1
00:00:02,610 --> 00:00:10,080
No. Yeah, we hear you, sir audio. Okay,
excellent. Good. Okay, well, thank you

2
00:00:10,080 --> 00:00:13,740
very much for inviting me all the way to
Paris, the trip was a lot easier than I

3
00:00:13,740 --> 00:00:18,060
anticipated. I've crammed quite a lot in
here. So we'll go with something of speed.

4
00:00:18,090 --> 00:00:23,400
I do have a meeting setup at the end to
discuss in more detail with anyone who has

5
00:00:23,400 --> 00:00:28,500
questions. You know, as theorists, we can
talk forever. So feel free to come with

6
00:00:28,500 --> 00:00:32,940
detailed questions, I'll be happy to
discuss them. This first half of the talk

7
00:00:32,940 --> 00:00:38,460
follows the title that was given, which is
a bit before the column and also follows

8
00:00:39,240 --> 00:00:44,700
this article here, which is a little bit
aged, but such as life sometimes just a

9
00:00:44,700 --> 00:00:49,140
little bit on where we're coming from and
why we're going to use effective field

10
00:00:49,140 --> 00:00:53,520
theory to parameterize what we should
expect from any new physics. And this is

11
00:00:53,520 --> 00:00:58,290
the exact same thing we did when we talked
about Fermi's for fermion interaction for

12
00:00:58,290 --> 00:01:04,020
weak decay. We actually still use even
that effective theory today, it's really

13
00:01:04,020 --> 00:01:08,310
good for capturing physics in a particular
energy regime where you're talking about

14
00:01:08,310 --> 00:01:12,690
processes at a much lower energy than the
mass scale of the new physics. In this

15
00:01:12,690 --> 00:01:17,880
case, energy over and w is counting that
you do. And the key point is that you need

16
00:01:17,880 --> 00:01:23,220
to be able to systematically improve your
theory predictions, at least in principle,

17
00:01:23,430 --> 00:01:28,650
in order for the FT to be useful. The goal
is to give you a new expansion that allows

18
00:01:28,650 --> 00:01:33,330
you to control your calculations in a new
way that you couldn't do in the full UV

19
00:01:33,330 --> 00:01:39,390
theory. So when you apply these tools to
the standard model, you get a nice long

20
00:01:39,390 --> 00:01:43,140
list of operators, which I'm sure any
expert has already seen, but I'll let you

21
00:01:43,140 --> 00:01:47,370
stare at them for a little bit just to get
nervous if nothing else, and then

22
00:01:47,400 --> 00:01:52,650
furthermore, I'll move on to the next
slide full of more operators to be nervous

23
00:01:52,650 --> 00:01:57,810
about. And in principle, what we would
really like to do in this model is the

24
00:01:57,810 --> 00:02:03,390
standard model EFT is to construct The
coefficients of all of these operators,

25
00:02:03,600 --> 00:02:09,090
these are all the operators at dimension
six. And if we knew where we have to land

26
00:02:09,150 --> 00:02:16,200
in coupling space at dimension six, we
could then have whatever region it is map

27
00:02:16,200 --> 00:02:21,930
that into every UV level of interest, and
get constraints on UV models for

28
00:02:21,930 --> 00:02:28,680
relatively zero work. That would be
awfully nice. Now, why am I going to start

29
00:02:28,680 --> 00:02:31,650
by telling you about loops of things?
Well, it's because we go and make all

30
00:02:31,650 --> 00:02:34,470
these electroweak measurements we've been
hearing about for the rest of the session

31
00:02:34,860 --> 00:02:37,560
with just ridiculous exquisite precision.

32
00:02:38,819 --> 00:02:40,829
So this is just left data here.

33
00:02:42,270 --> 00:02:45,900
And all these measurements are ridiculous.
And we're working to improve on these in

34
00:02:45,900 --> 00:02:52,830
various ways. I like this one step on all
of the right hand column, our to loop

35
00:02:52,830 --> 00:02:55,920
calculations in the standard model,
because that's the level of calculation

36
00:02:55,920 --> 00:02:59,340
you you need in order to really actually
test the standard model with data this

37
00:02:59,340 --> 00:03:03,480
precise If you did a tree level fit to the
experimental boundary, you would conclude

38
00:03:03,480 --> 00:03:08,820
this general model is unquestionably
wrong. But we know that's not the case. It

39
00:03:08,820 --> 00:03:11,940
turns out those loop corrections are
really important to understand what the

40
00:03:11,940 --> 00:03:17,970
theory is telling you. And the same thing
is true. If you predict new physics

41
00:03:17,970 --> 00:03:23,100
contributions to anything precisely
measured like this, you need to make sure

42
00:03:23,100 --> 00:03:26,880
that your theory prediction is as accurate
as the boundary you're trying to enforce.

43
00:03:26,940 --> 00:03:32,520
Otherwise, you're going to come to the
wrong conclusions. So what is the theory

44
00:03:32,520 --> 00:03:39,660
later on our tree level EFT effects? Well,
if we imagine that we have a loop that

45
00:03:39,660 --> 00:03:44,940
looks like a 1% defect. And actually, we
have an additional source of 1% defects

46
00:03:44,940 --> 00:03:51,360
here as well, from extensions in the Higgs
back over the new physics scale, and that

47
00:03:51,390 --> 00:03:55,140
if we're lucky, and there's new physics
that we actually have a chance of finding

48
00:03:55,140 --> 00:04:02,820
evidence of at some point is order 1% as
well. The numerical coefficients that show

49
00:04:02,820 --> 00:04:07,620
up in front of these two things are not
known a priori. The sum after

50
00:04:07,620 --> 00:04:12,960
normalization has been done now has been
rejected multiple different ways. You

51
00:04:12,960 --> 00:04:21,000
could do RG improvement to capture the the
leading log effect. If we have sort of Lh

52
00:04:21,000 --> 00:04:27,270
C scale or just trans Elysee scale in
physics, running from here to here, stops

53
00:04:27,270 --> 00:04:32,220
being as impressive as an effect as it is
and say supersymmetry, where the upper

54
00:04:32,220 --> 00:04:37,200
here is almost a POC scale. When we're
running just from here's where the new

55
00:04:37,200 --> 00:04:42,450
physics is, and now we're in the FT we're
going one or one and a half or two decades

56
00:04:42,480 --> 00:04:47,970
in energy to get down to the electroweak
scale, those laws stop being as huge as

57
00:04:47,970 --> 00:04:53,100
they normally are. And purifying that
effects can be comparable. So we really

58
00:04:53,100 --> 00:04:57,570
need the full loop result. If we want to
say that we have things at the level of

59
00:04:57,570 --> 00:05:01,830
accuracy that we're used to thinking
about. Now I'm going to present your

60
00:05:01,830 --> 00:05:06,630
results that happen in a particular limit
where I've only turned on two couplings

61
00:05:06,960 --> 00:05:10,860
that UCD coupling compares. But that's a
lot of work. These two are easy. So that's

62
00:05:10,860 --> 00:05:17,160
what we decided to do. It turns out that
doing anything with higher order gauge

63
00:05:17,160 --> 00:05:20,280
corrections in the presence of these
dimensions, six operators gives you

64
00:05:20,280 --> 00:05:25,500
additional subtleties in fixing and doing
your age fixing and your calculation

65
00:05:25,530 --> 00:05:32,340
becomes a lot more painful need to be
extra careful. Here, if you don't allow

66
00:05:32,340 --> 00:05:35,760
additional gauge couplings, then your
gauge independence is a trivial statement.

67
00:05:37,290 --> 00:05:40,980
This is a decent first step toward a full
analog treatment to the problem. And what

68
00:05:40,980 --> 00:05:45,930
I'm looking at here in particular is z
production. Okay, so z goes on couplings.

69
00:05:46,830 --> 00:05:51,210
Now, from all those operators, you can
identify classes that end up contributing

70
00:05:51,270 --> 00:05:56,160
at this order. You're going to find that
all of them have tops internal to the loop

71
00:05:56,190 --> 00:06:01,560
or Hicks's internal to the loop. And
that's because it's only tough Because in

72
00:06:01,560 --> 00:06:06,630
The Hague self coupling that we've turned
on, there's nothing deep about that. So

73
00:06:06,630 --> 00:06:11,160
for fermion operators that include the top
will give us corrections to those on

74
00:06:11,160 --> 00:06:17,040
couplings to other types of fermions
scalar thermionic. Current operators do

75
00:06:17,040 --> 00:06:23,490
two things. One, they change at the simple
tree level from the insertion of Higgs

76
00:06:23,490 --> 00:06:30,990
Bez, the Z coupling to fermions. And they
also give rise to new interactions that

77
00:06:30,990 --> 00:06:37,380
include both a scalar and a fermion
current. Both of these effects are

78
00:06:37,380 --> 00:06:42,600
important to get our one loop calculation
and cdks, right. And you also get effects

79
00:06:42,660 --> 00:06:49,080
from mixings that sort of correct your
Weinberg in your weak mixing angle. So one

80
00:06:49,080 --> 00:06:54,750
thing that can happen is you can have an
operator that talks to W and B and mixes

81
00:06:54,750 --> 00:06:59,970
them together with the Higgs loop, or
anything of dipoles. They give you access

82
00:07:00,000 --> 00:07:02,280
To a mixing of W and D again.

83
00:07:05,399 --> 00:07:09,989
One other place where you might not think
you need to worry so much is in

84
00:07:09,989 --> 00:07:14,189
calculating all of your input parameters.
It turns out, this actually becomes a

85
00:07:14,249 --> 00:07:19,799
pretty subtle problem already a tree level
in the Standard Model EFT. And it only

86
00:07:19,799 --> 00:07:25,109
gets worse at loop level, you need to
choose what inputs you're using to define

87
00:07:25,109 --> 00:07:30,149
your theory parameters. So for instance,
when I talk about the top Yukon, or the

88
00:07:30,149 --> 00:07:34,409
Higgs valve, those are theory parameters,
we have not measured those things.

89
00:07:34,439 --> 00:07:42,389
Instead, we've measured things like the
top mass, or the Higgs mass, or G Fermi is

90
00:07:42,389 --> 00:07:47,669
actually where we get the Higgs bed from
generically. And if you want to measure g

91
00:07:47,669 --> 00:07:52,319
for me, but the Standard Model EFT is
turned on, it's possible that what you

92
00:07:52,319 --> 00:07:55,829
thought was the Higgs there is some
combination of the Higgs verb and the new

93
00:07:55,829 --> 00:08:01,799
Wilson coefficient. So you need to work
hard to extract That would be standard

94
00:08:01,799 --> 00:08:08,429
model Hicks from the new EFT contribution.
And once you start calculating at one

95
00:08:08,429 --> 00:08:13,049
loop, you need to do that mapping that one
little border as well. So there's a

96
00:08:13,049 --> 00:08:17,099
certain amount of work that goes into
this. We treated all of our listening

97
00:08:17,099 --> 00:08:21,359
coefficients as though they were input
parameters as well, to try to get a hold

98
00:08:21,359 --> 00:08:26,729
of something where we could start making
predictions. And when you run all of these

99
00:08:26,729 --> 00:08:30,269
calculations through and you have three
different people do them and you all argue

100
00:08:30,269 --> 00:08:33,449
about who's wrong until you decide
everyone was wrong. And here's the right

101
00:08:33,449 --> 00:08:37,079
answer. and convince yourself that you
have it right, you get results that look

102
00:08:37,079 --> 00:08:43,739
like this. Now, this is meant mostly to be
a scare slide. These are all the different

103
00:08:43,739 --> 00:08:52,139
contributions to the left on cars, rb.
Okay, there's a massive quantity of them

104
00:08:52,229 --> 00:08:57,779
is the point that first line, the easy ish
looking line, that's the three level and

105
00:08:57,779 --> 00:09:00,839
the other two equations have to be
something to do. together to give you a

106
00:09:00,839 --> 00:09:08,999
group level vfg contribution to delta
delta RCP. Now, why am I showing you this?

107
00:09:08,999 --> 00:09:12,119
Well, at the end of the day, you don't
need to know all these coefficients to get

108
00:09:12,119 --> 00:09:17,429
the main point. The main point I'm trying
to make to you is count the number of

109
00:09:17,429 --> 00:09:22,049
listen coefficients that you need. Our
analog corrections have introduced

110
00:09:22,049 --> 00:09:28,409
dependence on a whole spate of operators.
And at this level of precision, we can

111
00:09:28,409 --> 00:09:33,269
measure only five or six if you want to
turn on a forward backward. I still worry

112
00:09:33,269 --> 00:09:36,899
a little bit about getting getting things
kicked around. But I've been told by

113
00:09:36,899 --> 00:09:40,949
experts including irises here, that I
should be able to turn that on and not

114
00:09:40,949 --> 00:09:46,859
worry too much. So the moral of the story
there is obviously I can't constrain all

115
00:09:46,859 --> 00:09:51,359
those rules and coefficients with that
small number of observables. In fact, at

116
00:09:51,359 --> 00:09:57,449
tree level, there already were flat
directions in the Z polar observables. And

117
00:09:57,449 --> 00:10:02,369
we were able to constrain those albeit
very weakly With the left triple gauge

118
00:10:02,369 --> 00:10:08,939
coupling measurements with this increase
in the number of parameters, all of our

119
00:10:08,939 --> 00:10:13,409
electroweak precision data is not enough
to constrain the EFT to live in some

120
00:10:13,919 --> 00:10:19,559
closed region of the relevant parameter
space, there will always be directions

121
00:10:19,559 --> 00:10:26,429
that you could travel arbitrarily far and
not in any way affect these observables.

122
00:10:27,419 --> 00:10:31,919
So the lesson here is that once you get to
precisions, where you think loops matter,

123
00:10:32,429 --> 00:10:37,829
which from those Numerix that I showed you
is really sort of an order 10% correction

124
00:10:38,219 --> 00:10:45,299
to the naive tree level prediction, then
you can't constrain things stronger than

125
00:10:45,299 --> 00:10:52,559
that by electronic precision data alone.
So you should never be imagining from your

126
00:10:52,559 --> 00:10:56,909
precision measurements, that you have
something that you could effectively turn

127
00:10:56,909 --> 00:11:01,949
off because of the vision that it lifts.
That can't be an issue. While they did not

128
00:11:01,949 --> 00:11:07,289
measure one listen coefficient for
separately, that's not possible

129
00:11:07,319 --> 00:11:12,479
generically, in EFT. So we cannot turn off
different operators. And they're going to

130
00:11:12,479 --> 00:11:16,139
give us different effects that led them
they did it left. So there'll be separable

131
00:11:16,139 --> 00:11:24,479
there, we hope. This is one example of
potential theory errors. Having calculated

132
00:11:24,479 --> 00:11:28,109
a one loop correction to something that we
had done at tree level, it gives us a

133
00:11:28,229 --> 00:11:31,859
handle on how big our theory errors were
at tree level.

134
00:11:33,720 --> 00:11:40,050
And just to give a little motivation,
about theory errors more generally, we

135
00:11:40,050 --> 00:11:44,340
oftentimes don't care very much about
uncertainties in our signals, at least I

136
00:11:44,340 --> 00:11:50,190
don't because I was trained as a VSM
phenomenologist. So we tend to turn on the

137
00:11:50,190 --> 00:11:55,380
arm physics model and say, Oh, yes, I know
exactly what this new physics will do. It

138
00:11:55,380 --> 00:11:59,280
will give me a bump over there and that
distribution, you guys should go and look

139
00:11:59,280 --> 00:12:04,050
for it. And the basic idea behind that was
that we can go back and clean things up in

140
00:12:04,050 --> 00:12:07,740
our VSM theory, once you found the
particle, then we can start saying, Oh,

141
00:12:08,040 --> 00:12:11,820
well, that bump was actually off from
where I thought it would be five, five gv.

142
00:12:11,880 --> 00:12:18,090
And I can explain that at one or two loop
order, that's fine. But if you're in the

143
00:12:18,090 --> 00:12:22,440
business of precision measurements, which
is where we find ourselves now, in this

144
00:12:22,470 --> 00:12:28,560
electroweak seminar, but also in the life,
the life cycle of the LFC, generally, I

145
00:12:28,560 --> 00:12:32,760
would say, we're now in the business of
making precision standard level

146
00:12:32,760 --> 00:12:37,980
measurements to a large extent and trying
to find small deviations there. ignoring

147
00:12:38,010 --> 00:12:45,540
your errors is never a good idea, even if
their signal errors. And here's the basic

148
00:12:46,140 --> 00:12:52,950
motivation for why we should not do that.
Just because historically. Sorry, you say

149
00:12:52,950 --> 00:12:57,750
five minute warning. Okay. Historically,
it's been the case that when we do EFT

150
00:12:57,750 --> 00:13:02,790
analyses, if you do them ignoring these
errors, Then model builders like Bhaskar

151
00:13:02,790 --> 00:13:07,380
here, who I use as a guinea pig, I really
like that start. This is a very honest

152
00:13:07,380 --> 00:13:13,680
statement from him about the history of
DFT analyses, he can build a model that

153
00:13:13,680 --> 00:13:18,180
violates the bounds that I have without
violating the data I got the bound from is

154
00:13:18,180 --> 00:13:24,030
the claim of Unix. And this is generically
true. So we need to fight back against

155
00:13:24,030 --> 00:13:28,140
that somehow. So for this little bit of
the talk, I want to pick based on these

156
00:13:28,140 --> 00:13:34,260
two articles, that 226 used to tell the
plus or minus one but now I've become more

157
00:13:34,260 --> 00:13:40,530
confident, so I took the robot off. So how
do we go about building a collider search?

158
00:13:41,400 --> 00:13:45,810
We try to run things through Monte Carlo
tools and constrain what comes out. There

159
00:13:45,810 --> 00:13:51,150
are nice tools for this in standard model,
EFT, there's a package that you can plug

160
00:13:51,150 --> 00:13:55,860
in directly into Minecraft and get out
your results. The greatest challenge to

161
00:13:55,860 --> 00:14:00,360
these searches is the concern about
whether the DFT is actually consistent Or

162
00:14:00,360 --> 00:14:06,270
not. The way to make sure that you are
using the FT honestly is to actually

163
00:14:06,270 --> 00:14:12,600
estimate your errors in that new
perturbation series though. Ensuring that

164
00:14:12,600 --> 00:14:16,860
this thing is consistent means making a
sample of what's going on at the next

165
00:14:16,860 --> 00:14:21,060
order in your perturbation series. And if
that's bigger than your signal, which is a

166
00:14:21,060 --> 00:14:26,100
first order, then you know you have a
problem. So I'll look at this in the

167
00:14:26,100 --> 00:14:32,370
context of leptons in particular, there
are two types of energy contributions to

168
00:14:32,370 --> 00:14:37,530
the electrons from the Standard Model DFT.
There are direct for fermion operators,

169
00:14:37,530 --> 00:14:40,890
they give you amplitudes that grow with
energy, which is what we're used to

170
00:14:40,890 --> 00:14:44,250
thinking about, but you can actually also
shift the z components like I was talking

171
00:14:44,250 --> 00:14:49,920
about earlier. We do a forward backward
separation. Everyone here is super

172
00:14:49,920 --> 00:14:55,560
familiar with this, you define forward as
more forward in either direction. And you

173
00:14:55,560 --> 00:15:00,270
have here the linear combinations of
Wilson coefficients that give you x Tree

174
00:15:00,270 --> 00:15:04,500
level forward and backward contributions
of two different types either energy

175
00:15:04,500 --> 00:15:08,850
growing at the top of the slide, or shifts
in the Z couplings and proportional

176
00:15:08,850 --> 00:15:15,660
external rates at the bottom of the slide.
The way to deal with your uncertainties

177
00:15:15,660 --> 00:15:21,120
from the EFT expansion is to be honest
about estimating your theory errors. We

178
00:15:21,120 --> 00:15:25,410
have completely unknown effects from
dimension eight operators in the Standard

179
00:15:25,410 --> 00:15:31,260
Model EFT they are in order one overland
to the fourth, and our signal function is

180
00:15:31,260 --> 00:15:36,720
at order one over lambda squared. Now, if
we want to estimate something and order

181
00:15:36,720 --> 00:15:40,020
one are willing to force that we don't
really know we can take something we do

182
00:15:40,020 --> 00:15:46,530
know in order to one overlain before and
yes, or just hope that that's a decent

183
00:15:46,530 --> 00:15:52,290
approximation to what happens generically
in order whatever line before, so I

184
00:15:52,290 --> 00:15:56,970
calculate the cross section from dimension
six squared, which is something that we

185
00:15:56,970 --> 00:16:02,490
get for free inside Smith sim For inside
really any Monte Carlo that's doing

186
00:16:02,490 --> 00:16:10,440
Standard Model DFT. And I change its
leading coefficient to lock myself up the

187
00:16:10,440 --> 00:16:14,820
effect not only of the dimension six
squared piece, but also have some number

188
00:16:14,820 --> 00:16:18,300
of dimension eight operators interfering
with the Standard Model amplitude here.

189
00:16:19,950 --> 00:16:23,970
We D correlate this between the bins of
our analysis because we don't want to

190
00:16:23,970 --> 00:16:27,990
assume that I mentioned six squared
describes the exact shape just gives us

191
00:16:27,990 --> 00:16:34,500
the overall normalization. And we insist
that our dimension a Wilson coefficient is

192
00:16:34,500 --> 00:16:41,610
at least the order one, or C six,
whichever is larger. And then we sub that

193
00:16:41,610 --> 00:16:45,990
in quadrature, with our other error
sources, and doing that we can mock up

194
00:16:46,050 --> 00:16:52,260
expectations of how far we should be able
to get. So these three curves, blue,

195
00:16:52,260 --> 00:16:58,500
orange, and green are showing us
100 303,000 inverse femto bonds of data

196
00:16:58,710 --> 00:17:02,940
and what we can learn There are four
directions in this parameter space that

197
00:17:02,940 --> 00:17:09,600
are interesting. We can with a large
amount of data close an ellipse in this

198
00:17:09,630 --> 00:17:14,220
pure forward and backward, which is to say
energy growing contributions to forward

199
00:17:14,220 --> 00:17:18,900
and backward distributions. The tight
constraint in this space is the total

200
00:17:18,900 --> 00:17:22,560
cross section direction which is almost
horizontal but not quite backward, it

201
00:17:22,560 --> 00:17:28,770
turns out as weaker than forward from this
left. And perpendicular to that here is

202
00:17:28,770 --> 00:17:34,320
the forward backward asymmetry direction.
Here you get some more interesting stuff

203
00:17:34,320 --> 00:17:37,920
with shifts, it turns out they're still a
direction you can run forever, your

204
00:17:37,920 --> 00:17:43,140
forward and backward asymmetry is more or
less this way. And you can see out here

205
00:17:43,140 --> 00:17:46,620
that at some point when you turn your
Wilson coefficients up way too high, and

206
00:17:46,620 --> 00:17:51,330
this is comically high for a listen
coefficient must be said. You lose your

207
00:17:51,330 --> 00:17:55,860
constraint again, because your theory and
errors become so large that you can't

208
00:17:55,860 --> 00:18:00,390
meaningfully constrain this distribution.
There's two other slices you can Easily

209
00:18:00,390 --> 00:18:06,810
taking the space as well. Both of the
forward directions, both of the backward

210
00:18:06,810 --> 00:18:12,960
directions energy growing and shrinking
perspective. So this is the kind of thing

211
00:18:12,960 --> 00:18:18,300
you can do if you're taking your EFT
seriously. And the nice thing is that

212
00:18:18,300 --> 00:18:24,090
these bounds once you've included that
theory, or should actually hold for UV

213
00:18:24,090 --> 00:18:28,620
models, which is to say they should be
useful to someone, our previous sort of

214
00:18:28,650 --> 00:18:33,960
EFT bounds gave us nice things that we
could constrain colliders. But ultimately,

215
00:18:33,960 --> 00:18:37,800
those constraints simply weren't
interesting to what should be our biggest

216
00:18:37,800 --> 00:18:41,490
consumers, which is the model builders,
who should be able to take our precision

217
00:18:41,490 --> 00:18:45,750
measurements and say, Do these constrain
my model or not in a relatively

218
00:18:45,750 --> 00:18:52,410
straightforward way without having to do a
complete recast of the analysis. So the

219
00:18:52,410 --> 00:18:56,160
moral of the story is that with all of the
excellent data that we have available, and

220
00:18:56,160 --> 00:19:00,000
all of the much, much more that we're
going to make, we really need to respect

221
00:19:00,000 --> 00:19:05,670
That data and do our new physics
predictions with as much precision as we

222
00:19:05,670 --> 00:19:13,800
can. The z poll data cannot close all of
the parameters that enter it at one loop

223
00:19:13,800 --> 00:19:18,570
order that this is a completely
unavoidable consequence of field theory.

224
00:19:19,320 --> 00:19:24,570
So, we need to go beyond that we need to
use a total real global analysis to

225
00:19:24,570 --> 00:19:30,270
properly constrain the Standard Model EFT
without making up assumptions. developing

226
00:19:30,270 --> 00:19:34,680
more observables that can be consistently
constrained in this way is a important

227
00:19:34,680 --> 00:19:39,630
future path here digests and die leptons
are basically done other directions are

228
00:19:39,630 --> 00:19:45,420
being worked on but there's an awful lot
to do this technology to to do these

229
00:19:45,420 --> 00:19:51,720
errors is 100% mad graph a minimal so this
could be picked up and done in house by

230
00:19:51,720 --> 00:19:57,210
experiments and actually it should, for
one important reason. If you take a look

231
00:19:57,240 --> 00:20:01,800
at the digest stuff in this in the backup
there information, you'll find that that

232
00:20:01,800 --> 00:20:05,760
analysis actually draws its constraining
power from a very different region of

233
00:20:05,760 --> 00:20:11,160
phase space than you guess if you don't
include theory errors. So I would

234
00:20:11,160 --> 00:20:16,710
recommend checking that out. It really
changes your search design significantly

235
00:20:16,710 --> 00:20:21,180
once you include these theory errors, this
is not just a trivial rescaling of your

236
00:20:21,180 --> 00:20:21,600
bound.

237
00:20:22,950 --> 00:20:26,910
The basic takeaway message from this talk
should be that setting shifts in

238
00:20:26,910 --> 00:20:31,320
electroweak observables to zero or even
worse setting operators that shift

239
00:20:31,320 --> 00:20:36,270
electroweak observables to zero for the
purpose of further searches, breaks model

240
00:20:36,270 --> 00:20:42,420
independence, it does not give you real
results. Neglecting theory errors in these

241
00:20:42,450 --> 00:20:47,700
EFT interpretations gets they're ignored
by model builders who really are meant to

242
00:20:47,700 --> 00:20:53,580
be the customers for this. So we must stop
ignoring those theory errors. We need to

243
00:20:53,580 --> 00:21:00,450
make results that they cannot evade by
using an honest theory error estimate. We

244
00:21:00,450 --> 00:21:05,430
need to then produce a sales pitch to them
that says, Look, this actually is better.

245
00:21:06,480 --> 00:21:11,160
And we need to push back against any
Cavalier claims that new models can always

246
00:21:11,160 --> 00:21:16,770
be built that vft results once we are
taking our error seriously. We need to go

247
00:21:16,770 --> 00:21:20,670
out there and say no, that's what it used
to be true. I understand where you're

248
00:21:20,670 --> 00:21:25,830
coming from. But we have fixed that
problem in our EFT now actually tells us

249
00:21:25,830 --> 00:21:30,240
something useful about whether your model
is there or not. In short, we need to make

250
00:21:30,240 --> 00:21:35,610
sure that best our ends up being wrong
about this in our new EFT interpretations,

251
00:21:35,790 --> 00:21:42,870
because otherwise, the CFT interpretations
simply are not very useful. Thank you very

252
00:21:42,870 --> 00:21:45,900
much for your attention, and I hope you'll
come and visit with me in the coffee

253
00:21:45,900 --> 00:21:46,290
break.

254
00:21:48,060 --> 00:21:51,210
Thanks a lot for the really illuminating
talk.

255
00:21:52,290 --> 00:21:54,720
Can I ask people to raise hands if they
have questions?

256
00:22:03,000 --> 00:22:05,250
Si RS has a question to go ahead.

257
00:22:05,970 --> 00:22:11,640
Yeah. Thanks. Could you walk me once more
through the plots you show that the end so

258
00:22:11,640 --> 00:22:16,650
so with the different colored curves? How
do I see the theory hours here?

259
00:22:17,520 --> 00:22:21,840
So one place where you can feel the theory
errors, they're not explicitly present in

260
00:22:21,840 --> 00:22:25,800
these plots, because these are just
constrained plots, but you can feel them

261
00:22:25,800 --> 00:22:30,540
happening at very large Wilson
coefficients. So this between these two

262
00:22:30,540 --> 00:22:37,230
green curves, for instance, is a region
that we will be able to constrain at large

263
00:22:37,380 --> 00:22:43,080
luminosity that's our 3% of our
constraint. So zeros you're not

264
00:22:43,080 --> 00:22:47,100
constrained because the standard model is
what we assumed as our signal. Sure, but

265
00:22:47,100 --> 00:22:50,850
very large Wilson coefficients also are
not constrained because the theory error

266
00:22:50,850 --> 00:22:52,530
has started to grow quadratically in

267
00:22:52,530 --> 00:22:57,810
that Wilson coefficient Oh, sure. I mean,
I understand it extremely Legion, but for

268
00:22:57,840 --> 00:23:02,100
for the Legion dead descended on zero. Do
you have any plots where you can compare

269
00:23:02,100 --> 00:23:06,420
the curves, whether you include those
dimension athili, Atlas or not.

270
00:23:06,990 --> 00:23:11,760
So on this curve, you can see the effect
of the theory errors very clearly in the

271
00:23:11,790 --> 00:23:17,490
in the non three inverse data markers. We
don't actually have a plot where I have

272
00:23:17,490 --> 00:23:22,980
ignored them. But you never get a behavior
like this where your urbane actually

273
00:23:22,980 --> 00:23:26,490
grows, as you get further away from the
standard model. If you don't include a

274
00:23:26,490 --> 00:23:31,800
theory, you will get a normal ellipse and
it will just be bigger. And so why does

275
00:23:31,800 --> 00:23:38,160
this Why me when theory error, because
eventually we do actually managed to get

276
00:23:38,160 --> 00:23:43,710
enough in a symmetry there. This is
actually still not an ellipse that has

277
00:23:43,710 --> 00:23:47,580
some whitening behavior out here. This is
not a perfect ellipse that you would get

278
00:23:47,580 --> 00:23:52,410
from just a linear fit. But we ultimately
get enough statistics that we can use the

279
00:23:52,410 --> 00:23:56,250
forward backward symmetry to close that to
some extent, if I assumed this plot

280
00:23:56,250 --> 00:24:02,010
layout, there's also an opening of green
down here again, Just like you see here at

281
00:24:02,010 --> 00:24:04,350
extreme regions. Thanks.

282
00:24:08,069 --> 00:24:09,569
Are there any further questions?

283
00:24:19,439 --> 00:24:25,559
If not, I propose that we thank William
again. And we can also thank all of our

284
00:24:25,559 --> 00:24:31,409
speakers. And again, I realized that with
the current setup, I can't ask everyone to

285
00:24:31,409 --> 00:24:35,009
unmute and give a round of applause. But
I'm sure that everyone is extremely

286
00:24:35,009 --> 00:24:39,269
grateful for the fantastic talks we've
heard today from all our speakers. So

287
00:24:39,269 --> 00:24:40,259
thank you very much.

288
00:24:40,980 --> 00:24:44,670
Sure, I speak for the other speakers as
well. When I say please do join the zoom

289
00:24:44,670 --> 00:24:47,490
rooms we made. If you have any more
detailed questions that you'd like to

290
00:24:47,490 --> 00:24:49,290
spend time on. We'll be happy to discuss
them.

291
00:24:51,060 --> 00:24:52,650
Yes, please, please do

292
00:24:54,420 --> 00:24:56,040
join in chat during the coffee

293
00:24:56,040 --> 00:25:02,730
break. And I'll just note that our next
Parallel session meeting is on Wednesday.

294
00:25:02,850 --> 00:25:04,260
So we look forward to