Textured metallic compression bonding

Abstract

integrated circuit chip-to-chip interconnections are made via gold pads on each chip that are bonded to corresponding gold pads on a silicon wafer chip carrier. The pads on the chips and/or the pads on the carrier are characterized by texturing (roughening) with a feature size of the order of a micrometer or less, so that each of the pads on the chip can be attached to each of the pads on the carrier by compression bonding at room temperature--i.e., cold-well bonding. In particular, the texturing of the gold pads on the silicon carrier is obtained by etching V-grooves locally on the surface of the underlying silicon carrier in the regions of the pads, thermally growing a silicon dioxide layer on the silicon career, and depositing the gold on the silicon dioxide layer.

Description

(1 11) (1 11) crystal plane. Subsequently the silicon dioxide layer 15 (FIG. 3) is thermally grown upon the top surface of the silicon wafer including upon the V-grooves.

Assembly of each chip onto the carrier--i.e., compression bonding of chip pads to carrier pads--is accomplished by cleaning and aligning the chip pads with the respective corresponding carrier pads and applying a mechanical pressure (compression) of about 20 to 40 kg-force/mm² of pad area to the chip and carrier at room temperature (with no applied heat) for a time interval of about 5 seconds, in order to press the bottom surfaces all chip pads in the chip simultaneously against the top surfaces of the respective carrier pads. Because the alignment is done at room temperature, it is ordinarily sufficient to align but two mutually diagonally situated pads, whereby all the other pads are automatically aligned. To clean the chips, before bonding them to the carrier, in particular, to clean them of photoresist, standard techniques are employed.

The resulting bonding of chip pads to carrier pads can be made stronger by applying ultrasonic waves to the pads during the compression bonding or by heating the pads with a laser beam focused on each pad after bonding. A carbon dioxide laser, which has a wavelength of about 10 micrometers (to which the silicon carrier is transparent), is useful for this purpose.

In the absence of the laser heating or the ultrasonic waves, the bonding of chips to carrier is reversible, in that one or more selected chips can be separately detached intact and removed from the carrier without damaging it simply by a mechanical pulling apart. This removability may be very advantageous in case one or more of the chips cease to function properly, in which case the improperly functioning chip(s) can be detached from the carrier simply by pulling apart such chip(s) from the carrier with an applied tensile stress of about 1 kg-force/mm² of total pad area or more without damaging the carrier, and replacing the thus detached chip(s) with properly functioning chip(s), again by means of cold-welded compression bonding of chip and carrier pads. Alternatively, some or all functioning chips bonded to a given carrier can be detached therefrom by pulling and then bonded to another carrier having a different wiring pattern, in order to use some or all of the same chips in a different chip-to-chip electrical interconnection configuration.

As shown in FIG. 3, an enlarged portion of FIG. 2 after assembly of chip to carrier by means of the mechanical pressure described above, as the chip 101 is pressed against the carrier 10, the gold of the pad 24 is squeezed and rubs along sloping portions of the V-shaped grooves (hereinafter "V-grooves"), thereby exposing fresh gold surfaces of both the chip pad and the carder pad, so that the result is a mechanically reliable cold-welded joining room-temperature bonding of chip pads to carrier pads.

As shown in FIG. 4, bottom gold surfaces of the chip pads can be textured with a feature size of about 1 micrometer, instead of (or in addition to) texturing the top surface of the carrier pads. Such texturing of the gold surfaces of the chip pads can be accomplished by photolithographic masking and etching of the gold or by electroplating gold on nickel--a process that automatically results in a textured gold surface.

The V-grooves as viewed from the top can also take the form of nested L's instead of squares, as shown in FIG. 5, where each L-shaped groove 44 has sides that slope downward to the bottom 45 of the groove. The width of the two elongated mask openings defining each L-shaped groove is typically about 1 micrometer, and the space between nearest adjacent L's is typically about 2 micrometers, whereby the L's are on about 3 micrometer centers.

More specifically, in the FIG. 1 embodiment, each of the interconnect lines 23 and 33 is typically made of a single layer of aluminum having a thickness of about one-half to one micrometer and a width of about two micrometers or less, or it can be made in the form of a layered structure of aluminum-titanium-platinum-gold or aluminum-titanium-platinum-gold-tin-gold likewise having a total thickness of about 0.10 micrometers, and a width of about one micrometer or less, with the titanium having a thickness of about 0.05 micrometers. The thickness of each of the chip pads 24 and 34 is typically about 3 micrometers of gold or more, and its width dimensions are typically about 10 by 10 micrometers square. It should be understood, of course, that the chip (and hence carrier) pad shapes as viewed from a vertical direction can be arbitrary: squares, rectangles, circles, etc. The metallization of each carrier pad is typically a gold layer having a thickness of about 0.3 micrometers on a layer of titanium having a thickness of about 0.05 micrometers, the titanium ensuring adhesion of the gold to the underlying silicon dioxide layer 15. The wiring 13 on the carrier 10 is typically made of the same material and thickness as those of the interconnect lines 23 and 33, but everywhere (including the regions of the carrier pads 12 and 14) the wiring 13 has a width of about 10 micrometers or slightly more, i.e., substantially the same width as that of the chips pads except for perhaps an added, relatively small safety margin in the width of the carrier pad.

The space between nearest adjacent chip pads is about 10 micrometers. The distance between centers of nearest neighboring chip pads thus is as little as about 20 micrometers or less. In this way a chip having a size of 1 cm by 1 cm--a periphery of 4 cm--can have as many as 2000 pads or more, i.e., one pad deep along the entire periphery of the chip, and can have many more pads if pads are also built at interior portions of the chip in addition to the periphery thereof. Pads thus located at interior portions have advantages in that thermal conductance and hence heat-sinking of the chip to the carrier is improved in magnitude and in uniformity, and in that the parasitics associated with long conductive paths on the chip from interior to periphery can be reduced. Moreover, in view of the relatively large number of pads, either non-electrically-functional ("dummy") pads or electrically functional redundant pads can be used for increased strength of attachment, increased electrical reliability, and improved heat-sinking.

It should be understood that the interconnection wiring on the carrier at portions thereof underlying the chips but removed from the pads, as well as between chips, can be fabricated with gold or other metals--such as aluminum. The wiring on the carrier can be fabricated on one or more planes ("metallization levels") that are insulated from one another by insulating layers, for example, of silicon dioxide or phosphorous doped glass, as known in the art. Accordingly, any desired wiring pattern including cross-overs can be fabricated on the carrier as known in the art.

Backside contact (not shown) of the chip to the carrier can be made by means of a fine gold wire which is bonded, after assembly of chip carrier, by means of silver-epoxy both to the backside of the chip and to a matching, typically smooth pad on the carrier.

Another specific embodiment is illustrated in FIGS. 6 and 7, in which the same reference numerals are used for elements corresponding to those shown in FIGS. 1-3. In this embodiment (FIGS. 6 and 7) an insulating spacing layer 25 separates the bottom surface of the chip 101 from a left-hand moiety of the chip pad 24. Typically, this layer 25 is hard baked photoresist or silicon nitride 3 micrometers thick, to which adherence of the gold of the chip pad is minimal or zero. A portion 30 of the top surface of the silicon carrier 10 underlying the right-hand moiety of the spacing layer 25 is smooth and is vertically indented ("sunken") beneath the original top surface of the silicon to a depth corresponding to the bottom of the V-grooves. Vertical indenting of the portion 30 can be obtained by photolithographic masking and etching at the same time as the photolithographic masking and etching of the V-grooves.

Starting with the situation shown in FIG. 6, mechanical compression is applied followed by a slight mechanical pulling which is sufficient to produce a vertical spacing y (FIG. 7) between the top left-hand surface of the chip pad 24 and the bottom left-hand surface of the spacing layer 25, and which is sufficient to produce vertical spacing between the chip pad 24 and the sunken surface portion 30. Typically, this spacing y is about 2 micrometers. Prior to electrical utilization of the circuitry of the chip 101, the layer 25 can be removed as by an oxygen plasma treatment of the photoresist material (but not if silicon nitride) thereof, in order to have greater compliance--i.e., greater leeway or margins in case of change in bowing of the bottom surface of the chip. This structure (FIG. 7) has the added advantage of relative freedom from strains induced by unequal thermal expansion of the chip 101 and the carrier 10 in the lateral direction due to unequal temperature changes. Hence this structure promises to withstand such strains, in case they indeed occur during operation, as might cause failures of the embodiment shown in FIGS. 1-3. For greater mechanical strength, nickel plated with gold can be used as the material for chip pad 24.

FIG. 8 shows another embodiment, wherein some or all chip-to-chip interconnection is obtained through a separate chip-to-chip interconnection wiring plate 201. The plate 201 is attached to the carrier 10 by means of the interconnection plate pads 64 and 74 and typically many others (not shown) which are bonded to career pads 44 and 54 and typically many others (not shown). The bonding is achieved in the same way that, for example, the chip 101 is attached to the carrier 10 by means of chip pad 24 bonded to carrier pad 14. In this way the interconnection wiring portions 13 stemming from chips 101 and 102 are interconnected--and hence pads 24 and 34 are interconnected--through pads 64 and 174 of the interconnection plate 201 plus interconnect lines 63, 73, and perhaps others (not shown) of the plate 201.

The interconnection plate 201 can simply take the form of an IC chip having no transistors, but having only wiring (typically multi-level) arranged for electrically interconnecting the various interconnection plate pads. Each of the plate pads is constructed in the same way as an IC chip pad. Moreover, chip-to-chip electrical interconnection may be modified by mechanically pulling and removing the plate 201 and replacing it with another plate having a different pattern of interconnect lines. Also, failures in the chip-to-chip interconnection plate can be repaired by similarly removing the failed interconnection plate followed by replacement of another operative plate.

Although the invention has been described in detail with reference to specific embodiments, various modifications can be made without departing from the scope of the invention. For example, etch pits other than V-grooves in silicon can be used for texturing the surface of the carrier pads. In silicon, for example etch pits with crystallographic (111) sides on (110) oriented silicon wafer surfaces could be fabricated having vertical sidewalls rather than V-grooves. Instead of such etch pits or V-grooves which are L-shaped or are square shaped as viewed from the top, other shapes such as elongated trench V-grooves can be etched into the silicon carrier by using correspondingly shaped masking for the etching. The surface of the gold itself of the carrier pads could be directly textured (without first texturing the underlying silicon) by means of photolithographically etching the top surface of the gold carrier pads rather than the underlying silicon, or by electroplating the gold on nickel. Moreover, gold-plated nickel has mechanically stronger greater breaking strength than pure gold. This greater breaking strength is especially desirable in the embodiment of FIGS. 6-7 where, in response to the mechanical pulling, before the bonded portion can detach as is desired, there is an undesirable tendency for the central portion of the chip pad to tear or break and thus prevent re-use of chip on another carrier. Instead of grooves, texturing of a surface can be obtained by forming a multitude of pillars (columns), pyramids, or other protrusions on the surface.

Also, texturing of a surface can be accompanied anisotropically etching the entire surface to a prescribed depth except for the top of grooves which are masked against the etching. Alternatively, texturing can be achieved by selectively depositing metal only in the regions of the pads, followed by formation of grooves in the deposited metal layer. A single chip can be attached to the carrier, instead of more than one chip, simply for the purpose of mechanically and electrically stable external access.

Instead of silicon, other materials for the carrier may be used, such as glass or ceramic; and the chips themselves can be crystalline gallium arsenide instead of silicon.

Moreover the welding procedure of this invention could be done at temperature above or below room temperature--the former (but not above about 300°C) for the purpose of stronger bonding, if desired, the latter for protecting the integrated circuitry by maintaining the temperature fairly low even in the presence of undesirable mounts of heat (if any) generated by the sliding of the chip pad surfaces along the textured carrier pad surfaces. The welding could also be done in an environment comprising a selected gas or liquid (such as for fluxing).

Finally the welding could be performed by the step of precleaning the surfaces of all pads by exposure to short ultraviolet light that generates ozone, i.e., light of wavelength equal to about 250 nanometers, followed by the step of compression bonding (cold-welding), with either or both steps being performed in a vacuum chamber.

Claims

1. In combination

(a) a semiconductor integrated circuit chip having an integrated circuit connected to a plurality of metallic chip pads located on a major surface of the chip;

(b) a carrier upon which are located metallic wiring interconnections having metallic carrier pads that are compression bonded to the chip pads, each of either the carrier pads, or the chip pads, or but not both, having at least a portion thereof that is textured prior to bonding with indentations whose depths are, or protrusions whose heights are, of the order of one micrometer.

2. The combination described in claim 1 in which each chip pad has a portion that is separated from the major surface of the chip by a portion of a localized layer of an insulating material, and in which each carder pad has a portion, underlying a complementary portion of the layer of insulating material, that is indented in the vertical direction and is smooth, the insulating material having relatively little or no adhesive tendency with respect to the chip pad.

3. The combination of claim 2 in which the carrier comprises a silicon wafer and in which the pads are bonded such that a tensile force of about 1 kg per mm² or more of pad area is required to pull them apart.

4. The combination of claim 3 in which the silicon wafer has a plurality of V-grooves at each of such portions thereof underlying the carrier pads.

5. The combination of claim 2 in which surfaces of the chip pads are essentially gold.

6. The combination of claim 5 in which surfaces of the carrier pads are essentially gold and in which the chip pads are nondestructively detachable from the carrier pads by means of a mechanical pulling apart.

7. The combination of claim 1 in which the chip carrier comprises a silicon wafer which is textured at the portions thereof underlying bonded to the carrier pads.

8. The combination of claim 7 in which the silicon wafer has a plurality of V-grooves at each of such portions thereof underlying bonded to the carrier pads.

9. The combination of claim 1 in which areas of the carrier pads to be bonded to chip pads are essentially gold and in which the chip pads are nondestructively detachable from the carrier pads by means of a mechanical pulling apart.

10. The combination of claim 10 9 in which the surfaces of the chip pads are essentially gold gold.

11. The combination of claim 1 in which surfaces of the chip pads are essentially gold.

12. The combination of claim 1 in which the indentations or protrusions have widths, as measured at the tops of the protrusions or bottoms of the indentations, of about 1 micrometer or less. 13. In combination:

a first body having a first metallic layer located contiguous with a first surface of the first body; and

a second body having a second metallic layer being compression bonded to the second metallic layer, either the first metallic layer or the second metallic layer, but not both the first and second metallic layers, having at least a portion thereof that is textured prior to bonding with indentations whose depths are of the order of one micrometer or less. 14. The combination of claim 13 in which the second body comprises a wafer and in which the first and second metallic layers are bonded such that a tensile force of about 1 kg per mm² or more of metallic area is required to pull them apart. 15. The combination of claim 13 in which a surface of the first metallic layer is essentially gold. 16. The combination of claim 15 in which a surface of the second metallic layer is essentially gold and in which the first metallic layer is nondestructively detachable from the second metallic layer by means of a mechanical pulling apart. 17. The combination of claim 13 in which the second body comprises a wafer which is textured at the portion thereof bonded to the second metallic layer. 18. The combination of 17 in which the wafer has a plurality of V-grooves at the portion thereof bonded to the second metallic layer. 19. The combination of claim 13 in which areas of the second metallic layer that are bonded to the first metallic layer are essentially gold, and in which the first metallic layer is nondestructively detachable from the second metallic layer by means of a

mechanical pulling apart. 20. The combination of claim 19 in which the surface of the first metallic layer is essentially gold. 21. The combination of claim 13 in which a surface of the

first metallic layer is essentially gold. 22. The combination of claim 13 in which the indentations have widths, as measured at the tops or at the bottoms of the indentations, of about 1 micrometer or less.