eshwen/phd-thesis

My PhD thesis, typeset in LaTeX

phd-thesis's Issues

Suggestions from Henning (21/05/2020)

@flaecher and I had a discussion about the overall structure of the thesis on 21/05/2020. The suggestions are the following (documenting as an issue as I'm not sure where else it can be put while still being clearly visible):

• Move results of previous Hinv searches from analysis chapter to the Hinv bit in theory chapter (done: 8363eea)
• Move definitions of variables (pt, MET, etc.) from theory chapter with the object definitions from the Hinv analysis chapter to a new object definitions chapter, placed between the detector and Hinv analysis chapters (done: 8363eea)
• Move discussion of production modes from Hinv analysis chapter to theory (done: 8363eea)
• For SVJ, could append my contributions to the theory chapter, or have a small, separate chapter describing my work (probably after detector, but before object definitions) (done: 5fa2054)
  • Make sure to clearly say that I worked on generator-level stuff to verify/provide alternative method for the signal samples for mature s-channel analysis, and that the code and stuff is being used in these new analyses like the boosted Z' and especially t-channel.
  • Can reference Giorgia's thesis or the paper or something for the full breakdown of the analysis
• For later sections in the Hinv analysis chapter, would have to figure out how to structure things for the background estimation/categorisation onward. For example (done: ab22f71)
  • ttH (categorisation, background estimation, results) -> VH (categorisation, background estimation, results) -> ggF (categorisation, background estimation, results)
  • Or categorisation (ttH, VH, ggF) -> background estimation (ttH, VH, ggF) -> results (ttH, VH, ggF)
• Don't need to mention that, in the main Hinv analysis, Boston are doing boosted VH and ggF/monojet
• Change thesis title
  • Something along the lines of "searches for dark matter/ptmiss with a focus on invisibly decaying Higgs bosons". Would probably have to include the words "LHC" or "CMS", and either/both of "sqrt(s) = 13 TeV" and "2016-18 data"
  • Maybe something a little more sensationalist: "Probing the Origins of Dark Matter: A search for invisibly decaying Higgs bosons in the ttH, VH and ggF channels at CMS at sqrt(s) = 13 TeV" (I only spent a few minutes on this, but it would be nice to have an attention-grabbing title, even if it's bit more toned down that what I've put above)
  • Since my introductory chapter and some theory is based around dark matter, I figure that should be in title somewhere. Otherwise, it doesn't flow particularly well from the title into the main text

Suggestions from Henning (28/05/2020)

@flaecher and I had our second thesis chat, again more of an overview discussion. Suggestions are the following:

• In the object definitions chapter, probably don't need to go into that much detail for things like electron volt
• In the Higgs to invisible analysis chapter, don't need to go into that much detail regarding smaller backgrounds (multiboson, ttbarX, etc.). Can describe the main backgrounds in detail, however
  • May not need to go into detail regarding Pythia tune, etc.
• Describe SFs and systematics for ttbar, b-tag, DeepAK8, JEC, JER in reasonable detail. Perhaps lumi, pileup, lepton ID as well
  • For the others, I can just briefly summarise in a couple of lines each
• Keep thinking about thesis title
  • If including SVJ in the thesis, probably don't need to mention it in the thesis title, but have it be something like "Dark matter searches with a focus on invisibly decaying Higgs bosons..."
• Will have to decide soon whether to include ggF in the thesis or just keep it focused on ttH and VH
  • Depends on how much extra work is required, and how quickly things progress with the analysis and thesis
  • If including ggF, might want to change all instances of where I've written "ggF" to "ggH" as that's the more widely-understood contraction
• Still need to figure out where semi-visible jets would fit in
  • Would probably just replicate, for the most part, the section I wrote in the AN that compares Pythia and MadGraph samples. Could remake the plots using better binning, axis limits, etc. Only include some of the highlight plots (MET, MT, min_dphi, maybe MET/MT, etc.)
  • Would also like to include t-channel distributions, similar to the s-channel to highlight the differences between them. Would need to talk to Kevin and get his fixes to my t-channel generation code
• For dark matter interpretations, it would be useful to have signal samples with different Higgs masses. The samples don't exist centrally, but I think have been requested. See this google doc for the dataset names and Higgs masses
• Note on the kind of plots that can be added to the Higgs to invisible chapter:
  • Before-and-after plots motivating certain cuts (min_omega_tilde, min_dphi). Do for inclusive categories to minimise bloat
  • Potentially some control region plots for inclusive categories (akin to https://github.com/eshwen/phd-thesis/tree/v0.5.2-forAPM/chapters/higgstoinv/figures/region_plots/2016to18)
  • Mountain range plots (pre-fit and post-fit)
  • Limits
  • Likelihood function for each production mode like page 19 of the latest Higgs paper (attached - 1809.05937(2).pdf)
  • Dark matter limit plots like page 21 of the latest Higgs paper (SI DM-nucleon cross section, kappa_V-kappa_F)

Suggestions from Henning (18/06/2020)

Henning and I had a thesis chat. Suggestions below regarding actual thesis material as well as the timeline:

• Remove the production modes from the thesis title
  • For semantics, ask Ben/Olivier on Mattermost. "Searches for dark matter with a focus on..." or "Search for dark matter with focus on...", "full Run-2 dataset from the CMS experiment" or "full Run-2 dataset of the CMS experiment"
• Ask Sudan about reading the thesis
• To ask for the 1 month extension (or perhaps longer if required), need to provide a detailed submission timeline so we know when to freeze the "thesis version" of the analysis, unblind, and write up. Kinda depends on when current studies are finished and validated
  • Henning to clarify with Krissy about the specifics of this extension. Since we previously asked for an extension to funding, rather than submission. But the form Terry and Krissy sent this week appears to be related more to a submission extension
  • Henning to also talk to Joel regarding the point above. Supporting evidence and other things may be required if pushing back the submission deadline
  • The studies on which scenario to use must be finished
  • Ideally, the studies on the DeepAK8 working points and the QCD estimation will also be finished. But that's not a necessity
  • Will need to discuss with the Bristol guys about the JES/JER corrections and systematics. Will need to run a new skim with the JES/JER up/down jet and MET collections, regardless. But NanoAODv7 also contains updated JME corrections, and Boston plan to switch to it if they haven't done so already. So we may benefit and save time by running the new skim with NanoAODv7 + new collections, rather than NanoAODv5 + new collections
  • Need to think about when we would go for pre-approval with the main analysis, since unblinding will take place/will have already taken place. If it's on the timescale of a less than a month, there will probably be little divergence between the main analysis and the one that ends up in the thesis
  • If the main analysis would otherwise take longer to go for pre-approval, Henning and I could unblind privately and then diverge from the main analysis. Would ideally only submit after main analysis is unblinded so timelines appear to work out
    • In this case, could initially perform fit to validation region in all 3 years to see if there's an excess. Bad if so because fit won't converge to get a limit
    • Could check Alex Titterton's thesis (2016 and 2017) plots to compare with ours, since some categories should share similarities
• For SVJ, specifically:
  • Figure out why Kevin's s-channel cross sections are so different from MadGraph
  • If wanting to plot s-channel and t-channel MadGraph on the same axes, consider using the ratio between s-channel Kevin and MadGraph to scale t-channel cross sections
  • Ask theorists how they got t-channel cross section?

Fix broken badge images

It's slightly annoying that Shields, who supply most of the inciteful badges at the top of the README, don't have an easy implementation for private repos. There are a couple of workarounds (badges/shields#4681, badges/shields#593 (comment)), but I haven't figured out how to implement either of them yet.

While the images are broken, the hyperlinks (e.g., for the downloads) still work, which is something