you can rotate it, et cetera. So just like when you're analyzing problems, for example, in quantum mechanics, where you want to use functions which respect the particular symmetries, so, for example, you use spherical harmonics, here you can derive similar partial waves which respect these particular symmetries of this jet substructure observable. So the exact form doesn't really matter. It's just given here.
And for this, these are two-dimensional conformal blocks. So once you have the symmetry structure understood, you can then look in this light ray operator product expansion for the particular operator that generates this cos2φ modulation. And so it turns out again... This is very easy. Because in QCD, there's a unique leading twist operator, which has these two gluon fields shown here, and so it's a purely gluonic contribution, as has been described in the previous talk.
And so once you have this operator combined with the... Its anomalous dimensions, and this partial wave, you can then just immediately get out the prediction that you want for this cos2φ modulation. So this may seem very different than how jets have been described in many of the previous talks. Because there's been no talk of splittings or what's been happening inside the jet. But if you just work out the prediction of what you get by kind of plugging things in, you see that it exactly agrees with this prediction from the PanScales Monte Carlo described in the previous talk, which is based on these splittings. And so it provides a different way of thinking about how to get these results out.
Which we think is very powerful for these higher point correlators. So for the particular case of this 3 point correlator, it turned out to be quite small, and I'll discuss more based on the questions why -- but this is also discussed nicely by Gavin et cetera, by the end of the discussion. Namely if you consider a gluon splitting to qq bar, you get the exact opposite sign, as for gluon splitting to two gluons. For this particular case, the azimuthal dependence largely cancels out, because you have almost the right number of flavors compared to gluons.
And so one thing we were... Which was, again, nicely mentioned by Gavin... Thinking about... Which also relates to previous measurements, in particular by ATLAS, is to maybe in light of these recent developments, both on the parton shower side and on the analytics side, to reconsider the case of glu to bb in Z+jet. And so the reason why you want to do Z+jet and not dijets is that dijet amplitudes are MHV and so you don't get interference, whereas in this case, you get this plus or minus gluon in here, which can interfere with itself, and by removing -- or if you can tag the b quarks, then you essentially remove this contribution here, and you just get a large order one -- as was emphasized by Gavin -- quantum interference effect, which I think would be very nice and also interplays with better understanding heavy flavor in jet substructure.
And so for the experimentalists, I kind of hope that I've just illustrated how much fun we're having, playing with these energy correlators, and that you can do some really neat things. But for practical applications, what I actually think may be more important is what was discussed yesterday in Yibei's talk, namely that for any of these energy correlator-based observables, we can immediately extend the calculation at the same perturbative accuracy to being on tracks. So, for example, these spin correlations and these higher point correlators, we can do them all on tracks, essentially the same easiness as without tracks.
And so my hope is that this will kind of open the door to precisely measuring some of these more complicated effects at very small angles, where the high angular resolution is required, or, for example, in heavy ion contexts, or wherever you want. And so... Based on... I quite liked Pedro's organization yesterday, where he just kind of winged it on responding to some of the questions. And so I thought I would just do the same, and just kind of go through some of the questions which were in the Google Doc. And I apologize that my writing is quite messy.
So the first... Which has already come up a number of times... Was related to exactly why these spin correlations are vanishing, and whether or not it's related to supersymmetry. And in general, these spin correlations will vanish in supersymmetric theories for the reason that you can often relate spinning operators to scalar operators. And so if... So essentially the reason why these jets are largely classical is because if you do not flavor tag, then QCD is approximately supersymmetric, and so there's kind of a priori -- there's no obvious reason why jets had to be so classical and not fully quantum. And so maybe a similar context, because these three point correlators are relatively complicated, you can understand this in the basic context of E+/E-, so if you look in your basic QFT books, you find it has this very famous 1+ cos² θ, the angle between the polarization vector and the direction of the produced quark.
And so if instead of just producing a quark, you measure just this one point function of the energy coming out, so you sum over whatever particles you would have, then again what you'll find is for each particle, it has exactly this cos² behavior, but with different coefficients. So you sum over quark going out and then you also have the case of a scalar going out. Or if you have a gluon, a gluon going out. And you'll find that whenever you sum over the appropriate fermion boson pairs in any supersymmetric multiplet, this will always cancel. It holds to all orders, just because the current -- which in QCD would be this shown here -- in a supersymmetric theory, this is always within a multiplet with the scalar, so you can always reduce this to the case of a scalar. So in the supersymmetric theories, the spin correlations would always cancel. This is why by doing the flavor tagging for glu to bb, you make it very not... You break the asymmetry between gluons and quarks. So the second question was: Can you use these high point correlators in hadronic physics? E.g. for Jψ or λ? I know a number of facts about this. This was proposed in 1978, by Efremov, who was interested in understanding parity. An interesting feature once you have the three point correlators is you now have a handedness. For example, if you order in increasing lengths, this way, of the sides of the correlator, they can now have two different orientations, which are related by parity transformation. So in principle, they don't have to be equal. But at this point, people couldn't really calculate these things. This was rejuvenated again in 1992 by Efremov and Nachtmann and 1995 by Dixon, and there's a talk later today which is also somewhat related. The short answer is I don't know if these can be useful, but I think now that we have these more modern techniques, both on the parton shower side and the analytic side, one can kind of start playing with these heavy flavor or just flavor in general... And trying to understand things in more detail. And then yeah, this was the third question. But I think this has now been quite nicely addressed by Gavin. And the real reason why the glu to bb is nice, is because in QCD you have almost exact cancellation between the minus sign for the fermions and the plus sign for the gluons, and so if you... And you need the intermediate gluon to have interference to get the cos2φ, so if you remove the gluon and just have the heavy flavor contribution, you get an order one spin asymmetry. So there there's no suppression of the quantum nature. So it is... Would be very interesting to study. And so I think that's all the questions.
ANDREAS: Thank you, Ian. Most questions already answered. Are there any further questions? Any hands raised? I see one question from Gavin. Or comment. Go ahead. Is that a hand raised from before?
GAVIN: I just wanted to say that... The spin correlation -- you mentioned me many times, Ian, but I think it's really Alexander, Rob, and Ludo.
IAN: Because you raised the question at the end to emphasize this point.
GAVIN: They did 90% of the work.
IAN: You raised this particular point of the cancellation in the discussion. So that... Stole some of the things I was going to point out. But yes.
GAVIN: And you... You were the first to put out something suggesting that those be measured. So...
IAN: I think the glu to bb... Once you can flavor tag, then order 1 quantum... It would be very interesting to measure.
GAVIN: Yeah, definitely.
ANDREAS: If there's no further questions here, I think the time of the session is anyway at the end. So let's thank all of the speakers quickly. And yes. See you again in 15 minutes for the next section.
ANDREY: I'm giving talks during these months... It's hard. But otherwise it's great.
ALBA: Ha-ha.
ANDREY: If not (inaudible) it would be even nicer.
ALBA: Right. I don't know if we have to start right away. Or the organizers want to say something.
MATTHEW: I was just typing you a message. You can start whenever you want. The speakers I guess know the format. The five minute length for talks and the rest for discussion. Whenever you're ready, Alba.
ALBA: Hello, everyone. I'm Alba and I will be the chair of the heavy ion session.
>> Recording in progress.
ALBA: And we are gonna follow this format throughout the conference. Five minutes short presentation with the highlights of each talk. And then we will read the questions in the Google Doc and finally open up the floor for questions from the live audience. And the first talk of the day is by Andrey from Santiago about ab initio coupling to collective flow.
ANDREY: Hi, all. First of all, let me thank organizers for the great workshop. It's a very curious format. Not simple, but definitely very curious. I'll go through my slides to highlight the results. You can find our very long paper. 80+ pages. Definitely five minutes talk is not a place where you can put everything. The idea is very simple. There is an old concept of jet tomography. You want to use jets to study how matter has evolved in space and time. Jets... Definitely complicated objects. Jets definitely look better to study the realtime evolution of matter and soft particles. They're less entangled. And they go through the plasma like a bullet through an apple. And see the evolution of plasma at different stages, et cetera.
And in our work, we are essentially trying to answer this question: Does a jet feel the flow? Basically a jet going through matter, matter is flowing as part of its dynamics. Can we use jets to see how... See how jets and flow interact with each other? And the answer is yes. So I will try to argue that... For the last four minutes I have. So all of those principles -- a principal approach to describe interaction of jets with nuclear matter essentially models the matter with classical... Not classical, but some collective potentials. And jet produced by some source, interact with the matter, and expansion... The BDMPS-Z approach -- essentially how you treat these interactions, whether you sum them or focus on a single interaction, you can try to keep more kinematics available, et cetera.
And all these approaches are in the original version. All of them 20 years old. Treat the sources of the potential, of the color potential, to be static, essentially. And what we do... We allow them to move. So if you start with the static case, the simple thing you can look at is jet broadening. So you look at the distribution of jets modified by interactions with matter, and ask how the transverse momentum is distributed for this sample. The two diagrams entering the calculation and the standard 20 years old answer. Here is the cross section, which depends on the model of the potential and the initial distribution, which you input in the calculation.
So what we did -- we extended the calculation to include the flow. If you have the flow, flow couples with the subleading order in reverse jet energy, essentially (inaudible) order. And it's a bit formal. It's probably hard to describe in five minutes. But one can look at the models of the distribution. And for the model distribution, which is δ function, distribution with transverse momentum, one can find velocity generate all the moments, the momentum, which are zero without velocity. And you can integrate them. It will depend on the choice of the potential. You have to put some model for the distribution function, et cetera.
And in the simplest example, you see that nothing depends on -- you know direction, et cetera, that's basically what you get. So that's the kind of observables you can study this way. So the flow bends jets, and that's how it looks like. Your jet going through matter if you're assuming matter is flowing one way. That is what will happen with the jet. Similarly we developed formulas to include effects of non-uniform flow and non-uniform matter. You have gradients of the density of the sources in this matter. They also bend the jet in a simple way. Also a little bit... Less trivial way. So the moments will look uglier. So similarly, one can go further and look at different observables. Look at different observables for, for instance, gluon emission spectrum. That's the next simple thing you can look at. That is the standard answer -- just here in the small x limit for broad source approximation, assuming these guys essentially are independent, and the connection we found... This factor is also appearing here, and one can figure out that it's correct boost of the answer. Also we do the thing. And it tells you that the jet emits subgluons and this limit is preferably along the transverse flow. And the illustration of the thing... And just stop here with the summary.
We developed the formalism of the coupling description in the expansion to include the flow and the flow effects and effects for non-uniform... Not uniform evolution of the matter. These effects generate odd moments of the momentum in the simplest observables we look at, but one can go further and study any other distributions, going to objects in the same way. And to look at them.
Okay. So there are some things about how the QCD results -- the coupling to the flow is not that sensitive to the effects of the theory. The same formalism can be extended to DIS context in a very similar way, displace of the moments by the sources -- you can try to probe the matter and its realtime evolution, say... Angular momentum, angular motion of partons on nuclei inside. Okay. These results can be applied to different plots in similar way, if we use the same kind of approach like expansion or BDMPS-Z. And they're ready to be included in simulations, like (inaudible) collisions. Thank you.
ALBA: Thank you, Andrey. Finished more or less on time. You have a question on the Google Doc that I don't know if you had time to read. But the question was: How extendible is this formalism? Is it straightforward to calculate multipoint correlation (inaudible) the energy correlator?
ANDREY: The first part is simple. It's as general as possible for this formalism. I'm not sure about what exactly is asked in the second part, because if we're not talking yet about the matter response or matter behavior... So you cannot ask how, say, energy density inside matter correlates between points. Else you can go there in the future. But it's a completely different story. On the other hand, you definitely can study any observable which can be studied with... (inaudible) approach, expansion of the (inaudible) for the jets in the medium. You just need to construct the proper distribution and the elements, which we developed -- which we described in our paper. This is the same. So you can just take that and construct whatever distribution you want.
ALBA: Thank you. The person is in the audience. And wants to ask further... Please feel free. Or anybody else.
FELIX: Thanks for the nice talk. Naive question -- can you comment on how your results are related to medium response? And so on. Which we usually do in simulations. Is that something you can do in this formalism?
ANDREY: You can definitely try to develop medium response. But we don't yet look at that. So at the moment, we ask how moving matter affects the jet. So it's a universal question. If you want to ask how the jet affects the matter, you need to go one step further and control what happens with the medium sources. You can do that.
FELIX: You think you could in principle do that?
ANDREY: I'm working on that right now. It's not simple. So we'll have to go and simplify things one by one. As usual. So there will be many simplifications. You probably have to go to linearized (inaudible) in a simplistic way. But you can.
FELIX: Cool. Interesting. Thanks.
ALBA: My question was more or less... If someone else wants to ask, I will stop, but in the mean time, it was more or less along the same lines. If you introduce gluonic spectrum that you have derived into a Monte Carlo, aren't you double counting? If you have a (inaudible) profile already, QR dependence, time dependence... Is this not part of what you are... Of course, this is a little bit more medium oriented. But is it not part of what your calculation will also do at the same time?
ANDREY: I don't think it's going to be double count. You just need to be careful... You need to decide whether this radiation by the leading parton is included into the jet or into the flow. So you need to go there and make... Think carefully about what you call a jet and what you include in the matter. On the one hand, in the simplest case, where you control the parton and it leaves the plasma, you say all this stuff is distributed into the flow. Then yes, it's part of the flow, no problems. Alternatively, you could say... It's more energetic, so it's small angle and part of the jet... So that's... Actually, that's part of the answer to Felix. That's why it's hard to study the medium correction. At least naively. So you need to carefully think about that. Or model that. But the good thing that all these coupling of the flow to the jet... You don't necessarily need to ask how jet changes the flow. Because if one jet doesn't change the flow too much... So we can try to play with cutoffs on the two sides. How my jets are affected by the flow. So look at something like... All the jets. In this work, we basically give you the answer how to do that. The simplest answer looks like that. But in the real world, the plasma depends on the... Is different... Along all the jet paths, et cetera, et cetera. And you can...
ALBA: You have another question by Gregory.
GREGORY: Yes, thanks. Sorry if there's some background noise here. Things are a bit hectic. What I was wondering is whether you thought about what would be the phenomenological imprint of this matter motion. Either on observable region of phase space or even some specific part of the gluon emission spectrum. Most affected by this. And sorry, I missed actually a fraction of the talk unfortunately because of said background noise.
ANDREY: No problem. I guess the simplest example is here. So basically if you have no flow, you have no direction to generate some odd moments of the jet momentum. As you calculate, you had averaged the p² per unit length. You can do the same thing for (inaudible), which is directional observable, and it's zero without flow. And then you can basically see that the jets are bent by the flow. Then you need to go and do event studies, and some of my collaborators are working on that in different ways. But that's not yet done in this paper. But for the phenomenology implication, the jets start... Essentially bent by the flow and bent by the gradient and then you need to look at how that happens. The opposite side, downside, these effects are weaker. Because all of them are in one or other way suppressed by the energy. So we have 1/e, and if your jet is 100 (inaudible), it's almost negligible. You go to 10 GeV jet, which is hardly called a jet. You definitely will see some... Like... Observable effects. So you need to try that. You need to somehow look at lower energy jets. Weak jets. Or you need to play with particular events, whether the flow is stronger at LHC and try again to choose lower energy jets.
GREGORY: Right, but the jet will be made of several sources. In fact, some of them will be lower energy, so you may have an imprint due to that, at the end of the day.
ANDREY: Yeah, but observables... That's kind of the things you want to ask.
GREGORY: Okay. Thanks.
ALBA: That's perfect timing, because we have to move on to our next talk. So thanks, Andrey, for the nice presentation, and we're moving to the next speaker. Wai, please.
>> Recording stopped.
ALBA: Wai, are you there?
WAI KIN: Sorry. Can you hear me?
ALBA: Yes. But we cannot see you.
WAI KIN: All right. Can you see clearly?
ALBA: We have a black bar. Maybe from one of your windows.
WAI KIN: Is it better? Thanks for the organizers... Is it okay?
ALBA: Yeah, go ahead.
WAI KIN: First thanks to the organizers for the opportunity to present this talk. It's about time reversal odd side of a jet, based on the work by Liu and Xing. So we knew that 3D structures of the proton were studied typically using either semi-inclusive hadron production, as well as jet production or hadron in jet.
And a pure jet was forced to be able to probe only a subset of the TMD PDFs. Namely, 4 out of 8 at leading twist. And this work investigates the possibility of probing all TMD PDFs with jet. So if you look at EIC... There's a lot of statistics at low pT in the forward direction. We typically look at low pT jet at EIC with pT of order Λ QCD... Unlike LHC, for which only jets with high pT are of interest.
The thing is, we can still get jets if we use the suitable jet algorithm at low pT. For example, the spherically invariant jet algorithm, which involves energy kt. So at low pT Q² we don't have a problem studying jets. From SCET, we have a factorization of the cross section of the jet production at low pT. So this is factorization theorem with hard function TMD PDF and jet functions. And we can look at the azimuthal asymmetries to determine the TMD PDF and TMD JFs. There are 8 TMD PDFs at leading twist for the PDF.
Traditionally, there are only four of them here. Which were accessible by jets. So traditionally on the jet side, we know that high pT jets are of interest. And production of high pT jets are perturbative. Since it's Λ QCD symmetric... When we consider low pT... Non-zero T odd jet function when the jet axis is different from the direction of the fragmenting parton. This is similar to Collins effect in fragmenting function. So there are a couple of energies of the jet function... The universality... If it's universal. Second is flexibility. That is the flexibility of choosing a jet algorithm, a jet recombination algorithm, which means defining the jet axis, provide us the opportunity to film the QCD non-perturbative dynamics. We'll demonstrate it later.
For the third is that we have high predictive power. For example, as the jet has a lot of hadrons, and so is more... It has more perturbatively calculable degrees of freedom than the fragmentation function, for instance, in the WTA scheme, we have total Z dependence determination. Like that. And for the non-perturbative predictability, we can follow the study of these authors. And jet can be -- jet function can be factorized into a product of a perturbative coefficient and a non-perturbative factor. And the non-perturbative factor is the operator definition. It is the vacuum matrix element which can be calculated on the lattice.
Like the TMD soft engine study on the lattice in the past... So it's quite different from the TMD fragmentation function which is an operator element that contains the hadron. We demonstrated azimuthal asymmetry at EIC, which probes the transversity, similar to Pythia, with the lately developed package which includes the spin interaction, measured jet charges to enhance the separation, use spherically invariant jet algorithm for low pT jets. So you see a effect of low pT here, with jet charge separation... Blue and red. For positive and negative jet charge.
So the WTA scheme is defined in this way. And that... The initiating parton has a direction different from the jet axis. If you go to E scheme, which is the traditional jet axis, the parton has -- the same direction as the jet axis. So this asymmetry vanishes at low pT. So you see this effect of changing from one axis to another, gives us an opportunity to film this non-perturbative physics at low pT. This demonstrates azimuthal asymmetry in e+ e-annihilation. Slide here. So the summary... We introduced the T-odd jet function, which is is relevant for low pT jets at EIC, the T odd jet function can probe all 8 TMD PDFs at leading twist, it has advantages of universality, flexibility, and high predictive power. T odd jet functions provide new input to global analysis of non-perturbative proton structure. Thank you.
ALBA: Thank you for the presentation. Again, there was one question in the Google Doc. Do we need an EIC to study the T odd jet function, or could it be proved at LHC for example where the detector is designed to capture forward physics?
WAI KIN: This is an interesting question. We are planning to study the pT collision... And we haven't done the simulation yet. But my impression is... I know that LHCb is very good in the forward detection. And it can detect for pT (inaudible) so I think this is kind of marginal. Although I haven't done the simulation for EIC yet. But if you look at EIC... Yeah. Typically pT of all the Λ QCD... If that is significant, at run 1 200 GeV or below. It's not conclusive. But it's my current opinion.
ALBA: Thank you. Any other questions? Felix?
FELIX: Hi. Can you hear me? Yeah. Thanks for the nice talk. I had a question. You mentioned the charge and you said it's not mandatory to introduce it. So... My understanding... Naive understanding was that a lot of these asymmetries vanish if you're not really sensitive directly to differences between, say, U and B quarks or something. But if you say you don't actually need to introduce such a sensitivity, then shouldn't a lot of these things vanish for jets, as opposed to, say, hadrons? Or how else do you introduce sensitivity to different flavors?
WAI KIN: Definitely I would say introduction of jet charges enhances the sensitivity to this asymmetry. There is definitely cancellation between... If we don't separate the flavor. Here it only means that the jet function, the formalism... Works no matter whether you do it with flavor separation. Yeah.
FELIX: You get a non-zero result no matter what? Is that what you're saying? Without performing any additional measurement on the jet, it's anyway non-zero?
WAI KIN: I'm not sure if it's non-zero for the asymmetry, but the jet function is still there. This is what I meant.
FELIX: Okay. I see. These are projections for measurements that you show there on that slide in the figure? What are the uncertainties that you show there?
WAI KIN: You mean this plot?
FELIX: On the other slide. Page 8. For example... Is this projections? Or data?
WAI KIN: These are from Pythia. Statistical uncertainty.
FELIX: Even though these are not implemented in Pythia directly?
WAI KIN: The jet function is just analytic stuff. But we demonstrated asymmetry at low pT for jet production. With different jet recombination schemes. This is purely a Pythia demonstration with the latest package. Which incorporates the spin interaction. And we do a jet production analysis.
FELIX: I see. Very interesting. Thank you.
ALBA: Thank you. Anything else from the audience? I wanted to ask you whether the energy of the collider plays a role or not. And I'm asking this because of the small x... This interpretation between the... (inaudible) factorization also. I don't know if you could comment a bit on whether this is (inaudible) QCD small x effects could be easily incorporated into a calculation of this observable. Or correlated to that.
WAI KIN: Excuse me. I can't hear. Can you say again?
ALBA: Probably. Yeah. I was asking about the energy of the collider. So do you have any constraints on the energy of the collisions? Or do you just want to...
WAI KIN: Oh, okay. I think that there is no constraint on the energy of collision. The thing is... There are several things we can look at. The low pT spectrum of the jet, or high pT dijet, or even the high pT jet with the different axis definition. So the collision energy... As long as the collision energy is perturbative, then it's fine. So we have a hard collision. Introducing a jet which is highly energetic.
ALBA: Okay. Thank you. I see no further questions. I would suggest that we move on to the next talk. Thank you very much.
>> Recording stopped. Recording in progress.
MRIGANKA: Shall I start?
ALBA: Yeah, please go ahead.
MRIGANKA: Okay. So I'll be briefly summarizing the work on this momentum and charge correlations within jets. The work has been done in collaboration with Chien, Deshpande, and Sterman. So we defined here a new charge-energy correlation and showed how its flavor dependence... We used recursive soft drop structure to show how these correlations behaved there. Hadronization is very well studied. But we are revisiting it in a unique way. We are seeing the two leading particles, and they're very special, in the sense they are leading and next to leading particles.
So they are selecting a non-perturbative region of interest. So we're choosing the leading and next to leading particle with rc. NCC is the number of cases where they are of same charges and NCC bar is of different charges. So the simple definition of rc. And that carries important information in string breaking picture. So if we think of a partonic final state of u and u bar which combines with neutral pairs d bar and d, they always form π+ and π-, opposite final charges, and rc=-1 in this equation. In random picture, rc=0. So it's a measure of string-like hadronization. We're using event generator Pythia and Herwig in these studies. They have different models of hadronization. So here we have plotting -- these observables with formation time. So formation time buildup with these information of leading and next to leading particle kinematics, so Kperp is relative transverse momentum. Depending on the region of where you choose for small Kperp, generally you're in non-perturbative dynamics, and that's large formation time. For very small formation time, these two particles decorrelated very early. So that's perturbative place.
So there is a transition region, and we see from Pythia there is a very strong flavor dependence. And in Herwig, they almost follow the same plane, but in spatial kinematic region, they are different. So specific case would be very interesting for a study with strange flavor tracking violating particle is π-, and next to leading particle is K+-, and case 2 is π+ and π+-. Let's consider a struck valance quark. To form π-and k+, there should be u bar and u and simple string can explain production. But in other case, π- in final state is u bar and u bar, which cannot explain production. So this case is favorable, and that leads to π-K+- to be much stronger for rc compared to the π+ case. So that's what we're seeing in simulation. Rate one is the favored case.
They're stronger than the other case. And in Herwig, it's very different. Okay. And such things can be measured very precisely at EIC with a small amount of the dataset, and most things are PID. That's the main advantage we have with the current data analyzing from H1 and STAR we don't have. LHC, LEP, and LC would be a good place for such measurements. So now we move to a place where we're seeing this observable with prong structure. So here we illustrate on top the case where in a first plate, we see this leading and next to leading particles are in the two separate prongs. So in case we say does n=1 is a resolved case.
And here the figure below, n=2, is resolved case. Leading and next to leading particles. It's in two separate prongs. So we're using recursive soft drop techniques for study, according to the hardest branch and using dynamic radius. So we are starting -- representing the partonic proxy. So we see clearly here n=1 is soft angle wide radiation. And for n=2 and higher, we see it's a relatively narrower and harder radiation.
Specific case... Sorry. N=1 and n=2 case, we see this red one and this green one and leading and next to leading particles is black one. So with formation time, we see different prong asymmetry, follow different -- behave differently. And need to be understood from data and from theory inputs.
So I move to summary. So the hadronization can be studied in a specific way at EIC, also from these cases, and we define a new observable. And saw significant differences in flavor dependence. And specific flavor tagged case can give significant knowledge on string fragmentation inspired models. So we need good particle identification for such studies. And we are understanding rc with C/A declustering tree. That's an alternative way of studying hadronizations. And Pythia shows distinct features of rc with formation time. So these need to be understood from data and theory. Thank you.
ALBA: Thank you very much. Let's go through the questions in the Google Doc. The first one is: How do neutral particles affect the interpretation of this observable? I guess in a sense it follows...
MRIGANKA: Yeah, it's a very interesting question. We started first also looking at the neutral combinations. But definitely from the perspective of measurements, we were mostly focused on charged correlations for the time being.
ALBA: So this person was arguing that it seems simplistic to ignore neutral particles in hadron formation. The effect of neutral particles expected to be overall scaling from isospin or (inaudible) flavor?
MRIGANKA: Generally, SU3 are for massless particles like UDS. And strong coupling doesn't distinguish flavor. Binding some u d and a system in strong interaction cannot distinguish certain flavor. So it might not be explained in that way.
ALBA: If the person is in the audience and wants to continue, feel free. If not, there is a second question by Simone that asks: Pythia and Herwig have different hadronization models but also different in parton showers. It would be interesting to repeat your study with Sherpa, for example, which would allow you to switch to different hadronization models while keeping the same parton shower.
MRIGANKA: Interesting. These two models have different shower, parton showers... So in that sense, it's not very appropriate to the way if we pinpoint hadronization using Pythia and Herwig... Sherpa is a good solution. Yeah.
ALBA: Any other questions from the audience? If not I ask in the mean time... Do you have any analytical insight on the calculation of rc from the recursive soft drop technique? Did you start looking how to compute it?
MRIGANKA: In recursive soft drop, when you have very wide soft radiation, gluon radiation... This correlation -- there shouldn't be much correlations between the leading and next to leading particles. Because they are independent. But in narrower angle, this correlation should be more stronger. I believe.
ALBA: What I meant is if you can compute this rc observable analytically.
MRIGANKA: I think my theoretical colleagues can put more light there. If (inaudible) is there, maybe he can say a bit more?
ALBA: Sure. If not, no worries.
YANG-TING: I'm around. So... Yeah. It's a very difficult question. We all know that the hadronization is non-perturbative. And a non-systematic way of calculating it. Except that method... For this work, on the theoretical object we identified to be associated with all this is the hadron fragmentation function and the perturbative properties of the objects is under study right now. And will be associated with the simulation studies in our various publications. At the moment, I just wanted to say that the properties of this non-perturbative objects will be presented in the upcoming application.
ALBA: Of course this was what I was asking. The partonic... The perturbative level, whether you can compute something. And of course... Yeah. You cannot do anything with it. And actually had another question. How do you define leading or next to leading based on pT or hadron or based on what?
MRIGANKA: It's momentum of the particles. Momentum-based.
ALBA: Transverse momentum?
MRIGANKA: No, the total momentum.
ALBA: Okay. Anybody else?
GREGORY: Just along the line of the calculability... I was wondering if you could... You could define a jet charge for jets, and would that help? You could build subjets, get the two leading subjets in that case, in pT, and compute the jet charge for both of them? Would that... Do you think that would help? In terms of trying to get an analytic handle on this?
YANG-TING: I guess that question is for me. Definitely. Promoting all these observables to subjets is the way to make things... Controls. And then one can study next to leading subjet correlations. We're going in that direction. Exactly, exactly. And that is a very interesting suggestion. In our simulation studies that we have shown here... We look into subjets, charged properties. But we don't have results to show here in this talk yet.
ALBA: Another question by Petar.
PETAR: Yes. Sort of from an experimentalist's standpoint, speaking of jet charge, I mean, I just want to sort of mention that in the B physics, in the 1990s and early 2000s, and even today, there have been various flavor tagging techniques that sort of relied on similar things. Same sign tagging and tagging from CDF and jet charge tagging from... Using the... SLD and CDF and LEP, and I think SLD also had the charged dipole tagging or something like this. And all of these techniques actually worked. In some sense, we know for a fact that you can actually see B0, B subs oscillations. We know that this model of string... String fragmentation works, and the charges of tracks are actually set in order, as we think they are. Now, these are done in the context of the b quarks, hadronizing. So maybe this is slightly different. But I just want to mention that there were actually experimental results which kind of... For a long time... Have pointed out that this picture is correct. So I think this is actually quite... Would be quite interesting to revisit these studies and study this... You know, make a whole program of sort of studying these things in slightly more detailed way. I think most of the studies that we have been doing with the substructure do not really care about charges or the relative charges of particles, and whether... So I think that's actually kind of an interesting twist here. And I think I want to... Thank the speakers for sort of bringing this up.
Because this is also quite... Points to a new direction of what can be measured.
MRIGANKA: Thanks. For feature... Yeah. EIC... Some places, we have limited particle identifications. For example, we have a lot of data from H1, and also from RHIC. But we have limited particle identification. So that puts some limits in our current measurements here.
ALBA: We are perfect on time. And I don't know who says goodbye in this session. If it should be me or... The organizers. But thanks to all the speakers in any case. For the talks.
MATTHEW: Thank you very much, Alba, for chairing the session and to our other session chairs and speakers today. I don't think we have anything again to say at the end of the day. We'll reconvene tomorrow for the last day of BOOST at the same time. 1500 CEST, or convert to your local time zone. Okay. Talk to you then if not sooner on the Gather.Town. Bye!