WEBVTT 00:25:31.000 --> 00:25:35.000 Let's have a try and do that or the early LHC data. 00:25:35.000 --> 00:25:48.000 But it wasn't seeing the because of the Higgs boson was predominately decaying into stuff, like some hidden sector states that decay into dark photons and missing momentum and so you've got these list of lifetime Jackson, you can be hiding them and the 00:25:48.000 --> 00:25:53.000 fact that you have these displays calling me that safe that people aren't specifically looking for. 00:25:53.000 --> 00:26:00.000 And while they're worth, you know, there was a description of UT models, some of these scenarios. 00:26:00.000 --> 00:26:14.000 The challenge here is that it becomes such a wide range of parameters that we can be looking for it signatures we can be looking for that can be a little bit hard to organize signatures compared to say around supersymmetry. 00:26:14.000 --> 00:26:28.000 I won't go into a ton of detail because I've looked at there's a lot of toxins workshop that are going to be touching on all of these different scenarios so I won't belabor the point here but I would say that a major turning point seems to be that as 00:26:28.000 --> 00:26:43.000 people moved away from some of the more canonical uniform a car parody conserving Suzy model. It was realized that actually you get along with particles and all sorts of bsm physics and that the one with particles are popping out, and asked for and that 00:26:43.000 --> 00:26:55.000 was really my own entree to the field and it seemed like no matter which problem I was working on whether it was neutrino ambassadors dark matter. Very agenda says, you just get logged with particles that are coming out of your theory. 00:26:55.000 --> 00:27:06.000 And so as we now know you know there are a lot of models that exhibit hidden sector confinement with dark red zones and blue balls like the drinking scenario and the neutral natural most ideas of error sectors. 00:27:06.000 --> 00:27:20.000 qc the axons where you play certain games to try and allow it to be heavier can show up as long with particles of course supersymmetry as I've mentioned but Dark Matter models inelastic Dark Matter reason Dark Matter co scattering dark matter. 00:27:20.000 --> 00:27:39.000 Heavy a neutral leptons so new MSM and various implementations of the seesaw mechanism, dark Higgs boson dark protons models of barrier Genesis where you again can motivate short, long lifetimes from cosmological consideration that accent like particles. 00:27:39.000 --> 00:27:52.000 And so I would say that in terms of the theory community if you look at head ch probably a substantial fraction of papers have long with particles and it's not because there's we pause it's because that's a genuine leak, with the UV dynamics is telling 00:27:52.000 --> 00:27:57.000 us. And furthermore, those loopholes are closing. 00:27:57.000 --> 00:28:04.000 And so, there have been so many really spectacular searches and developments that. 00:28:04.000 --> 00:28:19.000 For instance, have no results have shown that okay well we can't always have hide. And so for instance, you know, a few years ago there were these wonderful plots that were made for instance this is like Apple by Atlas with there are parody can RPC RPC 00:28:19.000 --> 00:28:32.000 RP document which shows glue, you know, constraints over the entire range of lifetimes and as you can see, ranging from very lucky very short, there is no more. 00:28:32.000 --> 00:28:45.000 And at that time, you had to rewrite had to live up to TV and so if you have theories that necessitates strong dynamics somewhere like Reno's or stops, then those loopholes are closing. 00:28:45.000 --> 00:29:00.000 However, there are still some gaps and lifetime searches so I'm showing here paper from some bear Genesis work that I've been doing, where you get particles that look a lot like stuff tons, and even here, we kind of had to be a little bit vague about 00:29:00.000 --> 00:29:17.000 where in the lifetime playing the prompt searches, give out Oh thank you, because there's not necessarily the same type of information that's given so maybe there's still a loophole here but there certainly don't exist the same ones that we had before. 00:29:17.000 --> 00:29:21.000 So I'll then move on to model coverage and reinterpretation. 00:29:21.000 --> 00:29:36.000 So I'm showing up here a plot taken from an atlas analysis in 2011, which shows a presentation of an early supersymmetry results and what some of the younger people might never have seen before, which is a plot and super constraint claim. 00:29:36.000 --> 00:29:47.000 And one of the reasons why I'm showing this is that even as late as this point that there was no simplified model interpretation, this was the only result that was shown in this paper. 00:29:47.000 --> 00:30:02.000 And that gives us a sense that in the grand scheme of things, how recently there was this pivot towards simplified models. And in the same year this wonderful and very extensive document came out with many contributions from members of the community. 00:30:02.000 --> 00:30:16.000 but I gotten to a point that was made if you look at the range of pages devoted to exotic, there's like one to two pages, you'll see that a single page covers things like a weird jets. 00:30:16.000 --> 00:30:28.000 Okay. And, you know, look at one of the paragraphs that won't read it out but just leaves the vertices from a resonance. And essentially, the entirety of displays vertex searches in terms of simplified models is contained within this paragraph. 00:30:28.000 --> 00:30:39.000 This is not a knock on these individuals it goes to show that, because there was so much work to be done, along with particles were kind of shunted to a later problem that needed to be tackled. 00:30:39.000 --> 00:30:53.000 And it was very unclear whether you could take results of along with particle search safer Model A and apply them to Model B, so an early example of interest was whether searches for displays digests that reconstructed a single particle could apply to 00:30:53.000 --> 00:31:06.000 for instance, along with particle became three jets and you know again to illustrate this a paper from 2013 you have things where you say, you know, and unfortunately, other important factors that obviously get the relative efficiency of this search, 00:31:06.000 --> 00:31:19.000 and this again is not a criticism, it's just that was the feeling at the time the community the theorists don't have the tools to be able to tell us whether these searches apply or not. 00:31:19.000 --> 00:31:31.000 I think it was pretty significant as theorists and again it was kind of a convergent evolution thing where there were three or four groups that all kind of did this at the same time, where we were like, what did we actually try and simulate this and take 00:31:31.000 --> 00:31:46.000 into account race detector effects can we got it to work. And we found that for certain examples the experiments had given enough information so for instance, in that exact case work I did with Dr Joshi, and around the same time john Muir, and Brock 3d 00:31:46.000 --> 00:32:06.000 had similar investigations of whether you know, read particle decay is going to be constrained by the direct searches and we showed that they definitely could and we even got similar results with our independent searches, but the information can be really 00:32:06.000 --> 00:32:15.000 For the first time, or at least in the first implementation the LDC and fears that necessarily know what we needed and so there was kind of a back and forth about this. 00:32:15.000 --> 00:32:29.000 Now, you know, when I look at the databases for longer particle searches and Atlas and CMS more of them than not have data, links, and these have data links can include very granular information at the function of the face space you know the mass of the 00:32:29.000 --> 00:32:35.000 truth vertex the number of tracks that caters to the angular information so on. 00:32:35.000 --> 00:32:51.000 And one of the major out sections of the galaxy along with particles white paper the James mentioned contain suggestions and examples of real world use of, you know how, when theorists try to do this mapping water issues that they ran into and that there's 00:32:51.000 --> 00:32:59.000 there's a real back and forth between the theory and experiment continues to do this, we're even seeing that passages like Delphia is. 00:32:59.000 --> 00:33:12.000 And actually don't know how this is pronounced it's modeled as models that you know the packages use the simplified models to constrain you the physics are including long with particles. 00:33:12.000 --> 00:33:19.000 I think we have to acknowledge the because long with particles necessarily live at the frontier of experimental information of what the detectors can do. 00:33:19.000 --> 00:33:30.000 We're always going to be a community where we're going to need to be trying things out and seeing whether it's the earth can accurately do it and having a robust conversation. 00:33:30.000 --> 00:33:41.000 Third, there's been this explosion in terms of dedicated LLP experiments and this has been said tongue in cheek that I've heard that you know nowadays every theorists have experiment and I think that this is true. 00:33:41.000 --> 00:33:53.000 This is led in part because in the 2000s there was sort of this renaissance of theorists proposing and collaborate with experiments proposing and even being like spokespeople of small scale experiments for hidden sectors. 00:33:53.000 --> 00:34:09.000 And then, you know, The first dedicated long with a particle detector at the LSC metal already began in 2010. And the last decade or so there's been this kind of blossoming of new ideas, many of which were either originated by theorists or done it very 00:34:09.000 --> 00:34:24.000 close collaboration proposing looking for, along with particles using dedicated detectors and I think that this is illustrated, it just how much theorists have become, you know I'm always an integral to the efforts, but now on the nitty gritty level of 00:34:24.000 --> 00:34:39.000 of thinking through backgrounds and you know what's feasible for entirely new experiments. And we're already seeing success in these dedicated the factors right sensibility and had a demonstrator run, which put new constraints and leading constraints 00:34:39.000 --> 00:34:53.000 on Mila charged particles and some parts of the face space taser which has been funded and, and was running in a test mode was able to detect the first tentative neutrino scattering events. 00:34:53.000 --> 00:34:56.000 So there's new physics being done already with these spirits. 00:34:56.000 --> 00:35:05.000 experiments. Finally, challenging signatures where were we then in now. 00:35:05.000 --> 00:35:22.000 And this list is a crude summary of where a lot of the gaps existed, that again being slightly vague but ironically decaying longer particles, especially at low mass and shorter lifetime, saying we produce long with particles for searches that relied 00:35:22.000 --> 00:35:24.000 on Paris. 00:35:24.000 --> 00:35:33.000 Even semi electronic, or in my particular case could you have challenging time passing the restrictions if for instance you have low masses like below 20 or 30 GV. 00:35:33.000 --> 00:35:44.000 Slightly displace leptons are displaced leptons not originating from just lights vertices this White House photon, I multiple cities combining hidden factors and Corky signature. 00:35:44.000 --> 00:35:58.000 and actually went back through some notes from the KTB workshop Andy Hoffman I work convener is up along with particles group and we were supposed to write a white paper which we didn't because it was too big of a task to kind of combining with these 00:35:58.000 --> 00:36:11.000 larger efforts. But when I looked at the Google Doc from our notes of the discussion the gaps that were identified in this 2015 workshop and that had been particularly the four years by many people before are very similar to the gaps in the 2019 LHC along 00:36:11.000 --> 00:36:16.000 particles white paper, which indicates the challenge of covering many of these signatures. 00:36:16.000 --> 00:36:32.000 I'll just show a few recent developments where, for instance, searches have been pushing to lower masses in terms of Medtronic the case of along with particles and pigs. 00:36:32.000 --> 00:36:45.000 that allow you to go to new phase, phase in terms of lifetime and coupling. and this is complemented for instance by searching FCP where the trigger is less of an issue but you have smaller acceptance. 00:36:45.000 --> 00:36:57.000 Electronic decay, there's been progress on both CMS and Atlas towards doing powerful new searches. And there was an earlier searches on CMS that we're doing this as well where you're looking at leptons, they're not forming a common vertex. 00:36:57.000 --> 00:37:05.000 But now, more flavor combinations are produced which allows us to put strong constraint that slept slept time like signatures. 00:37:05.000 --> 00:37:21.000 There's also in the case of electronic decay is moving to lower and lower masses using things like data scouting that allows to go into a space that would never allow an event to be fully triggered photon decades using similar strategies like zero plus 00:37:21.000 --> 00:37:36.000 to LPs the decade of photons was bad, and similar this allows us to go to lower masses of the particles to fade the photons like 30 GV that those who could never trigger on those photons directly we can trigger on z, and heavy neutral leptons so low mass 00:37:36.000 --> 00:37:41.000 displays vertices to came to electronic plus. 00:37:41.000 --> 00:37:58.000 Not even energetic enough to be jets but had runs and you can get sensitivity to having trouble icons as low as three GB and remaining of course of extreme importance, but dark showers and this whole kind of new frontier of high multiplicity along with 00:37:58.000 --> 00:38:08.000 particles and CMS had first search for emerging jets that came out a few years ago, looking at heavy state so that the trigger wasn't necessarily a problem. 00:38:08.000 --> 00:38:17.000 But, showing that you could put constraints using the internal structure and looking for high numbers of high impact parameter trap. 00:38:17.000 --> 00:38:31.000 And there's been a significant amount of work recently so I want to draw your attention to this Snowmass report, which gives a really extensive progress and its direction that was started already have a long way particles white paper that provides benchmarks 00:38:31.000 --> 00:38:44.000 and on logical models limitations of the current approaches that really we're now at the point where these things can start being more directly implemented and facing the hurdles to really expand searches here. 00:38:44.000 --> 00:38:51.000 So, you know, to be very crude about classifying the things I'd say that in this top category. 00:38:51.000 --> 00:39:03.000 I wouldn't say that there gaping loopholes and gaps and that really there's always more face states you can explore but there's some really powerful searches that are getting into the gaps for these ones down here there's still areas where there could 00:39:03.000 --> 00:39:19.000 be really wide open scenarios and in many of these cases triggers remain a challenge but there are lots of ideas for so for instance, you know, of course, along with particles white paper but also this interesting document from the community on using 00:39:19.000 --> 00:39:22.000 triggers for triggered on monolith particles. 00:39:22.000 --> 00:39:36.000 And finally I won't say much about this but first we're doing this to discover something or we should always expect the unexpected and 3.3 sigma global significance is not a discovery, but it is very cool to see along with particle search with the dots 00:39:36.000 --> 00:39:44.000 light up along with signal better than the background. And so we should be prepared for finding all sorts of really exciting things. 00:39:44.000 --> 00:39:56.000 So to summarize, we've really accomplished a lot and grown and changed over the past six to 10 years but that's really a continuation of an evolution that's gone back decades, but some things have never changed which is that LLP searches are challenging 00:39:56.000 --> 00:40:05.000 exciting. There are creative and bold ideas, and they've inspired the development of a collaborative technical community that it's been really meaningful to be a part of. 00:40:05.000 --> 00:40:11.000 So thank you and I'm happy to take any question. sorry for going a minute or two over. 00:40:11.000 --> 00:40:18.000 Thank you so much for this really nice overview of where we come come from and kind of where we're headed as well. 00:40:18.000 --> 00:40:31.000 So yeah, this is a very nice talk we have time for a few questions as well. I can see that we have a question, hand raised from Richard, so please go ahead, Richard. 00:40:31.000 --> 00:40:33.000 Hi, can you hear me. 00:40:33.000 --> 00:40:35.000 Okay, great. Hi Brian It's been a while. 00:40:35.000 --> 00:40:49.000 Great. Hi Brian It's been a while. I enjoyed this talk is nice to see kind of this big survey this big history of it that it particularly I want to focus on this. 00:40:49.000 --> 00:41:01.000 The last couple slides where you're talking about what are the current or hurdles for the LP community, and particularly the these little red points that you had maybe this is on slide. 00:41:01.000 --> 00:41:21.000 26, I believe. Yeah. So for these red points can you say specifically if these are analyses on the theory site that you that you think need to be done, or on the experimental side or motivation your first, for example, specifically displace cows are you 00:41:21.000 --> 00:41:35.000 hoping that what is the gap per se is this in the theory literature is is experimental churches or motivated models, can you say a bit on the the different breakpoints fish. 00:41:35.000 --> 00:41:42.000 Sure, that's a great question and I guess I would say, both in the sense that, um, 00:41:42.000 --> 00:41:55.000 you know there's motivation for experimental us to look for things where there is broad community interest, and I would say that, you know, there's broad interest in high multiplicity is but that's a really really hard thing to do both in terms of capturing 00:41:55.000 --> 00:42:09.000 theory and experiment and so I think that really explains, kind of, the slower progress there but of course there is progress that's happening and so this is more I think just that. 00:42:09.000 --> 00:42:22.000 actually have to be realized and made public and during developments Megan but their ideas about what needs to be done. Let's say for instance, in the case of towels, you know if the towels are energetic enough them they educate electrons and you can 00:42:22.000 --> 00:42:37.000 so for instance of these searches here, you can rely on electronic decades of towers, and because you're looking at high masses, that's fine. That starts becoming less fine when the, the massive along with particles in the 10s of gv say where the leptons 00:42:37.000 --> 00:42:50.000 become too soft to pass the trigger and acceptance. And you can potentially do better with hedonic towers but there's really not much that's done I don't think really in this theory literature there's probably a few papers, but that's something that I 00:42:50.000 --> 00:43:00.000 think could be explored and fleshed out more similar in the photons, there are some papers that have looked at for instance in elastic dark matter or photon jets are things like this. 00:43:00.000 --> 00:43:11.000 I think that this here really experimentally it's a very challenging thing to do because the thresholds are high, and if you don't have a lot of net or, for instance, in the case of the Higgs production, you can trigger on a z but if you don't have that, 00:43:11.000 --> 00:43:22.000 then it becomes challenging to do so I think we're seeing progress or do they can purchase it it's very very difficult to do. So I think they are just more work in general on the theory of you to be flushed out before we can really expect or experimental 00:43:22.000 --> 00:43:25.000 as colleagues to do that. 00:43:25.000 --> 00:43:32.000 But having a chat, maybe during the break about that to bring the rest of the workshop that answered my question. 00:43:32.000 --> 00:43:36.000 I might right you on the side but thank you. 00:43:36.000 --> 00:43:49.000 Really nice we can always continue with so unless most discussion, we have time. We have time for one more question I think before we move and those that hundreds from products, so please go ahead. 00:43:49.000 --> 00:44:01.000 Hi, Brian. Very nice overview. So, I had one question so how we have closed the gap for low mass had run into case 00:44:01.000 --> 00:44:02.000 closed the gap. 00:44:02.000 --> 00:44:18.000 But, you know, I think we need to distinguish between a gap where literally anything could be there and we just have no constraints and hedonic decay is where it's like, okay, we're now probing at the percent level for instance of Higgs to case that had 00:44:18.000 --> 00:44:28.000 drawn longer electrons which I didn't show because it's maybe that that search strategies a little bit more established. They can go to sub percent, and now sub point 1%. 00:44:28.000 --> 00:44:40.000 If the decay is further out in the detector because of being able to trigger on decay is in the H collar we want spectrometer. So here I would say like, yes they're going to be a lot of theories that are living down here, but it's not like we haven't 00:44:40.000 --> 00:44:56.000 looked at that there's intrinsic challenges the triggering maybe we can do somewhat better but it's, it's not like, oh, there's this whole unexplored face to face right something like non pointing or delayed photons without met I think that in many cases 00:44:56.000 --> 00:45:05.000 there's just not anything. And so then you could have something that even is like, you know, 30% branching branches or whatever the maximum is consistent with limits that could be there. 00:45:05.000 --> 00:45:14.000 So that drove that a little bit but of course this is really really well motivated and so this isn't just say we shouldn't be devoting a lot of effort to that. 00:45:14.000 --> 00:45:19.000 Okay. Thank you. 00:45:19.000 --> 00:45:38.000 Okay. Okay, well thanks again. 00:45:38.000 --> 00:45:41.000 Or at least I think I'm handing over to monitor. 00:45:41.000 --> 00:45:45.000 I can see your screen but I can't see any slides just share nice background. 00:45:45.000 --> 00:45:49.000 Oh, okay. I was waiting for you to if you wanted to say something in the trenches. 00:45:49.000 --> 00:45:51.000 Hi. 00:45:51.000 --> 00:45:58.000 Yeah. Laura jaunty is going to be giving a talk as well on the landscape. 00:45:58.000 --> 00:46:04.000 But from an experimental perspective. So, yes, do you want to go there you go. 00:46:04.000 --> 00:46:06.000 If you only get started. 00:46:06.000 --> 00:46:11.000 Hi, Good morning, good afternoon, and potentially Good evening. 00:46:11.000 --> 00:46:19.000 So I'm carrying on from that very nice talk that Brian gave and I'll be looking more at the experimental perspective. 00:46:19.000 --> 00:46:25.000 So we tried hard to not cover the same stuff so I think we've mostly succeeded in that. 00:46:25.000 --> 00:46:35.000 So the last six years, as we were targeted with from the experimental perspective happens to overlap very well with the analysis that comes out from run to. 00:46:35.000 --> 00:46:54.000 So my talk so somewhat focuses on what we've learned and innovated from run to, I will be mostly covering Atlas and CMS, and I won't be covering the dedicated detectors, as these have been covered by Brian and will be covered extensively dedicated session 00:46:54.000 --> 00:46:59.000 in this workshop, 00:46:59.000 --> 00:47:02.000 my slides there we go. 00:47:02.000 --> 00:47:21.000 So first, starting at someone an obvious place but it's worth stating. So, run two from experimental side has seen an extensive exploration of LPs LP face space so we have covered a large amount of unknown territory and that shapes, very much. 00:47:21.000 --> 00:47:27.000 The territory that we look forward to investigating in the future. 00:47:27.000 --> 00:47:35.000 So I'm going to start with a brief look at the experimental phase space that we have covered in to benchmark models. 00:47:35.000 --> 00:47:44.000 And I'm going to talk about the experimental challenges and innovations that we've developed in exploring that face space in these models and other models. 00:47:44.000 --> 00:48:01.000 So the two benchmarks that will be looking at is, first will be Higgs portal which is a very important portal that we recognize the importance after run one where the Higgs the case too long lived scalar which indicates fermion, and also be looking at 00:48:01.000 --> 00:48:13.000 a much heavier example which is long with cleaners for both of these, we gained his sensitivity and run to partly due to the increase energy data set that he gave us. 00:48:13.000 --> 00:48:28.000 But I want to take a moment just to recognize that that sensitivity gain is not for free, even if it's given to us in some sense by the accelerator. It's a huge amount of work to analyze the larger data set, it's not just returning a crank. 00:48:28.000 --> 00:48:37.000 There's a huge amount of surprises that come up in the data there's a huge amount of complications that come with just the data flow of a larger data set. 00:48:37.000 --> 00:48:45.000 And scientists we also can't help ourselves from innovating so when we redo and analysis we actually tend to improve things. 00:48:45.000 --> 00:48:52.000 So, even just when we say just again due to energy data set is actually a huge amount of work from the experimental side. 00:48:52.000 --> 00:49:06.000 Nonetheless, by there's been a lot of innovation in op searches. And that has enabled us to push in many directions so we've been pushing toward both shorter and longer lifetime is each of those has their respective challenges. 00:49:06.000 --> 00:49:16.000 We've been expanding our sensitivity to lighter lbs, which is also challenging for reasons that Brian mentioned due to trigger for example and back. 00:49:16.000 --> 00:49:21.000 We've been pushing deeper into cross section and branching ratios. 00:49:21.000 --> 00:49:35.000 And we've been trying to close gaps in coverage with a variety of approaches including new techniques, new signatures and reinterpretations of other searches that have sensitivity. 00:49:35.000 --> 00:49:40.000 And finally, we've been reaching sensitivity to significantly higher masses. 00:49:40.000 --> 00:49:55.000 So to start with the the Higgs LP benchmark. So this is the picture in 2015, roughly, at the time of the first LP community workshop, and what the data told us from run one. 00:49:55.000 --> 00:50:00.000 So you can see here at our most sensitive point in boat. See talent one meter. 00:50:00.000 --> 00:50:12.000 We were reaching to about the 1% branching ratio and cakes. And if we jumped to 2022, to where we are now. Note that the axes on the summary plots tend to change. 00:50:12.000 --> 00:50:20.000 We've actually reached the personal level so we've had a huge increase here in the scale and with that the the sensitivity. 00:50:20.000 --> 00:50:29.000 But I want to point out a few other things that we've improved and the summary plot. So we've been pushing toward longer and shorter lifetimes. 00:50:29.000 --> 00:50:42.000 Here, this is required to get sensitivity at the lower lifetimes required a dedicated machine learning technique to be able to discriminate against the center model background that's there. 00:50:42.000 --> 00:50:50.000 And then at the high lifetime this point comes from a reinterpretation of the Higgs to invisible search. 00:50:50.000 --> 00:51:05.000 We've also been extending our sensitivity to lighter LPs so that requires both optimizing sensitivity selection rather for those later LPs but also new techniques and one of those new techniques that allows us to really cover a wide variety of mass and 00:51:05.000 --> 00:51:19.000 lifetime ranges is by using different numbers of LP objects in our selection and trying to optimize the complementary coverage of one versus two versus multiple objects. 00:51:19.000 --> 00:51:32.000 And then finally, just to point out explicitly this impressive game in the depth of our coverage to look at a much heavier benchmark. So if we look at along with Gleaners in 2015. 00:51:32.000 --> 00:51:36.000 We had sensitivity at the shorter lifetimes. 00:51:36.000 --> 00:51:53.000 From a prompt search but then there was a gap from the displays for a text search we were up to about 1.5 TV, and then kind of intermediate coverage at the, at the higher lifetimes from from a number of searches looking for the direct detection of the 00:51:53.000 --> 00:51:57.000 the long with Gleaner, if we jump to 2022. 00:51:57.000 --> 00:52:12.000 You can see again that the scale of this summary plot has been increased by a TV which roughly covers the sensitivity that we've gained in these six or seven years which is you know quite a significant jump. 00:52:12.000 --> 00:52:19.000 I'm sure the theorists have lots to tell us but what that corresponds in terms of naturalists arguments. 00:52:19.000 --> 00:52:37.000 And so look at this a little bit more closely as well so we are closing gaps here, not only in jumping to higher masses but also you'll note here that the prompt to long life scenario is now fully, 00:52:37.000 --> 00:52:52.000 fully covered in terms of our sensitivity there. And we've also interpreted our longer signatures for the full lifetime range and so we're able to say more clearly what are coverages in the range here. 00:52:52.000 --> 00:53:05.000 And then just again to point out the the enormous gains we've made overall in this entire pot, expanding our sensitivity, up to almost two and a half TV. 00:53:05.000 --> 00:53:11.000 So in the rest of the talk, I'm going to be exploring some of the experimental innovations that have. 00:53:11.000 --> 00:53:25.000 We've developed in the last six years. I won't be covering explicitly existing gaps in our coverage as those will be covered by other talks in the workshop, or as I said by dedicated detectors. 00:53:25.000 --> 00:53:40.000 But I'm going to focus on new techniques we've developed new analyses, new challenges, new solutions. And this is not only to show off all the work that we've been doing as a community in the last six years or so, but also because these innovations are 00:53:40.000 --> 00:53:41.000 examples. 00:53:41.000 --> 00:53:55.000 Often from one or two analyses, and they can inspire progress and other analyses and this will will expect to see a lot more of these innovations, moving forward with the further analysis over into and within three. 00:53:55.000 --> 00:54:01.000 So I want to start with one of the most important 00:54:01.000 --> 00:54:16.000 considerations and LP searches which is the detector handles we use so we often use very unusual detector handles and many of these were developed and studied and, you know, in some sense, optimize and run one, but we've continued exploration and run 00:54:16.000 --> 00:54:28.000 to and I expect that we'll continue to see innovation moving in German three. So I selected, just two examples of new uses of the detector itself in run to. 00:54:28.000 --> 00:54:39.000 And so the first is from CMS, a very nice recent analysis in which they're actually using the forward and cap, as a sampling color emitter. 00:54:39.000 --> 00:54:51.000 So this has two advantages relative or give it complimentary coverage to the more traditional displays for tech searches, which look for signal similar signals here. 00:54:51.000 --> 00:54:57.000 So the first is that, because you're looking in the end cap and there's more shielding. 00:54:57.000 --> 00:55:12.000 This actually reduces a lot of the background and so if you're looking for. Neutral LPs decay, you actually can can relax the requirement of to displace vertices and only required one, which has corresponding acceptance gains. 00:55:12.000 --> 00:55:24.000 And also because you're using the moon detector as a sampling color meter, you're actually sensitive to the LP energy, rather than its mass and this makes the analysis relatively insensitive to the particle mass, where when you're looking at a displays 00:55:24.000 --> 00:55:34.000 for text analysis, your sensitivity tends to decrease with smaller mass because the opening angle decreases. 00:55:34.000 --> 00:55:51.000 When you're using it, the energy that that constraint doesn't apply and so here you can see very nicely that actually this analysis has sensitivity similar sensitivity, in terms of branching ratio reach for the different masses down to seven gv so this 00:55:51.000 --> 00:55:58.000 is a really nice innovation in using the detector, an innovative way. 00:55:58.000 --> 00:56:09.000 Another example from CMS is an analysis which, as far as I know is the first to use jet timing in the Ico explicitly as an analysis handle. 00:56:09.000 --> 00:56:19.000 So displays photon analyses, have used decal from both Atlas and CMS, but looking for hedonic decays This is the first emulation community. 00:56:19.000 --> 00:56:34.000 And this is possible due to the very precise timing resolution of the decal, which allows. As you can see in the bottom left plot here to actually separate very cleanly signal from from background, based on a tiny measurement. 00:56:34.000 --> 00:56:55.000 This allows the analysis to reduce background to only a few events, but retain high signal acceptance. And so this is again a nice compliment to the existing displays vertex techniques, and I can imagine and anticipate they will continue to see innovation 00:56:55.000 --> 00:56:56.000 in the timing front. 00:56:56.000 --> 00:57:02.000 Using detector timing in innovative ways, as in order entry. 00:57:02.000 --> 00:57:12.000 So generally we tend to think, especially my thinking of displaced tracks we tend to think about displace tracks and displace for disease as sort of synonymous. 00:57:12.000 --> 00:57:29.000 But run two has seen a pioneering use of displaced tracks without a vertex. So Brian mentioned this, but it was such a glaring death and it's so satisfying that we filled it but I will mention it as well, which is the display slept on analyses from both 00:57:29.000 --> 00:57:45.000 Atlas and CMS. So here you have to displace leptons but they don't come from a common vertex. And so, The background to displace track would be enormous to a single displays track but if you require a lot tonight ID for those displays objects you can 00:57:45.000 --> 00:57:57.000 actually get away without the advantages of a vertex which includes a mass reconstruction and or just requiring to displays objects which kills your background. 00:57:57.000 --> 00:58:10.000 So this, this feels an important death in signature space at the eulogy that was there for a long time, and it allows us to have significant sensitivity, for example to long lives loved ones. 00:58:10.000 --> 00:58:19.000 That was completely missing before. 00:58:19.000 --> 00:58:31.000 So, the number of LLP objects that we select in each event is is non trivial, it's not just like ordering off of a menu, and we often start with the low hanging fruit. 00:58:31.000 --> 00:58:47.000 Now the low hanging fruit differs depending on analysis that you're working with or the object. So sometimes the low hanging fruit is one object. If you can, if that object is really quite distinct from center mo background, or if you can easily tighten 00:58:47.000 --> 00:58:50.000 the object to reject background. 00:58:50.000 --> 00:59:02.000 But sometimes the low hanging fruit might be two objects per event or more. If you have significant background, and you need more than one unusual object to reject the background. 00:59:02.000 --> 00:59:23.000 So run two has seen the exploration of analyses that really study the relative optimization between one and two or more objects in one event. So here I've selected as an example of this the recent Atlas, nonprofit photon analysis, which in run one required 00:59:23.000 --> 00:59:45.000 to nonprofit photons to reject background. But in, in run two has really developed a very nice analysis, where they optimize over one and to displace objects in order to cover to gain acceptance, and to cover conceptually to cover more signatures. 00:59:45.000 --> 01:00:04.000 And in general, if these optimizations are done within analysis. So what analysis has to signal regions. This allows the analysis, design to exploit complimentary coverage to design the analysis regions to really be as complimentary as possible, and also 01:00:04.000 --> 01:00:06.000 to make sure that they. 01:00:06.000 --> 01:00:24.000 If a combination is envisioned that our second ality is naturally insured. So it's really, it really behooves the team to, who's thinking about more than one object to either communicate between teams or to lay out a pass in advance. 01:00:24.000 --> 01:00:31.000 It's not only how many objects, one has per LP objects one has for event. 01:00:31.000 --> 01:00:53.000 That's interesting. It's also the combination of LLP objects and more standard prompt handles. So, we've also seen a movement towards combining LP objects with prompt handles, which has a number of advantages which I will touch on in the next slide. 01:00:53.000 --> 01:01:02.000 So one example of this from Atlas has been was an early full run to results, looking for a displays for text me one. 01:01:02.000 --> 01:01:06.000 And the new and was not explicitly required to be displaced. 01:01:06.000 --> 01:01:14.000 So this analysis was targeting Long live stopped with an art parody violating decay. 01:01:14.000 --> 01:01:27.000 So here, one of the advantages of adding in a more standard or prompt object is you get an automatic trigger handle, which can really enhance the acceptance at the trigger level. 01:01:27.000 --> 01:01:45.000 It also allows the analysis to relax the requirements on the displays for text in this case, in order to increase acceptance to a signal and that's important if you want to push toward lower mass, or sometimes towards shorter lifetime. 01:01:45.000 --> 01:02:00.000 This adding in in general adding in various handles increases the sensitivity to diverse set of models. But of course when you add in one handle then you're in some sense, losing sensitivity to another model. 01:02:00.000 --> 01:02:04.000 So while you gain sensitivity in a specific model. 01:02:04.000 --> 01:02:16.000 You become less general and so the way to really exploit this requires a lot more analysis or a lot more signatures. As you optimize each for a particular phase space. 01:02:16.000 --> 01:02:26.000 And while this in principle sounds relatively trivial, you know, okay I have displays our text, let me just add a new on let me add an electron, tell me, cetera etc. 01:02:26.000 --> 01:02:41.000 It's often non trivial to do this with existing data flow patterns, the way that we've set up our data is is not always optimized in advance for non standard objects LLP objects. 01:02:41.000 --> 01:02:52.000 And so, adding in existing objects can require restructuring the data flow and this can take a long time to propagate, and then similarly for analysis tools. 01:02:52.000 --> 01:03:12.000 So there's often a lot of work behind the scenes to get new object combinations possible, and one goal, moving into run three is to simplify this so that it's actually less work for analyses two, two combined objects whether those are LLP objects or prompt. 01:03:12.000 --> 01:03:32.000 objects. Another nice example of combining a kind of traditional LLP objects with standard model or prompt objects is in the flavor of the disappearing track analysis so both CMS and Atlas have in rent to both expanded their disappearing track analysis 01:03:32.000 --> 01:03:47.000 and also extended it into a kind of non traditional scenario where you have a displace to displaced charging or along liturgy note that from the decay of long, from the case of a prompt school you know. 01:03:47.000 --> 01:03:50.000 So this is not the canonical disappearing text search. 01:03:50.000 --> 01:04:10.000 But by combining a district track with extra jet requirements from the prompt decay of the cleaner. You can actually significantly extend the sensitivity to interesting face space in this case to this more complicated decay. 01:04:10.000 --> 01:04:25.000 And this has its advantages so you can actually suppress the background by requiring prompt jets, you get an automatic trigger handle, which increases your acceptance for this particular model. 01:04:25.000 --> 01:04:31.000 It this again, as I said, is non trivial with existing data flow objects. 01:04:31.000 --> 01:04:46.000 But what's nice about this and I think what we should think about very carefully when we're designing analyses is that these can these more complicated decay chains can actually be very complimentary and their sensitivity with the direct LP production 01:04:46.000 --> 01:04:55.000 and so this is illustrated in this plot from Atlas on the left, where the lower green shaded region is the exclusion from the direct charging know decay. 01:04:55.000 --> 01:05:13.000 And you can see, if you were to have that charge you know, it would be excluded and so you can really focus your sensitivity when you're looking for guaino production of charging those on the upper half of the spot. 01:05:13.000 --> 01:05:28.000 Okay, so another avenue that we've been pushing across the sheet is increasing our sensitivity to short lifetimes. This is hard, because the background from the standard model is larger. 01:05:28.000 --> 01:05:39.000 And we're, but we've been trying to overcome this challenge in a couple of ways. So the first is by reinterpreting prompt searches which have natural sensitivity there. 01:05:39.000 --> 01:05:46.000 And the second is with dedicated searches for objects with smaller displacements. 01:05:46.000 --> 01:05:51.000 So the reinterpretation of searches has taken several forms. 01:05:51.000 --> 01:06:03.000 So this is often again more challenging than just writing a signal through the existing analysis, in particular, it requires dedicated treatment of the systematic uncertainties for nonprofit signals. 01:06:03.000 --> 01:06:11.000 So this is an example from Atlas of a reinterpretation we did have RPC Susie prompt decay. 01:06:11.000 --> 01:06:19.000 For a long lived Gleaner, which required studying that for example jet response as a function of displacement. 01:06:19.000 --> 01:06:38.000 Additionally care is needed to keep the sensitivity for in prompt analyses to Long live signals. So often, sort of standard requirements for example for event cleaning where you cut out jets that look like they're from detector noise that you know most 01:06:38.000 --> 01:06:49.000 prompt searches don't think twice about these can have somewhat catastrophic effects on the acceptance of Long live signals. And so, if you want to be able to efficiently. 01:06:49.000 --> 01:06:59.000 Well, if you want to be able to reinterpret a prompt search and have reasonable coverage, you need to think about the LPS sometimes when you're designing analysis. 01:06:59.000 --> 01:07:16.000 And so this can be quite natural. When the LP signal is considered while you're doing the prop search, but probably we should think about how more systematically to make sure we're not excluding sensitivity unduly in our prompt searches. 01:07:16.000 --> 01:07:32.000 The second way we can extend our sensitivity to short lifetimes is by having dedicated searches, which are optimized for those. And so this is a nice example from CMS of a displays protect search that really focused on the difficult region of, you know, 01:07:32.000 --> 01:07:50.000 sub millimeter displacement, and to try to gain sensitivity here really required developing new discriminates in this case, topological discriminates about the shape of pair produce displays vertices that allowed them to optimize from this regime and 01:07:50.000 --> 01:08:05.000 gain sensitivity, relative to both the prompt searches and the displays for text searches that look at displays vertices farther out. And so I think this is kind of inspiring that we can, it is possible to optimize for this regime. 01:08:05.000 --> 01:08:18.000 With dedicated signatures and dedicated work to reject the larger Standard Model background that's there. Okay. Thanks for the warning, I see it now. 01:08:18.000 --> 01:08:19.000 Okay. 01:08:19.000 --> 01:08:35.000 So machine learning has been a powerful tool that's been more fully exploited in prompt searches in run to that I run one, and there's also been a lot of developments in the world, which has different challenges and the prompts searches. 01:08:35.000 --> 01:08:42.000 So I picked out two examples of LP searches that use modern machine learning techniques. 01:08:42.000 --> 01:08:54.000 I'll say that here I think with LP searches the challenge is really are one of the challenges is that the simulation does not necessarily model, non standard signatures very well. 01:08:54.000 --> 01:09:06.000 So, how you train these is non trivial, and how you validate that your machine learning is actually going to be able to find your displace or nonprofit signature is also non trivial. 01:09:06.000 --> 01:09:26.000 So these two examples I've picked up handle this problem in two different ways. So this first search from CMS uses domain adaptation to make the jet classification which is trying to tag jets from LAPD case to make it invariant with respect to data or 01:09:26.000 --> 01:09:27.000 simulation. 01:09:27.000 --> 01:09:40.000 And you can see on the left here. The first plot is without domain adaptation comparison and a controller engine between data and Monte Carlo that doesn't decrease so well, which highlights the problem here in training on Monte Carlo. 01:09:40.000 --> 01:09:45.000 But what this domain, adaptation, you can see in the bottom seems to be much better. 01:09:45.000 --> 01:09:52.000 Atlas has done has tackled this similar problem but with a slightly different solution. 01:09:52.000 --> 01:10:01.000 So here, Atlas used an adversarial network to minimize the difference in the training between data and Monte Carlo. 01:10:01.000 --> 01:10:03.000 And again you can see on the left. 01:10:03.000 --> 01:10:12.000 The top plot shows what the data Monte Carlo agreement to the conclusion looks like without using the adversarial network and the agreement is really rather horrific. 01:10:12.000 --> 01:10:21.000 But when you use the adversary network you can see it dramatically improves and increases our confidence that this is a tool that can that can work very well. 01:10:21.000 --> 01:10:24.000 Even when you apply it to data. 01:10:24.000 --> 01:10:29.000 So, looking at dedicated triggers. This is a very exciting area. 01:10:29.000 --> 01:10:33.000 And it's a hot topic for development. 01:10:33.000 --> 01:10:46.000 I know that there's again, a dedicated session in this workshop so what I wanted to highlight here is that already in run two there's been a lot of work in pioneering LP triggers in particular for example topological triggers and the color matter in both 01:10:46.000 --> 01:11:01.000 Atlas and CMS tracking hlt from CMS and slow displace new ones as well. And so I think we have a lot to learn as a community from the experiences of the triggers that had been pioneered. 01:11:01.000 --> 01:11:14.000 And part of that is that, you know, designing a trigger is one thing but, ensuring its use an optimization through the full running is a completely other challenge. 01:11:14.000 --> 01:11:18.000 So Brian mentioned data scouting but I did want to mention as well. 01:11:18.000 --> 01:11:29.000 In addition to new triggers, new trigger level analysis is also an important frontier. This was pioneered in run to for long the particle searches by URL HTTP. 01:11:29.000 --> 01:11:33.000 So it's a very nice way to work around trigger limitations. 01:11:33.000 --> 01:11:53.000 And to access lower masses by storing limited event information at a higher trigger rate. In this case for them you in Paris. And this really allowed us up to access, much lower dark photon masses then they would by just using the then offline trigger 01:11:53.000 --> 01:12:07.000 rate, CMS has a similar analysis as well and this trigger level analysis has been explored for prompts searches as well but I think there's more potential for it to be explored in the world. 01:12:07.000 --> 01:12:12.000 And I'm going to end with some experimental challenges and some solutions that we've had, we found. 01:12:12.000 --> 01:12:24.000 So the first is that reconstructing displaced tracks, whether those tracks are in the inner detector or neon system is challenging do two huge component works at high pile up. 01:12:24.000 --> 01:12:26.000 And so this has been an ongoing problem. 01:12:26.000 --> 01:12:33.000 But we've been tackling this at least within Atlas with a very large effort to improve track quality. 01:12:33.000 --> 01:12:42.000 This allows us to reconstruct tracks with high signal acceptance but significantly fewer background tracks per event. 01:12:42.000 --> 01:12:58.000 So this not only reduces the background in each analysis, but more importantly simplifies the data flow, when you have fewer displace tracks reconstructed, you can afford to have those events in more places, and that that's been a lot of work in that 01:12:58.000 --> 01:13:00.000 list that setup for on three. 01:13:00.000 --> 01:13:14.000 This is a powerful tool that will allow us to reduce the threshold for reconstructing displaced objects, and that in turn will allow us to open up new face space and also to add a new prompt handles more easily. 01:13:14.000 --> 01:13:20.000 And so we can expect this to be a really active area, as we move it to run three. 01:13:20.000 --> 01:13:38.000 We've also been innovating on reconstructing new types of objects. And so, Atlas has a nice pub note out I encourage you to check it out if you haven't done it on a very challenging problem which is reconstructing soft displace objects for the example 01:13:38.000 --> 01:13:55.000 here is the soft pion. In the case of the disappearing track which is order 100 MTV. And so you're looking for 100 MTV displace track in a sea of pile up this trying to reconstruct This works against the flow of what's happening with more standard objects 01:13:55.000 --> 01:14:05.000 which is that as pilot increases, you tend to raise thresholds and tighten selection in order to keep event size and background under control and CPU usage under control. 01:14:05.000 --> 01:14:19.000 So trying to reconstruct something that soft and displaced, you know, requires a lot of work, to, to, to optimize around the constraints of keeping CPU and event size reasonable. 01:14:19.000 --> 01:14:32.000 But this has been shown preliminary in Atlas is feasible. And so I think this is a nice example of how we can continue to innovate, even in difficult reconstruction various. 01:14:32.000 --> 01:14:42.000 And finally, I wanted to just mention it's not only pile up. That's the problem. It's also integrated luminosity can be a challenge. 01:14:42.000 --> 01:14:47.000 So another way of saying this is that more data is always your friend. 01:14:47.000 --> 01:14:50.000 But sometimes, a fickle friend. 01:14:50.000 --> 01:15:02.000 So in particular for the inner detector integrated luminosity brings radiation damage, which then affects the charge collection, in particular in the pixel detectors. 01:15:02.000 --> 01:15:12.000 And this is challenging for tracking in general, but specifically tracking for ddx measurements which require efficient charge collection in those detectors. 01:15:12.000 --> 01:15:27.000 And so this plot on the right, you'll hear more about from an on tomorrow or Wednesday, but is a plot of the GD x from Atlas pixel detector as a function of integrated luminosity. 01:15:27.000 --> 01:15:41.000 And the fact that it's just going down is due to radiation damage. And so, this is an intrinsic challenge and there's multiple solutions to this. So the first detector solution is to just turn up the high voltage. 01:15:41.000 --> 01:15:49.000 You wanted an experimentalists perspective so I made sure I would talk about high voltage, at least once in the talk, but only once, but that has its limits. 01:15:49.000 --> 01:16:05.000 So an analysis solution, which we pioneered and this latest round of the GD x search, is it actually do data driven modeling data driven modeling of the DX for signal money color and everything else is done for background and therefore you basically say, 01:16:05.000 --> 01:16:16.000 I don't care about what this simulation says I'm just going to measure the effects of the radiation damage in data and use that for my data and for me, simulation. 01:16:16.000 --> 01:16:34.000 But there's also been work, a lot of work in Atlas on a simulation solution, which is to add the effects of radiation damage to the Monte Carlo, and that's a very interesting physics problem with lots of physics about silicon detectors and trapping and 01:16:34.000 --> 01:16:39.000 things. And so if you're interested in that I encourage you to check out the the paper here. 01:16:39.000 --> 01:16:53.000 So as we look forward. I agree with Brian LPs have always been exciting and I think they will continue to be exciting and particular run three is a very fertile territory to look for LPs. 01:16:53.000 --> 01:17:03.000 We have dedicated detectors coming online which is very exciting. We have a lot of work on dedicated triggers which will be interesting to follow their a new analyses. 01:17:03.000 --> 01:17:05.000 They're also new challenges. 01:17:05.000 --> 01:17:22.000 But I'm confident that will continue to innovate, that in ways that will allow us to push deeper into the Face Face and innovation will be key because the data set itself won't be increasing significantly and so all of these things new triggers new techniques 01:17:22.000 --> 01:17:27.000 will be important, and allowing us to continue to enhance our sensitivity. 01:17:27.000 --> 01:17:31.000 I just wanted to say one thing about the plot on the right. 01:17:31.000 --> 01:17:48.000 Because I often hear you know that LPs are hot now and you know there's the community is certainly growing, which is true but the, if you look at the number of papers we've produced as a function of time, it's more or less linear, and I think this is 01:17:48.000 --> 01:18:03.000 partly because we're doing more complex and sophisticated analyses and so even though we have more people in the community. We have more people per analysis, which also takes more time per paper and so the actual output of papers is slow. 01:18:03.000 --> 01:18:16.000 So I think this is reflective of the sophistication, actually, and the development, the innovation that we're doing in these analyses the fact that we have a larger community, but we're not necessarily putting out more papers. 01:18:16.000 --> 01:18:26.000 It's also probably reflective of the fact that the data set, Dublin time is increasing and so we have more time to work with the papers. 01:18:26.000 --> 01:18:28.000 Okay, thank you. 01:18:28.000 --> 01:18:35.000 Laura Thanks a lot, um, let's talk on the experimental landscape. Yes, with a thing. 01:18:35.000 --> 01:18:46.000 Does anyone have any questions for Laura. Oh, I see multiple hands, and I think I'm hoping that they're ordered. I guess the first one is product. 01:18:46.000 --> 01:18:52.000 Hi. Hello. Thanks for the very nice and useful experimental overview. 01:18:52.000 --> 01:19:05.000 So I had a question regarding elven triggers So, is there a possibility of having dedicated LLP triggered Elvin for CMS, like we have Kaldi should trigger from Atlas. 01:19:05.000 --> 01:19:12.000 I'm not an expert on CMS, So if someone from CMS would like to handle that. 01:19:12.000 --> 01:19:31.000 But I can, I can make a guess which is that, you know, in general, I would say, Atlas and CMS tend to have complimentary, you know, even if the implementation of our detector hardware is not the same we tend to have similar feasibility so I would guess 01:19:31.000 --> 01:19:39.000 yes but I'm going to see if someone from CMS wants to take that 01:19:39.000 --> 01:19:42.000 hey, I can say that yes we're we're working on it. 01:19:42.000 --> 01:19:46.000 But I don't want to spoil anything so stay tuned. 01:19:46.000 --> 01:19:53.000 Thank you. 01:19:53.000 --> 01:19:59.000 I think, I think the next from the cheetah. 01:19:59.000 --> 01:20:05.000 Thank you very much for this talk, there were a lot of tantalizing slides. 01:20:05.000 --> 01:20:19.000 So I actually have two very quick questions first was to understand the word handle that you used in the beginning. It's the first time I've come across this, a lot of new ll be handled so if you could explain what that is. 01:20:19.000 --> 01:20:23.000 And the second one is on slide 25. 01:20:23.000 --> 01:20:26.000 Okay, let me see if I can. Yes. 01:20:26.000 --> 01:20:36.000 So, the handle handle I meant there as a way of not repeating the word object. So when I use handle I meant. 01:20:36.000 --> 01:20:41.000 Yeah, or reconstructed object. 01:20:41.000 --> 01:20:58.000 Just going for me. In case I make a mistake later. Yeah. So on this slide, I was just looking at the red car for the RPC reinterpretation, and just looking at the dotted expected line, it seems that you might do a pretty good job just looking at exponential 01:20:58.000 --> 01:21:10.000 decay and counting how many of them decay within the fiduciary volume or something like this Do you think that's a good strategy for like a, like a quick estimate of how things would go. 01:21:10.000 --> 01:21:23.000 Yes, and for for reinterpretation of prompt searches you, there's a couple of things you should be careful of so if you're sure that you're not actually you're right. 01:21:23.000 --> 01:21:41.000 If you, if you think about an exponential decay right, even for a very long lifetime there's still there's still a lot of your LPs that decay right at the origin so you can you can you know look within the resolution of your prompt objects and just count 01:21:41.000 --> 01:21:59.000 the number that of your, your LP decays that that happened within that resolution from the the primary vertex. If you want to extrapolate out beyond that used to be very careful that you haven't that none of your object reconstruction is placing any bounds. 01:21:59.000 --> 01:22:08.000 So I talked here about the event cleaning which is one that on Atlas at least we found often can really kill nonprofit signals. 01:22:08.000 --> 01:22:23.000 When you try to reject jets that look like they're from noise. For example, But you can also think about left on requirements that that disfavor non prompt leptons and things like that. 01:22:23.000 --> 01:22:34.000 So if you want to go beyond the resolution and the decays that happened, of your nonprofit signal right at the origin, then I think you need to make sure that you're not hit by those acceptance effects. 01:22:34.000 --> 01:22:47.000 And then if you want a more sophisticated treatment you have to think about actually handling the systematic uncertainties, which can be significantly different than for prompt signals. 01:22:47.000 --> 01:22:52.000 Good, thank you. 01:22:52.000 --> 01:22:53.000 Michelle. 01:22:53.000 --> 01:22:58.000 I think Matt, you can go ahead and 01:22:58.000 --> 01:23:10.000 You said something about the work that's being done to make combination of objects both longer than stare model easier. Can you just say a few words about what kind of work that is. 01:23:10.000 --> 01:23:24.000 Let me see if I can remember what I said what I meant. Um, so one thing I didn't say in this talk that I kind of wanted to say but I didn't really have time was that in general, this isn't really answering your question but it's sort of related that you 01:23:24.000 --> 01:23:39.000 know there has been a lot of work on both Atlas and CMS, to increase the accessibility of our reinterpretation material for along with particles. So we have done a lot of work to improve the object that we provide the information that we provide to the 01:23:39.000 --> 01:23:40.000 general community. 01:23:40.000 --> 01:23:43.000 Okay, that's not what I was talking about. 01:23:43.000 --> 01:24:00.000 I can't remember where it was, but what what I was talking about when I talked about combining LP and non and prompt handles, at least from Atlas, I can say that for example, the work that we've done. 01:24:00.000 --> 01:24:23.000 To simplify, or improve the displace tracking reduces significantly the number of non standard tracks, you need to keep for each event. And this makes you know it possible to envision a world in which you can put those displace tracks into more data flow 01:24:23.000 --> 01:24:37.000 events and then you can combine them more easily with standard candles. So a lot of it comes down to very kind of experimentally things right of the past that you're the past that your different reconstruction takes and whether you need to have a separate 01:24:37.000 --> 01:24:48.000 stream for different types of non standard reconstruction. 01:24:48.000 --> 01:24:58.000 Hey, thanks a lot, not and, of course, and then I think we have maybe a very quick question from Julia, before we go and take our workshop photo. 01:24:58.000 --> 01:25:07.000 Yeah, sorry, um, I just wanted to actually make two really quick comments, so I think Laura you had said. 01:25:07.000 --> 01:25:19.000 For the display slept on searches that this was something we hadn't discussed that one without a common vertex, this is something we hadn't done before run to and I just wanted to push back on that a little bit. 01:25:19.000 --> 01:25:31.000 So CMS actually published a search on this I think in PRL, where you have displays electron and you are not coming from a common vertex, so I just wanted to make you aware of that sorry that's fair. 01:25:31.000 --> 01:25:46.000 No, I was aware of that I simplified a bit so right so if it had been electron and you on exactly and to reduce the background right you had to the to this similar flavors, what was what was New Year right is, I should have been more clear was the, the 01:25:46.000 --> 01:25:53.000 addition of the electric electronic me on the one channel. Absolutely. Yes, sir, with which are listed as well beautifully. 01:25:53.000 --> 01:26:12.000 And the other thing that I just wanted to bring up is that something else that I think that it has changed a lot in the last six years, or evidence of of this change is that in CMS at least going towards run three, there's 100 hertz at the trigger level, 01:26:12.000 --> 01:26:29.000 at hlt level dedicated now too long the particle triggers like new. Long live particle triggers, so I, in case I noticed that this didn't come up in your in your talk so I just wanted to mention it because I think it's really evidence of how times have 01:26:29.000 --> 01:26:51.000 changed and I think it's really nice thing. Thanks.