Skip to main content

Smart Technologies for eneRgy Efficient Active cooling in advanced Microelectronic Systems

Periodic Reporting for period 2 - STREAMS (Smart Technologies for eneRgy Efficient Active cooling in advanced Microelectronic Systems)

Reporting period: 2017-07-01 to 2019-06-30

The aim of STREAMS project is to bring Europe into the new leading thermal management paradigm and maintain EU position at the forefront of advanced nano-electronic technologies development. With a strong and focused consortium gathering complementary experts in the required domains, STREAMS will focus on the development of a generic active cooling thermal management solution, to keep nano-electronic devices and systems performances at their best, allowing EU to meet IC future challenges.
STREAMS generic system will be applied to newly manufactured servers and is foreseen to reach approximately 15 % of installed systems by 2020, increasing with hardware turnover and data centres strong development. As a consequence of STREAMS project innovative thermal management, cooling systems consumption in data centres is expected to be reduced drastically, lowering CFC consumption of central room air conditioning and therefore lowering greenhouse gases release in the atmosphere.
STREAMS propose to develop a generic SMART, ADAPTABLE, and EMBEDDED active cooling thermal management solution, targeting a 50 % decrease in power consumption, a 70 % decrease in footprint, while keeping the actual high efficiency of liquid cooling cold plate solutions.
Thus, three advanced functionalities will be developed in a Si-based interposer:
- Functionality 1: Versatile microfluidic actuation, to manage liquid cooling of varying thermal scenarios in an energy efficient way;
- Functionality 2: Anticipating thermal map, to enable the real-time spatial control of the thermal state of dies in complex package;
- Functionality 3: Thermal energy harvesting, to provide sufficient energy to power the developed active cooling thermal management solutions.
For the cooling part, efforts have been focused on the development of the two microfluidic solutions (self-adaptive fins and valves ). The performance of the proposed solutions has been numerically assessed and experimentally validated through the WP2 proof of concept. Indeed, in a first stage, the capacity of the Self-Adaptive fins to locally boost the heat transfer, demonstrated numerically, has been experimentally validated. In a second stage, a series of tests have been carried out at fixed flow rate to characterize the cell array cooling solutions with and without microvalves. The result demonstrated that the STREAMS cooling scheme allows to reduce drastically one of the major drawback of microchannel cooling devices: the pressure drops. Moreover, a time dependent and non-uniform heat load scenario has been applied. The analysis of these results demonstrated that all the objectives defined for the STREAMS cooling solution have been reached.
For the thermal map monitoring, the micro-thermal sensors (µTES) have been processed from two thermoelectric (TE) materials: SiGe (used as reference) and nanostructured materials (QDSL: Quantum Dots Superlattices, integrating TiSi2 nanoparticles inside a SiGe matrix). This is the first time that such nanostructured materials QDSL are integrated into a TE device. It was shown that all QDSL-based µTES offer higher performances than SiGe-based µTES. The main objective of WP3 dealed with the “integration of IC compatible passive heat flux sensors (sensitivity up to Se = 100 mV/K) at the interposer level to anticipate thermal map variation (time response ~ 200 ms and lateral spatial resolution ~ 500 μm)”. All results obtained from SiGe and QDSL-based µTESs fit with the objectives of project STREAMS: sensitivity higher than 100 mV/K, response time lower than 4 ms and a lateral spatial resolution equal to 500 µm. Finally, performances of µTESs developed in project STREAMS go over performances of the SoA. For example, the sensitivity density of STREAMS µTES is six times higher than the one of commercial µTES, making these µTES a reference in thermal sensors.
For the energy harvesting part, the objectives were to take advantage of the thermal gradient coming from chips to harvest thermal energy and to use this energy to make the active cooling system as much energy-autonomous as possible. Two sub-tasks have been achieved: to develop, implement and embed a high performance thermoelectric generator and to develop, implement and embed a power management interface circuit to power local functionalities. µTEGs and PMU are fully functionnal, but not together because of an issue not fully determined. For the µTEG, a generated power of 680 µW has been measured for a temperature difference corresponding to 8K, while the target was 1 mW. But it should be considered that the objectives of WP4 are also linked to the development and implementation of new cooling solutions, which are the objectives of WP2. At the beginning of the project, it was decided by the consortium, as well as by the reviewers, to focus on the success of this work package WP2. A traded-off between achieving high cooling performance (WP2) and high power generation by the µTEGs (WP4) thus needed to be performed. Due to this trade-off, optimal conditions for thermal energy harvesting and thus to achieve a temperature difference of 15K, i.e. a high temperature difference, are not feasible. Nevertheless, the power generated by the µTEG developed in this project is considerable. Indeed, compared to the state-of-the-art, our µTEG present the highest useful electrical power density compared to other Si-based µTEGs. Results obtained are very promising.
For the cooling part, the self-adaptive actuation of the cooling device, included in the Radar Innovation platform, has been demonstrated, improving the temperature uniformity of the cooled device while reducing the pumping power. The impact of the solution in other fields (CPV) has been assessed and a H2020 proposal is actually elaborated for a CPV device using this technology (Call RES-1). Finally, a project for the implementation of the cooling solution at a higher TRL has been granted. This work is developed in collaboration with recognized companies of both the electronic sector and the cooling one.
For the thermoelectric part,it was the first integration of SiGe-based nanostructured materials into thermoelectric devices. For thermal sensors, performances of µTESs developed in project STREAMS go over performances of the SoA. For example, the sensitivity density of STREAMS µTES is six times higher than the one of commercial µTES, making these µTES a reference in thermal sensors. For the energy harvesting part, µTEGs developed in project STREAMS present the highest useful electrical power density compared to other Si-based µTEGs. Only the commercial Bi2Te3-based device presents higher useful electrical power density, which is logical as this material is the best one at room temperature. But this material is not CMOS compatible and cannot be used for the applications targeted in this project. Results obtained are very promising and will have to be promoted with higher TRL projects.
For the circuit part, the objective was to advance the state of the art by developing and implementing a shared interface circuit for a multitude of energy harvesters of type voltage source. By the end of the project, a single and thus shared interface circuit was presented that harvests energy from four µTEGs. Its functionality and performance were evaluated using emulated µTEGs implemented by means of DC voltage sources and resistors.
STREAMS concept showing the integration of all functionalities