Method for Cooling of Chips Using a Plurality of Materials

The present invention relates generally to a new apparatus and method for customized cooling of chips using a plurality of thermally conductive materials. More particularly, the invention encompasses an apparatus and a method that provides customized cooling of a MCM (Multi-Chip Module) by using a plurality of thermally conductive fluids.

Semiconductor devices are becoming smaller and more dense with the evolution of new technology. However, increases in circuit density produce a corresponding increase in overall chip packaging strategies in order to remain competitive. Chip and chip carrier manufacturers are therefore constantly being challenged to improve the quality of their products by identifying and eliminating problems, reducing package size and weight, decreasing package costs, providing improved thermal efficiencies and by producing better and more advanced chips. Whereas significant improvements are being made to eliminate systematic problems by reducing process variability. Process improvements alone are not sufficient to eliminate all the problems which effect both performance and reliability.

One way to increase performance and reliability is to provide the shortest and most efficient thermal cooling path for the integrated circuit chips. This could be done by bringing the chip physically as close as possible to the heat sink. Another way would be to provide more efficient cooling of the chip. However, when the chips are brought closer to the heat sink, means also have to be provided to securely provide a thermal contact between the chip and the heat sink. In some cases highly thermally conductive epoxies have been used to provide a better thermal contact between the chip and the heat sink. However, this would create a problem if different types of chips are present on the MCM, because some chips may be "under cooled" while others may be "over cooled".

The introduction of MCMs has helped packages perform better electrically, but at the same time it has made them more difficult to cool. Moving chips from single chip modules (SCMs) to MCMs greatly affects the module heat flux density, even when chip flux densities are unchanged. The cooling of large MCMs has been the subject of many inventions, such as, for example, the use of pistons to cool the chips, and most recently FPC (flat plate cooling). While the previous approaches used tight tolerance mechanisms to provide a thermal/mechanical path for the heat to flow from the chip to the cooling fluid (air or water).

Flat plate cooling uses much fewer parts, and it uses a single thermal compound to provide the cooling path between the chip and the flat plate. In general, the concept of FPC is to install a flat thermally conductive "hat" just above, and parallel to the chip back side surfaces. The gap between the chips and the hat is typically filled with a single thermally conductive material. The single thermally conductive material (thermal compound, grease or paste) is installed in specific locations with sufficient quantity so that the back side of each chip is entirely covered. Although the compliant thermally conductive material has a thermal conductivity much lower than the metals used in previous inventions, it is able to provide better thermal performance at a lower cost because the thickness of the paste filled gap is small, there are fewer interfaces for the heat to cross, and it is insensitive to chip pitch.

A typical application of FPC would be to have an array of chips on a substrate, where each had the same size and height, cooled by FPC. In this application, each chip would have the same thermal resistance to the cooling hat.

One example of flat plate cooling is Umezawa, et al., U.S. Pat. No. 5,023,695, where a heat conducting compound is present between the chips and the flat cooling plate.

Research Disclosure, No. 270, Publication No. 27014 (October 1986), the disclosure of which is incorporated herein by reference, discloses a stick-on heat sink. A heat sink is attached to a module by sliding the module into the heat sink and where the edges of the heat sink snap close to secure the heat sink to the module. It is also disclosed that an adhesive or double sided tape could also be placed on the bottom surface of the heat sink to assure intimate contact between the module and the heat sink.

U.S. Pat. No. 4,092,697 (Spaight), the disclosure of which is incorporated herein by reference, discloses placing a film of thermally conductive material between the chip and the heat sink or heat radiator.

U.S. Pat. No. 4,233,645 (Balderes et al.), discloses placing a block of porous material which is impregnated with a suitable liquid between the chip and the heat sink to provide a thermally conductive path.

U.S. Pat. No. 4,849,856 (Funari et al.), the disclosure of which is incorporated herein by reference, discloses a direct chip to heat sink attachment process where a thermally conductive adhesive is used to directly secure the heat sink to the chip.

U.S. Pat. No. 4,939,570 (Bickford et al.), the disclosure of which is incorporated herein by reference, discloses another direct chip to heat sink attachment process where a thermally conductive adhesive is used to directly secure the heat sink to the chip.

U.S. Pat. No. 4,999,699 (Christie, et al.), the disclosure of which is incorporated herein by reference, discloses solder interconnection whereby the gap created by solder connections between a carrier substrate and semiconductor device is filled with a composition obtained from curing a preparation containing a cycloaliphatic polyepoxide and/or curable cyanate ester or prepolymer thereof; filler having a maximum particle size of 31 microns and being at least substantially free of alpha particle emissions.

U.S. Pat. No. 5,249,101 (Frey, et al.), the disclosure of which is incorporated herein by reference, discloses a coverless chip carrier which uses at least two encapsulants. The first encapsulant is used to provide flip-chip fatigue life enhancement. The second encapsulant is used to provide limited environmental protection. A third encapsulant is also required for carriers using peripheral leads to contain the second encapsulant prior to curing. Also disclosed is that the encapsulant have a CTE (Coefficient of Thermal Expansion) which is within 30 percent of the CTE of the solder balls.

U.S. Pat. No. 5,325,265 (Turlik, et al.) discloses that cushions could provide a thermally conductive path between an exposed back face of a chip and a heat sink. Also disclosed is the fact that a solder preform may be used for the cushions with the preforms being held in place using grooves or cavities or a combination thereof.

The inventors of this invention, however, are using an entirely different approach to solve this age old problem. They are customizing the thermal cooling area by providing a different thermal coefficient thermal fluid or paste or compound for each chip so that each chip could be cooled within its specific specifications. Also disclosed is that by varying the depth of the gap or the blind hole which is filled with the individualized thermal fluid one could also further fine tune the individual cooling regimes.

Furthermore, the structure and process of this invention offers several advantages over the prior art. For example, it provides a customized thermal path with a simplified modular construction which allows the ease of workability or repair of the assembled module.

This invention is a novel method and an apparatus for providing customized cooling by providing appropriate thermal fluid for the chip and also by varying the depth of the gap or the blind hole that is filled with the customized thermal compound. This becomes especially important for MCM where there are customized or different chips on the module or substrate or chip carrier.

Therefore, one purpose of this invention is to provide an apparatus and a method that will provide a direct thermal path to the heat sink or thermal cap.

Another purpose of this invention is to provide a customized cooling to individual chips on a chip carrier or substrate, and to ensure a secure thermal contact between the substrate and the heat sink or the thermal cap.

Still another purpose of this invention is to have a very economical and efficient thermally conductive path between a chip and a heat sink or a thermal cap.

Yet another purpose of this invention is to increase the available area on the substrate or the chip carrier for device joining, for example, active devices, such as, chips, or passive devices, such as, capacitors, etc.

Still yet another purpose of the invention is to provide a method and apparatus for ensuring the bond integrity between the substrate and the heat sink or the thermal cap.

Yet another purpose of this invention is the ability to rework or repair the completed or assembled module.

Therefore, in one aspect this invention comprises an apparatus to provide a customized thermally conductive path between a plurality of chips and a thermal cap, comprising at least a first thermally conductive material that is present between a first chip and said thermal cap and at least a second thermally conductive material that is present between a second chip and said thermal cap, and wherein the thermal conductivity of said first thermally conductive material is different than that of said second thermally conductive material.

In another aspect this invention comprises a method to provide a customized thermally conductive path between a plurality of chips and a thermal cap, comprising the steps of:

(a) securing said chips to a substrate,

(b) placing a first thermally conductive material on a first chip, and a second thermally conductive material on a second chip, wherein the thermal conductivity of said first thermally conductive material is different than the thermal conductivity of said second thermally conductive material, and

(c) securing said thermal cap to said substrate.

The features of the invention believed to be novel and the elements characteristic of the invention are set forth with particularity in the appended claims. The figures are for illustration purposes only and are not drawn to scale. The invention itself, however, both as to organization and method of operation, may best be understood by reference to the detailed description which follows taken in conjunction with the accompanying drawings in which:

FIG. 1, illustrates one prior art scheme for cooling chips on an MCM.

FIG. 2, illustrates another prior art scheme for cooling chips on an MCM.

FIG. 3, illustrates one preferred embodiment of this invention.

FIG. 4, illustrates another preferred embodiment of this invention.

FIG. 5, illustrates yet another preferred embodiment of this invention.

FIG. 6, illustrates still yet another preferred embodiment of this invention.

IBM's multilayered ceramic (MLC) electronic packages are among the most technically advanced electronic packages in the industry. This invention describes one way to reduce cost of such packages with an increase in their performance. Packaging methods which reduce cost advantageously increase the availability of such electronic packages in the marketplace. As a person skilled in the art knows that increased packaging density is typically achieved by optimum utilization of the real estate of the substrate or module or chip carrier.

It is also well known that chip temperatures impact device function, speed, and reliability. Therefore, it is important to have all the chips in a module operate inside a relatively narrow temperature window for best module performance. This window is set by the electrical engineer, early in the design process. It is the responsibility of the thermal engineer to design the cooling in such a way that makes all the chips operate within this temperature window at all times.

A multi-chip module that is designed to provide sufficient cooling to the most challenging chip in the module will often have much more cooling than is required at less challenging chip sites, causing them to be "overcooled". If this is not taken into account then some of the other chips could be running "hotter", while others could be running "cooler". Therefore, for a MCM having a variety of chips unless the cooling is customized, some of the chips would be operating at temperatures below or above their functional specification, unless chip power is reduced or is increased, or the temperature specification range is increased or reduced, or both.

It is not an easy task to make sure that all the chips operate within this relatively narrow temperature windows. The difficulty stems from the fact that chip operating temperatures are a function of many factors, including chip power, chip size, package design, placement within the package, and proximity of other chips. The primary driver is chip power. The chips within an MCM typically have a range of nominal powers, and their actual powers can vary significantly from their nominal power dissipation. Statistical variations in chip power (which is typically on the order of +/-30 percent) makes chip cooling customization more complicated as the thermal engineer must allow for this variation, and make sure that the window is met for the highest power chip, the lowest power chip, and the rest of the chips on the module. There are location and statistical variations associated with the cooling design itself. All this has to be taken into account in order to ensure that each chip will operate within the specified temperature window at all times.

It was determined by the inventors that the chip temperature rise over the temperature of the cooling hat is approximately proportional to the product of the chip power by the thermal resistance of the compound between the chip and the module or thermal cap. Mathematically, it is expressed as ##EQU1##

where, T is temperature,

P is power,

L is the depth of the gap between the chip and the module or thermal cap,

k is the thermal conductivity of the compound between chip and the cap, and

A is the chip area.

Therefore, the inventors have found that the chip temperatures could be adjusted by varying L, to vary the Delta T. The inventors discovered that this L could be varied by drilling blind holes in the thermal cap right above the chip where the chip temperature needs to be increased. These blind holes can be of any shape, such as, for example, circular, rectangular, triangular, polygonal, to name a few. The depth of the cut depends on how much the chip temperature needs to be increased or decreased.

Furthermore, k can be varied by using different thermal conductivity materials on the different chips within the MCM. For example, two or more different thermal compounds can be used in the same MCM. Higher power chips would use higher thermal conductivity compounds, and lower power chips would use lower thermal conductivity compounds. The conductivity of the compound selected is a function of how much the chip temperature needs to be increased. For each chip, there is a range of compound conductivities that can be used to get the chip to operate within its temperature specifications.

However, the results given by the above mentioned equation do not take into account how the power and/or temperature of one chip (or a group of chips) can impact the hat or thermal cap temperature, and therefore in turn the temperatures of other chips. This interaction between chips is due mainly to the heat spreading in the module cap (which is typically made of a good heat-conducting material) and can be accounted for in a three dimensional scheme. In the process of customization, the thermal engineer should use this type of scheme to account for the locations and powers of the chips inside the module.

FIG. 1, illustrates one prior art scheme for cooling chips on an MCM. As illustrated in FIG. 1, a plurality of semiconductor elements 20, such as, chips 20, are secured to a substrate or module 10, via a plurality of solder connections 22, such as, for example, C4 (Control Collapse Chip Connection), solder balls, solder column connection (SCC) or ball grid arrays (BGA) 22.

The substrate 10, could also have one or more electronic device(s), (not shown), such as, for example, decoupling capacitors, that are also electrically connected to the substrate 10, via solder connections 22. For some applications the solder balls 22, could be protected with one or more partial encapsulants (not shown), which encapsulants are well known in the art.

A thermally conductive material 23, typically a thermal compound or thermal grease or thermal paste 23, is usually applied over the exposed surface or back-side of the chip 20, such that a direct thermal contact is made between the chip 20, and the roof or thermal contact area 27, of the cap or cover or heat sink 25, when the cover 25, is placed over to cover and/or protect the chip 20.

The cap or cover 25, typically has an extension or ledge or wall 26, that provides the mechanical connection between the substrate 10, and the heat sink 25, which connection is well known in the art, for example, the extension 26, could be secured to the substrate 10, using solder or some sort of an adhesive 28. The substrate 10, typically has electrical connection means or I/O (Input/Output) means 12, such as, for example, pads, pins, solder connections, etc., to name a few.

The cap or cover 25, is typically a metallic or ceramic cap, that is placed over the chips 20, and is properly secured to the peripheral edge surface of the substrate 10. This is done primarily to prevent mechanical and chemical injury to the chip 20, solder balls 22, decoupling capacitors, any encapsulant that may be present, and any exposed metallurgy or circuitry on the substrate or module 10. The thermal cap 25, is also provided to be a heat sink for the chips on the chip carrier or substrate 10. It is also well known that a leak in the cap 25, or any misalignment of the thermal cap 25, may result in irrecoverable module yield losses. These losses could be substantial for an expensive module. A picture-frame type area on the top surface of the substrate 10, is required to specifically seal the cap 25, to the substrate 10, using techniques well known in the art, such as, for example, a cap sealant, solder or adhesive 28, to name a few. For a typical MCM this frame type surface could be between about 2 mm and about 6 mm wide. Therefore, the placement of all devices, such as, for example, chips 20, decoupling capacitors, are restricted to be within this picture frame area, which is typically only between about 50 percent and about 70 percent, of the area that would otherwise be available for additional or larger devices. Additionally, the cap or cover 25, typically adds between about 30 percent and about 50 percent, to the overall height of the module. It should be noted that thermal compound 23, must be placed between the chip 20, and the cap 25, to provide a most efficient heat transfer path, to the cap or heat sink 25. It is to be understood that the presence of the thermal cap or cover 25, adds additional weight to the completed or assembled module. It should also be noted that care needs to be so taken that the cover or cap 25, does not interfere with other electronic components that may be on or near the substrate 10.

FIG. 2, illustrates another prior art scheme for cooling chips on an MCM, where the heat sink or cap 35, has a plurality of extensions or partitions 34, that form uniform cavities or blind holes 33. The thermally conductive paste or grease 23, typically fills the uniform cavity 33, and provides the heat or thermal transfer path from the chip 20, to the heat sink or cap 35. Ledge or extension 36, provides securing means to secure the heat sink or cap 35, to the substrate 10.

With the advancement of MCM (Multi-Chip Module) technology there is a desire to now place chips on a module where each of the chips could be different. These chips could be different for a variety of reasons, such as, for example, in their size or power or performance, etc., and therefore their ability to conduct heat would normally be different.

FIG. 3, illustrates one preferred embodiment of this invention. For the purposes of illustration it shows a Flat Plate Cooling module with different thermal compounds. The FPC hat 25, cools the chips 41, 42, 43, 44 and 45, that have been secured to the substrate 10, by transferring the heat that is generated by these chips through the thermal compounds 51, 52, 51, 53 and 53, respectively, to the thermal cap 25. Chips 41 and 43, have a higher power and therefore they use the higher thermal conductivity compound 51. Whereas, chip 42, has medium power, and uses a medium thermal conductivity compound 52. While, chips 44 and 45, have lower power, and therefore use a lower thermal conductivity compound 53.

FIG. 4, illustrates another preferred embodiment of this invention where the heat sink or thermal cap 35, has a plurality of uniform blind holes or cavities 31, creating a thermal contact area 33. The blind holes 31, are above the chips. Typically, the uniform blind holes 31, above their corresponding chips are also customized to the real estate requirements or the physical requirements of the chips. Each of the chips 41, 46, 47 and 48, could be of the same size or could be of a different size. However, their power requirements would be different so that there is a need to customize their cooling profiles in order to effectively and efficiently cool these chips.

In order to provide the optimum thermal path appropriate thermally conductive material is placed between the chips and the thermal cap 35. For example, chip 41, could be a high performance chips and therefore may need to be cooled more. Such a chip 41, would require a high conductivity thermal grease or paste or compound 51. Similarly, chips 46, 47 and 48, would require their own customized thermal pastes or materials 55, 54 and 56, respectively. Thermal pastes or compounds 54, 55 and 56, could be thermal materials having a higher or lower thermal conductivity, or could have any thermal conductivity in between.

Ledge or lip 34, normally would provide a containment area for the thermally conductive fluids 51, 54, 55 and 56. Similarly, the ledge 34, would also contain any thermal fluid that may be used to fill the entire area between the substrate 10, and the thermal cap 35.

FIG. 5, illustrates yet another preferred embodiment of this invention. As illustrated in FIG. 5, the chips or semiconductor elements 61, 62, 63, 64 or 65, are somewhat different from each other, therefore, the heat sink or thermal cap 50, has to be customized to accommodate the thermal dissipation requirements for each of the different chips. The inventors of this invention have found a solution which is to customize thermal contact areas 71, 72, 73, 74 and 75, corresponding to the chips 61, 62, 63, 64 and 65, respectively.

As can be more clearly seen in FIG. 5, that the customized thermal contact area 71, is closer to the chip 61, while the customized thermal contact area 72, is further away from the chip 62. Similarly, the customized thermal contact areas 74 and 75, are each different for their corresponding chips 64 and 65, respectively. While, the customized thermal contact area 73, provides heat transfer to the chip 63, which is the same as or similar to the thermal contact area 71.

The amount of thermally conductive material, such as, for example, thermal fluid, thermal oil, thermal paste, thermal grease, to name a few, that is present or required for each of the chip shown in FIG. 5, could be the same or could be different.

As shown in FIG. 5, thermal paste 81, 82, 83, 84 and 85, provide a direct heat or thermal transfer path between the corresponding chips and the heat sink 50.

The thermal conducting paste that is between each chip and the customized thermal contact area could also be the same or it could be different pastes. For example, thermal paste 81, could be the same thermal paste as the thermal paste 84, while the thermal paste 81, could be different than the thermal paste 83.

Chips 61 and 63, have higher power, and therefore the thermal or heat transfer area 71 and 73, is so close that the gap between the chips and the heat sink is at a minimum. Chips 62 and 65, have medium power, and therefore the heat transfer path or the gap between the chips 62 and 65, and the corresponding heat transfer area 72 and 75, respectively, has increased. Chip 64, has lower power, and therefore it has a larger gap between the chip and the heat sink at the heat transfer location 74.

FIG. 6, illustrates another embodiment of this invention where a variety of chips 41, 42, 91, 92 and 93, have been mounted onto the substrate 10. Each of these chips have their own thermal specification and therefore in order to optimize the real estate on the substrate 10, and at the same time provide the most optimum cooling to these chips, it is desired to place chips 41 and 93, close to each other and use thermally conductive material 53, to provide the cooling path to the thermal cap (not shown in FIG. 6). Similarly, chips 42 and 92, are segregated together so that thermally conductive material 85, provides the cooling path. While chip 91, has a different thermal specification and therefore it was determined that thermally conductive material 55, would provide the most appropriate cooling.

This invention provides cooling customization whether thermal compound is placed only on the back side of the chips, e.g., with thermal grease, etc., or thermal compound entirely fills the module, e.g. with oil, etc. In this manner, high power chips can have a small customized gaps and/or high thermal conductivity material (for maximum thermal performance) and lower power chips can have a lower thermally conductive material and/or larger individualized gaps.

The optimum depth of the cut or the gap or the blind hole, and the thermal conductivity of the material used, is affected not only by the chip power, but also by the chip size, chip height, the placement of the chip in the module, the effect of the neighboring chips, etc. Optimum selection of gap thicknesses and/or the depth of the blind hole and/or the thermally conductive material will achieve cooling of the chip to just get the chip within the specification window. This of course will maximize module reliability.

Another option for cost reduction is to loosen the tolerances on the depth of the deeper blind holes, because larger gaps do not need to be as precise.

Using the apparatus and/or method of this invention all the chips on a MCM now will be sufficiently cooled, but not overcooled or undercooled.

The advantages of an electronic package or module such as the one disclosed in this patent application are many. Such as, the cost of the package is reduced due to (a) fewer process steps to assemble the module, (b) the customization of the heat transfer area above and near the chip, etc.

The window of acceptable thermal conductivities of the compound (for each site) is affected not only by chip power, but also by chip size, height, placement in the module, and neighboring chips. Final selection of compound conductivities will involve balancing the cost of higher numbers of thermal compounds against the benefits of optimized thermal performance.

Another advantage is the reworkability or repair of the module. The heat sink can be easily pulled-off from the substrate and the active devices, such as, a chip, or passive devices, such as, a decoupling capacitor, heat sink, etc., or module in its entirety could be reworked or repaired. Furthermore, any circuits on the surface of the substrate could also be reworked or repaired after the heat sink has been removed. Of course, the heat sink can be reattached once the rework or repairs have been made.

While the present invention has been particularly described, in conjunction with a specific preferred embodiment, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art in light of the foregoing description. It is therefore contemplated that the appended claims will embrace any such alternatives, modifications and variations as falling within the true scope and spirit of the present invention.