Connect through Zoom: https://uchicagogroup.zoom.us/j/170161490
password: 202021
Jorge de Blas - Overview of SMEFT fit in ESG and beyond
-------------------------------------------------------
ESG 2020 update
different working groups to assist Physics Preparatory Group
Higgs@Future Colliders WG formed by RECFA to evaluate physics potential of future
colliders in Higgs physics
main outcome: JHEP 01 (2020) 139 (arXiv:1905.03764)
some results summarized in EWK chapter of Physics Briefing
studies prepared in two scenarios: k-framework, and SMEFT-framework
k-framework useful to validate procedures/code
in SMEFT framework, performed fit
list of assumptions: dim-6 EFT only; no effect of 4-fermion operators (except for
the ones that affect muon decay - and GF); no dipole operators (could be relevant
for top)
two flavor assumptions: flavor universal (18 NP parameters) and flavor diagonal
(but no necessarily universal; 30 NP parameters)
for model building, rather cumbersome scenario
fit performed in gauge-invariant basis; chose parameterization basis-independent
to ease comparisons
defined effective Higgs couplings, converting EFT results (Wilson coefficients)
into pseudo-observable quantities (a bit similar approach as k-framework)
used HEPfit (hepfit.roma1.infn.it)
NP effects at linear order
table with summary of future colliders included (from HL-LHC to FCC-hh)
table with summary of inputs from different sectors: Higgs, aTGC, EWPO, top EW
Higgs: mostly inclusive measurements (signal strengths for different production
and decay modes)
multi-bosons: mostly in format of aTGC searches
EWPO: runs at Z pole; assuming GigaZ option for ILC and CLIC
top EW: no from CEPC (no run at 500GeV?)
performed fit, comparison of sensitivity of different colliders to single-H
couplings
C. Grojean's talk during EF01 kick-off meeting has more details
list of all numbers and assumptions
sensitivity to NP in EW interactions: EW Zff couplings: can get to 0.01%
different assumptions on size of modeling uncertainties lead to quite different
expectations (shown conservative and aggressive) -> useful to reach agreement
about systematics
theory requirement in context of oblique parameters
see that current status of calculations (full 2-loop) are not enough -> made
extrapolations assuming 3-loop EW and QCD will be available
technically challenging, but feasible
high-energy probes
new effective operator will induce interference w/ SM, and have effect in tail of
distributions
look at m(ZH) in pp->ZH->Zbb, or pT(WZ) in WZ production
studied with SILH-like effective Lagrangian
for WW production, received only aTGC limits; no full EFT study available
also, projections based only on differential angular distributions, w/o
correlations among bins
JHEP 12 (2019) 117: new sensitivity study using full info about each event
supposed to give maximal statistical sensitivity, but nothing about systematics
method provides interesting increase in sensitivity, would be interested in
applying to analysis, add realistic experimental systematic, and see
future directions, some examples:
- interest in experimental input on WW OO method
- top sector explored superficially
- BSM interpretations
Sven Heinemeyer: taking inputs from experiments and using them in fits, any
assessment of whether they were all on equal footing? should be critical to check
what is behind each input, and if each input is justified, or too optimistic
A: tasked with including all, and used both most aggressive and most conservative
scenarios; this would be good moment for doing a more detailed assessment/comparison
between experiments.
Michael Peskin: group did optimal statistical analysis of e+e-; OO analysis
improves observables connected with e+e-, and has implications with other
processes: e.g., link with Higgs couplings. EF01 interested in H soft couplings,
and this improvement (and error quoted here) makes big impact; people shall look
at experimental determination of parameters at e+e-, and see what is best
(stat+syst) we can do; not done in Snowmass before, shall be done
Junping: was there EFT treatment of top measurements?
A: were basically limited to EW couplings of top (but no 4f operators), and using (for those) the measurements from FCC-ee; some old paper available from ILC
Junping: addition of bbbar data would help
A: if adding, need to add 4-fermion operators; something important to be done?
William Shepherd - Theory errors in EFT analyses
------------------------------------------------
emphasize that talk in context of doing BSM, and should never look at one operator
at a time
if trying to build BSMEFT, then really need to consider theory errors
links to two papers, and one in preparation
in pheno study, generally do not care much about theory errors; signatures would
not change drastically because of small theory error
in precision measurements, which is current business, not looking for new striking
signals: neglecting errors is never correct
nice quote from Bhaskar Dutta about models in which one or few operators are
considered only, and theory errors are neglected
historically, quark compositeness is source for this
posited quark contact interaction
not an EFT, it is a UV model
kept all dim-6 operators, also squared ones; one can make angular search, with
good sensitivity
challenge is EFT consistency: what happens at next-order? (EFT is new perturbation
theory)
need to expand _observable_ (the cross section), not the amplitude: would end up
with un-cancellable effects in squares of loops
need also to include all dim-6 operators; in jet study for quark compositeness,
there are only two contributing independently to angular distributions
dim-6 squared easy to get, it is 1/lambda^2, and it is good proxy for the dim-8
operators (the part we are neglecting in expansion, and gives size of uncertainty)
need to abandon idea to fit for QCD normalization: errors would swamp analysis
looking at un-normalized distribution
can get bounds by doing analysis at different luminosities
issue is that narrow angle in coupling space can be constrained
dileptons and SMEFT
- Z couplings can be shifted by SMEFT operators
- direct 4-fermion operators give amplitudes which grow with energy
- these produce preferably forward e+/e-
theory error treatment
dim-8 effects are order 1/lambda^4, signal is 1/lambda^2 -> compare to dim-6
squared, also 1/lambda^4
model theory error using Wilson coefficient of dim-6 squared; consider bins
uncorrelated, require dim-8 coupling not to be small
plots with coupling shifts
generic behavior: not really an ellipse; at large Wilson coefficients constraint
becomes looser (series not converging any more)
moral:
- need _global_ analysis for BSM
- develop more off-shell observables is future path
- dijets and dileptons first step, need to find more directions
- concerns apply, in EF04, to di-boson production
take-away:
- setting EW observables shifts to zero does not give model-independent results
- should stop neglecting theory errors, or model builders will ignore
Christophe Grojean: when one increases L, theory error increases in plot?
A: it goes down as expected (some plots have a small bug, will be fixed soon)
in the region of interest, but basically stays same in the region where c_6 are very large and unbounded.
Christophe: those new directions can likely be closed in future by including more observables.
A: hopefully, but there are many directions to constrain.
Christophe: agree on the need for more global analyses in future
Laura Reina: search for un-normalized distributions; how are bands estimated?
A: total error; red and black are MadGraph output normalized to fixed luminosity
(NLO + some k-factor); difference red-black calculated; error on top has no QCD;
experimental errors in tail, increasing luminosity reduces them in tails, while
theory error does not change; theory error from formula in previous slide (using
dim-6 squared uncertainty as proxy for missing dim-8)
Pietro Govoni - Future perspectives for EFT studies at LHC
----------------------------------------------------------
disclaimer: personal opinions on the subject; experience in CSM w/ Higgs searches
and multi-boson VBS processes
some material to trigger discussion on set of questions
EFT at LHC: performed since beginning of LHC
first plots on Wilson coefficients dim-6 from 2013
most results expressed in terms of anomalous couplings; NP summarized as
multiplicative modification of one coupling in single vertex
either anomalous triple or quartic gauge coupling
historically associate TGC to di-boson, and QCG to tri-boson and VBS
also in Higgs sector: limits on SMEFT operators HH, HB, HG...
Higgs usually described in kappa framework
standard template cross section (STXS): way to parameterize cross sections in
bins, at different levels of refinement (so-called versions)
EFT fits also performed; list of some analysis doing fits in basis of higher-order
operators
note that analyses use different basis: from effective amplitude variation, to
Warsaw basis, to SILH, HEL, STXS/k-framework/HEL: each on their own...
top sector: look for FCNC, or charged lepton flavor violation, or search for
general anomalies
several pieces already on table toward global fit
there are different approaches on how to model BSM effects, connected to what
operators get tested in each final state
different choices on how to treat EFT unitarity issues
set of questions!
need to move from anomalous coupling to fully-blown EFT study
overcome traditional separation dim-6 vs dim-8: lower terms always more important
in general
need phenomenological studies for guidance (in particular for sophisticated final
states, such as VBS)
example of VBS analysis being more sensitive to dim-6 operators, while
traditionally considered for dim-8
identify which base to use, or how to translate from one to another to compare
cannot do blindly, need common reference
issue with how to treat loop-induced processes (important for Higgs)
important to avoid absorbing NP effects in fit (e.g., proton PDF)
what can we learn from BSM?
should not simply look for all operators, when UV-complete models may suggest that
some operators should be irrelevant
important to get guidance from theory community
do's and do not's
how should EFT models be used? how mix dim-6 and dim-8 in fits?
how to cure unitarity problem?
breakdown at some critical energy
what technique to use to avoid having too optimistic results? how to combine
several analyses?
how to simulate samples?
generating samples in hypervolume of Wilson coefficients is very costly
question: can we generate linear, interference, quadratic terms separately?
up to what extent can we use event weights?
how much we lose using only differential distributions or STXS?
interplay with NxLO calculations
QCD and EW higher-order corrections could be of same size as EFT effects
need to agree on generation tools, prescriptions for combining calculations and
uncertainties
experimental fits
several tools exist, global fits involve thousand of events, hundreds of NPs
new tools may need to be developed
summary
rapidly increasing interest on systematic EFT fits
important to involve large number of final states
important interplay theory/experiments
William Shepherd: question on slide 14: some of these interferences (of two contributions with a dim-6 insertion each) are understood, and we know we need to make a choice: if you add these terms, need to add all dim-8 (not happening soon...)
translating from one base to another possible in theory community, and will change
the meaning of term included here
at any time, full EFT analysis with dim-8 must include all dim-6, but also all
dim-8 operators, and this does not seem to be happening soon
last term in expression is ill-defined; also, UFO w/ all dim-8 does not exist, and
people still working on all directions for dim-6
WS: slide 15: unitarity issue automatically handled by error prescription
described in my talk: those errors blow up at high energy, so that safe region is
always consistent within uncertainty (in other terms, high-energy region in which
unitarity fails has so large errors that it brings no information)
WS: slide 16: MadGraph allows this type of split production, so that one can use
weights
A: had talk from MadGraph authors, and indicated that MadGraph phase space work
only if one generates full process
WS: worth searching, but I suspect they mean that it is very inefficient in
producing, but not incorrect
Ayres Freitas: could produce events with large computing time (because produced few),
still better than re-doing full process simulation multiple times
A: question on slide 14; if one does double-insertions then results change
depending on base; do not see why should not be allowed to do?
WS: analogous to do 1-loop graph and interference with tree-level, get wrong
result and infinity if not properly done; EFT sorts of shields the infinity part,
which clearly tells us there is a problem
Philip Chang: double-insertions, what is implication to VBS?
WS: if doing VBS, can do double-insertion if you add only all dim-6, otherwise one
gets the wrong calculation
PC: does not seem we have the model with all of these; does this mean we cannot do
this properly?
WS: until we have the whole dim-8 story should only keep the Ci(6)/lambda^2 piece
PC: what can we say about dim-6 in VBS?
WS: only include terms linaer in dim-6 operators -- this can then be combined with all other dim-6 searches