Left: relative trigger yields as a function of instantaneous luminosity, normalised to $\mathcal{L} = 2\times 10^{32}\text{ cm} ^{-2} \text{ s} ^{-1} $. Right: rate of decays reconstructed in the LHCb acceptance as a function of the cut in $ p_{\mathrm{T}}$ of the decaying particle, for decay time $\tau > 0.2\text{ ps} $.

Layout of the upgraded LHCb detector.

Electronics architecture of the upgraded LHCb experiment.

The readout system located in the modular data centre and the front-end electronics in the underground cavern are connected through long-distance optical fibres installed in the PM85 shaft.

Left: {\gls*{rms}} (under the mylar protection foil at the left side) and {\gls*{plume}} (inside the scaffolding). The arrow indicates the position of {\gls*{bcm}} which is hidden behind the {\gls*{plume}} scaffolding. Right: upstream {\gls*{bcm}} detectors inside their kapton-insulated support surrounding the beam pipe.

Schematic view of a {\gls*{plume}} elementary detection module. The module is 153 $\text{ mm}$ long and has a diameter of 24 $\text{ mm}$ .

Left: {\gls*{plume}} test beam setup. Right: charge collected by the first {\gls*{pmt}} in the test beam for all events (black line) and events producing a simultaneous signal in the two {\gls*{pmt}} {s} and the trigger (red line). The scale is in nVs where 1 $\text{ nVs}$ = 20 $\text{ pC}$ .

Left: a 3D view of the upgraded {\gls*{velo}} , with cut-out. Some of the new items are highlighted, such as the Side C pixel modules and readout electronics (brown), the Side A RF box (red), the internal gas target system with a storage cell (green), the upstream beam pipe with a sector valve (cyan). Right: data rate per pixel {\gls*{asic}} in $\text{ Gbit/s}$ for the most active module. The numbers in parenthesis are the number of traversing tracks per {\gls*{lhc}} bunch crossing for an average number of interactions per crossing equal to 7.6. Arrows indicate the readout direction.

Left: schematic top view of the $z-x$ plane at $y = 0$ (left) with an illustration of the $z$-extent of the luminous region and the nominal LHCb pseudorapidity acceptance, $2<\eta<5$. Right: sketch showing the nominal layout of the {\gls*{asic}} {s} around the $z$ axis in the closed {\gls*{velo}} configuration. Half the {\gls*{asic}} {s} are placed on the upstream module face (grey) and half on the downstream face (blue). The modules on the Side C are highlighted in purple on both sketches.

Top left: microscope image showing the elongated sensor pixels above the inter- {\gls*{asic}} region. Top right: image highlighting the ion-etched round corner of the sensor. Bottom: schematic of the sensor tile, showing the overall dimensions of the sensor and {\gls*{asic}} . The pixel layout is shown only under {\gls*{asic}} 2. There are $256 \times 256$ active bonded pixels (only every fourth pixel is shown in the figure). An additional row of pixels identified in dark red provides a connection between the {\gls*{asic}} ground and the innermost guard ring of the sensor. Three corners, encircled in red, are shown in detail on the left (A, B, C).

Left: illustration of the silicon microchannel coolers and fluidic connector. Right: the parallel lines represent the etched microchannels, which can be seen in the X-ray image.

Left: Image of a fluidic connector aligned in the horizontal plane with the corresponding solder layer on the silicon cooler, indicated with the red shadow. Right: X-ray of the solder joint attaching the fluidic connector to the microchannel cooler. No solder has entered the two inlet regions, and no large voids are seen in the solder layer.

Photo of the (left) upstream and (right) downstream faces of a fully-assembled {\gls*{velo}} module.

From left to right: bare module; module with tiles; then with hybrids too.

Cross-shaped markers used for module metrology. Left: on the microchannel cooler; centre: on the {\gls*{velopix}} {\gls*{asic}} ; right: on the {\gls*{asic}} -side of the sensor.

Tile metrology results showing $x$ and $y$ positions and angles for each module.

Block diagram showing the main parts of the {\gls*{velo}} electronics system.

The {\gls*{fend}} (left) and {\gls*{gbtx}} (right) hybrids.

Left: custom-made vacuum feedthrough board. Right: a populated vacuum flange.

The Side C half, with 26 modules, ready for the installation into the vacuum vessel.

Illustration of the {\gls*{velo}} halves showing modules on the module support bases and the LHCb acceptance as a transparent pyramid. On the left, the flexible electronic cables are shown leading to the vacuum feedthrough boards and {\gls*{opb}} boards in their custom frame. On the right, the flexible construction of long cooling loops is shown as well as the interface between the secondary and isolation vacua, in which sits an array of valves.

Left: the open {\gls*{velo}} vessel (seen from upstream) during the installation of the RF boxes. Right: view inside the Side A RF box showing the module slot structure.

Sketch of a tubular storage cell of length $L$ and inner diameter $D$. Gas is injected at the centre with flow rate $\Phi$, giving a triangular density distribution $\rho (z)$ with maximum $\rho_0$ at the centre.

Left: view of the storage cell (blue) supported from the {\gls*{velo}} RF box flanges (in green) in the closed {\gls*{velo}} position. Two flexible wakefield suppressors (orange) provide the electrical continuity. Right: storage cell in the open position (without showing {\gls*{velo}} elements).

Picture of the {\gls*{smog}} storage cell system installed into the {\gls*{velo}} vessel.

The four assembly groups of the {\gls*{smog}} {\gls*{gfs}} are the gas supply (with 4 reservoirs), the main table, the pumping station (PS) and the feed lines to the {\gls*{velo}} vacuum vessel. The various components are described in the text.

Drawing of the four {\gls*{ut}} silicon planes with indicative dimensions. Different colours designate different types of sensors: Type-A (green), Type-B (yellow), Type-C and Type-D (pink), as described in the text.

A 3D view of the {\gls*{ut}} system.

Left: a completed stave. At the bottom, the end of the cooling tube is visible with the high voltage and signal connections. In orange are the dataflex cables and in brown the hybrids. The reflective areas are the silicon detectors. Right: an exploded view of an instrumented stave showing the individual components described in the text.

Design features of the {\gls*{ut}} silicon sensors (see text). Left: Type-D sensor cut-out region. Right: Embedded pitch adapter.

Block diagram of the 128-channel {\gls*{salt}} {\gls*{asic}} .

Left: Two fully visible VERA hybrids, not yet equipped with {\gls*{asic}} {s}, embedded in a carrier 8-hybrid panel before cutting. Right: SUSI hybrid mounted in a final {\gls*{ut}} Type-D module.

Picture of one Short and one Medium dataflex cable.

Left: pigtails in two different shapes; middle: pigtail section showing the 3 subcables; right: installed pigtails.

Schematic of LV, HV, data, fast and slow control signal distribution to {\gls*{ut}} {\gls*{fend}} electronics.

Left: Illustration of the usage of dataflex cables on the different stave variants. Short cable shaded in blue, Medium in orange and Long in red. The modules hosted by the cable are highlighted with a coloured contour. The numbers 3, 4, 5 indicated the number of wire-bonded e-ports per chip on the given module. Right: Arrangement of a single plane of the {\gls*{ut}} hybrids into power groups sharing a common {\gls*{lv}} output channel. Only one quadrant of the plane is shown. Boxes represent hybrids and the numbers (1 to 4) the power group.

Simulation of the thermal behaviour of a central stave (variant C) at an inlet $\mathrm{ CO_2}$ temperature of \SI{-25}{ $ ^\circ$ {\text{C}} }.

Test beam results for a Type-A unirradiated sensor for tracks with normal incidence. Left: Distribution of collected charge (in ADC counts) at 300 $\text{ V}$ bias. The data are fitted with a Landau distribution convoluted with a Gaussian function. Right: Most probable value of the Landau fit result and hit efficiency versus the applied bias voltage.

The raw and common mode subtracted noise measured in a single hybrid with only that hybrid powered and configured (blue and red curves, respectively) compared with the noise measured in the same hybrid when all hybrids in the stave were operating (black and cyan).

Material scan through the {\gls*{ut}} detector, with thickness given as a fraction of a radiation length. Left: thickness map in $xy$ plane for normal track incidence. Right: thickness as a function of pseudorapidity ($\eta$) as seen from the interaction point.

Expected {\gls*{ut}} strip occupancies for minimum bias events in the sensors near the detector midline ($y=0$).

Front and side views of the 3D model of the {\gls*{sft}} detector.

Map of the total expected ionising dose in $\text{ kGy}$ for an integrated luminosity of 50 $\text{ fb} ^{-1}$ at the T1 station of the {\gls*{sft}} from Fluka simulations of the LHCb detector.

Left: view of a fibre mat with a microscope before and after the side fibres are cut away. Right: a photo of a fibre mat with the polycarbonate end-pieces and {\gls*{sipm}} alignment holes. Figures from Ref. [84].

A cross-section and projections with dimensions (in $\text{ mm}$ ) of a scintillating fibre module and its components. The outlines of the fibre mats and light injection system (LIS) are shown with dashed lines. The nominal gap between fibre mats is shown in a zoomed inset on the left.

The traversed amount of material, averaged over azimuthal angle $\phi$, in units of (left) hadronic interaction length and (right) radiation length as a function of $\eta$ for a sample of simulated $ B ^0_ s \rightarrow \phi\phi$ decays. The total average is also shown for $\eta$ between 2.2 and 4.5 (range limited to the acceptance of the {\gls*{sft}} shown with dashed lines)

An H2017 {\gls*{sipm}} array bonded to a flex cable. The white stiffener is visible on the lower side of the flex cable.

Left: correlated noise probabilities for an H2017 detector as a function of $\rm \Delta V$ . Right: dark-count rate for an irradiated {\gls*{sipm}} as a function of temperature for three overvoltages.

Left: picture of assembled Master, Clusterisation and PACIFIC boards. Right: corresponding schematics of signal data routing.

PACIFICr5q Channel block diagram.

Left: the simulated shaper amplitude as a function of time for a single 10 photoelectron signal at \SI{4}{ $\text{ V}$ } above breakdown. Right: the track-and-hold output values as a function of signal arrival time. The dashed lines indicate the nominal threshold value with respect to the maximum value for two settings (pz5 and pz6, see text).

Left: an example threshold calibration for one comparator of one channel, showing the ratio (number of data above threshold to the total number of data) as a function of the threshold value. The red curve is the result of a fit. The vertical dashed lines through the steps highlight the discrete photoelectron amplitudes. Right: a diagram of the clustering algorithm.

A cutaway view of the cold-box fixed to the fibre module.

The survival probability (attenuation) map of direct and reflected photons in the {\gls*{sft}} after 50 $\text{ fb} ^{-1}$ from simulation.

Simulated number of thermal-noise clusters per 25 $\text{ ns}$ clock cycle as a function of the {\gls*{dcr}} . The curves stem from a power law fit to the data points. The comparison is made for two different models of the PACIFIC settings (in blue the pz5 settings, in red the pz6). The three threshold values used are 1.5, 2.5 and 4.5 photoelectrons.

Left: the single hit inefficiency of the five most efficient fine time bins of all channels measured at several positions across the module. The distribution is fit to a log-normal distribution. The mean and standard deviation of the log-normal distribution is also shown. Right: an example hit position residual distribution fitted with a single (dashed curve) and double (solid curve) Gaussian function. Gap regions have been excluded.

Schematic view of the (left) RICH1 and (right) RICH2 detectors.

Left: the {\gls*{mapmt}} {s} selected for the upgraded RICH detectors with the 2-inch model on the left and the 1-inch model on the right. Right: scheme of the internal structure of the {\gls*{mapmt}} {}.

Left: typical signal amplitude spectra for a pixel as a function of the {\gls*{hv}} value. Right: {\gls*{qe}} curves for a batch of 1-inch {\gls*{mapmt}} {s} from the production: the ultra bi-alkali photocathode allows to reach excellent {\gls*{qe}} values.

Left: CLARO {\gls*{asic}} with its packaging. Right: block schematic of a CLARO channel. The purpose of the dummy amplifier is to give each channel a differential structure, improving the power supply rejection ratio and allowing DC-coupled input to the discriminator.

Exploded view of the {\gls*{ecr}} .

Counting efficiency as a function of the longitudinal magnetic field for an edge pixel, at different values of {\gls*{hv}} , for an {\gls*{ecr}} (left) without and (right) with the magnetic shield.

Schematic view of the {\gls*{ech}}

Left: picture of a fully populated {\gls*{pdmdb}} -R board. Right: schematic view of the {\gls*{pdmdb}} -R board main components. The {\gls*{pdmdb}} -H differs by having one less {\gls*{fpga}} and {\gls*{dtm}} with respect to the {\gls*{pdmdb}} -R.

Pictures of (left) a completed RICH1 column (front view) and (right) a RICH2 column (side view, {\gls*{ec}} {s} on the left), with the photon detector chain and the complete set of services.

The optical geometries of (left) the original and (right) the upgraded RICH1 .

Left: Side view CAD layout of the RICH1 gas enclosure. Right: photo of the RICH1 gas enclosure after its installation in the LHCb cavern.

Pictures of (left) the spherical and (right) the bottom flat mirror assemblies.

Picture of RICH1 columns while being inserted into their support structure, the lower {\gls*{mapmt}} {} chassis. The rails and alignment structures are visible on the left side. The chassis is mounted to the soft-iron magnetic shielding that surrounds the {\gls*{mapmt}} {} region. The push connector for the copper-carried services is visible as well in the left-most column.

Fully assembled and commissioned RICH2 photon detector array. Left: CAD view; right: photograph taken in the assembly area

RICH2 photon detection system inside its enclosure. Left: CAD view; right: photograph

Typical output of the DAC scans procedure. On the left, the calibration of a single CLARO channel with offset bit enabled and no attenuation is shown. The charge (in units of electron charge) corresponding to a threshold DAC code (th) is determined by the linear relation $Q=Q_0 + Q_\text{th}\cdot\text{th}$. On the right, the distribution of the charges (in units of electron charge) corresponding to one threshold step ($Q_\text{th}$) for a RICH2 column is shown. The linearity of the threshold setting as a function of the injected charge is found to be excellent for all attenuation and offset values.

Left: distribution of RICH2 thresholds of the CLARO comparator converted into absolute charge (black). The mean and standard deviations of the distribution are $(207.58 \pm 0.16)\times 10^3$ electrons and $(39.64 \pm 0.10)\times 10^3$ electrons. The threshold settings can be compared to the pixel gain at 900 $\text{ V}$ (red), 950 $\text{ V}$ (green) and 1000 $\text{ V}$ (blue). Right: RICH1 simulated photon detector hit time distribution showing the signal peak (S) and a possible time gate in the front-end electronics.

Left: average quantum efficiency of the {\gls*{mapmt}} {s} used in the RICH detectors. Right: a typical {\gls*{pid}} performance of the kaon identification obtained from the LHCb software for the configuration described in the text (red). A corresponding curve for the Run 2 conditions (prepared using the simulation with LHCb Run 2 geometry and luminosity, as reported in Ref. [118]) is shown for reference (black).

Average expected occupancy per channel for different {\gls*{mapmt}} {s} in the (left) RICH1 and (right) RICH2 detector.

Lateral segmentation of (left) the {\gls*{ecal}} and (right) the {\gls*{hcal}} . One quarter of the detector front face is shown.

Schematic of an {\gls*{ecal}} cell.

Schematic of an {\gls*{hcal}} cell.

Picture of the {\gls*{feb}} .

Block diagram of the ICECAL {\gls*{asic}} .

Schematic of the calorimeter crate.

Control board connection scheme with the new {\gls*{gbt}} {} protocol.

{\gls*{feb}} test beam results. Left: integrated charge (ADC counts) as a function of hit time phase with respect to the internal 40 $\text{ MHz}$ clock. Right: spillover measured in different clock cycles as a function of the beam energy.

Left: energy values measured with the {\gls*{feb}} prototype at a beam test with respect to the reference charge integrator values. Right: the nonlinearity deviation is shown to be less than 1%.

Noise of all channels of a {\gls*{feb}} , measured in laboratory conditions.

{\gls*{node}} communication scheme.

{\gls*{fend}} electronics communication scheme.

{\gls*{node}} block diagram

{\gls*{node}} board layout

Schematic view of the {\gls*{nsync}} architecture and its interface with the GBT chipset [139].

{\gls*{nsbs}} block diagram

Left: functional representation of the {\gls*{nsbs}} {\gls*{ecs}} path. Right: functional representation of the {\gls*{nsbs}} {\gls*{tfc}} path

Finite state machine hierarchy scheme (example for the Side A ).

Mechanical drawing of the tungsten shielding around the beam pipe.

Left: M2R2 new pad chamber prototype. Right: M2R1 new pad chamber prototype.

Typical recovery procedure for a gap (data are from a chamber in region M5R3). The plots on the left show (top) the current and (bottom) the {\gls*{hv}} setting during a period of about three days around the first appearance of the {\gls*{hv}} trip and the subsequent start of {\gls*{hv}} training. The plots on the right show (top) the current and (bottom) the {\gls*{hv}} setting during the full recovery procedure, which lasted about two weeks. The nominal {\gls*{hv}} setting for this gap is 2600 $\text{ V}$ . in normal conditions, the average current in presence of colliding beams is about 0.6 $\text{ \muA}$ .

Current in the {\gls*{mwpc}} during the Malter-effect recovery training: default mixture (full circles) is compared with a mixture containing $\sim$2% of oxygen (open circles).

Upgraded LHCb online system. All system components are connected to the {\gls*{ecs}} shown on the right, although these connections are not shown in the figure for clarity.

View of the {\gls*{pciefty}} board.

{\gls*{tellfty}} firmware architecture. Common blocks in blue; specific subdetector blocks in red.

Logical architecture of the upgrade {\gls*{tfc}} system.

Schematic view of the packing mechanism to merge {\gls*{tfc}} and {\gls*{ecs}} information on the same {\gls*{gbt}} links towards the {\gls*{fend}} electronics. {\gls*{gbt}} words are subdivided into small e-links.

Scope of the {\gls*{ecs}} .

Simplified {\gls*{ecs}} architecture.

{ {\gls*{fend}} }- {\gls*{ecs}} interface

LHCb Run Control panel.

Aggregated data-rate in the LHCb event-builder network as a function of the number of builder units for various nominal event-fragment sizes.

Online data flow [171].

Baseline {\gls*{hltone}} sequence, updated from [177]. Rhombi represent algorithms reducing the event rate, while rectangles represent algorithms processing data.

Track types in the LHCb detector bending plane.

Shapes used for the upgrade calorimeter reconstruction, referred to as cross (left), $\mathit{2\times 2}$ (centre) and $\mathit{3\times 3}$ (right).

Schematic view of the real-time alignment and calibration procedure starting at the beginning of each fill.

Memory consumption of a prototype {\gls*{hltone}} application when run in (left) multijob and (right) multithread modes. The application was run on 3000 events per thread on a reference server node with 20 physical cores and a factor 2 hyper-threading. Note that the y-axis scale of the right plot does not start at zero.

Evolution of the throughput of the CPU-based {\gls*{hltone}} prototype application between autumn 2018 and summer 2019, as measured on a reference server.

Schematic representation of the LHCb upgrade data flowand the related LHCb application, with an emphasis on simulation.

Schematic structure of the Gauss application.

Graphical representation of the dependencies in the simulation software stack in (left) Run 1-2 and (right) in the upgrade, where the experiment agnostic package Gaussino decouples Gauss from Pythia and Geant4 .

Throughput and memory scaling for the generation of {\gls*{prpr}} collisions producing at least a $ D ^0$ meson with beam conditions as found in the 2016 data-taking period in LHCb. Shown are the curves for a shared (P8) and a thread-local (P8MT) interface to Pythia8 , followed by the Geant4 -based simulation. The contribution of the simulation phase, as obtained by reading the generated events from file ("Sim only"), is also shown. Figure is reproduced from Ref. [217].

A flow-chart representing the Lamarr project as a pipeline of parametrisations.

Offline data flow. Figure taken from Ref. [171].

Diagram of a DIRAC release components. The concepts of (top) horizontal and (bottom) vertical extensibility are illustrated.

The LHCb offline data processing workflow.

Throughput of the HLT1 application on a selected subset of current generation {\gls*{gpu}} cards and a representative modern CPU server.

Throughput of the {\gls*{hlttwo}} application and fraction of {\gls*{hlttwo}} resources used by different parts of the reconstruction and selections, measured on a representative {\gls*{hlttwo}} server. A total of 1111 selection algorithms were executed as part of this test.

Long track reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for long reconstructible electrons (blue squares) and non-electron (black dots) particles from $ B $ decays within $2<\eta <5$. Shaded histograms show the distributions of reconstructible particles.

Long track reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for long reconstructible particles from $ B $ decays within $2<\eta <5$. Shaded histograms show the distributions of reconstructible particles.

Ghost rate of long tracks reconstructed by the forward and match tracking algorithms as a function of momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices.

Downstream track reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for reconstructible particles from long-lived particle (marked as strange in the legend) decays within $2<\eta <5$ that have no hits in the {\gls*{velo}} . Shaded histograms show the distributions of reconstructible particles.

Downstream track reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for reconstructible particles from $ B $ / $ D $ decays within $2<\eta <5$ that have no hits in {\gls*{velo}} . Shaded histograms show the distributions of reconstructible particles.

Ghost rate of downstream tracks reconstructed by the forward- and match-tracking algorithms as a function of momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ and number of primary vertices.

Seeding track-reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for long reconstructible electrons (blue squares) and non-electron (black dots) particles within $2<\eta <5$. Shaded histograms show the distributions of reconstructible particles.

Seeding track reconstruction efficiency versus momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices for long reconstructible particles from $ B $ decays within $2<\eta <5$. Shaded histograms show the distributions of reconstructible particles.

Ghost rate of standalone seeding tracks as a function of momentum $ p$ , transverse momentum $ p_{\mathrm{T}}$ , pseudo-rapidity $\eta$ , and number of primary vertices. The prominent peak in the ghost rate at low transverse momentum (top-right panel) results from a combination of geometric and kinematic effects

Relative resolution of the momentum of reconstructed tracks as a function of momentum $ p$ , and pseudo-rapidity $\eta $.

Resolution of the $x$ projection of the impact parameter, $\sigma_{\mathrm{IPx}}$ (left) and $\sigma_{\mathrm{IPy}}$ (right) as a function of the inverse of transverse momentum $1/ p_{\mathrm{T}} $. A minimum bias sample is used for the {\gls*{ip}} resolution study.

Reconstruction efficiency of (left) {\gls*{velo}} and (right) long tracks as a function of the occupancy in the vertex detector and {\gls*{sft}} , respectively.

Top: track reconstruction efficiency as a function of the $z$ position of the primary vertex ($PV_z$), for simulated samples with stand-alone (blue) $ p \text{He}$ and (green) {\gls*{prpr}} , and overlapping (red) {\gls*{prpr}} + $ p \text{He}$ and (orange) {\gls*{prpr}} + $ p \text{Ar}$ collisions. The distribution of $PV_z$ for reconstructible {\gls*{pv}} {s} is also shown (shaded histogram, arbitrary units). Bottom: corresponding rate of fake reconstructed tracks as a function of track momentum $ p$ .

Primary vertex reconstruction (top) efficiency and (bottom) resolution as a function of the $z$ coordinate measured on simulated samples with stand-alone (green) {\gls*{prpr}} , (blue) $ p \text{He}$ and overlapping (red) {\gls*{prpr}} + $ p \text{He}$ and (orange) {\gls*{prpr}} + $ p \text{Ar}$ collisions.

{\gls*{ecal}} cluster reconstruction efficiency versus energy $E$, transverse energy $ E_{\mathrm{T}}$ and $x$ and $y$ position in the {\gls*{ecal}} for reconstructible photons from $ B ^0 \rightarrow K ^{*0} \gamma $ decays.

{\gls*{ecal}} -cluster (left) $x$ position and (right) $y$ position resolution versus energy for reconstructible photons from $ B ^0 \rightarrow K ^{*0} \gamma $ decays.

Merged $\pi ^0$ (left) $x$ position and (right) $y$ position resolution versus energy for ${\pi ^0 \rightarrow \gamma \gamma }$ from $ B $ decays.

Main electron {\gls*{pid}} variables for the {\gls*{ecal}} : distributions for signal and background separately for the variables (left) EcalE/ $ p$ and (right) matching $\chi^2$ of a bremsstrahlung cluster candidate to a track. The distributions of the bremsstrahlung matching $\chi^2$ are conditional on having a cluster candidate in a $3\times 3$ cell grid around the bremsstrahlung track extrapolation.

Performance of the {\gls*{hltone}} inclusive selections as a function of (left) parent-particle transverse momentum and (right) parent-particle decay time. The top row plots are the single-track selections, while the bottom row plots are the two-track displaced vertex selections. The signal topologies are indicated in the legend above each plot. The decay time plots are drawn such that the $x$ axis is binned in units of the lifetime for each hadron in its rest frame.

Performance of the {\gls*{hltone}} muon selections. The signal topologies are indicated in the legend above each plot. In the top row the performance of the dimuon selections is plotted as a function of (left) parent-particle transverse momentum and (right) parent-particle decay time. The decay time plot is drawn such that the $x$ axis is binned in units of the lifetime for each hadron in its rest frame. In the bottom row the performance of the single high- $ p_{\mathrm{T}}$ muon selection is plotted as a function of parent transverse momentum. The shaded histograms indicate the distribution of the parent particle prior to any trigger selection.

Performance of the {\gls*{hlttwo}} inclusive selections. The signal topologies are indicated in the legend above each plot. In the top row the performance of the inclusive displaced-vertex selections is plotted as a function of (left) parent-particle transverse momentum and (right) parent-particle decay time. The decay-time plot is drawn such that the $x$ axis is binned in units of the lifetime for each hadron in its rest frame. In the bottom row the performance of the single high- $ p_{\mathrm{T}}$ muon selection is plotted as a function of parent transverse momentum. The shaded histograms indicate the distribution of the parent particle prior to any trigger selection.

Performance of example {\gls*{hlttwo}} exclusive selections as a function of (left) parent-particle transverse momentum and (right) parent-particle decay time. The signal topologies are indicated in the legend above each plot. The decay-time plots are drawn such that the $x$ axis is binned in units of the lifetime for each hadron in its rest frame. The shaded histograms indicate the distribution of the parent particle prior to any trigger selection.

Animated gif made out of all figures.

Specifications of the upgraded {\gls*{velo}} compared to those of the original version.

Summary of the {\gls*{velopix}} capabilities.

VELO sensor specifications.

Summary of {\gls*{ut}} detector components.

Main design parameters for the {\gls*{ut}} silicon sensors. For Type-D the given length is for outside the cut-out region.

Summary of the specifications of the {\gls*{salt}} {\gls*{asic}} .

Material budget contributions from the scintillating fibre module components. Densities and radiation lengths are taken from Ref. [85].

The average number of clusters per event occurring in the current bunch crossing for different PACIFICr5q models based on a sample of $B_s\rightarrow \phi\phi$ events generated at the given delay from the current crossing.

Simulated performance of the upgraded RICH detectors. For RICH2 , the values are given for the inner detector regions populated with the 1-inch {\gls*{mapmt}} {s}.

Summary of the requirements for the calorimeter analog {\gls*{fend}} .

Maximum output bandwidth ( $\text{ Gbit/s}$ ) per {\gls*{pcie}} interface in the muon system stations at two different luminosity values when zero-suppression is applied; for comparison, also the output rate with no zero-suppression is reported.

Number of {\gls*{tellfty}} per station and of input links per board.

Extrapolated throughput to tape and disk for the \texttt{FULL} , \texttt{Turbo} and \texttt{TurCal} streams.

Summary of the LHCb upgrade computing model requirements. Top section: main assumptions of the model. Bottom section: indicative resource requirements