HIP-2016-07/TH

Dark-Photon searches via Higgs-boson production at the LHC

Sanjoy Biswas,1 Emidio Gabrielli,2, 3 Matti Heikinheimo,4 and Barbara Mele5
1

arXiv:1603.01377v1 [hep-ph] 4 Mar 2016

KIAS, 85 Hoegi-ro, Dongdaemun-gu, Seoul 130-722, Republic of Korea

2
Dipart. di Fisica Teorica, Universit`
a di Trieste, Strada Costiera 11,
I-34151 Trieste, and INFN, Sezione di Trieste, Via Valerio 2, I-34127 Trieste, Italy
3
NICPB, Ravala 10, 10143 Tallinn, Estonia
4
Helsinki Institute of Physics, P.O.Box 64, FI-00014 University of Helsinki, Finland
5
INFN, Sezione di Roma, P. le A. Moro 2, I-00185 Rome, Italy
(Dated: March 7, 2016)
Dark photons mediating long-range forces in a dark sector are predicted by various new physics
scenarios, and are being intensively searched for in experiments. We extend a previous study of a
new discovery process for dark photons proceedings via Higgs-boson production at the LHC. Thanks
to the non-decoupling properties of the Higgs boson, BR(H
) values up to a few percent
are possible for a massless dark photon, even for heavy dark-sector scenarios. The corresponding
signature consists (for a Higgs boson at rest) of a striking monochromatic photon with energy
E = mH /2, and similar amount of missing energy. We perform a model independent analysis at
the LHC of both the gluon-fusion and VBF Higgs production mechanisms at 14 TeV, including
parton-shower effects, and updating our previous parton-level analysis at 8 TeV in the gluon-fusion
channel by a more realistic background modeling. We find that a 5 sensitivity can be reached in the
gluon-fusion channel for BR(H
) ' 0.1% with an integrated luminosity of L ' 300 fb1 . The
corresponding VBF reach is instead restricted to 1%. Such decay rates can be naturally obtained
in dark-photon scenarios arising from unbroken U (1)F models explaining the origin and hierarchy
of the Yukawa couplings, strongly motivating the search for this exotic Higgs decay at the LHC.

I.

INTRODUCTION

Although long awaited, no conclusive signal of New

Physics (NP) at the TeV scale showed up in the first
run of LHC at 7 and 8 TeV, and in the initial phase of
Run 2 at 13 TeV either. As a consequence, consents are
growing up around the idea that a new and unexplored
dark (or hidden) sector, weakly coupled to the standard
model (SM), is responsible for the observed dark matter
(DM). The latter, which is five times more abundant in
the Universe than ordinary baryonic matter [1], remains
still a mystery, with its constituents and detailed properties yet unknown. A dark sector could then have its internal structure and interactions, in complete agreement
with present astroparticle and cosmological observations.
It is also conceivable that a hidden sector could contain an extra long-range force mediator among the dark
particles. The most simple example is provided by a new
unbroken U (1) gauge group, predicting a dark (or hidden) photon in its spectrum [2]. Dark-photon scenarios
have been extensively considered in the literature in the
framework of NP extensions of the SM gauge group [3][7].
In cosmology, dark photons may help to solve the
small-scale structure formation problems. Massless dark
photons interacting with dark matter [8] can lead to the
formation of dark discs of galaxies [9], analogously to the
galaxy structure formation in the ordinary universe, or
to the collisional behavior of dark matter in mergers of
galaxies and galaxy clusters [10]. In astroparticle physics,
dark photons may induce the Sommerfeld enhancement
of DM annihilation cross section needed to explain the
PAMELA-Fermi-AMS2 positron anomaly [11] , as well

as assisting light-DM annihilations to make asymmetric

DM scenarios phenomenologically viable [12]. In some
scenarios, massive dark photons have also been considered as potential dark-matter candidates, with dedicated
experiments looking for their direct detection in the mass
range from a few eV up to 100 KeV [2],[13].
Most of present astrophysical and accelerator constraints apply to massive dark photons, and can be
evaded in the case of a massless dark-photon scenario, allowing for potentially large dark-photon couplings in the
dark sector. Indeed, in the massless case, on-shell dark
photons can be fully decoupled from the SM quark and
lepton sector [3], which is not true for the massive case
due to the potential tree-level mixing with ordinary photons of massive dark photons. This property can lead to
observable new signatures at colliders for massless dark
photons, provided there is a messenger sector letting the
SM and dark sector communicate.
Recently, a massless dark photon scenario has been
foreseen in the framework of a theoretical proposal aimed
to naturally solve the flavor hierarchy problem [14]. This
model predicts a new Higgs-boson decay channel into a
photon () and a massless dark-photon (
)
H ,

(1)

which is induced at one-loop. The final gives rise to

missing energy and momentum in the detector, leading to an exotic resonant mono-photon signature at the
LHC. The latter features a distinctive photon transversemomentum (pT ) distribution peaked around mH /2, same
/ T ) distribution, and
for the missing transverse-energy (E
a
transverse-mass distribution peaked around mH .
This exotic signature has been recently analyzed for the

2
first time in [15], in a model independent way. In particular, a parton-level analysis at the 8-TeV LHC has been
performed for the main Higgs-boson production channel,
namely the gluon fusion process. Using the full 8-TeV
LHC data set, a 5 sensitivity for a Higgs H
branching ratio (BR) down to 0.5% has been obtained.
These results have been worked out under assumptions
that might underestimate one of the main reducible backgrounds, given by a photon plus jet (j), and did not include parton-shower effects.
The purpose of the present paper is twofold. On the
one hand, we upgrade our previous 8-TeV analysis of the
H
decay in the main Higgs production channel
[15] by including parton-shower effects to the previous
parton-level Montecarlo study of the signal and of SM
backgrounds. We also consider a more realistic background modeling, based on recent experimental studies of events with a photon plus missing energy at the
LHC [16]. We then extend the analysis to the upgraded
nominal LHC energy of 14 TeV. On the other hand, we
analyse for the first time an alternative signature coming
from the H
decay for a Higgs boson produced via
the Vector-Boson-Fusion (VBF) mechanism. The gluonfusion channel will turn out to be the most sensitive to
BR(H
) (BR ). Nevertheless, we will see that the
VBF process could significantly contribute to either a
measurement or a determination of upper bounds of the
decay rate of the exotic Higgs decay into a dark photon,
possibly giving an independent confirmation of the signal in case of a positive observation in the gluon-fusion
process.
The plan of the paper is the following. In Sec. II, we describe a theoretical framework that might give rise to the
H
signature with observable rates. In Sec. III (A),
we study the potential of the gluon-fusion Higgs production mechanism at the LHC for constraining the H

rate, by a detailed analysis of both the signal and main

backgrounds. The same is done for the VBF production
mechanism in Sec. III (B). In Sec. IV, we summarize our
results and conclude.

II.

THEORETICAL FRAMEWORK

We now provide a model-independent parametrization

of the amplitude for the H
channel, and then discuss the corresponding BRs range that can be expected
in a class of NP models that might explain the origin and
hierarchy of the Higgs Yukawa couplings.
The H
amplitude can be parametrized in a
model-independent way by requiring gauge invariance,
as follows
M =

1
T (k1 , k2 )1 (k1 )2 (k2 ),

(2)

where is the effective scale associated to the NP,

T (k1 , k2 ) g k1 k2 k2 k1 , and 1 (k1 ) and 1 (k1 )

are the photon and dark-photon polarization vectors, respectively. The corresponding decay width is given by
(H
) = m3H /(32 2 ).

(3)

A massless dark photon does not couple to SM particles

at tree level. One can then assume that the effective
amplitude in Eq. (2) arises at one loop by the exchange
inside the loop of dark and messenger fields, the latter
being charged under both SM and extra U (1)F gauge interactions. By naive dimensional analysis, one expects
the scale be proportional to the mass of the heaviest particle running in the loop, presumably related to
the dark-sector. If this were the case, the chances of observing this process at the LHC would be dramatically
limited to a light dark-sector scenario, which is a quite
strong requirement. On the contrary, due to the nondecoupling properties of the Higgs boson, this scale could
be proportional to the Higgs vacuum expectation value
(vev) (similarly to what happens for the H , Z, gg
decay rates), which would allow for potentially large rates
regardless of the characteristic mass scale of the dark
sector. This crucial property turns out to hold in the
framework of the model proposed in [14], as has been
explicitly verified in [15]. This framework can then be
used as a benchmark model for computing all the relevant quantities for predicting the Higgs decay rates into
dark photons.
In [14], the Flavor and Chiral Symmetry breaking
(ChSB) are assumed to be generated in a dark sector,
and transferred to the Higgs Yukawa sector at one loop
via Higgs-portal type scalar-messenger fields. A new exact U (1)F gauge symmetry in the dark sector produces
via a nonperturbative mechanism an exponential spread
in the Yukawa couplings Yi (with i a flavor index), providing a natural solution to the SM Flavor hierarchy problem. Apart from the gauge boson of the unbroken U (1)F
gauge group (the massless dark photon), the dark sector
consists of SM-singlet massive dark fermions, Qi , a sort
of rescaled replica of SM fermions. The requirement that
the gauge sector is unbroken allows dark fermions, which
have U (1)F charges, to be stable and thus potential darkmatter candidates. In addition to the dark sector, there
are scalar messenger fields (with the same quantum numbers as the squarks and sleptons of supersymmetric models), which communicate the ChSB and Flavor breaking
from the dark sector to the Yukawa couplings.
By restricting, for instance, only to the contribution
of colored messenger fields, the effective scale can
then be exactly derived in the low energy limit [15]. In
particular, for a universal average messenger mass m,
one
obtains (up to corrections of order o(m2H /m
2 ))

1
R
2
=
,
(4)

6v 1 2
P3
where v is the Higgs vev, R = Nc i=1 (eU qUi + eD qDi ),
with qUi , qDi the U (1)F charges in the up and down sec-

3
tors, eU = 32 , eD = 13 the corresponding e.m. charges,
the EM fine structure constant, and Nc = 3 is the number
of colors. Also, = /m
2 , with = S v parametrising the left-right mixing of the messengers scalars, and
S is the vev of a singlet scalar field. The latter spontaneously breaks the H H parity symmetry needed
to forbid Higgs Yukawa interactions at tree level, since
Yukawa couplings are generated radiatively [14].
The non-decoupling properties of the Higgs boson
clearly show up in Eq.(4). Indeed, the effective scale
turns out to be proportional to the Higgs vev, that is
it tends to a finite value in the limit m
(for fixed
mixing parameter < 1). As stressed in [15], this is a
general property of the Higgs boson, and does not depend
on the peculiar structure of the model in [14], provided
a messenger sector letting the SM and the dark sector
communicate exists.
The same off-shell fields contributing to the H

decay amplitude at one loop can induce the H

transition to two dark photons (that increases the invisible Higgs decay width), and also give extra contributions
to the H , Z, gg SM decay rates. By parametrizing these effects in a model independent way, BR values up to 5% can be allowed, while respecting all other
LHC constraints [15]. Such large BR values for H

are natural in the framework of the model in [14] (see

also [17] for further more model-dependent predictions).
Such high decay rates strongly motivated the study
of the Higgs production followed by the H
decay at the LHC Run-1 energy and integrated luminosity
[15]. The corresponding signature is indeed quite distinctive, with an almost monochromatic and massless invisible (dark-photon) system and equally monochromatic
photon, jointly resonating at the Higgs mass.
In the present study, we will extend our previous analysis to the 14-TeV LHC setup, upgrading different aspects
of the study of the main gluon-fusion production channel,
and including VBF Higgs production in order to improve
the final sensitivity to the H
signature.

III.

PHENOMENOLOGICAL ANALYSIS
A.

Gluon-fusion channel

We start by extending our previous LHC analysis of

the gluon-fusion process at 8 TeV [15] to 14 TeV, improving the treatment of both the signal and the main
SM backgrounds. A crucial point in the refinement of
most important backgrounds will be the use of recently
published experimental data by the CMS collaboration
[16], where the relevant SM backgrounds are measured
and reported. We will model our background accordingly. All this will result in a higher reliability of our
signal and background estimates, that will anyhow substantially confirm our previous results on discovery potential based on a more naive analysis.

The process pp H
, where the Higgs is produced in the gluon-fusion channel, is characterized by
a single photon recoiling against missing transverse momentum. In our previous analysis we outlined a search
strategy for this process, based on the following requirements (now slightly updated to take into account smearing effects discussed in the following):
one isolated (R > 0.4) photon with pT > 50 GeV,
and | | < 1.44;
/T >
missing transverse momentum satisfying E
50 GeV;
transverse mass in the range 100 GeV < MT <
130 GeV;
no isolated leptons.
The transverse-mass variable is defined as MT =
q
/ T (1 cos ), where is the azimuthal dis2pT E
tance between the photon transverse momentum pT , and
/T.
the missing transverse momentum E
The main SM background for the above selection criteria is pp j, where the missing transverse momentum can arise from a) neutrinos following heavy-flavor
decays in the jet, b) mismeasurement of the jet energy,
and c) very forward particles escaping the detector. To
the latter channel contributes also pp jj, whenever
one of the jets is misidentified as a photon. We assume
the corresponding mis-tagging probability to be 0.1%.
Also, a photon identification efficiency of 90% is adopted
throughout this analysis. In our previous study [15], the
hadronic SM background was estimated at parton level in
a quite crude way, by treating any parton with || > 4.0
as missing energy.
The CMS analysis of the data set at 8 TeV in [16] assumes event selection criteria quite similar to the above,
in order to search for an exotic three-body decay of the
Higgs boson into a photon and two invisible particles.
Unfortunately, the CMS analysis imposes an upper limit
of 60 GeV on the photon transverse momentum, cutting
away an important fraction of the signal region for the
two-body decay of interest here (for which pT <
mH /2).
However, due to the similarity of the residual event selection criteria in the two analysis, the continuous SM
backgrounds are expected to be comparable. As a consequence, we decided to model our QCD background according to the CMS measured distributions, benefitting
from the highly optimized experimental procedure for the
missing transverse-momentum determination. This will
lead to a much improved reliability of our background
estimate in the gluon-fusion channel.
We started by simulating the j and dijet backgrounds
with MadGraph5 aMC@NLO (v2.2.2) [18], interfaced
with PYTHIA (v6.4.28) [19], hence including initial- and
final-state radiation (ISR and FSR), hadronization and
detector-resolution effects in the present updated analysis. We have generated event samples both at 8 TeV and

4
14 TeV. We have then matched our 8-TeV samples to
the event yield corresponding to the SUSY benchmark
event selection criteria reported in the CMS analysis [16].
This matching results in k-factors connecting our simulated samples to experimental data at 8 TeV. We find
k = 0.11 for the j background, and k = 0.058 for the
j background. The order-of-magnitude reduction in
the background estimate reported by CMS as compared
to our simulation is to be understood as a result of CMS
advanced strategies for reducing event yields arising from
mis-measured missing transverse momentum in hadronic
events, as detailed in [16]. It is beyond the scope of this
work to attempt to exactly reproduce the CMS analysis.
Instead, we assume that the CMS optimization strategy
works with comparable efficiency also in 14-TeV collisions, and that the corresponding reduction of the 14TeV hadronic SM backgrounds is reliably captured by
rescaling our simulated samples with the same k factors
obtained from the 8-TeV matching.
We also upgraded the simulation of H
signal events by including the ISR effects. Accordingly,
we simulated Higgs production in association with either one or no jets with ALPGEN (v2.14) [20], interfaced with PYTHIA for jet-parton matching, hadronization and detector-resolution effects (see Sec. III (B) for
the jet definition and other simulation details).
The corresponding smearing in the pT and MT spectra for the H
signal is shown in Figure 1. There,
the two categories corresponding to no extra jets and
one extra jet accompanying the Higgs signal are shown
separatly, along with the distributions for the hadronic
backgrounds coming from j production, and dijet production followed by j mistagging. The latter distributions are obtained with a nominal cut on the photon
transverse momentum, pT > 10 GeV, and pjT > 10 GeV
on fake jet in the dijet analysis.
Besause of initial-state-radiation and detectorresolution effects, a better sensitivity for the signal is
obtained by relaxing the maximum value of the photon
transverse-momentum cut, and increasing the transverse
mass window from 100 GeV < MT < 126 GeV to
100 GeV < MT < 130 GeV with respect to [15].
The main electroweak background consists of the channels pp W e, where the electron is misidentified as
a photon, pp W ( `), for ` outside charged-lepton
acceptance, and pp Z( ). We have simulated
these processes at parton level according to the analysis
in [15] , using a e conversion probability of 0.005 for
the first process.
In Table I, one can find a summary of the cross sections
times acceptance (in fb) for the signal and backgrounds at
8 TeV and 14 TeV for the gluon-fusion process, assuming
BR =1%, and obtained as discussed above.
With the 20 fb1 data set at 8 TeV, our improved analysis gives a 5 discovery reach at BR ' 4.8 103 ,
compatible with our previous estimate [15]. The present
more-realistic event simulation was expected to deteriorate the capability of separating signal from background.

0.5

H (no jet)
Hj
j
jj

0.45
0.4
0.35
0.3
1 dN
N dp

0.25
0.2
0.15
0.1
0.05
0
0

50

100

150

200

250

pT

0.3

H (no jet)
Hj
j
jj

0.25
0.2
1 dN
N dM T 0.15

0.1
0.05
0
0

50

100

150

200

250

MT

FIG. 1: Photon pT (upper plot) and transverse-mass (lower

plot) distributions for the H
signal in the gluon-fusion
/ T final
process, and for SM backgrounds, for inclusive + E
states with no isolated leptons. The effect of extra radiation
on the signal events is also depicted. All distributions are
normalized to unity.

This effect has been actually mostly compensated by the

advanced optimization experimental strategies recently
applied to the missing transverse-momentum data, on
which we have now modeled our background simulation.
Assuming an integrated luminosity of 100 (300) fb1
at 14 TeV, and extrapolating the effect of these optimization technique to higher energies, we find a 5 discovery
potential for BR down to 1.6 103 (9.2 104 ). At
the High-Luminosity LHC (HL-LHC), with an integrated
luminosity of 3 ab1 , the 5 reach is extended down to
2.9 104 .

B.

VBF channel

We now turn our focus on the Higgs production in

the VBF channel. This presents a lower production rate
with respect to the gluon-fusion channel. On the other
hand, it is in principle more controllable due to its strong
kinematical characterization. In particular, the process
pp Hjj
jj, where the Higgs boson arises from a

5
A [8 TeV] A [14 TeV]
H
(BR = 1%)
44
101
j
63
202
jj j
59
432
e
55
93
W ( `)
58
123
102
174
Z( )
total background
337
1024

0.6

V BF (H
)
+ 2jets
+ 3jets
+ Z + 2jets

0.5
0.4
1 dN
N dp

0.3
0.2
0.1

TABLE I: Cross section times acceptance A (in fb) for the

gluon-fusion signal and backgrounds at 8 and 14 TeV, assuming BR = 1%, with the selection pT > 50 GeV, | | < 1.44,
/ T > 50 GeV, and 100 GeV < MT < 130 GeV.
E

0
0

50

100

150

250

V BF (H
)
+ 2jets
+ 3jets
+ Z + 2jets

0.9

W (Z)-pair fusion, results mostly in two forward jets with

opposite rapidity, one photon and missing transverse momentum.
We started by simulating the signal by PYTHIA, by
including both the Higgs VBF production and its subsequent decay into a
final state. The main SM backgrounds are given by the production of QCD multi-jets,
+jets, and + Z( )+jets. The +jets background
has been simulated using ALPGEN. We have generated
j, jj, and jjj samples with pT > 10 GeV and
| | < 2.5 for photons, and pjT > 20 GeV and | j | < 5
for jets. An isolation of R > 0.4 between all pairs of
objects is required. We have then interfaced ALPGEN
and PYTHIA, and incorporated the jet-parton matching,
according to the MLM prescription [21]. Events containing hard partons are generated in ALPGEN with a cut on
the transverse momentum (pT > 20 GeV), and on the rapidity (|| < 5.0) of each parton, along with a minimum
separation (R > 0.4) between them. These events are
then interfaced with PYTHIA for showering, to take into
account soft and collinear emission of partons. All partons are then clustered using a cone jet algorithm with
pT > 20 GeV, and a cone size of R = 0.6 (the latter
used only for matching purposes, not for the jet definition
in the event selection). An event is said to be matched
if there is a one-to-one correspondence between jets and
initial hard partons. An event with an extra jet which is
not matched to a parton is rejected in case of exclusive
matching, while is kept in case of inclusive matching for
the highest jet-multiplicity samples.
For the QCD multi-jet process and the +Z+ jets process we have used MadGraph 5 interfaced with PYTHIA.
In case of the QCD multi-jet process, the most central
jet is assumed to be mistagged as a photon with a corresponding faking probability of 0.1%. The ISR and FSR
effects, parton shower, hadronisation and finite detector
resolution effects have also been implemented for the signal and all backgrounds. We have then assumed a photon
identification efficiency of 90%. The distributions are obtained with a nominal cut on the photon transverse momentum, pT > 10 GeV, and pjT > 10 GeV on fake jet in

200

pT

0.8
0.7
0.6
1 dN
N dE
/T

0.5
0.4
0.3
0.2
0.1
0
0

50

100

150

200

E
/T

FIG. 2: Photon pT (upper plot), and missing transverseenergy (lower plot) distributions for the signal and SM backgrounds in the VBF process. The final state in this case is
/ T + ( 2)jets with no isolated leptons. All distributions
+E
are normalized to unity.

the QCD multijets analysis.

In Figures 2 and 3, we plot a few kinematic distributions which are useful to separate the signal from the
backgrounds.
On this basis, we propose to select the events according
to the following criteria:
(basic cuts) one isolated photon with pT > 30 GeV
and | | < 2.5, and two or more jets with pjT > 20
GeV and | j | < 5.0, and angular separation R >
0.4 between all objects;
/ T > 30 GeV;
(basic cut) missing transverse energy E
(basic cut) no isolated leptons;
(rapidity cuts) rapidities of the two highest pT jets
obey j1 j2 < 0 and | j1 j2 | > 4.0;
(MT cuts) transverse mass of the photon and invisible system satisfying 100 GeV < MT < 130 GeV
(as above, the upper bound has been extended with
respect to mH to take into account the smearing of
the MT distribution, cf. Figure 3).

6
0.025

Cuts

V BF (H
)
+ 2jets
+ 3jets
+ Z + 2jets

0.02

Signal +jets + Z+jets QCD multiijet

Basic cuts

17.7

266636

1211

72219

Rapidity cuts

8.8

8130

38.1

33022

MT cuts

5.0

574

6.5

3236

0.015

TABLE II: Cross sections times acceptance A (in fb) for

the VBF signal and backgrounds at 14 TeV, after sequential
application of cuts defined in the text, assuming BR =1%.

1
dN
N d| j1 j2 |

0.01

0.005

0
0

| j1 j2 |

0.3

H (no jet)
Hj
j
jj

0.25

1 dN
N dM T 0.15

0.1
0.05
0
50

100

150

200

2.67

84.2

1.84

758

1.6

1.82

6.9

2.16

37

4.6

both cuts

1.21

1.2

0.67

19

4.5

TABLE III: Cross sections times acceptance A (in fb)

for the VBF signal and backgrounds at 14 TeV, assuming
BR =1%. The first and second row corresponds to the sep/ T ) cuts, respectively, after
arate effect of the y and (ji , E
applying all the cut sequence in Table II. The last row represents the combined effects of the two cuts. The last column
shows the signal significance for an integrated luminosity of
L=300 fb1 .

0.2

Signal +jets + Z+jets multijet L=300 fb1

Cuts
y < 1.0
/ T ) >1.5
(ji , E

250

MT

FIG. 3: Rapidity gap between the two forward jets (upper

plot), and transverse-mass (lower plot) distributions for the
/ T +( 2)jets final state
signal and SM backgrounds in the + E
with no isolated leptons. The = | j1 j2 | distribution is
obtained with a cut pT > 30 GeV, for pjT > 30 GeV on the
/ T > 30 GeV. The
fake jet in the QCD multijets analysis, and E
transverse mass distribution is obtained with the additional
cuts j1 j2 < 0 and | j1 j2 | > 4.0. All distributions are
normalized to unity.

In Table II, we present the cross sections for the signal

and dominant SM backgrounds after the sequential application of basic cuts, rapidity cuts on the two forward
jets, and transverse-mass cut on the photon plus missing
transverse-energy system.
In order to better control the missing transverse energy arising from jet energy mis-measurements, we have
/ T ) > 1.5
also imposed an azimuthal isolation cut (ji , E
/ T direction
(with i = 1, 2) on the angles between the E
and the transverse momenta of the two highest-pT jets.
Furthermore, we studied the effect of a selection cut
occasionally applied for searches in the VBF channel
(see, e.g., the W ` analysis in VBF in [22]). This
is the y < 1.0 cut on the Zeppenfeld variable defined
as y = |y H 12 ( j1 j2 )|, where the Higgs rapidity
y H is reconstructed from the photon momentum and the
missing transverse energy as described in [23]. X systems produced via VBF are in fact characterized by a
smaller y value, with respect to other X+2-jet back-

/ T ) and y cuts have

grounds. The values of the (ji , E
been separately optimized in order to increase the signal
significance.
Table III presents the independent effect of the y and
/ T ) cuts, applied after the set of cuts listed in
(ji , E
Table II. The combined effect of these two cuts is also
/ T ) cut
shown in the last row of Table III. The (ji , E
turns out to be much more effective in separating the
signal from background. We then dropped the y cut in
our final selection.
/ T ) distribution is asymmetric in the
Since the (ji , E
exchange of the first and second highest-pT jets, we have
also tried to optimize the signal significance by assuming
/ T ), that is by applying
an asymmetric cut on (ji , E
different cuts on the first and second highest-pT jets. We
anyway found that the best signal to background ratio
/ T ) > 1.5 on
is obtained with the symmetric cut (ji , E
both jets.
Finally, assuming an integrated luminosity of 300 fb1 ,
in the last column of Table III we present the estimated
VBF signal significancesfor BR =1%. For this setup,
the signal significance S/ S + B approaches the 5 level.
For 100 fb1 , the 5 reach in branching ratio is about
BR ' 2%. With the HL-LHC integrated luminosity of
3 ab1 , the 5 reach can be extended down to BR =
3.4 103 .

IV.

SUMMARY AND CONCLUSIONS

We have studied the prospects for discovering an exotic Higgs-boson decay into a SM photon and a new neutral massless
vector boson, a dark photon, at the LHC

with S = 14 TeV. We have updated our previous anal-

7
L=100 fb1 L=300 fb1 L=3 ab1
Significance
3
5
3
5
3
5
BR (VBF) 1.1
1.9 0.65 1.1 0.21 0.34
BR (ggF ) 0.096 0.16 0.055 0.092 0.017 0.029
BR (%)

TABLE IV: Reach in BR (in percentage) for a 3 exclusion

or a 5 discovery at the 14 TeV LHC, in the VBF and gluonfusion channels, for different integrated luminosities L.

by CMS for the suppression of the SM hadronic back/ T signature can be very effective even
grounds to the E
for relatively low transverse-momentum final states, possibly resulting in experimental sensitivities for branching
ratios well below the permil level. Similar methods could
actually be applied (once the corresponding experimental
analyzes will be available) for suppressing the SM multijet background to the VBF channel, possibly increasing
the relative weight of the VBF analysis in the search for a
H
signature, hence expanding the LHC potential.

ysis of the gluon-fusion channel at 8 TeV by a more

reliable treatment of both the signal and hadronic SM
backgrounds, and extended this approach to 14-TeV collisions. We also explored for the first time the possibility
of detecting the exotic H channel in the VBF Higgs
production.
A summary of our findings is presented in Table IV,
where we show the predicted reach in detectable BR
for both exclusion (at a 3 level) and discovery (at a
5 level), assuming 100, 300 and 3000 fb1 of data at
14 TeV. The gluon-fusion potential turns out to be definitely higher, extending the BR reach with respect
to the VBF channel by more than one order of magnitude. In particular, according to the present analysis,
the full LHC program will allow to discover (exclude) a
BR value down to less than 1 103 (6 104 ), while
the HL-LHC phase will be sensitive to BR as small as
3 104 (2 104 ). We recall that BR values up to
5% are allowed in realistic BSM frameworks [15].
In light of the projected discovery reach and of the
theoretical interest in dark-photon models, we urge the
ATLAS and CMS experiments to perform a dedicated
/ T signature in two-body fianalysis of the H + E
nal states. The event selection criteria used in the CMS
analysis [16], by imposing an upper limit of 60 GeV on
pT , considerably restrict the signal phase space for the
two-body decay mode. Nevertheless, the methods used

After the recent observation at the LHC of an excess

in the di-photon spectrum around an invariant mass of
about 750 GeV [24, 25], it would be also advisable to ex/ T final states to higher invariant
tend the search for + E
masses of the
pair. Indeed, the observed features of
the would-be 750-GeV resonance might require new
degrees of freedom in a hidden sector in order to give rise
to effective couplings to photons (and gluons) (see,e.g.,
[26]). The latter degrees of freedom could well be portals
to a massless dark photon, in case they are also charged
under an extra unbroken U (1)F . Since a large U (1)F coupling might be naturally allowed [17], the corresponding
rate for a
resonance at 750 GeV could already be sizable with the present data set. This possibility has also
been envisaged in [27].

[1] P. Ade et al. [Planck Coll.], arXiv:1303.5076.

[2] For a review and references see, R. Essig et al.,
arXiv:1311.0029 [hep-ph].
[3] B. Holdom, Phys. Lett. B 166, 196 (1986).
[4] S. Chatrchyan et al. [CMS Coll.], Phys. Rev. Lett. 108,
261803 (2012); G. Aad et al. [ATLAS Coll.], Phys. Rev.
Lett. 110 (2013) 011802.
[5] G. Aad et al. [ATLAS Coll.], Phys. Rev. D 87, no. 11,
112003 (2013); S. Chatrchyan et al. [CMS Coll.], JHEP
1310, 164 (2013) [arXiv:1309.1117 [hep-ex]].
[6] D. Curtin et al., arXiv:1312.4992 [hep-ph]. J. F. Kamenik and C. Smith, Phys. Rev. D 85, 093017 (2012)
[arXiv:1201.4814 [hep-ph]]. H. Davoudiasl, H. -S. Lee,
I. Lewis and W. J. Marciano, Phys. Rev. D 88, no.
1, 015022 (2013) [arXiv:1304.4935 [hep-ph]]. Y. Sun
and D. -N. Gao, Phys. Rev. D 89, 017301 (2014)
[arXiv:1310.8404 [hep-ph]].
[7] A. Falkowski and R. Vega-Morales, arXiv:1405.1095.
[8] L. Ackerman, M. R. Buckley, S. M. Carroll and

M. Kamionkowski, Phys. Rev. D 79, 023519 (2009)

[arXiv:0810.5126 [hep-ph]].
J. Fan, A. Katz, L. Randall and M. Reece, Phys. Rev.
Lett. 110, no. 21, 211302 (2013) [arXiv:1303.3271].
M. Heikinheimo,
M. Raidal,
C. Spethmann
and H. Veerme, Phys. Lett. B 749 (2015) 236
doi:10.1016/j.physletb.2015.08.012
[arXiv:1504.04371
[hep-ph]].
N. Arkani-Hamed, D. P. Finkbeiner, T. R. Slatyer and
N. Weiner, Phys. Rev. D 79, 015014 (2009).
K. M. Zurek, Phys. Rept. 537, 91 (2014).
H. An, M. Pospelov, J. Pradler and A. Ritz, Phys. Lett.
B 747, 331 (2015) [arXiv:1412.8378 [hep-ph]].
E. Gabrielli and M. Raidal, Phys. Rev. D 89, 015008
(2014) [arXiv:1310.1090 [hep-ph]].
E. Gabrielli, M. Heikinheimo, B. Mele and M. Raidal,
Phys. Rev. D 90, no. 5, 055032 (2014) [arXiv:1405.5196
[hep-ph]].
V. Khachatryan et al. [CMS Collaboration],

In case the di-photon signature will be confirmed at

/T
the LHC, the search for new structures in the + E
transverse-mass distributions at 750 GeV would provide
extra invaluable insight about the nature of the NP behind it.
Acknowledgments.
We thank Daniel Fournier,
Jean-Baptiste de Vivie de Regie, and Rachid Mazini for
useful discussions. E.G. would like to thank the TH division of CERN for its kind hospitality during the preparation of this work. The work of M.H. has been supported
by the Academy of Finland project number 267842.

[9]
[10]

[11]
[12]
[13]
[14]
[15]
[16]

8
arXiv:1507.00359 [hep-ex].
[17] S. Biswas, E. Gabrielli, M. Heikinheimo and B. Mele,
JHEP 1506, 102 (2015) [arXiv:1503.05836 [hep-ph]].
[18] J. Alwall et al.,
JHEP 1407,
079 (2014)
doi:10.1007/JHEP07(2014)079 [arXiv:1405.0301 [hepph]].
[19] T. Sjostrand, S. Mrenna and P. Z. Skands, JHEP
0605, 026 (2006) doi:10.1088/1126-6708/2006/05/026
[hep-ph/0603175].
[20] M. L. Mangano, M. Moretti, F. Piccinini, R. Pittau and
A. D. Polosa, JHEP 0307, 001 (2003) doi:10.1088/11266708/2003/07/001 [hep-ph/0206293].
[21] S. Hoeche, F. Krauss, N. Lavesson, L. Lonnblad,
M. Mangano, A. Schalicke and S. Schumann, hepph/0602031.
[22] CMS Collaboration, Measurement of the cross section of
the electroweak production of a W boson with two jets in

pp collisions at sqrt(s) = 8TeV, CMS-PAS-SMP-13-012.

[23] D. L. Rainwater, R. Szalapski and D. Zeppenfeld, Phys.
Rev. D 54, 6680 (1996) doi:10.1103/PhysRevD.54.6680
[hep-ph/9605444].
[24] The ATLAS collaboration, Search for resonances decay
ing to photon pairs in 3.2 fb1 of pp collisions at s =
13 TeV with the ATLAS detector, ATLAS-CONF-2015081.
[25] CMS Collaboration, Search for new physics in high mass
diphoton events in proton-proton collisions at 13 TeV,
CMS-PAS-EXO-15-004.
[26] R. Franceschini et al., What is the resonance at
750 GeV?, arXiv:1512.04933 [hep-ph].
[27] Y. Tsai, L. T. Wang and Y. Zhao, Faking The
Diphoton Excess by Displaced Dark Photon Decays,
arXiv:1603.00024 [hep-ph].