<p>Dilepton interpretation</p>

<ul>
	<li>CMS does same as ATLAS: provide limits for different width hypotheses</li>
	<li>Assumptions: linear interpolation is OK (triangular contour finding)</li>
	<li>Three potential problems:
	<ul>
		<li>How does xsec scale inclusively? -&gt; we understand, solvable problem</li>
		<li>What is the effect of the width? -&gt; we have our list of limit values so that’s kind of fixed</li>
		<li>3rd issue: the 650 GeV window. do we have an acceptance effect from this? Probably only at 10%ish so we could just specify it as a concern and leave it at that</li>
	</ul>
	</li>
</ul>

<ul>
	<li>Assumptions about model in there? Just tune couplings til you get that width? What do you use for signal template in the limit setting</li>
	<li>Then changing couplings should just change cross section line, and we pick intersection points with the most appropriate width curve</li>
	<li>Large grid of NLO reference cross sections: for everything?</li>
	<li>OK using LO/LO * NLO xsec?</li>
	<li>So what would be involved in trying a new set of couplings? New NLO generation?
	<ul>
		<li>Hope is no - just take a ratio * NLO xsecs and say it should be good enough.</li>
	</ul>
	</li>
	<li>Oleg: why the 650 GeV window anyway? CMS team: historical, and hasn’t been a real problem yet</li>
	<li>Oleg: We’re limited to below 10% in width? Team: yes</li>
	<li>Phil: can we use an analytica form for NLO cross sections? Or do we really need to generate this grid? Andreas: are you asking about the necessity of doing this for coupling plot, or mass-mass plot? Phil: using full width and leading order ME ought to do OK, both off-diagonal and along the diagonal. Only issue is it’s a LO scaling and so you kind of need the k-factor to be consistent</li>
	<li>Phil: in case of dijet it’s just a bump</li>
	<li>Phil: production kinematics change as a function of widht for mono-X so you can’t just correct for that.</li>
	<li>Felix: topologies are different than for dijets. Phil: doesn’t matter though. If we have a k-factor for dilepton we use that just once.</li>
	<li>Felix: about the mass range. Could it just be replaced by something that scales with the mediator mass? Like larger at larger mass
	<ul>
		<li>David: yeah, makes a lot more sense to me</li>
	</ul>
	</li>
	<li>Felix: do leptonic tau decays have any effect in here? Might give you some wider peak. Andreas and David: think it’s completely ignored, not been looked at. Felix: but if you want to look at very wide mediators, might you not have to deal with this? Andreas: it should be a not massive effect given the BR… Oleg: you have even a bigger suppression factor because you’d need both taus decaying to the same flavour.</li>
	<li>Oleg: when you say you provide limits for different widths, is this for this model or something idealised? Andreas: the signal shape is analytical (Gaussian?) yes. Compared DMSimp to standard MC used for signal shapes and found they looked the same, so it was concluded it was close enough
	<ul>
		<li>Signal shape used is actually BW-crystal ball convolution</li>
	</ul>
	</li>
	<li>Andreas: for the purposes of this paper though, all these distinctions boil down to something we can generalise. So we should be aiming for some kind of common recipe</li>
	<li>David: how about distribution? What do we want?
	<ul>
		<li>Andreas: let’s make a list of all the individual pieces that we want. Library of functions giving width for a set of parameters.</li>
		<li>Let’s do it in GitHub issue tracking</li>
	</ul>
	</li>
</ul>

<p>&nbsp;</p>

<p>Spin-0</p>

<ul>
	<li>Phil: For the resonant search, the branching ratio for the scalar mediator to DM when the DM is light, is like 99%. So there’s no resonance search in that region. The resonance search does play a role off-shell but the searches all change once you’re above the ttbar mass and you start to get 4-tops and stuff. Below that it’s just a resonance to bbbar as dominant channel. So in that area we can just do the same rescaling as dijet.</li>
	<li>When we’re looking at invisible decays, it’s just going to be a triangle. There’s basically no width effects and basically always 99% decay to DM. So 1D should be fine.</li>
	<li>The one time we’d want to do 2D is if we wanted to do MET+X and di-b together on one plot. For the intrest of astrophysicists</li>
	<li>Couplings need to be really large to get the right relic density when the mediators are really light. And we’ve already excluded that.</li>
	<li>Technology: 1D plot for dijet and monojet. From there we can make a 2D plot that looks nice</li>
	<li>Andreas: but the 2D plots are just two independent triangles so that’s pretty easy. Phil: Yep. And then the fun loopy line from the relic.</li>
	<li>Re plot on slide 4: Phil says there really shouldn’t be structure here. It should be basically flat cause monojet is not sensitive to 5% width effects. I could only see this varying once you go past ttbar, and then we need to think about it some more. (Ask a theorist?)</li>
	<li>Oleg: would be nice to cast this onto these plots.</li>
	<li>Oleg: on the comment about the two triangles, it would still be interesting for showing what channels are relevant. Phil: yeah, the 2D plot is the only way to do that. Oleg: wouldn’t it then make sense to use something like the mass difference as one of the axes? Andreas: really depends on how interesting you think the diagonal is, and in his opinion that isn’t that interesting. Oleg: it just kind of shows anything in the triangle is trivial. But can do either way. Phil: mass difference plot is interesting in SUSY if it causes something interesting - here it’s not the case. Phil: you can get coannihilation along the diagonal but it’s a really fine tuned statement.</li>
	<li>Felix: what happens with one-loop-induced gg or diphoton coupling? Phil: you could have the diphoton. Starts pretty high though, so only sensitive for really big couplings. Felix: there are diphoton searches around the Higgs mass though…. Phil: yeah but the BR is suppressed quite a bit compared to Higgs. Should revisit it though. gg effect is there. Talked about going to NLO but it’s pretty new. Felix: you’d turn on all the other diboson at 1-loop too. Oleg: isn’t this stuff (EF operator limits) usually included in Higgs papers anyway? Phil: yeah but those are focused at 125. This would let you vary the mass of the scalar. [Higgsy discussion a bit over Kate's head ensues.]</li>
</ul>

<p>&nbsp;</p>

<p>Other:</p>

<ul>
	<li>How do CMS deal with ttbar width effects? David: just don’t interpret above the point where it matters. For full Run 2, planning to address it similarly to dilepton. Oleg: you can’t use Gaussian limits though. David: use the MC. Oleg: lots of complexities to that though. We will all have to look into it
	<ul>
		<li>
		<p>cut off above 5% for 36/fb</p>
		</li>
		<li>
		<p>extend above 5% for full Run 2, take MC mass</p>
		</li>
	</ul>
	</li>
	<li>Oleg: could we have our next meeting on ttbar interpretations in spin-1? We aren’t sure this applies here. But we can set up a meeting if useful</li>
</ul>

<p>&nbsp;</p>

<p>&nbsp;</p>

<p>Follow-up</p>

<p>&nbsp;</p>

<p>Who wants to write the dijet section? We should use a CMS example to balance out the ATLAS example in the mono-X section.</p>

<ul>
	<li>Andreas and David volunteer - thanks!</li>
</ul>

<p>Combined plots?</p>

<ul>
	<li>Boyu happy to do them</li>
</ul>

<p>Andreas: who cares about code and wants to do it? Most likely all of us in the meeting.</p>

<ul>
	<li>He’ll start a Github issue and we’ll go from there.</li>
</ul>