US 20060255473 A1
A solder mask for flip chip interconnection has a common opening that spans a plurality of circuit elements. The solder mask allows confinement of the solder during the remelt stage of interconnection, yet it is within common design rules for solder mask patterning. Also, a substrate for flip chip interconnection includes a substrate having the common opening that spans a plurality of circuit elements. Also, a flip chip package includes a substrate having a common opening that spans a plurality of circuit elements.
1. A solder mask for flip chip interconnection having a common opening that spans a plurality of circuit elements.
2. A flip-chip package substrate, comprising a patterned metal layer on a die attach side of a dielectric substrate layer, the metal layer including interconnect sites, the substrate including a solder mask having an opening spanning a plurality of the interconnect sites.
3. The substrate of
4. The substrate of
5. The substrate of
6. A method for making a flip chip interconnection, comprising providing a substrate including a solder mask having an opening spanning a plurality of the interconnect sites, and mounting the chip onto the substrate.
This application claims priority from U.S. Provisional Application No. 60/594,885, filed May 16, 2005, titled “Solder confinement integrated circuit package system”, which is incorporated by reference herein.
This invention relates to semiconductor packaging and, particularly, to flip chip interconnection.
Flip chip packages include a semiconductor die mounted onto a package substrate with the active side of the die facing the substrate. Conventionally, interconnection of the circuitry in the die with circuitry in the substrate is made by way of bumps which are attached to an array of interconnect pads on the die, and bonded to a corresponding (complementary) array of interconnect pads (often referred to as “capture pads”) on the substrate.
The areal density of electronic features on integrated circuits has increased enormously, and chips having a greater density of circuit features also may have a greater density of sites for interconnection with a package substrate.
The package is connected to underlying circuitry, such as a printed circuit board (e.g., a “motherboard”) in the device in which it is employed, by way of second level interconnects (e.g., pins, solder balls) between the package and the underlying circuit. The second level interconnects have a greater pitch than the flip chip interconnects, and so the routing on the substrate conventionally “fans out”. Significant technological advances have enabled construction of fine lines and spaces; but in the conventional arrangement space between adjacent pads limits the number of traces than can escape from the more inward capture pads in the array, and the fan out routing between the capture pads beneath the die and the external pins of the package is conventionally formed on multiple metal layers within the package substrate. For a complex interconnect array, substrates having multiple layers may be required to achieve routing between the die pads and the second level interconnects on the package.
Multiple layer substrates are expensive, and in conventional flip chip constructs the substrate alone typically accounts for more than half the package cost (about 60% in some typical instances). The high cost of multilayer substrates has been a factor in limiting proliferation of flip chip technology in mainstream products.
In conventional flip chip constructs the escape routing pattern typically introduces additional electrical parasitics, because the routing includes short runs of unshielded wiring and vias between wiring layers in the signal transmission path. Electrical parasitics can significantly limit package performance.
In some conventional processes, flip chip interconnection is made by contacting the bumps or balls on the die with corresponding interconnect sites on the substrate circuitry, and then heating to reflow the fusible portion of the solder bumps (or to reflow the solder bumps in their entirety) to make the electrical connection. In such processes the melted solder may flow from the interconnect site along the metal of the circuitry, depleting the solder at the connection site; and where the bumps are collapsible under reflow conditions the bumps may contact adjacent circuitry or nearby bumps, resulting in electrical failure. To avoid these problems, typically in conventional flip chip packages the solder is confined by a “solder mask”, consisting of a layer of dielectric material overlying the patterned metal layer at the die mount surface of the substrate, and having openings each exposing an interconnect site on the underlying circuitry. Process limitations in patterning the solder mask prevent reliably forming well-aligned and consistently dimensioned openings and, accordingly, where a solder mask is employed, substrates having fine circuitry feature dimensions as would be required for finer pitch interconnection are not attainable.
The interconnect pitch in conventional flip chip interconnects is limited in part by the dimensions of the capture pads on the substrate (typically the capture pads are much wider than the circuit elements connecting them). Recently flip chip substrate circuitry design has been disclosed, in which reliable interconnection is made on narrow circuit elements on the substrate, as for example in “bond-on-narrow pad interconnections” (BONP), as described generally in copending U.S. application Ser. No. 11/388,755, filed Mar. 26, 2006; and as for example in “bump-on-lead interconnections” (BOL), as described generally in copending U.S. application Ser. No. 110/985,654, filed Nov. 10, 2004, both incorporated herein by reference. Where a conventional solder mask is to be employed, limitations in the process for patterning the solder mask can limit pitch reduction even in some BONP or BOL substrate configurations. The exposed bondable surface of the lead may be contaminated by or covered by solder mask residue, resulting in an imperfect solder joint; or, the bondable surface of the lead may be inconsistently or only partially exposed at the interconnect site, resulting in an unreliable and inconsistent trace structure.
Generally according to the invention, a solder mask for flip chip interconnection has a common opening that spans a plurality of circuit elements. The solder mask allows confinement of the solder during the remelt stage of interconnection, yet it is within common design rules for solder mask patterning.
In one general aspect the invention features a flip-chip package substrate, including a patterned metal layer on a die attach side of a dielectric substrate layer, the metal layer including interconnect sites, the substrate including a solder mask having an opening spanning a plurality of the interconnect sites.
In some embodiments the plurality of interconnect sites comprises interconnect sites arranged in a row, and the opening comprises an elongate opening spanning the row.
In some embodiments the opening has an irregular shape.
In some embodiments the interconnect sites are arranged in an array of rows, and the opening comprises an elongate opening spanning one of the rows of interconnect sites.
In some embodiments the invention features a method for making a flip chip interconnection, comprising providing a substrate including a solder mask having an opening spanning a plurality of the interconnect sites, and mounting the chip onto the substrate.
The invention will now be described in further detail by reference to the drawings, which illustrate alternative embodiments of the invention. The drawings are diagrammatic, showing features of the invention and their relation to other features and structures, and are not made to scale. For improved clarity of presentation, in the FIGs. illustrating embodiments of the invention, elements corresponding to elements shown in other drawings are not all particularly renumbered, although they are all readily identifiable in all the FIGs.
All patents and patent applications referred to above and below are incorporated herein by reference.
The conventional flip chip interconnection is made by using a melting process to join the bumps (conventionally, solder bumps) onto mating surfaces of corresponding interconnect sites on the patterned metal layer at the die attach surface of the substrate. Where the site is a capture pad, the interconnect is known as a “bump-on-capture pad” (“BOC”) interconnect; where the site is a lead, or a narrow pad (which may be a modest widening of the circuitry) the interconnect is known as a “bump-on-lead” (“BOL”) or “bump on narrow pad” (BONP) interconnect. In the BOC design a comparatively large capture pad is required to mate with the bump on the die. In some flip chip interconnections, an insulating material, typically known as a “solder mask” is required to confine the flow of solder during the interconnection process. The solder mask opening may define the contour of the melted solder at the capture pad (“solder mask defined”), or the solder contour may not be defined by the mask opening (“non-solder mask defined”); in the latter case the solder mask opening may be significantly larger than the interconnect site (capture pad, narrow pad, or lead). As noted above, the techniques for defining solder mask openings have wide tolerance ranges. Consequently, for a solder mask defined bump configuration, the capture pad must be large (typically considerably larger than the design size for the mask opening), to ensure that the mask opening will be located on the mating surface of the pad; and for a non-solder mask defined bump configuration, the solder mask opening must be larger than the capture pad. The width of capture pads (or diameter, for circular pads) is typically about the same as the ball (or bump) diameter (which may be measured at the attachment of the bump with the pad on the die), and can be as much as two to four times wider than the trace width. This results in considerable loss of routing space on the patterned metal layer at the die attach surface of the substrate. In particular, for example, the “escape routing pitch” is much bigger than the finest trace pitch that the substrate technology can offer. This means that a significant number of pads must be routed on lower substrate layers by means of short stubs and vias, often beneath the footprint of the die, emanating from the pads in question.
Significantly finer pitch interconnects may be obtained by employing BOL or BONP design. Various BOL structures, and methods for making them, are described for example in U.S. application Ser. No. 10/985,654, filed Nov. 10, 2004; and various BONP structures, and methods for making them, are described for example in U.S. application Ser. No. 11/388,755, filed Mar. 24, 2006, both of which are incorporated herein by reference.
In embodiments as in
As shown in
Referring particularly now to
The BOL interconnection structure such as is shown by way of example in
Methods employing a no-flow underfill to confine the solder during the remelt stage are described, for example, in U.S. Application No. [Atty Docket No. CPAC 1024-2], by Rajendra D. Pendse et al., filed May 15, 2006, titled “Flip chip interconnection”, which is hereby incorporated herein by reference.
Alternative bump structures may be employed in the bump-on-lead interconnects having no solder mask. Particularly, for example, so-called composite solder bumps may be used. Composite solder bumps have at least two bump portions, made of different bump materials, including one which is collapsible under reflow conditions, and one which is substantially non-collapsible under reflow conditions. The non-collapsible portion is attached to the interconnect site on the die; typical conventional materials for the non-collapsible portion include various solders having a high lead (Pd) content, for example. The collapsible portion is joined to the non-collapsible portion, and it is the collapsible portion that makes the connection with the lead according to the invention. Typical conventional materials for the collapsible portion of the composite bump include eutectic solders, for example.
As outlined above, methods for forming flip chip interconnection having high density have been proposed. However, the density of flip chip interconnection in which a solder mask is desired is limited by process capability of the solder mask patterning process.
According to the invention, a solder mask configuration allows confinement of solder in high-density (fine pitch) flip chip interconnects, yet is within design rules for solder mask patterning.
According to the invention, a solder mask is provided having at least one opening spanning two or more (usually a larger number in a row) circuit elements such as, for example, leads or narrow pads. The opening has a generally elongated shape, and is oriented so that its longer dimension spans the circuit elements, and the shorter dimension limits the extent of exposure of the lengths of the circuit elements. Accordingly, the flow of fusible material that is melted during the reflow step in the interconnection process is limited along the length of the circuit elements (leads, pads) by the width of the solder mask opening, and the number of interconnect sites on which the flow of melted bump material is so limited is determined by the length of the solder mask opening (and, therefore, by the number of pads or leads that are spanned by the opening).
An idealized example is shown in
As shown in
The width (narrow dimension) of the elongated solder mask serves to limit flow of solder away from the interconnect site along the solder-wettable lead. The width (narrow dimension) of the elongated solder mask opening according to the invention may in some embodiments be determined by the limit of the design rules for patterning the solder mask; it may, for example, approximate the width (or diameter) of a conventional solder mask opening. The width may have a nominal mask width in the range about 80 um to 90 um or less, but it can be 100 um or more. Solder mask materials can be resolved at such pitches and, particularly, substrates can be made comparatively inexpensively with solder masks having 90 um openings and having alignment tolerances plus or minus 25 um. In some configurations laminate substrates (such as 4 metal layer laminates), made according to standard design rules, are used.
According to the invention, the feature sizes required for the solder mask can be made coarser; because the elongate solder mask opening spans a number of leads, the alignment of the mask openings with the interconnect sites can be significantly relaxed. Risk of partial exposure of bondable areas of leads at interconnect sites is practically avoided. Solder run-off along the length of the circuit features at the interconnect sites is confined by the opening (width dimension). And runoff toward adjacent circuit features is mitigated (at least) because the dielectric material of the substrate dielectric is not wettable by the solder.
In some embodiments the interconnect includes a bump, metallurgically joined to an interconnect site (e.g., lead or narrow pad); this may include solder fillets formed along the surrounding surface and exposed sidewalls of the lead.
In some embodiments the interconnect is formed in two broad steps: the bump is thermo-mechanically joined to the lead without melting; and a no-flow underfill is cured to a gel stage; thereafter the bump is melted in a reflow operation to form a reliable interconnection. This confines the joint to a relatively small volume and minimizes the risk of solder bridging to an adjacent circuit element.
Solder paste can be provided at the interconnect sites on the leads, to provide a fusible medium for the interconnect. The paste is dispensed, for example by a standard printing process, then is reflowed, and then may be coined if necessary to provide uniform surfaces to meet the balls. The solder paste can be applied in the course of assembly; or, a substrate may be provided with paste suitably patterned prior to assembly. Other approaches to applying solder selectively to the interconnect sites may be employed in the solder-on-lead embodiments, including electroless plating and electroplating techniques. The solder-on-lead configuration provides additional solder volume for the interconnect, and can accordingly provide higher product yield, and can also provide a higher die standoff.
For interconnection of a die having high-melting temperature solder bumps (such as a high-lead solder, conventionally used for interconnection with ceramic substrates) onto an organic substrate, the solder mask of the invention ca be employed to limit the flow of fusible solder paste along the circuit element near the interconnect site. The solder paste can be selected to have a melting temperature low enough that the organic substrate is not damaged during reflow. To form the interconnect in such embodiments the high-melting interconnect bumps are contacted with the solder-on-lead sites, and the remelt fuses the solder-on-lead to the bumps. Where a noncollapsible bump is used, together with a solder-on-lead process, no preapplied adhesive is required, as the displacement or flow of the solder is limited by the fact that only a small quantity of solder is present at each interconnect, and the noncollapsible bump prevents collapse of the assembly.
In other embodiments the solder-on-lead configuration according to the invention is employed for interconnection of a die having eutectic solder bumps.
Packages according to the invention, employing no-flow underfill techniques, can be made for example as follows. A substrate is provided, having at least one dielectric layer and having a metal layer on a die attach surface. The metal layer is patterned to provide circuitry, particularly traces or leads and including sites for interconnection, on the die attach surface. The substrate is supported, for example on a carrier or stage, with a substrate surface opposite the die attach surface facing the support. A die is provided, having bumps attached to die pads on the active side. The bumps include a fusible material which contacts the mating surfaces of the leads. A quantity of an underfill (filled encapsulating resin adhesive) is dispensed over the die attach surface of the substrate, covering at least the interconnect sites on the leads; or over the active side of the die. A pick-and-place tool including a chuck picks up the die by contact of the chuck with the backside of the die. Using the pick-and-place tool, the die is positioned facing the substrate with the active side of the die toward the die attach surface of the substrate; and the die and substrate are aligned and moved one toward the other so that the bumps contact the corresponding interconnect sites on the traces (leads) on the substrate. Then a force is applied to press the bumps onto the mating surfaces at the interconnect sites on the leads. The force must be sufficient at least to displace the adhesive from between the bumps and the mating surfaces at the interconnect sites on the leads. The bumps may be deformed by the force, breaking the oxide film on the contacting surface of the bumps and/or on the mating surface of leads. The deformation of the bumps may result in the fusible material of the bumps being pressed onto the top and over the edges of the lead. The adhesive is caused to cure at least partially, as shown at, as for example by heating to a selected temperature. At this stage the adhesive need only be partially cured, that is, only to an extent sufficient subsequently to prevent flow of molten solder along an interface between the adhesive and the conductive traces. Then the fusible material of the bumps is melted and then is re-solidified, forming a metallurgical interconnection between the bump and lead, and the adhesive curing is completed, to complete the die mount and to secure the electrical interconnection at the mating surface (now an interconnect interface).
In the plane of the sectional view shown in
Other solder mask opening configurations are within the invention. Particularly, the elongated opening may expose interconnect sites on two or more adjacent circuit features; in some embodiments the opening exposes a row of interconnect sites, which may be a row on an array of interconnect sites. The row of interconnect sites exposed need not be in a straight line and, accordingly, the opening need not be rectangular: the opening may have an arcuate shape, or may be irregular. Where the elongated opening has a shape of a regular polygon, such as a rectangle for example, the elongated opening need not necessarily be oriented parallel to a row of interconnect sites or to the die margin.
Where interconnect is formed by a no-flow underfill process, the no-flow underfill adhesive can be pre-applied to the die surface, or at least to the bumps on the die surface, rather than to the substrate. The adhesive can, for example, be pooled in a reservoir, and the active side of the die can be dipped in the pool and removed, so that a quantity of the adhesive is carried on the bumps; then, using a pick-and-place tool, the die is positioned facing a supported substrate with the active side of the die toward the die attach surface of the substrate, and the die and substrate are aligned and moved one toward the other so that the bumps contact the corresponding traces (leads) on the substrate. Such a method is described in U.S. Pat. No. 6,780,682, Aug. 24, 2004, which is hereby incorporated by reference. Then forcing, curing, and melting are carried out as described above.
The adhesive may be referred to as a “no-flow underfill”. In some approaches to flip chip interconnection, the metallurgical interconnection is formed first, and then an underfill material is flowed into the space between the die and the substrate. The “no-flow underfill” according to the invention is applied before the die and the substrate are brought together, and the no-flow underfill is displaced by the approach of the bumps onto the leads, and by the opposed surfaces of the die and the substrate. The adhesive for the no-flow underfill adhesive according to the invention is preferably a fast-gelling adhesive—that is, a material that gels sufficiently at the gel temperature in a time period in the order of 1-2 seconds. Preferred materials for the no-flow underfill adhesive include, for example, so-called non-conductive pastes, such as those marketed by Toshiba Chemicals and by Loktite-Henkel, for example.
Alternative bump structures may be employed in the bump-on-lead interconnects according to the invention. Particularly, for example, so-called composite solder bumps may be used. Composite solder bumps have at least two bump portions, made of different bump materials, including one which is collapsible under reflow conditions, and one which is substantially non-collapsible under reflow conditions. The non-collapsible portion is attached to the interconnect site on the die; typical conventional materials for the non-collapsible portion include various solders having a high lead (Pd) content, for example. The collapsible portion is joined to the non-collapsible portion, and it is the collapsible portion that makes the connection with the lead according to the invention. Typical conventional materials for the collapsible portion of the composite bump include eutectic solders, for example.
Other embodiments are within the following claims.