Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
The LHCb VELO: Performance and radiation damage H.L. Snoek Nikhef, P.O. Box 41882, 1009 DB Amsterdam, The Netherlands
On behalf of the LHCb VELO group art ic l e i nf o
a b s t r a c t
Available online 11 June 2014
LHCb is a forward spectrometer experiment dedicated to the search for New Physics in the decays of beauty and charm hadrons produced by the proton–proton interactions at the Large Hadron Collider (LHC) at CERN. The measurement of the flight distance of these hadrons is critical for the physics program. The VErtex LOcator (VELO) is the silicon detector surrounding the LHCb interaction point and provides excellent resolution of charged tracks and vertex positions. The VELO has been run successfully since installation. The sensors have the first sensitive strips at a radius of 8.2 mm and are exposed to 1 delivered integrated luminosity. maximum radiation doses of 0:6 1014 1 MeV neq =cm2 per fb The performance of the VELO during the first LHC run is described, together with methods to monitor radiation damage. Results from the radiation damage studies are presented showing interesting features, such as an unexpected charge coupling to the second metal layer routing lines after irradiation. The radiation damage has so far no impact on the track reconstruction performance. & 2014 The Author. Published by Elsevier B.V. All rights reserved.
Keywords: LHCb VELO Silicon radiation damage Detector performance Silicon detector
1. Introduction
2. The VELO detector
The LHCb detector [1] is a single-arm forward spectrometer that studies beauty- and charm-flavoured hadron decays at the LHC to search for New Physics phenomena beyond the Standard Model. The analysis of these decays relies on excellent vertex and momentum resolution and particle identification. The silicon VErtex LOcator (VELO) is an LHCb sub-detector positioned around the proton–proton interaction point and extends from 30 cm upstream to 75 cm downstream with respect to the interaction point. It provides excellent vertex and impact parameter resolution with excellent efficiency. Additionally it provides fast pattern recognition for triggering purposes through the identification of displaced vertices that are identifiers for beauty and charm decays. The good performance of the VELO is vital for the LHCb experiment. The LHC has started its proton–proton collisions in Fall 2009 and currently is undergoing its first long shutdown for planned maintenance. The first data run has been very successful. In this period LHCb has received an integrated luminosity of 3.4 fb 1 with increasing beam energies up to 4 TeV. It is planned to resume operations in 2015.
The VELO consists of two detector halves, each containing 21 modules with half-disc shaped sensors. A module consists of two, 300 μm thick, silicon strip sensors, one which measures the radial distance (R-type) from the beam and the other, the azimuthal angle (ϕ-type). The inner radius of the sensors is 7 mm away from the beam axis with the sensitive area between 8.2 and 42 mm radius. All sensors are n þ onn type with the exception of a module with n þ onp type sensors placed most upstream, to test this technology in an operation environment for future upgrade projects. Each sensor contains 2048 strips, with pitches ranging from 38 to 102 μm. Each strip implant is capacitively coupled to a first metal layer that is following the implant along its full length. The first metal layer is connected to a routing line in a second metal layer that carries the signals to one of the 16 Beetle chips [2] located at the outer radius of the sensors. The chips sample the signal at 40 MHz and provide analogue information. The module produces about 20 W and is cooled with a bi-phase CO2 cooling system [3,4]. During normal operations, the silicon sensors are cooled to 7 1C, when the electronics are powered off to about 30 1C, and there are exceptional periods at room temperature when interventions are made to the cooling system. The sensors are operated in a secondary vacuum separated from the primary beam vacuum by a 300 μm thick foil. The foil is corrugated to allow the sensors of the two detector halves to overlap for alignment and efficiency purposes.
E-mail address:
[email protected] http://dx.doi.org/10.1016/j.nima.2014.06.009 0168-9002/& 2014 The Author. Published by Elsevier B.V. All rights reserved.
H.L. Snoek / Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
The two detector halves are retracted by 3 cm to prevent damage from the proton beams during injection and non-stable beam operations. Once the proton beams are declared stable, there is a fully automated procedure to close the detector safely and efficiently. During the procedure, the proton–proton interaction positions are individually reconstructed by the two detector halves. The information of these calculations is used to centre the detector around the beams, both vertically and horizontally. The closing of the detector takes 210 s at the start of each LHC fill.
0.3
3. Performance
45
Projected angle 0-4 degrees Projected angle 7-11 degrees
40
Binary Resolution
Resolution [μm]
35 30 25 20 15
0.931 ± 0.011 0.677 ± 0.009 -0.017 ± 0.002
0.15 0.1
0
LHCb preliminary s = 8 TeV 5
10
15
20
25
30
35
40
N Fig. 2. Resolution along the beam axis (z) of the reconstructed primary vertex versus the number of tracks used in the reconstruction. The measured data points have been fitted with a function in terms of the track multiplicity: σ PV ¼ A=N B þ C.
100
s = 8 TeV
90
2012 Data
80 σ(IPX) [μm]
70 60 50 40 30 20
LHCb VELO Preliminary σ = 11.6 + 23.4/p
10
T
0 0
0.5
1
1.5
2
2.5
3
1/p [c/GeV] T
Fig. 3. Resolution along the x coordinate of the track impact parameter versus the 1/pT of the used tracks.
4. Radiation damage The VELO is subjected to severe radiation due to its close proximity to the LHC beams, leading to damage of the silicon bulk. The radiation fluence scales roughly as 1=r 2 with the radial distance to the beam line. This implies a strongly non-uniform irradiation over the area of a single sensor, with the fluence received by the inner and the outer radius differing by over an order of magnitude. Also the location of the sensor with respect to the beam interaction point influences the received radiation dose. Fig. 4 shows the fluences predicted from simulations as a function of sensor position and radial distance from the beam axis. Up to 5 1013 1 MeV neq =cm2 per delivered fb 1 is expected in the hottest regions of the VELO detector, with an assigned error of 8%, for more details see Ref. [6]. Several methods are in place to monitor the radiation damage in the VELO. We will briefly summarise them here, more detailed information can be found in Ref. [6]. 4.1. Current–temperature scans
10 5 0 30
A (mm) B C (mm)
0.2
0.05
The VELO has operated stably and reliably throughout the first run period of the LHC. The hit resolution has been measured as a function of the local strip pitch and for different projected track angles, see Fig. 1, by measuring the residuals between the reconstructed hit positions and the track projection on the sensor [5]. There is a clear benefit in the resolution from charge sharing over multiple readout strips which can be seen by comparing the hit resolution for different impact angles. For the most favourable conditions (inner sensor region and higher impact angle), the hit resolution obtained is 4 μm, which is a clear manifestation of the excellent performance, tuning and alignment of the VELO. The precise reconstruction of the primary vertex (PV) position, and separating it from secondary vertices in the subsequent decay of particles, is a crucial component of many of the LHCb physics analysis performed. Fig. 2 shows the resolution of the reconstructed PV along the beam axis (z^ ) versus the number of tracks associated to that PV. The vertex resolution has been measured by splitting the track sample associated to a single PV arbitrarily in two, and measuring the difference between the primary vertices reconstructed from the two subsamples [5]. For a typical PV with 25 associated tracks the measured resolution is σ x;y 13 μm perpendicular to the beam axis ^ and vertical (y) ^ directions, and σ z 90 μm along in the horizontal (x) the beam axis. This shows the good performance of the VELO. The impact parameter (IP) is the distance of closest approach of a track to the PV and is widely used in signal selections and the LHCb trigger. The main contributions to the IP resolution are the single hit resolution and the amount of multiple scattering before the first hit measurement. Fig. 3 shows the momentum dependent IPx resolution. It has been measured to be 11:6 þ 23:4=pT μm, this is a clear demonstration of the capabilities of the VELO.
z
0.25 Resolution [mm]
36
LHCb Preliminary 40
50
60 70 80 Strip Pitch [μm]
90
100
Fig. 1. Hit resolution for an R-type sensor versus the local strip pitch for two categories of projected impact angles. The dashed line indicates the binary resolution where no charge sharing information is available.
In the relevant temperature range the bulk current is expected to scale according to IðTÞ p T 2 expð Eg =2kTÞ, where T is the temperature in Kelvin, k is the Boltzmann constant and Eg is the band-gap energy with an expected effective value of 1.21 eV [7]. Fig. 5 shows the current versus temperature behaviour for two sensors (one in each plot) before and after (some) irradiation. Before irradiation one of the sensors has a surface dominated
H.L. Snoek / Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
37
Fig. 4. Left: The Monte Carlo predicted fluence (per cm2) from 1 fb 1 of delivered luminosity as a function of radius for two VELO sensors. Between 1 and 2 fb 1 is expected at the LHCb interaction point each year. Right bottom: The fluence (per cm2) at the innermost radius of the sensor against its distance along the beam axis from the nominal proton interaction point (z). Each point represents an individual VELO sensor. Right top: The fitted exponent for each sensor, where the fluence as a function of radius is fitted with the function Ar k. The coefficient A is a constant and r is the radial position on the sensor in cm. The distribution of the fluence across the sensor varies as a function of z becoming flatter further away from the interaction region.
Fig. 5. The current versus temperature for two VELO sensors operated at the nominal 150 V bias. Before irradiation the sensor in the left plot shows a small exponential contribution while the sensor in the right plot shows a large surface current contribution that is constant versus temperature. After irradiation a large bulk-dominated exponential component is seen for both sensors, the surface current has been largely annealed.
current with a large linear behaviour. After some irradiation both sensors show bulk dominated currents with exponential temperature dependence, it is assumed that the surface defects have largely annealed reducing the surface current component. The effective band-gap energy of the VELO sensors is extracted from the distributions, the weighted average of the measured values is Eg ¼ 1:16 7 0:03 70:04 eV, compatible with the expected value. The largest source of uncertainty is related to the unknown exact operation temperature of the silicon.
initial phase a fraction of the sensors shows large surface currents, while most are bulk dominated. Under the influence of irradiation, the bulk currents increase typically with 1:9 μA per 100 pb 1. The spread in the currents is dominated by variations in operation temperature but is also partly due to the differences in fluence due to the sensor positions. Short annealing periods can be identified that correspond to periods where the cooling system is switched off. The mean of the measured currents is in good agreement with the predicted currents from simulations [8].
4.2. Current versus time
4.3. Effective depletion voltage
The leakage currents of each VELO sensor have been measured regularly since the start of operations. The raw measured currents are shown in Fig. 6. The sensors were operated at the nominal 150 V bias voltage and at a mean temperature of approximately 7 1C. In the
Before the sensor assembly, the depletion voltage of each VELO sensor was measured performing a CV scan [9]. After installation the CV scans can no longer be performed, a technique was developed to monitor the depletion voltage behaviour over time.
T [°C]
H.L. Snoek / Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
20 0 6000
Sensor Temperatures
L [pb-1]
38
4000
Delivered Luminosity
2000
Currents [mA]
0.08 Individual Sensors
0.06
Mean - Measured Mean - Predicted
0.04 0.02 LHCb VELO Preliminary
0 Feb 2011
Apr 2011
Jun 2011
Aug 2011
Oct 2011
Dec 2011
Feb 2012
Apr 2012
Jun 2012
Aug 2012
Oct 2012
Dec 2012
Fig. 6. Currents measured in the VELO for each sensor as a function of time (bottom). The delivered luminosity and the average sensor temperature is shown for the same time period (middle and top). In general, after initial surface current annealing, the current increases with increasing delivered luminosity. The mean of the sensors (excluding the surface dominated sensors) shown in green, agrees well with predictions from simulation shown in magenta. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this paper.)
In special data-collection periods, every fifth module, referred to as test module, is operated at a specific voltage ranging between 0 and 150 V. The remaining modules are maintained at the nominal 150 V and used for track reconstruction. The pedestal corrected charge is summed in the five strips near the intercepts of the track and the test station. For each applied bias voltage of the test sensor the most-probable value (MPV) of the charge distribution is found. At a certain bias voltage the MPV reaches a plateau as the sensor is fully depleted. The effective depletion voltage (EDV) is defined at the bias voltage where the MPV is at 80% of the level of the plateau. This 80% level is chosen as it agrees with the results of the CV scan. The two methods agree within 10 V for unirradiated sensors. EDV values below 20 V cannot be measured with this method due to limited integration time in the readout electronics. Five dedicated data-collection periods have been taken between April 2010 and October 2011 corresponding to delivered luminosities of 0, 0.04, 0.43, 0.80 and 1.22 fb 1. As the fluence delivered to the sensors varies significantly with sensor radius, each sensor is divided into 5 radial regions such that the fluence does not change by more than a factor of two across a region. The results of the measurements are summarised in Fig. 7. Initially the EDVs decrease with the predicted fluence followed by a stable low period after which the EDV starts to rise again. For all n þ onn type sensors the rise in EDV occurs at approximately the same fluence of ð10–15Þ 1012 1 MeV neq . It is assumed that the sensors go through type-inversions near the EDV minimum. The measurements are compared to the Hamburg model [8] that models the irradiation-induced change in the depletion voltage of n þ onn type sensors as a function of time, temperature and fluence. The model predicts three components: short-term annealing, stable damage and reverse annealing. The predictions of the Hamburg model are overlaid on the measured EDV values in Fig. 7 and good agreement is found for lower and higher fluences. The EDV trend of the n þ onp type sensors follows, in general, the same behaviour of the n þ onn type sensors; after an initial decrease there is an increase of the EDV. The underlying physical cause of this understood to be oxygen induced removal of boron interstitial acceptor sites [10,11]. The turning point is at 2 1012 1 MeV neq , much earlier than the n þ onn type sensors. If the EDV increase rates maintains the same at higher fluences, it is expected that the n þ onn type sensor will reach the VELO hardware limit of 500 V after receiving approximately 380 1012 1 MeV neq , this corresponds to between 7 and 8 fb 1 for the most inner region of the closest sensor with respect to the interaction point . Under the same assumptions, the n þ onp
Fig. 7. The measured EDV against expected fluence for regions of n þ onn and n þ onp type sensors of VELO sensors grouped per initial EDV. The EDV from the data is, for each group, compared to depletion voltages predicted by the Hamburg model [8] (in green), with good agreement observed prior to, and after sensor typeinversion. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this paper.)
type sensors will reach the VELO hardware limit at approximately 35 1012 1 MeV neq less fluence than the n þ onn type sensors. 4.4. Cluster finding efficiency All physics analyses at LHCb rely on efficient track reconstruction that is based on the clusters found in the VELO sensors. A cluster is defined as one or a group of adjacent strips with charge above a defined threshold. The data samples described in Section 4.3 are used to measure the Cluster Finding Efficiency (CFE), by looking for clusters on a test sensor at the location of the intercept with a propagated track. Fig. 8 shows the measured CFE versus the expected local fluence after a delivered luminosity of 3.4 fb 1. Before irradiation the CFE was greater than 99% [5], after irradiation the CFE in many sensors decreased significantly. Furthermore, an efficiency loss was visible in the R-type sensors that will be explained in more detail in Section 4.5. 4.5. Charge loss to the second metal layer An unexpected phenomenon has been observed in the R-type sensors. A significant CFE decrease is seen for those sensors after receiving more luminosity, in particular in the outer regions, see Fig. 9. A large data sample has been used to measure the CFE
H.L. Snoek / Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
39
Cluster Finding Efficiency [ % ]
100 98 96 94 92 90 88 0
Cluster Finding Efficiency [ % ]
Fig. 8. The cluster finding efficiency versus the expected local fluence for different regions of the VELO sensors.
8
12 16 20 24 28 32 36 40
~ 2000
RL distance [ μm ]
Fig. 12. The cluster finding efficiency as a function of the distance of the track intercept and the nearest routing line, for several bins of the distance between the particle intercept and the closest strip implant.
100 99 98
LHCb VELO 150 V
97
Delivered luminosity:
96
0.04 fb
95
0.43 fb 0.80 fb 1.22 fb
94
5
10
15
20
25
30
35
40
45
Sensor radius [ mm ]
Fig. 9. The cluster finding efficiency for an R-type sensor as a function of radius for different amounts of received luminosity. 50
100
40
98
94 20
92
10
90
0
88
CFE [%]
96
30 X [mm]
4
86
-10 -50 -40 -30 -20 -10
0
10
20
30
40
50
Y [mm]
Fig. 10. The cluster finding efficiency in small spatial regions of an R-type sensor.
inside to the outside of the sensors are unaffected by the loss in CFE. These areas are known to have no second metal layer. Furthermore, there is a pattern where the loss in CFE increases for larger radii where the density of strip implants decreases (due to the larger strip pitch). The density of the second metal routing lines is almost constant over the sensors, apart from the regions where no second metal layer is present. It is hypothesised that, due to irradiation, modifications to the field line structures have been induced causing charge to flow to the second metal layer that would have otherwise flown to the first metal layer. The reconstructed charge distributions, see Fig. 11, show a deformation of the Landau shape and a growing peak at small charges. The effect is linked to the charge loss to routing lines in the second metal layer of the sensors. Fig. 12 shows the CFE as a function of the distance between the particle trajectory intercept and the second metal routing line. For particles that traverse the sensor underneath a strip implant there is virtually no CFE loss. The highest reduction is when the particle traverses the silicon far away from a strip and close to a routing line. No structures are seen in the CFE spatial distribution for the ϕ-type sensors. The second metal in the ϕ sensors always is situated right on top of the strip implants following in parallel the strip to the outside of the sensor. In the R-type sensors the strips are always perpendicular to the direction of the routing lines. Hence, for the ϕ sensors no CFE loss is present due to charge drifting to the second metal layer. The VELO tracking efficiency has not been impacted by the CFE loss. The induced fake charges in the routing line do not contribute to the construction of fake tracks as they have no correlation to the clusters in adjacent sensors. However, the evolution with further irradiation will be closely monitored.
5. Conclusions The LHCb VELO has operated smoothly in the first data taking period of the LHC. Its excellent efficiency and resolution have been a key ingredient in the physics output of LHCb. Radiation damage has been observed in the detector and is in general consistent with the expectations. The highest fluence areas of the n-type silicon sensors have already type inverted, as expected. A reduction in the cluster finding efficiency has been observed in connection to the second metal layer and is being monitored. No visible impact on track reconstruction efficiency has yet been observed. Fig. 11. The ADC spectrum of all clusters seen in the R-type sensors for data at three integrated luminosities.
in small spatial regions on a sensor at nominal bias voltage. The result is shown in Fig. 10. There is a clear structure visible. The innermost region and narrow regions that stretch from the
References [1] A. Augusto Alves Jr., et al., Journal of Instrumentation 3 (2008) S08005. [2] S. Löchner, M. Schmelling, The Beetle Reference Manual—Chip Version 1.3, 1.4 and 1.5, Technical Report, CERN, 2006.
40
H.L. Snoek / Nuclear Instruments and Methods in Physics Research A 765 (2014) 35–40
[3] M. van Beuzekom, A. Van Lysebetten, B. Verlaat, CO2 cooling experience (LHCb), PoS VERTEX2007, 2007, p. 009. [4] E. Jans, Operational aspects of the VELO cooling system of LHCb, PoS VERTEX2013, 2013, p. 038. [5] K. Akiba, et al., Performance of the LHCb vertex locator, Journal of Instrumentation, in preparation. [6] A. Affolder, et al., Journal of Instrumentation 8 (2013) P08002 arXiv:1302. 5259.
[7] A. Chilingarov, Generation Current Temperature Scaling, Technical Report PHEP-Tech-Note-2013-001, CERN, Geneva, January 2013. [8] R. Wunstorf, et al., Nuclear Instruments and Methods in Physics Research Section A 315 (1–3) (1992) 149. [9] P. Turner, T. Bowcock, G. Patel, VELO Module Production: Sensor Testing, Technical Report, CERN, 2007. [10] F. Lemeilleur, et al., Nuclear Physics B—Proceedings Supplements 32 (1998) 415. [11] M. Lozano, et al., IEEE Transactions on Nuclear Science NS-52 (2005) 1468.