A Forward Multiparticle Spectrometer for the LHC: Hadron spectra and Long-lived particle search
AA Forward Multiparticle Spectrometer for the LHC:Hadron spectra and Long-lived particle search
Michael G. AlbrowE-Mail: [email protected]
Fermi National Accelerator Laboratory, Batavia, IL 60510, USA
Presented at the Workshop of QCD and Forward Physics at the EIC, the LHC, and Cosmic Ray Physics inGuanajuato, Mexico, November 18-21 2019
Abstract
I describe a possible Forward Multiparticle Spectrometer (FMS) that could be installed down-stream of the superconducting recombination dipole D1 in Run 4, between z = 96 m - 126 m tomeasure multi-TeV hadron spectra in low luminosity pp collisions at √ s = 14 TeV, as well as p+Oand O+O collisions as relevant for cosmic ray showers. Light antinuclei and charmed hadrons athigh Feynman x F can be measured, both of importance for astrophysics. At the full high luminos-ity HL-LHC a search for new long-lived neutral particles (LLPs) decaying in a 20 m long, 70 cmdiameter vacuum pipe to visible decay modes (including γγ, e + e − , µ + µ − , τ + τ − , c ¯ c and jets) canbe made. The FMS is especially well suited for LLPs with 1 GeV < M ( X ) <
10 GeV and lifetimes cτ from about 10 m to several km.I discuss this as a possible addition to CMS but it has no formal approval yet, therefore thetalk is not given “on behalf of CMS”. The sparcity of accelerator data on particle production in the forward direction above √ s = 63GeV at the CERN Intersecting Storage Rings (ISR), its importance for understanding cosmic rayshowers, and the possibility of measurements at the LHC was addressed in Ref. [1].We have been developing a forward multiparticle spectrometer, FMS, that could be added asa new subsystem to CMS for Run 4 (2027+). A schematic overview of the spectrometer is shownin Figure 1. The main detectors are situated at z = 116 m - 126 m and surround the beam pipebetween radii R in = 12 cm and R out = 35 cm . The LEFT + RIGHT and UP + DOWN azimuthal regionshave distinct physics motivations and operational modes; hadron spectroscopy and a new long-lived particle (LLP) search respectively. The detectors can use the same techniques as the CMSEndcap upgrade planned for Run 4, with silicon tracking and calorimetry with precision timing,followed by a magnetised toroid with GEM layers for muon measurement. Transition radiation All dimensions are provisional and subject to optimisation. a r X i v : . [ phy s i c s . i n s - d e t ] J un etectors (TRD) for TeV hadron identification are being developed [2]; these are the only detectorsnot included in the CMS upgrade plans. They are essential for the hadron mode, but optional forthe LLP mode. The area is only about 0.3 m , which is less that 1% of the future Endcap.An earlier talk on the hadron mode is given in Ref.[3]. Forward spectra of π ± , K ± , p and ¯ p havenot been measured above √ s = 63 GeV at the ISR [4, 5] but are important to understand cosmic rayshowers. At the LHC with √ s = 14 TeV we will be 220 times higher in √ s . In fixed target terms, asappropriate for cosmic ray showers, E BEAM is about 50,000 times higher. The ISR energy is wellbelow the famous knee in the cosmic ray spectrum; the LHC energy is well above. An excess ofmuons is observed in very high energy showers compared with expectations; forward spectra atthe LHC may shed light on this, as well as being relevant for atmospheric neutrinos, which are abackground to cosmic neutrinos as seen in ICECUBE.Event generators such as
PYTHIA have not been tuned for this region since there is little data.Low p T physics being non-perturbative QCD is theoretically more challenging than high p T andis worthy of more attention; this is the ”low- Q frontier” of QCD. At the LHC only leading pro-tons with Feynman- x F (cid:38) π and neutrons) at θ = 0 ◦ have beenmeasured [6], demonstrating the very large spread in cosmic ray shower Monte Carlos.When planning future high energy hadron colliders, such as the √ s = 100 TeV FCC (p+p mode,as well as with ions) predictions for radiation levels in the forward direction should benefit fromimproved knowledge of these cross sections. The straight section downstream of IR1 (ATLAS) and IR5 (CMS) between the end of the new su-perconducting D1 dipole at z = 81 m and the entrance to the TAXN absorber at 126.5 m is mostlyfree of equipment, with a new straight beam pipe presently planned to have R = 7.5 cm at thefront increasing to 12.5 cm at the back. (The regions downstream of LHCb and ALICE are wherethe proton beams are injected and are more complicated.) We propose to change the design ofthis pipe to have an enlarged radius, nominally R = 40 cm over at least 20 m, from z = 96 m to116 m. Immediately downstream of D1 there is to be a cold diode structure (in DFBX) parallel tothe beam pipe which, if it cannot be repositioned, limits the beginning of the proposed detectorsystem, shown schematically in Fig. 1.The first new element is an iron (ASII 1010 low-carbon steel) toroid (I thank V. Khashikhin,Fermilab, for the study), a cylinder of length 3 m, R in = 8 cm and R out = 40 cm. It is constructed intwo halves for easy assembly/disassembly and allowing separation of top and bottom halves forbakeout of the beam pipe. Two water-cooled copper coils, both in the bottom half, with currentsof 5 kA each, give a circular field in the iron varying from 1.9 T at the inner radius to 1.5 T atthe outer. All charged particles (mostly muons) exiting the toroid steel are measured in a counterhodoscope mounted on the back of the steel, followed by track chambers, e.g. a pair of GEMsor silicon strip layers, separated by 1 m. The field deflects charged particles emerging from theback of D1 inwards or outwards, reducing the flux of muons at the detectors downstream. Thisis predicted by FLUKA to be 0.9 (0.65) per bunch crossing even with 140 interactions (HL) without(with) the toroid powered.The field at the center of the beam pipe is less than 3 Gauss; both incoming and outgoingbeams are inside the pipe but not centered. If necessary, the field inside the pipe could be reducedby a thin iron shield around it. In the charged hadron spectroscopy mode, the main role of thistoroid is additional background reduction; the particles to be measured pass through the centralhole.Immediately after the toroid+tracker the pipe transitions to a wide pipe, similar to the 25.7 mlong pipe at ALICE in LSS2. NEG-coated liners inside an 80 cm pipe leave a clear aperture of diam- igure 1 : Schematic layout of proposed FMS spectrometer (side view). Dimensions are subject tooptimization, in particular the length budget. The start in z could be earlier if the LHC cold diodecan be displaced, and the space allocation for the main detectors can be increased at the expenseof vacuum volume.eter 70 cm. (I thank V. Baglin, CERN, for discussions on the beam pipe.) The main difference fromthe ALICE pipe is that the transition at z = 116 m to the small pipe should be such as to minimizeinteractions and especially multiple scattering. A 1 mm thick steel window perpendicular to thepipe axis gives a multiple scattering angle θ ◦ = 3 × − for 100 GeV/c particles, decreasing like1/ p . An option is to have a thinner window with strengthening ribs. To avoid the beam “seeing”a sharp change in pipe diameter an internal inclined wire grid or similar can be employed .After exiting the steel window the main elements of the spectrometer could use identical tech-nology to the CMS Endcap upgrade, namely silicon pixel tracking, followed by electromagneticand hadron calorimetry based on silicon pads with tungsten/copper and steel plates. Precisiontiming ∼
25 ps is planned, and is important.The need for precise tracking is very different for the hadron mode and the LLP mode. Inthe former case (L and R quadrants, hadrons of 1 - 3 TeV) we need to measure the momentaof particles coming directly from the collision region (or from charm decays) using the transportmatrix through the magnet lattice, and also to reject beam halo and tracks coming from interactionsin the upstream pipe and other material. In the LLP mode the detected particles, from a decay inthe vacuum, have not traversed any magnetic field; the essential need is to project the tracks backto a vertex and ensure it is inside the vacuum. Also we need to ensure that the neutral parent points A detailed design will require an Engineering Change Request from CMS, which has not yet formally considered thisproject. ack to the collision region through the steel (making allowance for missing neutrinos in any τ + τ − events!). Particles with momenta as low ( sic ) as 50 GeV/c may be of interest. One may dedicateabout 3 m of space for tracking, with σ ∼ µ m resolution giving angular resolution < − .Transition radiation detectors (needed especially for hadron spectroscopy, but less essential forthe LLP search) can incorporate tracking, so one may consider combining them and dedicating upto perhaps 5 m for both silicon strips or pixels and the TRD.Transition radiation detectors, sensitive to γ = E/m , have usually been used to help distiguishelectrons from pions at low energies. Anatoli Romaniouk and the ATLAS TRD Group have beendeveloping detectors that could distinguish π, K, and p in the TeV region [2]. Cherenkov countersare ineffective as β is too close to 1.0. X-rays are emitted from transitions between media of dif-ferent dielectric constants, or plasma frequencies, with a small probability which rises with γ andthen saturates. One can select radiator materials and thicknesses and gap widths to optimise for aselected range. Tests have been done at the SPS using layers of xenon-filled straw tubes betweendifferent foils with electrons, muons and pions. The yields, X-ray energy spectra and angular dis-tributions are very well predicted by detailed simulations. Interestingly the typical emission angledecreases like 1/ γ , and high granularity silicon or GaAs pixel detectors measuring the emissionangles of even a few X-ray photons may improve the separation power.The imaging calorimeter (HGCAL) layers will be sensitive to muon tracks, and fast timingwill be incorporated (perhaps with LGADs) to help with background reduction. The EM partof the calorimeter has very good measurement of high energy shower directions, addressing thechallenge of locating the decay point of an X → γγ decay within the vacuum pipe. This has nophysics background; probably the main backgrounds are photons from material before the decayregion with a mis-measured vertex, or from two unrelated photons that appear to come from avertex. The tracker + imaging calorimeter combination identifies γ, e, µ and hadrons, and the TRDcan provide some distinction between π + π − and K + K − . This capability would be especiallypowerful in searching for X → c ¯ c and X → τ + τ − , even X → jet + jet. The FMS is particularlywell suited to LLPs in the M(X) = 1 GeV to 10 GeV range (fixed target experiments have higherluminosity and are more sensitive to lighter particles, e.g. dark photons with M(A’) < 500 MeV).Behind the calorimeter we propose another iron toroid, identical to the plug at the front endexcept that it is subdivided longitudinally with a few gaps allowing insertion of muon trackinglayers (e.g. GEMs). With 1.5 T the bending angle for a 100 GeV/c muon is 13.5 mrad, comparedwith the multiple scattering θ rms = 2 mrad. The sensitivity to X → µ + µ − needs a full simulation,but the mass resolution is probably a minor issue, since there is no physics background from K decays (B.R < − ) or any other SM particles. The back of this toroid is shielded from backgroundcoming from behind by the TAXN absorber.Like the front toroid, the back detectors cover full azimuth but can be separated into top andbottom halves, or quadrants.Installation of an FMS in both outgoing beams is technically possible and would give twicethe data and LLP sensitivity for less than twice the cost. When the ISR came into operation in 1971 Feynman proposed that forward hadron spectra shouldscale with energy √ s when plotted as a Lorentz-invariant cross section at fixed p T vs. x F = p z /p beam ; this is Feynman scaling. It was based on the parton model, pre-QCD, and while it isa good approximation in the ISR energy range for light particles at low- p T [4, 5], QCD has scalingviolations, heavy flavors have thresholds, etc. Feynman scaling should not hold over the largeenergy range from √ s = 63 GeV to 14,000 GeV! igure 2 : Fluxes of primary charged pions at z = 116 m per 100 fb − (M. Sabate-Gilarte). Thecentral grey disk is the outgoing beam pipe. To convert to numbers per collision divide by 8 × .The regions above and below the pipe are clear of primary particles. The outer radius is nowplanned to be larger than shown, namely 35 cm or 40 cm.The new 35 Tm beam recombination dipole D1, ending at z = 81 m, is here used as a spectrom-eter magnet, deflecting charged particles into right (R) and left (L) quadrants. The beam crossingangle (planned to be vertical for CMS, 250 µ rad half-crossing angle) and the quadrupoles affectthe distributions, as shown in Fig. 2. A large beam pipe, R = 40 cm, from z = 96 m to 116 m at theend of which is a steel vacuum window about 1 mm thick, allows charged particles to enter thespectrometer, where they can be measured in short low luminosity p+p, p+O, and O+O runs. Theacceptance for primary charged particles is approximately p z = 1 - 3 TeV/c. Higher p z particlesremain within the pipe. Fig. 2 shows the spatial distribution of primary charged particles at z =116 m (M. Sabate-Gilarte and F. Cerutti).Hadrons from fragmentation of diffractively excited protons, p → p ∗ populate this region, andin low pileup data it would be interesting to study in combination with a leading proton in theopposite direction if there are suitable Roman pots.The FMS can also measure light nuclei and antinuclei ( ¯ d, ¯ t, ¯ He ) which are relevant for un-derstanding γ -rays from the galactic center and a possible dark matter annihilation signal. It willhave acceptance for J/ψ → µ + µ − and charmed hadrons, specifically D → K − π + , ¯ D → K + π − and Λ + c → pK − π + at x F > ∼ c ¯ c in the proton wavefunction, can givea large cross section [7]. The challenge of seeing these narrow charm signals on a large combinato-rial background drives the need for excellent tracking (traversing only vacuum from the collisionpoint to the window at 116 m) and good π/K/p separation. Prompt muons can also be measured,subtracting the spectra from π ± and K ± decays (which will be known) as another measure of c -and b -production. Note that the mean decay length for a 2.5 TeV charged pion(kaon) is 139(18.5)km, and for a 5 TeV D it is 33 cm!The expected fluxes of charged particles as well as charmed hadrons have been calculatedusing different cosmic ray Monte Carlos by H. Menjo (priv. comm.) and by M. Sabate-Gilarte(priv. comm. and [8]) using FLUKA with
DPMJET including upstream interactions. There is no igure 3 : Left: Spectra in x F of D and ¯ D from FLUKA, p + p at √ s =
14 TeV. Top right: Spatialdistribution of K + and π − from ¯ D decays at z = 116 m. Bottom right: Distribution in the p T : x F plane of ¯ D with both K and π in FMS acceptance. (M. Sabate-Gilarte)space here for details but e.g. the expected flux of µ ± within R = 30 cm at z = 116 m is only0.9 per bunch crossing with 140 interactions, reducing to 0.65 with the front toroid powered, andnearly all of these have p µ < 50 GeV/c. Most pp collisions produce no direct hadrons in the FMSacceptance (the average is about 0.2) and measurements of the inclusive charged hadron spectracould be made with some pileup, but for multiparticle states like D decays the signal:backgroundmay be unacceptable unless there is not much pileup.If there is a zero-degree calorimeter (ZDC/LHCf) between the beam pipes downstream of theFMS to detect neutrons and π , we will be able to study coincident events, e.g. p → nπ + π bydiffraction dissociation. In the absence of a discovery of high mass dark matter particles, searches are turning to the pos-sibility that they are light (e.g. M(X) <
10 GeV) but weakly interacting. There may be “portals“that couple SM particles to dark matter particles, that are weak enough to penetrate a lot of matterbut then decay to known Standard Model particles such as γγ, e + e − , µ + µ − , τ + τ − , c ¯ c and b ¯ b . TheFMS can search at full luminosity for penetrating neutrals with all of these decays occurring insidethe vacuum pipe. A FLUKA calculation by M. Sabate-Gilarte predicts, exiting the front toroid steel,per bunch crossing with 150 inelastic interactions, 0.6 photons, 0.45 neutrons, 0.15 antineutrons,and 0.12 K above 50 GeV/c. Above 200 GeV/c the fluxes are much less, see Figure 4. The upper(U) and lower (D) quadrants are devoid of primary charged particles since D1 acts as a sweepingmagnet, and the detector area is out of the angular range for direct neutral particles. The low occu-pancy in these quadrants provides an excellent opportunity to search for BSM long-lived neutralparticles (LLPs) from the primary collisions that penetrate 35 - 50m of steel (> 190 λ int ) in theQ1-Q3 magnets and D1, and decay in the vacuum of the large pipe. The decay products, be theyphotons, electrons, muons or charged hadrons, can be measured in FMS during high luminosityrunning. Excellent tracking to show that the vertex is inside the 20 m-long vacuum region, andnot initiated by a charged particle (e.g. µ ), should eliminate backgrounds; Standard Model LLPs I thank Francesco Cerutti for the following numbers of interaction lengths, for a straight track from the IP to the end ofthe D1 cold mass at 81 m: 320 λ int at y = 15 cm, 220 λ int at y = x = 15 cm due to yoke holes at 45 ◦ , 300 λ int at y = 20 cmdue to the smaller section of the multipole correctors, and 190 λ int at y = 10 cm due to the part of the path in the vacuum. .S. Gilarte Figure 4 : Fluxes of neutral particles emerging from the back of the front toroid (M. Sabate-Gilarte).such as K and Λ are recognized in the spectrometer which has tracking, calorimetry and muonchambers, and can be reduced using mass and lifetime information.One may ask about of the sensitivity of the FMS to semi-weakly interacting LLPs that do not decay in the vacuum pipe or interact in the steel absorber but interact inside the calorimeter. Atfirst sight it may seem that backgrounds are overwhelming, but the HGCAL will provide sensi-tivity to single charged particles, detailed shower starting point and directional information, pre-cision timing and energy measurement. The muon chambers immediately behind the calorimeterhave information about possible muon content in the shower. Combining time-of-flight and en-ergy measurement gives M(X); look for a peak! For example, if M(X) = 2 GeV/c and p = 50 GeV/c,the flight time to the calorimeter is 200 ps later than that of a neutron; for M(X) = 5(10) GeV/c with p = 100 GeV/c it is more than 0.4(2.0) ns later. While the time resolution on a shower shouldbe ∼
20 ps, the time spread of the collisions themselves, projected in the forward direction, may bea limiting factor. Also for decaying LLPs the time-of-flight from collision to signals in the trackerand HGCAL can be useful information if γ (cid:46)
20. Selecting showers from interacting neutral parti-cles emerging from the back of the 20 λ int absorbers, and (critically) pointing back to the collisionregion, one might reduce background from scattered high energy neutrons to an acceptable level.This unique capability of FMS merits a detailed study. There are several other experiments searching for penetrating then decaying long-lived particles.Here I only compare with that most similar to FMS, FASER, which is approved for a first run inRun 3 with a decay volume length only 1.5 m and radius 10 cm, but is planned to be upgraded to5 m and 1 m radius for Run 4. It is much further downstream (of IR 1) at z = 480 m - 485 m, afterabout 100 m of rock absorber. It is centered on the collision axis, and with a radius R = 0.1(1.0) mhas pseudorapidity η above 9.2(6.9) neglecting beam crossing angle effects.The probability of a particle that enters a decay volume decaying in it is: F = e − z in / ( γcτ ) − e − z out / ( γcτ ) (4.1) igure 5 : Spectra calculated using FLUKA for π and D for FMS ( . < | η | < . )and FASER( η > . (Run 3).For FASER(5m) this exceeds − for γcτ between 130 m and 20 km, and has a maximumwhen γcτ = 480 m at which F = 3 . × − . The FMS is both closer to the IP and is much longer,and F exceeds 1% between γcτ = 24 m and 1.85 km, with a maximum of 6.9% at γcτ = 116 m.Since FMS(LLP) with . < | η | < . is at a larger polar angle than FASER, we should alsocompare the fluxes of particles as a function of momentum. These have been calculated by F.Cerutti and M. Sabate-Gilarte with FLUKA and two examples are shown in Figure 5. Withouthaving predictions for the production of LLPs of various masses, we assume the spectrum of anA’ light enough to come from π decays to be similar to that of π themselves, and that of an LLPwith M(X) ∼ to be similar to that of a D . So these are only indicative, but show thatthe FASER flux is higher for π with momenta above about 1 TeV, but heavier particles have largermean p T and the charm flux is much higher in the FMS(LLP) angular region. For the FASER Run 4proposal their charm flux will be higher, although this prediction has very large uncertainty sincethere is no data. However FMS in the hadron mode will measure very forward charm, largelyresolving this issue.Of course if an A’ or similar BSM particle is discovered before Run 4, the FMS should be ableto study it in a novel way.The FASER- ν extension is to measure neutrino interactions in an emulsion stack at the samelocation. Since the spectra of charged pions, kaons and charmed hadrons at large x F presentlyhave an order of magnitude uncertainty, the neutrino cross sections cannot be measured withoutknowing those spectra. FMS (hadron mode) will measure these up to x F ∼ Hadron interactions
In the hadron mode, behind the TRD there will be a flux of identified hadrons of known momentabetween about 1 and 3 TeV/c. One could insert thin foil targets, e.g. of carbon and polyethylene(C H ) with some pixel tracking a few meters behind for a special short run. By counting tracksfrom a vertex in the foil one could get a measure of σ inel and the N charged distribution for thedifferent beam particles (including light nuclei and antinuclei) on both protons and carbon. Tomake longitudinal space for this the calorimeter may need to be displaced. In the hadron spectroscopy mode the ideal running condition would be to have an average ofabout one inelastic collision per bunch crossing ( µ = 1), with a level-one trigger based on oneor more tracks or EM-calorimeter signals. The full CMS detector would be read out to studycorrelations, and the single-track rate would need pre-scaling. To maximize statistics for charmetc., ≥ -track triggers could include a fast processor selecting candidates. While the single-particleinclusive spectra could be measured at higher pileup, the charm signal:background would becomeworse; this needs a study.In the LLP mode at high luminosity, when a candidate in FMS is accompanied by a largenumber of inelastic collisions, there does not appear to be any value in reading out the full centraldetector with an FMS trigger. The DAQ could then have an FMS-only data stream, with a triggerselecting events with charged particles, or an anomalous calorimeter signal, behind the big pipe,with no corresponding charged particle entering at the front. A powerful multiparticle spectrometer could be installed in the 30 m straight section between theD1 dipole and the TAXN for physics in Run 4. In the L and R quadrants hadron spectra can bemeasured in a few days of low luminosity running. In the U and D quadrants a search can be madeat full luminosity for new long-lived particles decaying in a large diameter 20 m-long vacuumpipe. The detectors system uses the techniques of the CMS Endcap upgrade (but with about 1%of the area) with novel transition radiation detectors. A longer write-up is in preparation; newparticipants are welcome!
I thank the organizers of the Workshop at Guanajuato inviting me and for allowing me to includein this write-up the LLP search mode, developed since that workshop. I am a CMS member, butthis paper is not on behalf of CMS but ad personam since the project does not yet have any officialstanding in CMS. I hope that will change. I especially want to thank Hiroaki Menjo, AnatoliRomaniouk (TRD development), and CERN staff on LHC: Francesco Cerutti, Marta Sabate-Gilarteand Vincent Baglin for information and calculations showing the feasibility of the project.
References [1] M.G. Albrow,
Accelerator Data for Cosmic Ray Physics ,Contribution to ISVHECR 2010, eprint: 1009.4178[hep-ex].[2] N. Belyaev et al. , Development of transition radiation detectors for hadron identification at the TeV energy scale ,J.Phys - Conf.Ser p.1 (2019).3] M.G. Albrow,
A very forward hadron spectrometer for the LHC and cosmic ray physics , PoS EDSU2018 (2018)048, eprint: 1811.02047 [physics.ins-det].[4] M.G. Albrow et al. (CHLM Collaboration),
Negative particle production in the fragmentation region at theCERN ISR , Nucl. Phys.
B56 (1973) 333 - 345.[5] M.G. Albrow et al. (CHLM Collaboration),
Positive particle production in the fragmentation region at theCERN ISR , Nucl. Phys.
B73 (1974) 40 - 56.[6] E. Berti et al. (LHCf Collaboration),