WO2012038843A1 - Electrochemically powered integrated circuit package - Google Patents

Abstract

The invention is notably directed to an integrated circuit package (10c). Said package has a layer structure with ICs and electrodes (17) arranged in electrical connection with a layer (16) of the layer structure. The package further comprises one or more fluid circuit sections (19), each meant to receive a respective electrolyte solution (or two distinct solutions, see the dual flow redox mode described below). Each solution involved has soluble electroactive species. A fluid section is designed to receive and allow an electrolyte solution to contact corresponding electrodes, such as to supply power to the ICs, in operation. As electrodes are integrated to the package, electrical power can be supplied close to the ICs, thereby improving efficiency of the power supply. Finally, as a liquid is involved in-situ, suitable heat removal can be contemplated, it being noted that electrical power delivery and heat removal needs are congruent.

Description

ELECTROCHEMICALLY POWERED INTEGRATED CIRCUIT

PACKAGE

FIELD OF THE INVENTION

The invention relates to electrochemically powered integrated circuit packages, e.g., as provided in computer systems such as a datacenter. In particular, it relates to an integrated circuit package powered via electrolyte solutions with soluble electroactive species, such as to supply power to integrated circuits of the package. Consistently, the invention further concerns a computer system equipped with such circuit packages and a method of operating the same.

BACKGROUND OF THE INVENTION

Increasing the integration density in planar integrated circuits (ICs) leads to a reduction of feature sizes and denser packing of components. While transistor switching speed benefits from this evolution, the delay time due to on-chip wiring increases markedly and limits the overall performance of ICs.

Three-dimensional (3D) integration of ICs significantly reduces the wiring length by providing vertical pathways for signal and power transmission, as known. The stacking approach is highly modular and enables the integration of dissimilar technologies in a single cube, and provides massive bandwidth improvement by stacking cache on processor units. On the other hand, 3D integration requires higher power per unit area and connection pins. Another issue is the high cooling demand per unit projected area. Through- silicon vias (TSVs) are an important aspect of 3D integration; they provide vertical signal and power transmission but reduce the active silicon surface area and cause wiring congestions. The extent of vertical integration is severely limited by the accumulated power density such that more than two layers of stacked high-performance logic cannot be easily cooled nor can sufficient power be delivered. Scalable solutions for cooling large numbers of stacked processors have been demonstrated. However, there remains the problem of a viable approach for the supply of electrical power. More generally, powering and cooling are two concerns associated with IC chips.

Prior art solutions to these problems include those of:

US20090092862 ("Integrated Self Contained Sensor Assembly"), that generally relates to methanol fuel cells with a porous proton-exchange membrane. In particular, it discloses a self-contained sensor assembly including a hybrid power module, a transceiver and one or more sensors or detectors. The hybrid power module of the sensor assembly includes a fuel cell and an electronic storage device that may be charged by the fuel cell. The fuel cell membrane and the micro-fuel cell can be directly integrated into an electronic device;

"Microfabricated Fuel Cells to Power Integrated Circuits", Christopher W. Moore, PhD Dissertation, Chemical Engineering, 2005, p. 210,

Georgia Institute of Technology. This publication notably discloses, Figure 2.9, a schematic diagram of an IC with an integrated micro fuel cell. This design shows the fuel being vaporized by waste heat a process that also serves to cool the integrated circuit.

There remains a need for improving known solutions for supplying electrical power to IC chips or within the chip stacks.

Other disclosures, like those of WO2008133655, US20070148527, and Bakir MS, Huang G, Sekar D, King C. "3D Integrated Circuits: Liquid Cooling and Power Delivery". IETE Tech Rev 2009; 26:407-16, illustrate related aspects of the background art.

BRIEF SUMMARY OF THE INVENTION

According to a first aspect thereof, the present invention provides an integrated circuit package, comprising: a layer structure with: electrodes arranged on a layer thereof; and integrated circuits in electrical connection with the electrodes, and one or more fluid circuit sections, each configured to receive at least one respective electrolyte solution with soluble electroactive species therein, and allow said solution to contact at least some of the electrodes, such as to supply power to the integrated circuits, in operation.

In embodiments, the said integrated circuit packaging may comprise one or more of the following features:

at least one of the one or more fluid circuit sections are designed in accordance with a respective electrolyte solution to substantially cool down the integrated circuits in operation;

- at least one of the one or more fluid circuit sections is configured to allow a respective solution to flow along a layer of the layer structure;

at least one of the one or more fluid circuit sections is configured to allow a respective solution to flow between two layers of the layer structure, the said two layers being preferably two layers of integrated circuits arranged as a 3D stack of layers of integrated circuits;

at least some of the electrodes are designed as spacers constraining layers of the layer structure; at least some of the electrodes are arranged on one side of a layer of the layer structure, along which side a solution is allowed to flow by the at least one of the one or more fluid circuit sections;

at least some of the electrodes are arranged on each of the said two layers between which an electrolyte solution of a respective one of the one or more fluid circuit sections is allowed to flow;

electrodes arranged on one side of one of the said two layers comprise both cathodes and anodes;

the layer structure comprises: a printed wiring board; a substrate interconnect; and an integrated circuit chip comprising at least some of the integrated circuits, and wherein at least some of the electrodes and at least one of the one or more fluid circuit sections are arranged on one side of one of: the printed wiring board; the substrate interconnect; or the integrated circuit chip;

the one or more fluid circuit sections is configured to receive a single electrolyte solution, and wherein the electrodes comprise selective cathodes and anodes, said one or more fluid circuit sections configured to allow the said single electrolyte solution to contact the selective cathodes and anodes, forming therewith a single flow redox system;

the one or more fluid circuit sections is configured to: receive two electrolyte solutions, and preferably comprises a membrane arranged to separate the two electrolyte solutions in the one or more fluid circuit sections, and allow the two electrolyte solutions to contact respective subsets of the electrodes, and one of the subsets comprises cathodes, and another one of the subsets comprises anodes, thereby forming a dual-flow redox system;

- the one or more fluid circuit sections are each filled with a respective electrolyte solution, wherein: said respective electrolyte solution comprise a redox couple which is soluble in both its oxidized form and reduced form, a supporting electrolyte which preferably does not exhibits redox processes in a potential range used for supplying power to the integrated circuits, and additives for tuning a redox potential and / or reversibility of the redox couple; and

said respective electrolyte solutions comprises: any one of the following redox couples, or a derivative thereof: Fe²⁺/Fe³⁺, V²⁺/V³⁺, V0²⁺/V0₂ ⁺, Ce³⁺/Ce⁴⁺, Co²⁺/Co³⁺, Cr²⁺/Cr³⁺, Ti³⁺/TiOH³⁺, Cr³⁺/Cr₂0₇ ^2~, BH₄7B0₂, OH7H₂0₂, BrVBr^3", Mn²⁺/Mn³⁺, Ru²⁺/Ru³⁺; and an additive such as acetate, o-phenanthroline, methylphenanthroline, dimethylphenanthroline, bipyridine, ethylenediamine, and said respective electrolyte solutions preferably comprises: any one of the following supporting electrolytes: H₂S0₄, HC1, Na₂S0₄, NaCl, NaOH, K₂S0₄, KOH, and water as a solvent. According to another aspect, the invention is embodied as a computer system comprising: at last one integrated circuit package according to the invention; and an electrochemical power delivery unit in fluid communication with one or more fluid circuit sections of said at last one integrated circuit package, the electrochemical power delivery unit further comprising a convection unit configured to regulate convection of one or more electrolyte solutions in the fluid circuit sections of said at last one integrated circuit package, in accordance with power supply needs thereof, in operation.

According to a final aspect, the invention is embodied as a method of operating a computer system, comprising a step of: providing a computer system comprising an integrated circuit package according to the invention; supplying power to the integrated circuits by forcing at least one electrolyte solution in a respective fluid circuit section of said at last one integrated circuit package to contact at least some of the electrodes of said at last one integrated circuit package.

Methods, devices and systems embodying the present invention will now be described, by way of non-limiting examples, and in reference to the accompanying drawings.

BRIEF DESCRIPTION OF SEVERAL VIEWS OF THE DRAWINGS

- FIG. 1 shows a conventional single-chip package (prior art);

FIGS. 2 - 15 exemplify schematically integrated circuits according to various embodiments of the present invention;

FIG. 16 is a perspective view of a 3D stack of IC chips wherein embodiments of the invention can be implemented;

- FIGS. 17 - 19 depict fluid circuit sections filled with electrolyte solutions that contact electrodes to supply power to IC package, as involved in embodiments;

FIG. 20 is a flowchart depicting steps of a method according to an embodiment of the invention;

FIG. 21 shows a computer system equipped with an electrochemical power delivery unit, according to an embodiment of the invention;

FIG. 22 illustrate a conventional datacenter architecture including an online uninterruptible power supply (prior art); and

FIG. 23 is a scheme for a modified datacenter architecture including an online uninterruptible power supply, reflecting embodiments of the invention.

DETAILED DESCRIPTION OF THE INVENTION

As an introduction to the following description, it is first pointed at a general aspect of the invention, directed to an integrated circuit package. Said package has a layer structure with ICs and electrodes arranged in electrical connection with a layer of the layer structure. The package further comprises fluid circuit sections, each meant to receive a respective electrolyte solution (or two distinct solutions, see the dual flow redox mode described below). Each solution involved has soluble electroactive species. A fluid section is designed to receive and allow an electrolyte solution to contact corresponding electrodes, such as to supply power to the ICs, in operation. As electrodes are integrated to the package, electrical power can be supplied close to the ICs, thereby improving efficiency of the power supply. A high electrical power density can furthermore be achieved, owing to the forced convection of the electrochemical solution contacting the electrodes. Finally, as a liquid is involved in-situ, suitable heat removal can be contemplated, it being noted that electrical power delivery and heat removal needs are congruent. In this respect, the fluid circuits can be optimally designed to substantially cool down the ICs. Thus, a combined solution can be achieved which simultaneously solves the problems of supplying electrical power and cooling.

Such a solution is particularly well suited for 3D ICs, in which interlayer cooling combines with electrochemical power delivery. For instance, a bifunctional water-based coolant can be provided with electroactive redox couples which remain soluble at all stages of the power delivery process. Aqueous redox couples are known which can provide a potential difference of e.g., 1 V between the negative and positive terminals. Heat removal at rates above 200 W/cm can furthermore be achieved by means of forced convective interlayer cooling in e.g., 3D silicon stacks with pins.

In ICs, virtually all electrical power is converted to heat. Thus, as noted above, local cooling and power requirements are congruent, which favors a combined cooling and power delivery. Both heat dissipation and current density provided by an electroactive coolant flow (e.g., pressure-driven) benefit from optimized convective mass transport and increased temperature. Owing to the present approach, critical resources can be freed. For example, in a 3D stack, the number of through-silicon- vias (TSVs) allocated to power delivery (power vias) can be significantly reduced, thus freeing valuable chip area, reducing wiring congestions and minimizing macro redesign. An increasing number of signal vias can be introduced, thereby improving communication bandwidth. Overall, power-related wiring is furthermore simplified due to the need for on-chip wiring only, avoiding interconnects beyond the wafer level.

In addition, the above solution advantageously applies to server operation as used in datacenters, in which the integration of reservoirs for the electroactive coolant can provide uninterruptible power supply (UPS) functionality. The autonomy can easily be scaled by varying the size of the reservoirs. By modifying existing UPS design, power delivery, and voltage conversion in datacenters, significant efficiency improvements are possible.

A further benefit of the electrochemical power supply is the elimination of the need for decoupling capacitors which frees more TSVs as well as valuable space in the processor stacks.

Other features of the invention shall be illustrated in reference to the embodiments of FIGS. 2 - 15. A conventional single-chip package 10' is first described, in reference to FIG. 1, for the sake of understanding. Such a chip has a layer structure, as known. The various layers depicted may for example respectively represent:

- a printed wiring board (or PWB) 11

- A solder balls layer 12;

- A substrate interconnect 13;

- Solder bumps 14 and underfill layer 15; and

- The IC chip itself 16.

Such ingredients are known per se. Electrical power is supplied to the IC through the layers, i.e. from the PWB 11 via solder balls 12, substrate interconnect 13, and the solder bumps 14. Note that, for IC microprocessors, typically the majority of solder balls and bumps are allocated to power delivery (power/ground), while a smaller fraction is allocated to signal transmission.

An improved electrical power supply can be achieved as illustrated in the following embodiments.

First, in reference to FIG. 2, a modified integrated circuit package 10a is shown, which still has a layer structure, evoking the device of FIG. 1. The package may for instance include the same successive layers as in FIG. 1, namely, the PWB 11, solder balls, substrate interconnect 13, solder bumps and underfill, and the chip 16.

However, electrodes 17 are now provided in electrical connection with a layer of the structure, e.g., directly arranged on such a layer. In the example of FIG. 2, electrodes 17 are arranged on PWB 11, which extends on one side of the remaining layer arrangement of the device 10a. ICs are thereby connected to the electrodes (here indirectly, through successive layers 11 - 15).

In addition, a fluid circuit section 19 is provided, which can be filled with an electrolyte solution with soluble electroactive species. This section 19 is configured such as to allow the solution to contact electrodes 17 arranged on top of PWB layer 11 (e.g., an electrode array 17), whereby electrical power can be supplied to ICs layer 16, in operation. In this example, the fluid section 19 is merely a distribution manifold, with an opening facing electrodes and which is connectable to a fluid circuit dispensing the solution, as symbolized here by inlet/outlet arrows 19i, 19o. Suitable fluid circuit sections and connections to a fluid circuit can be obtained following known techniques of e.g., microfluidic sciences. Details of the electrochemistry shall be discussed later in reference to FIGS. 17 - 21.

For example, the manifold could be designed as a cavity in a first silicon block, open on one or more faces thereof (e.g., on a bottom face). Fluid circuits can be provided as grooves in such a silicon block, e.g., open on the top face. Manifolds can else typically be built out of ceramics, metals or hard polymers, etc. A second silicon block can be brought in contact with the first one to close the open grooves. The first block is then laid on top of the structure layer where electrodes are arranged. Many other implementation techniques are possible. For instance, channels for fluid distribution can be machined into metal heat sinks with high thermal conductivity, whereby the heat sinks can be placed on top of the IC packages and the electrodes are isolated from the heat sink material by an insulating layer. Another implementation may include the micromachining of silicon to provide microfluidic channels, with possible arrangements of the electrodes being discussed later in FIGS. 9 - 15.

To summarize the embodiment of FIG. 2, a conventional single-chip package is provided, yet with neighboring electrochemical (EC) power conversion unit and electrode array for power delivery placed on PWB. Such a solution allows for in situ power supply. In addition, it participates to heat removal inasmuch as the solution circulating through the manifold 19 shall capture heat transferred from the chip to the PWB.

The next figures correspond to other embodiments of a modified integrated circuit package 10b - 10η, still having a layer structure as in FIG. 1 or 2. Yet, not all numeral references are repeated, for conciseness.

For example, FIG. 3 shows a conventional single-chip package 10b, wherein a neighboring power conversion unit 19 (e.g., an electrolyte solution manifold in fluid communication with a fluid circuit) and electrodes 17 are provided for power delivery, this time on the common substrate interconnect 13, the latter arranged on top of solder balls and PWB layers, which together extend laterally from the rest of the package. In this example, electrical power is supplied closer to the final consumer 16. Accordingly, better heat removal can be expected, in comparison with the example of FIG. 2.

Next, FIG. 4 depicts another single-chip package 10c with a fluid distribution manifold 19 on top of an electrode array 17 arranged directly on the IC chip 16 for power delivery and cooling. Such an arrangement enables power supply directly to the IC layer; it further allows for the electrolyte solution to substantially capture heat produced at the IC layer, and this, more efficiently than the embodiments of FIG. 2 or 3.

FIG. 6 is similar, i.e., it shows a device lOe with a fluid distribution manifold 19 on top of the layer structure, arranged in fluid communication with an electrode array 17 for power delivery and cooling. Yet, the device represented is now a vertically integrated multi-chip package. For instance, prior art solutions exist which allow for the manufacture of the multilayer in such a multi-chip package. A difference here resides in the integration of electrodes 17 and power conversion unit (fluid section 19) suitably arranged on top of a chip 16.

A skilled person may realize that in the above embodiments (FIGS. 2 - 4, and 6), the fluid circuit sections 19 are configured such as to allow the electrolyte solution to flow along one layer (i.e., 11, 13 or 16) of the layer structure of the IC package 10a - c. Such a configuration is advantageous inasmuch as it allows for designing the power conversion unit merely as an add-on, placed on top of a layer of the structure 10a - c. A skilled person may further realize that such a design does not require substantial changes in the fabrication process of the chip. In addition, in each of the above embodiments (FIGS. 2 - 4, and 6), a single- or dual-flow redox system can be contemplated. In a single flow redox system, the fluid circuit section 19 contains only one electrolyte solution, the latter contacting an array of selective electrodes. If a dual-flow redox system is contemplated, the section 19 has to be filled with two fluids, e.g., separated by a membrane, and contacting respective sets of electrodes, suitably arranged to that aim.

Next, other embodiments can be contemplated, wherein a fluid circuit section is configured to allow an electrolyte solution to flow between two layers of the layer structure, e.g., two IC layers in a 3D stack of ICs. Such embodiments shall be discussed in reference to FIGS. 5, 7- 15. Such arrangements enable power supply close to the IC layers, if not directly. In addition, they allow for the electrolyte solution(s) to more efficiently remove the heat produced at the IC layers.

For instance, FIG. 5 shows a single-chip package lOd with fluid distribution integrated into substrate interconnect 13' for power delivery and cooling. Here the substrate interconnect 13' is configured as two suitably separated layers, e.g., constrained by spacers 23, as known per se. Hence, the double-layer configuration of the substrate interconnect 13' replaces the previous fluid circuit sections and provides a fluid path 19' that is advantageous in many respects (efficient power supply and heat removal). If needed, the dimensions of the double-layer can be adapted to the nature of the electrolyte solution, in order to favor convection thereof and/or prevent undesired capillary effects. Several variants to the above embodiments can be contemplated. For instance, FIG. 7 shows a vertically integrated multi-chip package lOf with dual power delivery and cooling, which actually combines the concepts and advantages of FIGS. 4 and 5 within a multi-chip package.

A still different concept of power delivery and cooling is provided in the embodiment of FIG. 8, proposing a vertically integrated multi-chip package lOg with spacers 23 for interlayer fluid distribution 19'. Numeral reference 19' is hereafter adopted for interlayer fluid circuit section, as opposed to reference 19 for single-layer sections. Here, a housing 21 (e.g.., silicon lateral walls) can be provided which confines the fluid distribution close to the multiple layer structure. The electrolyte solution can therefore be flown between layers of the layer structure, e.g., IC layers. Accordingly, an efficient, combined scheme for power supply and cooling can be achieved for a vertically integrated multi-chip package. A corresponding perspective view is shown in FIG. 16. In FIG. 16, numeral references 24, 25 respectively corresponds to through-silicon vias (TSV) and solder balls. The structure as a whole defines multiple fluid circuit sections 19' that ensures efficient power supply and cooling. While a single general inlet 19i and a single general outlet 19o is depicted in this embodiment, we note that suitable variants may use a dual flow inlet and single outlet, or a dual flow inlet and a dual flow outlet, etc.

Accordingly, summarizing the embodiments of FIGS. 5, 7, or 8 (and 16), it can be realized that at least one fluid circuit section 19' can be designed to allow an electrolyte solution to flow between layers of the device, e.g., two IC layers of a 3D stack. Note that the electrode arrangements pertaining to the interlayer fluid sections are not depicted in these figures, for the sake of intelligibility. In fact, any one of the electrode configurations discussed in reference to FIGS. 9 - 15 can be applied, as to be discussed now. The embodiments of FIGS. 5, 7, or 8 (and 16) can each accommodate single or dual redox flow solutions. Minor changes are nonetheless required for implementing a dual redox flow solution, such as to allow for proper distribution of electrolyte solutions within one or more fluid sections 19'.

Summarizing now the embodiments of FIGS. 2 to 8, it can be realized that at least one fluid circuit section 19, 19' can be designed such that each of the sections 19, 19' may receive and allow one (or two) electrolyte solution(s) to flow along or between layers of the layer structure of the devices 10a - lOg.

FIGS. 9 - 15 show partial views of chip devices lOh - 10η. FIG. 9 depicts an on-chip electrode arrangement 17 for a single electrolyte flow in the interlayer configuration. The layers 13, 16 at stake can e.g., be a dual substrate interconnect (such as 13' in FIG. 5 or 7) or two parallel IC chips 16 as in FIG. 8, suitably constrained by spacers 23. A possible electrode arrangement is an array of alternated electrodes mounted on the inner-side of each of the two layers involved (i.e., the side in contact with the electrolyte flow). Electrodes 17al, 17cl may for instance correspond to anodes and cathodes, respectively, arranged on a first layer 13, 16. Electrodes 17a2 and 17c2 are their counterpart on the second layer 13, 16 (subscript "a" is for "anode" (white), "c" is for "cathode" (black), "1" and "2" correspond to first and second layers, respectively). As evoked above, such an electrode vs. electrolyte configuration requires an adapted selectivity for the electrodes, since a single flow is involved here. Descriptions of single flow or mixed reactant systems have been given previously (e.g. Priestnall 2002, Sung 2007, Mano 2003, see the final reference list). This configuration can for instance be used in the embodiments of FIGS. 5, 7 and 8. The electrode selectivity may, for instance, be implemented by coating one electrode with a porous material, e.g. a zeolite, which allows for the passage of the smaller ions only. Further, electrocatalytic materials may be introduced on the electrodes which promote oxidation or reduction of the electroactive species, respectively.

FIG. 10 shows a similar arrangement, yet with a somewhat different electrode arrangement. Here cathodes 17c 1 are distributed on the inner-side of the first (top) layer, while anodes 17a2 are mounted on the second layer. This configuration again requires electrode selectivity. While it allows for easier manufacture than the configuration of FIG. 9 (only one type of electrodes is provided on a given layer), the electrical circuit obtained is less easily closed, implying somewhat less power efficiency, and requires high electrode selectivity. This interlayer configuration can yet be used in the embodiments of FIGS. 5, 7 and 8.

FIG. 11 shows another electrode arrangement, wherein the electrode functionality is now combined with that of the spacers 23. For instance, an alternated arrangement of anodes 17a and cathodes 17c can be contemplated. The material of the spacers is accordingly adapted. Suitable processes are available which allow to selectively adapt the material used on the spacer surface. The electrode material may be applied to the spacer surface by means of known microfabrication techniques, such as evaporation, sputtering, vapor deposition or plating in combination with anisotropic material removal using inverse sputtering or ion etching. The electrode materials used may be selected according to their compatibility with the solution and activity toward the electrochemical reactions, and may notably comprise carbon and noble metals such as Au, Pt, or Pd. Further electrode materials may include metallic alloys, electronically conductive silicides or carbides such as SiC and B₄C. Selectivity and catalytic activity may be imparted to the electrode materials as mentioned previously. A variant such as that of FIG. 11 advantageously allows for optimizing the spacer functionality and free some valuable area on the inner sides of the layers involved. This configuration can still be used together with embodiments of FIGS. 5, 7 and 8.

In each of the embodiments of FIG. 9, 10, or 11, a single overall electrolyte flow is contemplated (selective electrodes are then needed). Yet, although one speaks here of a "single" flow, the corresponding fluid circuit can decompose into one (FIG. 5) or multiple interlayer fluid circuit sections 19' (e.g., as in FIG. 8), or still comprise an on-top circuit section 19 in combination with an interlayer section 19' (FIG. 7). In other words, one or more fluid circuit sections 19, 19' are involved, yet in combination with a single electrolyte solution. In each case, the said fluid sections are suitably configured, such that the single electrolyte solution may contact selective cathodes and anodes (single flow redox solution).

At variance with the embodiments of FIG. 9, 10, or 11, the embodiments of FIGS. 12 - 15 involve interlayer fluid circuit sections 19' meant to receive distinct electrolyte solutions (not shown). The solutions can e.g., be separated by a membrane 20 or another type of separation, see below. The integration of membrane materials for microfluidics has been described previously (de Jong 2006) and may in particular comprise porous silicon (Moghaddam 2010, Tjerkstra 2000). In each case, the electrolyte solutions are accordingly confined within respective fluid circuit sections 19₁', 19₂', such as to contact respective subsets of the electrodes (i.e., cathodes or anodes), thereby forming a dual-flow redox system.

For instance, FIG. 12 shows an on-chip electrode arrangement for two electrolyte flows confined within respective fluid circuit sections 19₁', 19₂, in an interlayer configuration. The two flows are separated in the plane parallel to the chip surface (dashed line), which denotes either a membrane 20, a diffusion zone 20' or an intermediate layer, or still another type of separation which allows passage of non-active ions in order to close the electrical circuit. The electrode arrangement is otherwise similar to that of FIG. 10.

In FIG. 13, the electrode arrangement 17a2, 17cl is now applied to spacers, as in FIG. 11, except that each spacer is now designed to serve as both a cathode and anode (i.e., on each side of a separation line).

Next, in FIG. 14, another on-chip electrode arrangement for two electrolyte flows in the interlayer configuration is shown. The flows are now directed out of the plane of the drawing and are separated in the plane perpendicular to the chip surface (dashed line), which again may denote a membrane, diffusion zone, intermediate layer or still another type of separation. The interlayer spacers are not shown for clarity. The two flows are confined within respective anodic or cathodic sections 19a', 19c'. Numeral references 27 denote fluid outlets out of the plane of the drawing. Finally, FIG. 15 is similar to FIG. 14, except that the electrode arrangement is now applied to spacers, as in FIG. 11 or 13. Here, a spacer on one side of the separation 20, 20' serves as cathode 17c, the other (on the other side) serving as anode 17a. Again, the flows are confined within respective anodic or cathodic sections 19a', 19c'.

At present, more shall be said about the electrolyte solution(s), in reference to FIGS. 17 - 21.

FIGS. 17 - 19 show details of fluid circuit sections filled with electrolyte solutions 31, 32 that contact electrodes for supply power and cooling. FIGS 17 and 18 pertain to a dual-flow redox system, as previously discussed, while FIG. 19 corresponds to a single-flow redox system. In all cases, Red and Ox represent reduced and oxidized electroactive species, and electron transfer at the electrode/solution interface is indicated by arrows. The flow direction of the electrolyte solutions is from top to bottom.

In FIG. 17, the two electrolyte solutions 31, 32 are separated by a membrane

20. FIG. 17 evokes a conventional redox flow battery configuration with an ion- selective membrane separating electrolyte solutions and identical electrode materials. On the contrary, FIG. 18 invokes a membraneless redox electrochemistry solution, relying on membraneless colaminar flow configuration with identical electrode materials. FIG. 19 pertains to single-flow architecture with selective electrodes (implying different electrode materials which selectively enhance the oxidation and reduction reactions). Note that in each case, substantial heat removal can be obtained via the circulation of the electrolyte solutions, in addition to EC power supply.

Both the dimensions and flow rates of the fluid circuit sections may be adapted according to the actual electrolyte solution used and power needs, e.g., in order to optimize power density and heat removal. Care should notably be taken of capillarity phenomena, i.e., minimal interlayer dimensions may need to be considered. On the contrary, capillarity effects can be exploited to favor convection, e.g., using capillary pumps. Note that such considerations are absent when dealing with gas instead of liquids. However, the heat removal capacity provided by a liquid is much larger than that of a gas.

The electrolyte solution generally consists of an active redox couple, which participates in electronic charge transfer with the electrode; a supporting electrolyte, which provides ions which contribute toward the ionic conductivity of the solution but not to electronic charge transfer; and a solvent, which is capable of dissolving appreciable amounts of both the active redox couple and the supporting electrolyte.

The active redox couple should preferably be present in dissolved form both in the reduced and the oxidized state, and may therefore be selected from couples comprising Fe²⁺/Fe³⁺, V²⁺/V³⁺, V0²⁺/V0₂ ⁺, Ce³⁺/Ce⁴⁺, Co²⁺/Co³⁺, Cr²⁺/Cr³⁺, Ti³⁺/TiOH³⁺, Cr³⁺/Cr₂0₇ ^2~, BH₄7B0₂, OH7H₂0₂, BrVBr^3", Mn²⁺/Mn³⁺, Ru²⁺/Ru³⁺. It is understood that the redox couple may be introduced into the solution in the form of a salt or any suitable derivative, such as a sulfate, chloride, hydroxide, or carbonate. The concentration of the salt should be high enough to provide a high density of electroactive species, preferably 1 mol/L or higher. The concentration may be lower for miniaturized electrode dimensions due to enhanced rates of diffusion, as is known for microelectrode arrays.

The supporting electrolyte should be chosen to match the salt containing the active redox couple, e.g. H₂S0₄, HC1, Na₂S0₄, NaCl, NaOH, K₂S0₄, KOH. The concentration of the supporting electrolyte should be high enough to minimize the resistance of the solution while still avoiding ion association or too high viscosities, preferably 0.5 mol/L or higher. In variants, concentrations of 2 mol/L or higher are preferred, yielding appreciable performances.

The solvent should enable high solubilities of the salt containing the active redox couple and of the supporting electrolyte. For the species listed above, water is a suitable solvent.

In addition to ingredients listed above, further additives may be introduced into the solution to improve performance of the system. In particular, ligand species may be added to adapt the redox potential of the active redox couple, as is known in the science of electrochemistry (e.g. Chen 1981, Chen 1982, Murthy 1989).

In the dual-flow configuration (FIGS. 17 and 18), each of the two solutions preferably contains a different active redox couple. The potential difference between the redox couples is chosen to lie close to the voltage required for powering the IC.

In the single-flow configuration, the single solution contains both active redox couples, with the electrodes exhibiting a high selectivity as described previously.

The separation distance between the electrodes may be arbitrarily chosen, but should preferably be minimized in order to obtain low ionic resistance of the solution between the electrodes. In the interlayer configuration with the electrodes positioned on opposing layers (FIGS. 8, 10, 12, 16), the spacing between the electrodes is defined by the height of the spacers. Preferably, this height lies in the range of 10 μιη

Next, FIG. 20 is a flowchart depicting steps of a method according to an embodiment of the invention. Basically, the method aims at supplying power to an IC package as discussed above by forcing convection of an electrolyte solution. In practice, a computer system is provided with several chip packages (i.e., according to embodiments above), and further equipped with an electrochemical (EC) power unit, step 100. The EC unit is in fluid communication with the chip's fluid sections, such as to enable convection of the electrolyte solution(s). Preferably, the actual power needs can be monitored (step S300) and the fluid convection accordingly regulated, S200, such as to supply suitably adapted electrical power. A simple feedback loop would then be required.

In that respect, FIG. 21 shows a computer system equipped with an electrochemical power delivery unit 110, according to an embodiment of the invention. The proposed redox flow power delivery merely consists of inserting a computer system 100 (e.g., a datacenter comprising hundreds of chip packages such as discussed above, such as 3D silicon stacks with pins) on the electrolyte flow paths of a conventional redox flow battery. Such a battery likely comprises two fluid circuits 111, 112, with respective electrolyte tanks 121, 122. Flow rates in the circuits are regulated by pumps 141, 142. The cells 151, 152 are furthermore contacted by respective electrodes 131, 132 (here for charge only) and are separated by an EC membrane 25, onto which voltage is applied by an additional power supply 120 for re-loading the solutions. As discharge occurs at the computer system 100, the configuration of FIG. 21 actually enables continuous operation. Depending on actual implementation details, such a battery may further allow for high flow rates, high power density, elevated temperature operation and coolant functionality. Interconnection of the computer system 100 on the electrolyte flow paths of the battery may use techniques imported from microfluidic sciences (e.g., tubings and/or PDMS machining, manifolds, etc.), which are known per se.

Using water as liquid coolant, heat removal at rates above 200 W/cm can be achieved by means of forced convective interlayer cooling in 3D stacks with pins (Brunschwiler 2009). Power delivery is achieved by providing electroactive species in the electrolyte solutions which electrochemically discharge at electrodes of the integrated circuit package, as discussed above.

A large number of redox couple pairs exist in the field of aqueous electrochemistry which enables potential differences in the order of 1 V across an electrode pair (CRC Handbook of Chemistry and Physics 2010). Such a voltage is sufficient for supplying present transistors with power. It can further be realized that electrochemical power delivery using aqueous solutions is even more suited for current trends in supply voltage, i.e., geared toward lower values, e.g., 0.6 V.

The benefits of such a technology are substantial, e.g., improvements both on the level of packaging and computer system (e.g., datacenter) infrastructure are expected.

First, concerning packaging benefits: in a high-performance 3D IC, power- carrying TSVs (power vias) are currently estimated to account for 66% of all TSV- occupied area. By transferring the entire power delivery to a bifunctional coolant, this area is freed for enhanced design freedom or integration of additional signal vias. Since electrochemical discharge occurs locally (e.g., on-chip), no interfacing with components beyond the chip level is required. All power-related wiring can thus be simplified and restricted to the chip level.

Second, concerning infrastructure benefits: power delivery is guaranteed by electroactive species in the coolant. By increasing the absolute amount of electroactive species available, the autonomy of the system is enhanced. In that respect, and as represented in FIG. 21, a simple way to provide this autonomy is by introducing reservoirs 121, 122 for the solution into the cooling loop 111, 112. In the case of power failure from the utility side, the electroactive coolant can still be provided until the volumes have been depleted. It can further be measured that the voltage delivered to the chip drops with the depth of discharge of the solution. The optimum operational range of the electrochemical system is at depths of discharge between 10 % and 90 , where the supply voltage of an electrochemical power delivery system consisting of one-electron redox couples varies roughly linearly and predictably by about 0.1 V. A significant loss of supply voltage is only achieved at a depth of discharge approaching 100%. The autonomous operation of the system is thus essentially governed by the size of the reservoirs, which can easily be scaled independently of the local on-chip electrode configuration.

Thus, an uninterruptible power supply (UPS) can be provided. For coupling such an UPS to a computer system, a redesign of current power delivery infrastructure may be proposed with the aim of improving efficiency.

Schemes of a modified vs. conventional datacenter architecture including an online UPS are shown in FIGS. 22 - 23. More specifically, a comparison is made between conventional power delivery architecture in a datacenter (FIG. 22) with combined cooling/electrochemical power delivery (FIG. 23). The energetic efficiencies for components included in the conventional architecture are based on reported values for heavy loads (Pratt 2007) while the energetic efficiencies for the electrochemical system are best estimates. Thin lines represent current flow, while electroactive coolant flow is represented by thick lines. The abbreviations used are standard: Uninterruptible Power Supply (UPS), Alternating Current (AC), Direct Current (DC), Power Distribution Unit (PDU), Power Supply Unit (PSU), Voltage Regulator (VR), Electrochemical Cell (EC).

The percentages given in FIG. 22 correspond to fractions of electrical power remaining after successive conversion steps of the electrical power entering the datacenter. Here, only the electrical power needed to directly supply the processors is shown as an example. In the conventional datacenter architecture, the power delivery to the processors is estimated to be only 36% efficient within the datacenter, mostly due to conversion losses.

By incorporating electrochemical power delivery (FIG. 23), transformation from utility-level 400 V AC to server-level 1 V DC can be performed centrally at the beginning of the conversion chain. For copper-based wires, this voltage transformation would require 400 times more copper to carry the corresponding current with acceptable ohmic loss. However, by converting the electrical energy to chemical energy (DC/EC electrochemical cell in FIG. 23) at a central location in the facility, such excessive wiring is not required. The charged electroactive coolant can be distributed to the rack level at a flow rate of 10 L/min, which corresponds to the transport of about 16 kA of electrical current assuming an exchange of one electron per redox couple pair and 1 mol/L active redox couple concentrations for each of two electrolyte solutions. Following this stage, power is provided in the form of chemical energy, and the rate of energy transport is dictated by the flow rate of the electroactive coolant. Since power requirements for cooling are not included in FIG. 23, the pumping power involved in transporting the electroactive coolant is considered part of the cooling infrastructure and is not included in the efficiency comparison. Further, for large currents in the order of several kA, charge transport by forced convection of ions through pipes of suitable diameters is more energy- and cost-efficient than electronic conduction through metallic wires. At the rack level, the coolant flow is distributed through a manifold to the server level at a lower flow rate to avoid significant pressure drops in narrower channels, until the final conversion of chemical energy to electrical energy occurs in close proximity to the IC, e.g. in the 3D IC interlayer (processor level in FIG. 23).

The estimated efficiency of power delivery to the processors is 77 % using the electrochemical power delivery concept in a datacenter, which provides dramatic improvements compared to conventional architectures. The main reasons for the better performance of the present electrochemical power delivery concept are the removal of the conventional online UPS system and the ability to distribute power at a low voltage level, in the form of chemical energy instead of electrical energy, thus avoiding multiple transformation steps in the power delivery chain.

The storage of electrical energy as chemical energy in the form of soluble electroactive species and its conversion back to electrical energy using forced convection can be related to the field of redox flow batteries (RFBs) or reversible fuel cells. Scientific reviews of the class of galvanic cells are available, whose commercial applications have hitherto been primarily geared toward energy storage in decentralized power grids as well as peak-shaving and load-leveling purposes (Bartolozzi 1989, de Leon 2006). Among the most well-studied and established systems are the iron-chromium (Thaller 1979) and all-vanadium (Skyllas-Kazacos 1987) RFBs.

As noted already, in contrast to the proposed embodiment of electrochemical power delivery of FIGS. 21 and 23, classical RFB systems execute the charge and discharge operations in the same electrochemical cell (i.e., which would correspond to cell 151, 152 in FIG. 21), enabling only intermittent power delivery.

The transfer of RFB electrochemistry to microfluidic applications has only been studied relatively recently, featuring membraneless configurations based on co- laminar flow at low Reynolds numbers and fuel cell concepts with membranes in integrated microreactors (Maynard 2002, Ferrigno 2002).

It is noted that maximum fuel utilization is a key performance indicator in conventional RFB technology, particularly in microfluidic implementation (e.g. Kjeang 2007). In this context, fuel utilization is generally defined as the fraction of chemical energy converted to electrical energy by electrochemical processes upon a single circulation of the electroactive solutions.

In the present case, it can be realized that fuel utilization is not a primary concern compared to power density, since pumping power needs to be invested in order to drive the cooling circuit in any case. For this reason, convective mass transport can be forced to a significantly higher degree than usually considered in microfluidic RFBs. Note that both the electrochemical performance as well as the heat removal performance of the solutions benefit from enhanced mass transport. Further, the greater extent of miniaturization in ICs compared to conventional microfluidic RFBs is favorable in terms of the ionic resistance of the integrated galvanic cell, which enables higher power densities due to reduced ohmic losses.

In summary, the proposed combined cooling/electrochemical power delivery strategy carries a significant potential for exploiting a liquid-cooling infrastructure by simultaneously addressing the power delivery to ICs. Major potential benefits of this concept include an improved bandwidth and efficiency due to simplification of the power delivery infrastructure on the chip and facility scales.

Yet, a disadvantage of the embodiment proposed in FIG. 21 is the need for two separate electroactive species for the anode and the cathode reaction and the need for two storage tanks. This is not a major obstacle but this increases the complexity of the setup. Another problematic component of the redox flow battery is the need for a semi-permeable membrane (de Jong 2006).

Thus, in other embodiments, the membrane is eliminated, the implementation of the invention relies instead on parallel laminar flows that are separated again at the end of the conversion zone (see FIG. 18). This approach is, however, restricted to strictly colaminar flow (Kjeang 2009) and may be somewhat difficult to implement in TSV compatible arrays.

As noted earlier, in other embodiments, the electrodes are functionalized to allow specific anode and cathode reactions to take place selectively at the electrodes from within the same solution (see FIG. 19). This approach may however be difficult to implement for 1 V reactions, due to limited catalytic selectivity (Mano 2003). It becomes easier to implement for lower voltages.

Last but not least, the beneficial effects of the electrochemical power supply can be increased when the systems are operated at elevated temperatures as it is needed for direct thermal re-use. Re-use of thermal energy in, e.g., heating applications further improves the energetic efficiency of the computer system. A 40

K temperature increase increases electrochemical reaction kinetics when combined with efficient mass transport improvements by three to four orders of magnitude.

Such a combination appears particularly advantageous for those applications where very high power-densities are needed to supply enough power to the chips in the chip stacks.

Computer program code required to implement at least parts of the above invention (e.g. the fluid convection regulation) may be implemented in a high-level (e.g., procedural or object-oriented) programming language, or in assembly or machine language if desired; and in any case, the language may be a compiled or interpreted language. Suitable processors include general and special purpose microprocessors. Note that operations that the device, the terminal, the server or recipient performs may be stored on a computer program product tangibly embodied in a machine-readable storage device for execution by a programmable processor; and method steps of the invention may be performed by one or more programmable processors executing instructions to perform functions of the invention.

More generally, the above invention may be at least partly implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations of them. Generally, a processor will receive instructions and data from a read-only memory and/or a random access memory. Storage devices suitable for tangibly embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices, such as EPROM, EEPROM, flash memory devices or others.

While the present invention has been described with reference to certain embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the scope of the present invention. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the present invention without departing from its scope. Therefore, it is intended that the present invention not be limited to the particular embodiment disclosed, but that the present invention will include all embodiments falling within the scope of the appended claims. For example, the electrodes in contact with the electrolyte solutions may be structured using conventional microstructuring techniques in order to achieve a surface area enlargement as well as enhanced mass transport by diffusion. Additional features such as so-called turbulence promoters may be structured close to the electrodes in order to promote mass transport to the electrodes by convection, which is favorable for high power densities.

References:

Brunschwiler et al., Microsystem Technologies 15, 57 (2009).

Pratt and Kumar, The Datacenter Journal, March 28 (2007).

Bartolozzi, Journal of Power Sources 27, 219 (1989).

de Leon et al., Journal of Power Sources 160, 716 (2006).

Thaller, DOE/NASA/1002-79/3 (1979).

Skyllas-Kazacos and Grossmith, Journal of the Electrochemical Society 134, 2950 (1987).

Kjeang et al., Journal of Power Sources 186, 353 (2009).

Maynard and Meyers, Journal of Vacuum Science and Technology B 20, 1287 (2002).

Bazylak et al., Journal of Power Sources 143, 57 (2005).

Kjeang et al., Electrochimica Acta 52, 4942 (2007).

de Jong et al., Lab on a Chip 6, 1125 (2006).

Mano and Heller, Journal of the Electrochemical Society 150, A1136 (2003).

Sung and Choi, Journal of Power Sources 172, 198 (2007).

Priestnall et al., Journal of Power Sources 106, 21 (2002).

Moghaddam et al., Nature Nanotechnology 5, 230 (2010).

Tjerkstra et al., Journal of Microelectromechanical Systems 9, 495 (2000).

Chen et al., Journal of the Electrochemical Society 128, 1461 (1981).

Chen et al., Journal of the Electrochemical Society 129, 63 (1982).

Murthy and Srivastava, Journal of Power Sources 27, 119 (1989).

CRC Handbook of Chemistry and Physics, 90th edition, 2009-2010.

Ferrigno et al., Journal of the American Chemical Society 124, 12930 (2002).

Claims

1. An integrated circuit package (10a - 10η), comprising:

- a layer structure with:

- electrodes (17) arranged on a layer (11, 16) thereof; and

- integrated circuits (16) in electrical connection with the electrodes, and

- one or more fluid circuit sections (19, 19', 19₁ ' , 19₂'), each configured to receive at least one respective electrolyte solution (29, 29', 29") with soluble electroactive species therein, and allow said solution to contact at least some of the electrodes (17), such as to supply power to the integrated circuits, in operation.

2. The integrated circuit package of claim 1, wherein at least one of the one or more fluid circuit sections are designed in accordance with a respective electrolyte solution to substantially cool down the integrated circuits in operation.

3. The integrated circuit package of claim 1 or 2, wherein at least one of the one or more fluid circuit sections is configured to allow a respective solution to flow along a layer of the layer structure.

4. The integrated circuit package of claim 3, wherein at least one of the one or more fluid circuit sections (19') is configured to allow a respective solution to flow between two layers (16) of the layer structure, the said two layers being preferably two layers of integrated circuits arranged as a 3D stack of layers of integrated circuits (10g).

5. The integrated circuit package (lOj, 101, 10η) of claim 4, wherein at least some of the electrodes (17a, 17c, 17a2, 17cl) are designed as spacers constraining layers (13, 16) of the layer structure.

6. The integrated circuit package of claim 3, 4 or 5, wherein at least some of the electrodes are arranged on one side of a layer of the layer structure, along which side a solution is allowed to flow by the at least one of the one or more fluid circuit sections.

7. The integrated circuit package of claim 4 or 5, wherein at least some of the electrodes are arranged on each of the said two layers between which an electrolyte solution of a respective one of the one or more fluid circuit sections (19') is allowed to flow.

8. The integrated circuit package (lOh, lOj, 10m, 10η) of claim 7, wherein electrodes arranged on one side of one of the said two layers comprise both cathodes and anodes.

9. The integrated circuit package (10a - 10c, lOe, lOf) of any one of claims 1 to 3, wherein the layer structure comprises: a printed wiring board (11); a substrate interconnect (13); and an integrated circuit chip (16) comprising at least some of the integrated circuits, and wherein at least some of the electrodes and at least one of the one or more fluid circuit sections (19) are arranged on one side of one of: the printed wiring board (11); the substrate interconnect (13); or the integrated circuit chip (16).

10. The integrated circuit package of any one of claims 1 to 9, wherein the one or more fluid circuit sections (19, 19') is configured to receive a single electrolyte solution (31), and wherein the electrodes comprise selective cathodes and anodes, said one or more fluid circuit sections configured to allow the said single electrolyte solution to contact the selective cathodes and anodes, forming therewith a single flow redox system.

11. The integrated circuit package of any one of claims 1 to 9, wherein the one or more fluid circuit sections (19, 19', 19₁', 19₂') is configured to:

- receive two electrolyte solutions (31, 32), and preferably comprises a membrane arranged to separate the two electrolyte solutions in the one or more fluid circuit sections, and

- allow the two electrolyte solutions to contact respective subsets of the electrodes, and wherein one of the subsets comprises cathodes, and another one of the subsets comprises anodes, thereby forming a dual-flow redox system.

12. The integrated circuit package of any one of the previous claims, wherein the one or more fluid circuit sections are each filled with a respective electrolyte solution, wherein:

- said respective electrolyte solution comprise a redox couple which is soluble in both its oxidized form and reduced form, a supporting electrolyte which preferably does not exhibits redox processes in a potential range used for supplying power to the integrated circuits, and additives for tuning a redox potential and / or reversibility of the redox couple.

13. The integrated circuit package of claim 12, wherein said respective electrolyte solutions comprises:

- any one of the following redox couples, or a derivative thereof: Fe²⁺/Fe³⁺, V²⁺/V³⁺, V0²⁺/V0₂ ⁺, Ce³⁺/Ce⁴⁺, Co²⁺/Co³⁺, Cr²⁺/Cr³⁺, Ti³7TiOH³⁺, Cr³⁺/Cr₂0₇ ^2~, BH₄7B0₂, OH7H₂0₂, BrVBr^3", Mn²⁺/Mn³⁺, Ru²⁺/Ru³⁺; and

- an additive such as acetate, o-phenanthroline, methylphenanthroline,

dimethylphenanthroline, bipyridine, ethylenediamine,

and wherein said respective electrolyte solutions preferably comprises: any one of the following supporting electrolytes: H₂S0₄, HC1, Na₂S0₄, NaCl, NaOH, K₂S0₄, KOH, and water as a solvent.

14. A computer system comprising:

- at last one integrated circuit package (10a - 10η) according to any one of the previous claims; and

- an electrochemical power delivery unit (110) in fluid communication (111, 112, 19i, 19o) with one or more fluid circuit sections (19, 19', 19₁', 19₂') of said at last one integrated circuit package, the electrochemical power delivery unit further comprising a convection unit (141, 142) configured to regulate convection of one or more electrolyte solutions (31, 32) in the fluid circuit sections of said at last one integrated circuit package, in accordance with power supply needs thereof, in operation.

15. A method of operating a computer system, comprising a step of:

- providing (SI 00) a computer system comprising an integrated circuit package according to any one of claims 1 to 13;

- supplying power (S200 - S300) to the integrated circuits by forcing at least one electrolyte solution in a respective fluid circuit section of said at last one integrated circuit package to contact at least some of the electrodes of said at last one integrated circuit package.