- Email: [email protected]
Micro channel evaporative CO2 cooling for the upgrade of the LHCb vertex detector
Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Contents lists available at SciVerse ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Micro channel evaporative CO2 cooling for the upgrade of the LHCb vertex detector J. Buytaert a,n, P. Collins a, R. Dumps a, E. Greening b, M. John b, A. Leflat c, Y. Li d, A. Mapelli a, A. Nomerotski a,b, G. Romagnoli e, B. Verlaat a,f a
CERN, CH-1211 Genève 23, Switzerland University of Oxford, Particle Physics, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK c Skobelitsyn Institute of Nuclear Physics, Moscow State University, Moscow 119991, Russian Federation d Department of Engineering Physics, Tsinghua University, Beijing 100084, China e Department of Mechanical Engineering, University of Genoa, Via all’Opera Pia 15, Genoa 16145, Italy f NIKHEF, Science Park 105, 1098 XG Amsterdam, Netherlands b
art ic l e i nf o
Keywords: Silicon sensor Vertex detector CO2 evaporative cooling Micro channel
a b s t r a c t Local thermal management of detector electronics through ultra-thin micro-structured silicon cooling plates is a very promising technique for pixel detectors in high energy physics experiments, especially at the LHC where the heavily irradiated sensors must be operated at temperatures below −20 1C. It combines a very high thermal efficiency with a very low addition of mass and space, and suppresses all problems of CTE mismatch between the heat source and the heat sink. In addition, the use of CO2 as evaporative coolant liquid brings all the benefits of reliable and stable operation, but the high pressures involved impose additional challenges on the micro channel design and the fluidic connectivity. A series of designs have already been prototyped and tested for LHCb. The challenges, the current status of the measurements and the solutions under development will be described. & 2013 Elsevier B.V. All rights reserved.
1. Introduction LHCb is a dedicated flavor physics experiment at the Large Hadron Collider for precision measurements of the decays of charm and beauty hadrons. The Vertex Locator (‘VELO’) plays a key role in identifying secondary vertices, thanks to its excellent track impact parameter accuracy of 13 mm+25 mm/pT. The VELO detector has pioneered the use of evaporative CO2 cooling for high energy physics since its installation in 2007 [1]. The current VELO has 42 double sided modules, each side equipped with 1 sensor and 16 readout ASICs dissipating around 16 W. The silicon sensors are positioned just 7 mm from the colliding beams and must be kept cool (−10 1C) to limit the effect of significant radiation doses. For this reason the cooling system must not only be radiation resistant but low in mass since the cooling components of the modules are within the LHCb particle acceptance. Also, the unique environment, within the LHC vacuum, requires a perfectly leak tight and robust system. In 2018 the VELO will be upgraded along with the rest of the LHCb experiment, to operate with a 5-fold increased luminosity (2 1033 cm−2 s−1) leading to a higher integrated radiation dose n
Corresponding author. Tel.: +41 7675737. E-mail address: [email protected] (J. Buytaert).
(8 1015n1 MeV cm−2) after collecting 100 fb−1 [2]. The temperature of the sensor nearest to the beam must remain well below −15 1C to avoid thermal runaway. Moreover, the new readout ASICs (VeloPix [3]) will lead to higher total heat dissipation per module ( 36 W) located on top of the sensor. To achieve the lower temperature and increased cooling power a new and efficient thermal management of the module is required and a solution is presented which aims to combine the recent advances in micro channel cooling technology with the experience already gained with CO2 evaporative cooling.
2. Evaporative CO2 cooling The CO2 liquid, brought very near to the boiling condition, enters a channel where the heat taken from the environment makes it boil. The temperature of this liquid–vapor mix remains constant as long as liquid is available. A ‘dry-out’ situation can occur when all liquid has been evaporated, because the mass flow is insufficient to absorb the heat flux. CO2 can boil on any point on the liquid–vapor saturation curve between the triple and critical point in the (P,T) phase diagram and therefore, by controlling the pressure between 5 and 73 bar, the boiling temperature can be set between −56 1C and 31 1C. With a CO2 cooling system one can
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.03.069
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
2
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
obtain a very predictable, stable and uniform temperature at the heat exchanger.
Obviously, no extra thermal interfaces to the cooling pipe are
2.1. Micro channel cooling plates Micro channels etched in silicon for high heat fluxes were first realised 30 years ago [4] and the use of two-phase cooling with micro channels was proposed recently [4]. A system of miniature channels, together with an inlet and outlet manifold, are DRIEetched in a silicon wafer and are sealed by bonding a second silicon wafer (direct fusion bonding) or Pyrex (anodic bonding) on top (Fig. 1). This last process flow is fairly standard and was done by a CERN PH/DT team at the Center of Micronanotechnology (CMI), EPFL lausanne. These embedded micro channels lead to many advantages:
needed. These interfaces tend to be, despite all efforts, a source of significant thermal resistance and usually represent a nonnegligible contribution to the total mass. Many parallel small channels represent a large surface for heat exchange. One can layout the channel exactly underneath the heat sources. As a consequence, no heat must flow ‘in-plane’ and the substrate will have smaller temperature gradients and less thermal stress or deformation. With silicon detectors and ASICs, the module is all-silicon and no CTE mismatch occurs between parts.
This integrated approach is very attractive for high energy physics detectors and is being pursued in a number of other domains [5,6]. 2.2. The new VELO pixel module The upgraded VELO will have 25 identical stations positioned along a 1 m long beam collision region. Fig. 2 shows a conceptual layout of one station. A station consists of two identical modules,
Fig. 1. 50 mm 50 mm micro channels in Si-Pyrex.
Fig. 3. Drawing and photograph of the “snake” design etched into the silicon substrate.
Fig. 2. Conceptual layout of one upgraded VELO station. The left diagram shows the cross section with the micro channels embedded in the substrate underneath the ASICs.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
which carry four silicon sensors. The sensors are mounted on both sides of a substrate as shown in the cross section. Each sensor is read out by three VeloPix ASICs. The substrate must be both mechanically stable and thermally very conductive to evacuate the total heat ( 36 W) dissipated by 12 ASICs. The temperature of the Si nearest to the beam must be kept below −20 1C to avoid thermal runaway caused by the irradiation damage. A 400 mm thick silicon plate with micro channels is proposed to obtain a cooling performance of at least 3 W/cm2.
2.3. Prototype of microchannel cooling plates The prototype LHCb samples discussed below were prepared at EPFL (Lausanne) employing anodic bonding [7] of a 0.38 mm thick silicon and 0.5 mm thick Pyrex glass wafers. The sample has a rectangular size of 40 mm by 60 mm. Fig. 3 shows a drawing and a photograph of the sample taken from the glass side making the micro channel network clearly visible. The design has 15 parallel channels, 70 μm deep and 200 μm wide, covering the complete surface and with total length varying between 195 and 229 mm. The inlet and outlet are located in the same corner. All etching is performed on the silicon wafer only. The first part (43 mm) of a channel has a reduced width of 30 μm and constitutes the main flow impedance of the channel. These restrictions ensure equal distribution of the liquid flow across all channels [9]. Figs. 4 and 5 shows SEM photographs of the transition region between these restrictions and the main evaporation channels. The inlet and outlet are 2 mm diameter holes as shown in photographs in Fig. 3. The layout of the distribution and collection area around the in- and out-let must be carefully designed, since the large liquid pressure (up to 73 bar) will result in considerable forces, making these the most critical points for pressure resistance. At the inlet, the restricted channels can be directly connected to the hole avoiding any distribution manifold. At the outlet, small pillars are added in the collection manifold reducing
3
the area and hence the pressure force and improve the bonding strength. These designs demonstrate the great flexibility that can be achieved by the silicon etching process. 2.4. Simulation model A private simulation package “CoBra” [9] for round tubes was used to determine the optimal channel dimension for a given set of input variables, namely the type of liquid, the channel length, the absorbed heat, the temperature of the liquid at the inlet and the vapor quality at the outlet. A volumetric heat transfer conduction can be defined as Λv ¼ Q/(VΔT), where Q¼ total absorbed heat, V¼volume of the channel and ΔT ¼highest temperature difference between channel wall and liquid. A plot of Λv versus channel diameter and for various liquids is shown in Fig. 6. Clearly, CO2 is the optimal liquid for micro channels. This is mainly due to its low viscosity, high latent heat and the high vapor pressure. Ethane was not considered, as we have no operational experience, it is flammable and we can reuse our existing CO2 cooling plant. Since round tubes cannot be etched as micro channels in the Si substrate, they are ‘converted’ to rectangular channels of equivalent cross section. The optimal rectangular channel would be 70 670 mm2, given the restriction on depth imposed by the substrate thickness. The width was limited to 200 mm for pressure resistance up to 100 bar. 3. Experimental setup The measurements were performed at CERN employing an inhouse built CO2 cooling plant ‘TRACI’ [10] and a specialized vacuum chamber with visible and infrared view ports. The TRACI is capable of circulating liquid CO2 at temperatures down to −40 1C with a heat load of 400 W. Nanoport connectors made of PEEK plastic [8] are glued to the outside surface of the silicon wafer for connection to the cooling tubes with diameter of 1.5 mm. The inlet and outlet pressure were monitored at the outside of the vacuum
Fig. 4. SEM photographs of the transition region between the 30 μm wide restrictions and the 200 μm wide evaporation channels.
Fig. 5. Photographs of inlet (left) and outlet (right).
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
chamber. The vacuum of about 1 mbar avoids ice formation and is also the final detector environment. Fig. 7 shows a photograph of the test stand and an assembled micro channel sample with cooling connectors, a metallized silicon heater attached by glue and pt100 temperature sensors.
3.1. Test results 3.1.1. Pressure resistance The rupture pressure as a function of the thickness of the silicon cover and the channel width has been estimated through a structural analysis with ANSYS 2D, as shown in Fig. 8. The target pressure is 100 bar, which limits the maximum channel width to 700 mm for a Si cover of 200 mm. Initial pressure test has failed because of a quality problem of the direct Si–Si bonding. It was decided to temporarily continue testing samples with Si-Pyrex (2 mm) anodic bonding. A sample with channels of 200 mm width and 70 mm depth and a Si cover with a thickness of 200 mm was measured to withstand a pressure of at least 69 bar.
Fig. 6. Volumetric heat transfer conduction Λv as a function of the tube diameter for a channel length of 12 cm at 1.5 W cooling power and −20 1C.
3.1.2. Cooling power To determine the maximum cooling power, the voltage across the heater was ramped up until the dry-out condition was reached. A typical time sequence of the monitored temperatures is shown in Fig. 9. The steep increase of the temperatures occurs when dry-out is reached after 80 s. The power was immediately turned OFF and turned ON somewhat later. Note the very small temperature increase of the Si substrate during the initial 30 s of normal cooling operation. Also, the inlet temperature remains unaffected during the dry-out. After 80 s, the cooling remained critical and the power and CO2 flow were stopped. The maximum power measured before the dry-out was 1.9 Wcm−2 at T ¼0 1C and 0.5 Wcm−2 at T ¼−27 1C. These results are at present limited by a pump failure which degrades significantly the achievable mass flow. Much higher cooling power is expected with increased mass flow, well beyond the required 3 Wcm−2.
3.2. Special fluidic connector The Nanoport connectors used so far are not rated for high pressure (o100 bar) nor for vacuum applications. We designed a metal connector consisting of two parts, attached by three screws that clamp the cooling substrate in between, as illustrated in
Fig. 8. Pressure resistance (P) as a function of the width (W) of the channel and the thickness (tp) of the cover.
Fig. 7. Left: The TRACI cooler and the vacuum vessel. Right: A test sample with heater, temperature sensors and Nanoport connectors.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 9. Temperature variations during a heating sequence.
5
Fig. 11. Concept drawing of a full VELO module with micro channel cooling.
4. Conclusions Evaporative CO2 cooling in silicon micro channels would meet the stringent requirements for the upgraded LHCb VELO detector, regarding thermal efficiency, low temperature (−27 1C), high power density (12.9 W) and low mass. Temperature gradients of less than 4 1C can be achieved in the module. We are addressing the issue of high pressure resistance and reliability to operate in the detector vacuum.
Fig. 10. Drawing of the new connector.
Fig. 10. The holes in the connector are aligned with the inlet and outlet holes of the module and two O-rings ensure leak-tight joints. Two metal pipes are welded into channels in the lower halve.
Acknowledgments We are thankful to P. Lau for preparing the “snake” design drawings; to J. Noel and T. Singh for discussions and preparation of the samples used in these measurements; R.N. Collins for help with preparing the diagrams and to N. Nakatsuka for help with measurements. References
3.3. Future development. New, smaller (15 mm 30 mm) micro channel prototypes have been designed specifically to do pressure tests beyond 100 bar on a large number of samples (4100) with high quality Si–Si direct fusion bonding. A commercial supplier (LETI, Grenoble) has been selected for the production and the samples will be delivered in December 2012. We will also explore the quality assurance aspect, using scanning acoustic microscope techniques and thermal cycling fatigue tests. A possible layout of micro channels for the upgraded VELO module is shown in Fig. 11. It is essentially a combination of two similar layouts as described previously, rotated by 90. It would have a single inlet and two outlets. The two outlets will be recombined inside the connector.
[1] B. Verlaat et al., CO2 cooling for the LHCb-VELO experiment at CERN, CDP 16T3-08, in: Proceedings of the 8th IIF/IIR Gustav Lorentzen Conference on Natural Working Fluids, Copenhagen, Denmark, 2008. [2] The LHCb Collaboration, Letter of Intent for the LHCb Upgrade, CERN-LHCC2011-001, 2011. [3] M. Van Beuzekom et al., NIM A, in: Proceedings of the VeloPix ASIC Development for LHCb VELO Upgrade. [4] D.B. Tuckerman, R.F.W. Pease, IEEE Electron Device Letters EDL-2 (1981) 126. [5] S.G. Kandlikar, Heat Transfer Engineering 26 (8) (2005) 5. [6] A. Mapelli et al., Micro channel cooling for high-energy physics particle detectors and electronics, in: Proceedings of the 13th IEEE Intersociety Conference onThermal and Thermo-mechanical Phenomena in Electronic Systems (ITherm), 2012, pp. 677–683. [7] G. Wallis, D.I. Pomerantz, Journal of Applied Physics 40 (1969) 3946. [8] IDEX Health and Science, 〈www.idex-hs.com〉. [9] B. Verlaat, J. Noite, Design considerations of long length evaporative CO2 cooling lines, GL-2012, in: Proceedings of the 10th IIR Gustav Lorentzen Conference on Natural Working Fluids, Delft, The Netherlands, 2012. [10] B. Verlaat et al., TRACI, a multipurpose CO2 cooling system for R&D, GL-208, in: Proceedings of the 10th IIF/IIR Gustav Lorentzen Conference on Natural Working Fluids, Delft, The Netherlands, 2011.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
Contents lists available at SciVerse ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Micro channel evaporative CO2 cooling for the upgrade of the LHCb vertex detector J. Buytaert a,n, P. Collins a, R. Dumps a, E. Greening b, M. John b, A. Leflat c, Y. Li d, A. Mapelli a, A. Nomerotski a,b, G. Romagnoli e, B. Verlaat a,f a
CERN, CH-1211 Genève 23, Switzerland University of Oxford, Particle Physics, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK c Skobelitsyn Institute of Nuclear Physics, Moscow State University, Moscow 119991, Russian Federation d Department of Engineering Physics, Tsinghua University, Beijing 100084, China e Department of Mechanical Engineering, University of Genoa, Via all’Opera Pia 15, Genoa 16145, Italy f NIKHEF, Science Park 105, 1098 XG Amsterdam, Netherlands b
art ic l e i nf o
Keywords: Silicon sensor Vertex detector CO2 evaporative cooling Micro channel
a b s t r a c t Local thermal management of detector electronics through ultra-thin micro-structured silicon cooling plates is a very promising technique for pixel detectors in high energy physics experiments, especially at the LHC where the heavily irradiated sensors must be operated at temperatures below −20 1C. It combines a very high thermal efficiency with a very low addition of mass and space, and suppresses all problems of CTE mismatch between the heat source and the heat sink. In addition, the use of CO2 as evaporative coolant liquid brings all the benefits of reliable and stable operation, but the high pressures involved impose additional challenges on the micro channel design and the fluidic connectivity. A series of designs have already been prototyped and tested for LHCb. The challenges, the current status of the measurements and the solutions under development will be described. & 2013 Elsevier B.V. All rights reserved.
1. Introduction LHCb is a dedicated flavor physics experiment at the Large Hadron Collider for precision measurements of the decays of charm and beauty hadrons. The Vertex Locator (‘VELO’) plays a key role in identifying secondary vertices, thanks to its excellent track impact parameter accuracy of 13 mm+25 mm/pT. The VELO detector has pioneered the use of evaporative CO2 cooling for high energy physics since its installation in 2007 [1]. The current VELO has 42 double sided modules, each side equipped with 1 sensor and 16 readout ASICs dissipating around 16 W. The silicon sensors are positioned just 7 mm from the colliding beams and must be kept cool (−10 1C) to limit the effect of significant radiation doses. For this reason the cooling system must not only be radiation resistant but low in mass since the cooling components of the modules are within the LHCb particle acceptance. Also, the unique environment, within the LHC vacuum, requires a perfectly leak tight and robust system. In 2018 the VELO will be upgraded along with the rest of the LHCb experiment, to operate with a 5-fold increased luminosity (2 1033 cm−2 s−1) leading to a higher integrated radiation dose n
Corresponding author. Tel.: +41 7675737. E-mail address: [email protected] (J. Buytaert).
(8 1015n1 MeV cm−2) after collecting 100 fb−1 [2]. The temperature of the sensor nearest to the beam must remain well below −15 1C to avoid thermal runaway. Moreover, the new readout ASICs (VeloPix [3]) will lead to higher total heat dissipation per module ( 36 W) located on top of the sensor. To achieve the lower temperature and increased cooling power a new and efficient thermal management of the module is required and a solution is presented which aims to combine the recent advances in micro channel cooling technology with the experience already gained with CO2 evaporative cooling.
2. Evaporative CO2 cooling The CO2 liquid, brought very near to the boiling condition, enters a channel where the heat taken from the environment makes it boil. The temperature of this liquid–vapor mix remains constant as long as liquid is available. A ‘dry-out’ situation can occur when all liquid has been evaporated, because the mass flow is insufficient to absorb the heat flux. CO2 can boil on any point on the liquid–vapor saturation curve between the triple and critical point in the (P,T) phase diagram and therefore, by controlling the pressure between 5 and 73 bar, the boiling temperature can be set between −56 1C and 31 1C. With a CO2 cooling system one can
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.03.069
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
2
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
obtain a very predictable, stable and uniform temperature at the heat exchanger.
Obviously, no extra thermal interfaces to the cooling pipe are
2.1. Micro channel cooling plates Micro channels etched in silicon for high heat fluxes were first realised 30 years ago [4] and the use of two-phase cooling with micro channels was proposed recently [4]. A system of miniature channels, together with an inlet and outlet manifold, are DRIEetched in a silicon wafer and are sealed by bonding a second silicon wafer (direct fusion bonding) or Pyrex (anodic bonding) on top (Fig. 1). This last process flow is fairly standard and was done by a CERN PH/DT team at the Center of Micronanotechnology (CMI), EPFL lausanne. These embedded micro channels lead to many advantages:
needed. These interfaces tend to be, despite all efforts, a source of significant thermal resistance and usually represent a nonnegligible contribution to the total mass. Many parallel small channels represent a large surface for heat exchange. One can layout the channel exactly underneath the heat sources. As a consequence, no heat must flow ‘in-plane’ and the substrate will have smaller temperature gradients and less thermal stress or deformation. With silicon detectors and ASICs, the module is all-silicon and no CTE mismatch occurs between parts.
This integrated approach is very attractive for high energy physics detectors and is being pursued in a number of other domains [5,6]. 2.2. The new VELO pixel module The upgraded VELO will have 25 identical stations positioned along a 1 m long beam collision region. Fig. 2 shows a conceptual layout of one station. A station consists of two identical modules,
Fig. 1. 50 mm 50 mm micro channels in Si-Pyrex.
Fig. 3. Drawing and photograph of the “snake” design etched into the silicon substrate.
Fig. 2. Conceptual layout of one upgraded VELO station. The left diagram shows the cross section with the micro channels embedded in the substrate underneath the ASICs.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
which carry four silicon sensors. The sensors are mounted on both sides of a substrate as shown in the cross section. Each sensor is read out by three VeloPix ASICs. The substrate must be both mechanically stable and thermally very conductive to evacuate the total heat ( 36 W) dissipated by 12 ASICs. The temperature of the Si nearest to the beam must be kept below −20 1C to avoid thermal runaway caused by the irradiation damage. A 400 mm thick silicon plate with micro channels is proposed to obtain a cooling performance of at least 3 W/cm2.
2.3. Prototype of microchannel cooling plates The prototype LHCb samples discussed below were prepared at EPFL (Lausanne) employing anodic bonding [7] of a 0.38 mm thick silicon and 0.5 mm thick Pyrex glass wafers. The sample has a rectangular size of 40 mm by 60 mm. Fig. 3 shows a drawing and a photograph of the sample taken from the glass side making the micro channel network clearly visible. The design has 15 parallel channels, 70 μm deep and 200 μm wide, covering the complete surface and with total length varying between 195 and 229 mm. The inlet and outlet are located in the same corner. All etching is performed on the silicon wafer only. The first part (43 mm) of a channel has a reduced width of 30 μm and constitutes the main flow impedance of the channel. These restrictions ensure equal distribution of the liquid flow across all channels [9]. Figs. 4 and 5 shows SEM photographs of the transition region between these restrictions and the main evaporation channels. The inlet and outlet are 2 mm diameter holes as shown in photographs in Fig. 3. The layout of the distribution and collection area around the in- and out-let must be carefully designed, since the large liquid pressure (up to 73 bar) will result in considerable forces, making these the most critical points for pressure resistance. At the inlet, the restricted channels can be directly connected to the hole avoiding any distribution manifold. At the outlet, small pillars are added in the collection manifold reducing
3
the area and hence the pressure force and improve the bonding strength. These designs demonstrate the great flexibility that can be achieved by the silicon etching process. 2.4. Simulation model A private simulation package “CoBra” [9] for round tubes was used to determine the optimal channel dimension for a given set of input variables, namely the type of liquid, the channel length, the absorbed heat, the temperature of the liquid at the inlet and the vapor quality at the outlet. A volumetric heat transfer conduction can be defined as Λv ¼ Q/(VΔT), where Q¼ total absorbed heat, V¼volume of the channel and ΔT ¼highest temperature difference between channel wall and liquid. A plot of Λv versus channel diameter and for various liquids is shown in Fig. 6. Clearly, CO2 is the optimal liquid for micro channels. This is mainly due to its low viscosity, high latent heat and the high vapor pressure. Ethane was not considered, as we have no operational experience, it is flammable and we can reuse our existing CO2 cooling plant. Since round tubes cannot be etched as micro channels in the Si substrate, they are ‘converted’ to rectangular channels of equivalent cross section. The optimal rectangular channel would be 70 670 mm2, given the restriction on depth imposed by the substrate thickness. The width was limited to 200 mm for pressure resistance up to 100 bar. 3. Experimental setup The measurements were performed at CERN employing an inhouse built CO2 cooling plant ‘TRACI’ [10] and a specialized vacuum chamber with visible and infrared view ports. The TRACI is capable of circulating liquid CO2 at temperatures down to −40 1C with a heat load of 400 W. Nanoport connectors made of PEEK plastic [8] are glued to the outside surface of the silicon wafer for connection to the cooling tubes with diameter of 1.5 mm. The inlet and outlet pressure were monitored at the outside of the vacuum
Fig. 4. SEM photographs of the transition region between the 30 μm wide restrictions and the 200 μm wide evaporation channels.
Fig. 5. Photographs of inlet (left) and outlet (right).
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
chamber. The vacuum of about 1 mbar avoids ice formation and is also the final detector environment. Fig. 7 shows a photograph of the test stand and an assembled micro channel sample with cooling connectors, a metallized silicon heater attached by glue and pt100 temperature sensors.
3.1. Test results 3.1.1. Pressure resistance The rupture pressure as a function of the thickness of the silicon cover and the channel width has been estimated through a structural analysis with ANSYS 2D, as shown in Fig. 8. The target pressure is 100 bar, which limits the maximum channel width to 700 mm for a Si cover of 200 mm. Initial pressure test has failed because of a quality problem of the direct Si–Si bonding. It was decided to temporarily continue testing samples with Si-Pyrex (2 mm) anodic bonding. A sample with channels of 200 mm width and 70 mm depth and a Si cover with a thickness of 200 mm was measured to withstand a pressure of at least 69 bar.
Fig. 6. Volumetric heat transfer conduction Λv as a function of the tube diameter for a channel length of 12 cm at 1.5 W cooling power and −20 1C.
3.1.2. Cooling power To determine the maximum cooling power, the voltage across the heater was ramped up until the dry-out condition was reached. A typical time sequence of the monitored temperatures is shown in Fig. 9. The steep increase of the temperatures occurs when dry-out is reached after 80 s. The power was immediately turned OFF and turned ON somewhat later. Note the very small temperature increase of the Si substrate during the initial 30 s of normal cooling operation. Also, the inlet temperature remains unaffected during the dry-out. After 80 s, the cooling remained critical and the power and CO2 flow were stopped. The maximum power measured before the dry-out was 1.9 Wcm−2 at T ¼0 1C and 0.5 Wcm−2 at T ¼−27 1C. These results are at present limited by a pump failure which degrades significantly the achievable mass flow. Much higher cooling power is expected with increased mass flow, well beyond the required 3 Wcm−2.
3.2. Special fluidic connector The Nanoport connectors used so far are not rated for high pressure (o100 bar) nor for vacuum applications. We designed a metal connector consisting of two parts, attached by three screws that clamp the cooling substrate in between, as illustrated in
Fig. 8. Pressure resistance (P) as a function of the width (W) of the channel and the thickness (tp) of the cover.
Fig. 7. Left: The TRACI cooler and the vacuum vessel. Right: A test sample with heater, temperature sensors and Nanoport connectors.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i
J. Buytaert et al. / Nuclear Instruments and Methods in Physics Research A ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 9. Temperature variations during a heating sequence.
5
Fig. 11. Concept drawing of a full VELO module with micro channel cooling.
4. Conclusions Evaporative CO2 cooling in silicon micro channels would meet the stringent requirements for the upgraded LHCb VELO detector, regarding thermal efficiency, low temperature (−27 1C), high power density (12.9 W) and low mass. Temperature gradients of less than 4 1C can be achieved in the module. We are addressing the issue of high pressure resistance and reliability to operate in the detector vacuum.
Fig. 10. Drawing of the new connector.
Fig. 10. The holes in the connector are aligned with the inlet and outlet holes of the module and two O-rings ensure leak-tight joints. Two metal pipes are welded into channels in the lower halve.
Acknowledgments We are thankful to P. Lau for preparing the “snake” design drawings; to J. Noel and T. Singh for discussions and preparation of the samples used in these measurements; R.N. Collins for help with preparing the diagrams and to N. Nakatsuka for help with measurements. References
3.3. Future development. New, smaller (15 mm 30 mm) micro channel prototypes have been designed specifically to do pressure tests beyond 100 bar on a large number of samples (4100) with high quality Si–Si direct fusion bonding. A commercial supplier (LETI, Grenoble) has been selected for the production and the samples will be delivered in December 2012. We will also explore the quality assurance aspect, using scanning acoustic microscope techniques and thermal cycling fatigue tests. A possible layout of micro channels for the upgraded VELO module is shown in Fig. 11. It is essentially a combination of two similar layouts as described previously, rotated by 90. It would have a single inlet and two outlets. The two outlets will be recombined inside the connector.
[1] B. Verlaat et al., CO2 cooling for the LHCb-VELO experiment at CERN, CDP 16T3-08, in: Proceedings of the 8th IIF/IIR Gustav Lorentzen Conference on Natural Working Fluids, Copenhagen, Denmark, 2008. [2] The LHCb Collaboration, Letter of Intent for the LHCb Upgrade, CERN-LHCC2011-001, 2011. [3] M. Van Beuzekom et al., NIM A, in: Proceedings of the VeloPix ASIC Development for LHCb VELO Upgrade. [4] D.B. Tuckerman, R.F.W. Pease, IEEE Electron Device Letters EDL-2 (1981) 126. [5] S.G. Kandlikar, Heat Transfer Engineering 26 (8) (2005) 5. [6] A. Mapelli et al., Micro channel cooling for high-energy physics particle detectors and electronics, in: Proceedings of the 13th IEEE Intersociety Conference onThermal and Thermo-mechanical Phenomena in Electronic Systems (ITherm), 2012, pp. 677–683. [7] G. Wallis, D.I. Pomerantz, Journal of Applied Physics 40 (1969) 3946. [8] IDEX Health and Science, 〈www.idex-hs.com〉. [9] B. Verlaat, J. Noite, Design considerations of long length evaporative CO2 cooling lines, GL-2012, in: Proceedings of the 10th IIR Gustav Lorentzen Conference on Natural Working Fluids, Delft, The Netherlands, 2012. [10] B. Verlaat et al., TRACI, a multipurpose CO2 cooling system for R&D, GL-208, in: Proceedings of the 10th IIF/IIR Gustav Lorentzen Conference on Natural Working Fluids, Delft, The Netherlands, 2011.
Please cite this article as: J. Buytaert, et al., Nuclear Instruments & Methods In Physics Research A (2013), http://dx.doi.org/10.1016/j. nima.2013.03.069i