And, already, we come to the last parallel session of the conference. I am once again bouncing around sessions.
4:30 pm: Matrix Element Techniques, Kontantin Matchev
My apologies to Konstantin for not recording his jokes.
Phenomenologists are the ones who answer questions about Monte Carlo simulations. The Theory to Experiment chain involves a long chain of tools. Despite being very standard, automated and user-friendly, knowledge of these tools is essential. These tools go from matrix elements to event (likelihoods); but they can be run in reverse, going from data to matrix elements! This is what we are talking about.
The Matrix Element Method (MEM) uses all available kinematic information. It also has a physical meaning. The main problem-the difficulty in computation-is now solved by these modern tools. Technically, despite name we are using the likelihood; Neyman-Pearson lemma says this is the best possible discriminator for any signal vs background test.
MEM can also be used for measurements, e.g. mass measurements in SUSY-like events, where there is missing information (the invisible mass).
Another good use is for general searches for unknown NP. In this case use background ME only.
Finally, consider scanning a complex parameter space. This is challenging for typical NP models. Various techniques, but an efficient one is to use a reweighting of fully simulated event sampled by ratios of ME. The biggest problem is that you need to store "true" event information (extra memory). But this cost is still less than the computational cost of rerunning the simulation chain.
Questions
Can systematics be incorporated? Yes, though so far have focused on simpler cases.
5:00 pm: Searching for New Collider Resonances through Topological Models, Mohammad Abdullah
The search for NP has two extreme limits. Specific signatures are too limiting, fully general searches too broad. Instead seek intermediate approach. Pick a final state, and survey all possible topologies that lead to resonance features. Construct simple models that lead to such topologies and perform the analysis. Systematic, easy, complementary and model independent but limited to visible resonances only.
Case study: 2l2j. For simplicity take a Z' initiator. Then four possible cases, though one (leptoquarks) already well studied. Take (ll) plus (jj) resonance structure (for time reasons). Construct (prospective) limits.
Biggest advantage is that model is simple, so can easily be applied to future models.
Questions
Plans for expansion to other resonances? No.
5:15 pm: Discovering New Physics with Voronoi Tessellations, Jamie Gainer
New methods for finding features. In particular, we want a useful tool to find something even if we don't know what it is, and that can work in multi-dimensional parameter space.
Voronoi tessellations based on splitting region into cells, defined as set of points closest to various seed points. Interpret these as data points in possible space of values. Use properties of the cells to identify features. Example: search for some edge in data. Best test statistic: standard deviation of areas of neighbour cells, relative to mean area.
Interesting trick to improve analysis: replace data points by centroids of associated cell. This is Lloyd's algorithm, a smoothing algorithm. However, if overdone this can actually remove any features; only iterate a small number of times.
Questions
Publically available? Yes, works in standard Python (& Root).
Why is Lloyd's algorithm better than other smoothing? Unknown yet.
Real data? Not yet, would be fun.
5:30 pm: Heavy Type III Seesaw Leptons at NLO in QCD, Richard Ruiz
NLO in BSM is pretty much standard these days. More accurate rate predictions, but especially for distributions. Other seesaws looked at at higher orders, but not Type III.
Typical heavy fermions in this model ~ TeV, assuming neutrino and electron Yukawas similar. One new fermion is charged (SU(2) triplet). It is pair production of charged or charged+neutral that we are interested in here.
Use phase space slicing: divide radiative corrections into regions, so can focus on divergences and non-divergent pieces separately.
Results: see cross sections enhanced by about 20 to 40%. Scale dependence improved compared to LO.
5:45 pm: Vector Dark Matter via Higgs Portal, Anthony Difranzo
Vector Higgs portal is dimension 4, but non-renormalisable. Observables then have different dependence on vector mass than might be expected. UV completion: Higgs portal with scalar charged under new gauge symmetry. Well explored, so leave it be. Instead, focus on radiatively generated interaction.
UV complete theory should be anomaly free and forbid kinetic mixing. Can be done with two (SM) vector-like doublets, plus a SM Dirac singlet. Charge conjugation symmetry forbids kinetic mixing. Must assume Higgs portal coupling is small.
Questions
Relic Density: calculation in progress.
Direct detection limits from Higgs invisible width as vector mass goes to zero. EFT similar to loop model, thought it was breakdown of EFT? Not sure.
6:00 pm: Dark Matter Explained through two distinct ideas related to the Higgs, Shreyashi Chakdar
Trying to present two models, this will be tight.
Model 1: Asymmetric matter (Mirror SM?) with higher QCD scale in hidden sector. Communication only through Higgs portal. Use Pati-Salamm gauge group to forbid kinetic mixing. Also assume hidden sector colder to meet BBN constraints. Consider two Higgs split by only 100 MeV, look for at ILC. Might be possible.
Model 2: WDM in 2HDM. Specifically inert doublet model to get DM.
6:15 pm: Scalar Dark Matter mediated via Colour Scalar, Gaurav Mendiratta
Gaurav has been delayed, so this is a Skype talk!
Scalar somehow does not have Higgs portal coupling, but mediates through a new coloured scalar. Further allow mediator to couple to right-handed up-quarks. This means mediator has flavour symmetry, so assume MFV.
DM mass required to be heavier than mediator by relic density (unsurprising, only two body annihilation channel). Some other limits.
No comments:
Post a Comment