FIELD OF THE INVENTION
The present invention relates generally to the field of microelectronics, and more particularly to methods for compensating for differences in the coefficient of thermal expansion (“CTE”) in different materials used in wafer-level and chip scale packaging of semiconductor devices, especially high frequency (RF) semiconductor devices.
BACKGROUND OF THE INVENTION
As semiconductor technology advances, and more active semiconductor functionality is packed into smaller silicon real estate, the packaging and passives costs, especially in RF devices, increase compared to the overall cost of the device. To decrease the cost, wafer-level packaging (“WLP”) is becoming a more practical and more cost-effective approach. However, the cost savings associated with wafer-level packaging comes at the expense of performance, especially in RF devices.
In some approaches, RF passives are either built sequentially over the surface of the wafer, either on the usable semiconductor substrate area, or on another dielectric substrate deposited over the semiconductor wafers; or built on another substrate in parallel, and then joined over the semiconductor wafer containing active circuitry. Regardless of the method used to realize semiconductor devices such as those disclosed herein, CTE mismatch between the semiconductor substrate and passive materials, such as dielectrics and metals, is a major issue. Usually, matching of the CTE causes sacrifices in the quality factor (Q factor) and dielectric properties for high frequency devices. Traditional or current approaches to integrate antennas directly with microchips involve planar circuits, multi-layer glass stacks, BCB, polyimide or other technologies either bonded to or directly integrated on the semiconductor substrate. These options may create their own CTE challenges.
In cases where dissimilar microelectronic substrates are to be joined together at large die format or wafer-to-wafer level format, the CTE mismatch between adjacent substrates can render this effort either impossible or create significant additional cost due to part damage or failure including substrate deformation. Similarly, for a sequential build of metallic circuitry on a low CTE semiconductor wafer or substrate, such as a PolyStrata® coaxial transmission line build, this CTE mismatch is problematic.
SUMMARY OF THE INVENTION
In one of its aspects the present invention relates to methods and structures to allow cost effective wafer level packaging of semiconductor die to turn the structures into chip scale packages at the wafer level. In addition, the present invention can address the CTE mismatch problem that occurs in applications ranging from wafer level packaging for memory or other chips, to antenna circuitry made in a primarily copper PolyStrata® process that is formed on, or attached to, semiconductors, such as to SiGe circuits on silicon substrates, at scales normally not possible due to the CTE mismatch between these two relatively rigid formations. (Examples of PolyStrata° processing/technology are illustrated in U.S. Pat. Nos. 7,948,335, 7,405,638, 7,148,772, 7,012,489, 7,649,432, 7,656,256, 7,755,174, 7,898,356 and/or U.S. Application Pub. Nos. 2010/0109819, 2011/0210807, 2010/0296252, 2011/0273241, 2011/0123783, 2011/0181376, 2011/0181377, each of which is incorporated herein by reference in their entirety, hereinafter the “incorporated PolyStrata® art”). This may be achieved by applying flexible CTE compensating regions between unit cells into the primarily copper formations to reduce the propagation of stress across large CTE mismatched areas of two or more bonded regions. This approach may work independent of the structures being formed separately and joined, or formed one upon the other. A three-dimensional copper spring structure is envisaged by using PolyStrata® fabrication technology and other manufacturing techniques.
In another of its aspects, the present invention may provide an microelectronic structure having CTE compensation for use in wafer-level and chip-scale packages. The microelectronic structure may include a plurality of substrate tiles each having a generally planar upper surface, the upper surfaces of the tiles disposed within a common plane to provide a generally planar grid of the tiles. Each respective pair of adjacent tiles may have a gap disposed therebetween and a spring structure spanning the gap and connecting the adjacent tiles. The spring structure may be configured to permit movement of the adjacent tiles relative to one another to provide compensation for thermal expansion or contraction of the tiles. The substrate tiles may comprise a semiconductor material and/or metal. The spring structure may comprise metal, and may be provided in the form of a membrane and/or a 2-dimensional serpentine structure. In one particular configuration, the spring structure may be a u-shaped membrane having a longitudinal axis that is disposed parallel to an edge of the upper surface of the selected tile to which the u-shaped membrane is attached. In addition, a dielectric material may be disposed on the common plane and within the gap. A device layer may be attached to the common plane, and the layer may include one or more of resistors, capacitors, inductors, and three-dimensional metal/dielectric structures. The three-dimensional metal/dielectric structures may include one or more of coaxial waveguides, antennas, Wilkinson combiner/dividers, Gysel combiner/dividers, and filters. The plurality of tiles may each comprise multiple layers of metal disposed parallel to the upper surface, which may result from formation using a multilayer build process, such as the PolyStrata° process. Microelectronic structures of the present invention may include antenna arrays having a plurality of antenna radiator structures disposed above the upper surfaces of the tiles.
In yet another of its aspects, the present invention may provide a method of forming a three-dimensional microstructure by a sequential build process, comprising disposing a plurality of layers over a substrate, wherein the layers comprise one or more layers of a conductive material and one or more layers of a sacrificial material, thereby forming a structure above the substrate. The structure may include a plurality of conductive tiles formed of the conductive material, each tile having a generally planar upper surface, the upper surfaces of the tiles disposed within a common plane to provide a generally planar grid of the tiles, each respective pair of adjacent tiles having a gap disposed therebetween and a spring structure spanning the gap and connecting the adjacent tiles. The spring structure may be configured to permit movement of the adjacent tiles relative to one another to provide compensation for thermal expansion or contraction of the tiles. In addition the method may include removing the sacrificial material and/or the substrate. The step of disposing a plurality of layers over a substrate may include disposing one or more layers of a dielectric material. In addition, the method may include electrically connecting a device layer to the common plane. The device layer may comprise one or more of resistors, capacitors, inductors, and three-dimensional metal/dielectric structures, such as coaxial waveguides, antennas, Wilkinson combiner/dividers, Gysel combiner/dividers, and filters.
In still a further of its aspects, the present invention may provide a method of forming a three-dimensional microstructure by a sequential build process, such as the PolyStrata® process. The method may include providing a substrate having opposing upper and lower surfaces, with a trench extending from the lower surface to a region proximate the upper surface; providing a stretcher layer on the trench and lower surface of the substrate; expanding the stretcher layer linearly along a direction parallel to the lower surface of the substrate to separate the substrate into first and second substrate portions divided at the location of the trench providing an expanded trench region; and providing a dielectric material in the expanded trench region. The step of providing a dielectric material may include providing the dielectric material on the upper surface of the first and second substrate portions. A plurality of conductive vias may be provided through the dielectric material, and the vias may be in electrical communication with the upper surface of at least one of the first and second substrate portions. A device layer may be provided over the dielectric material, the device layer including at least one electronic component in electrical communication with a selected one of the vias.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing summary and the following detailed description of exemplary embodiments of the present invention may be further understood when read in conjunction with the appended drawings, in which:
FIGS. 1-4 schematically illustrate cross-sectional views of some of the common approaches currently used in wafer fabrication;
FIGS. 5-10 schematically illustrate cross-sectional views of an exemplary process and wafer configuration having reduced CTE mismatch in accordance with the present invention, in which
FIG. 5 schematically illustrates a wafer with a connecting region between dice and a thinned substrate region on an opposing surface where the dice will be separated,
FIG. 6 schematically illustrates FIG. 5 after a material is deposited which may have selective adhesion to differing regions of the wafer but keeps the wafer together after the substrate is fragmented into elements,
FIG. 7 schematically illustrates FIG. 6 after the wafer material is separated (made discontinuous) between die,
FIG. 8 schematically illustrates FIG. 7 after the wafer is expanded and die separated as the material 400 is stretched along the plane of the wafer,
FIG. 9 schematically illustrates FIG. 8 after the wafer is overcoated with a material to create a spacer region between die, which overcoat may be of a different CTE than the wafer, and
FIG. 10 schematically illustrates the overcoat material of FIG. 9 after vias are formed therethrough and passives or contact pads are created thereon;
FIGS. 11-12 schematically illustrate cross-sectional views of an exemplary spring-like formation in accordance with the present invention that may be provided at the wafer level with a substrate portion removed proximate the formation, with FIGS. 11A-11F showing an exemplary process for the making thereof;
FIG. 13 schematically illustrates a typical existing chip-scale package with the contacts redistributed from the front surface of the die to a back surface of a substrate in the form of solder balls for surface mounting;
FIG. 14 schematically illustrates a cross-sectional view of wafer level chip scale packaging in accordance with the present invention using the die configuration of FIG. 8, where a back surface of a substrate includes solder ball contacts for surface mounting;
FIGS. 14A-14G schematically illustrate cross-sectional views of an exemplary method in accordance with the present invention for forming the wafer level chip scale package of FIG. 14;
FIG. 15 schematically illustrates a simplified top-isometric view of a three-by-three array of ground plane tiles connected by U-shaped spring features and a detailed view of the U-shaped spring structures in accordance with the present invention;
FIG. 16 schematically illustrates a bottom isometric view of the array of ground plane tiles of FIG. 15;
FIG. 17 schematically illustrates a top view of the array of ground plane tiles of FIG. 15;
FIG. 18 schematically illustrates a cross-sectional view of an antenna array joined using CTE-decoupling features in accordance with the present invention, and associated
FIGS. 18A-18O schematically illustrate cross-sectional views of an exemplary method in accordance with the present invention for creating the structure of FIG. 18;
FIG. 19 schematically illustrates the antenna array of FIG. 18 attached to a semiconductor chip or wafer;
FIG. 20 schematically illustrates a top view of two-dimensional-spring connected ground plane tiles in accordance with the present invention;
FIG. 21 schematically illustrates cross-sectional views of a version of the structure of FIG. 18 with staggered features to increase overall radiation shielding effectiveness; and
FIGS. 22A, 22B schematically illustrate cross-sectional views of a singulation technique in accordance with the present invention showing its use in creating an antenna array similarly structured to that of FIG. 18.
DETAILED DESCRIPTION OF THE INVENTION
Referring now to the figures, wherein like elements are numbered alike throughout, in one of its aspects the present invention provides structures and methods for alleviating CTE mismatch between semiconductor substrates or wafers and components mounted or formed thereon. For example, FIGS. 1-4 illustrate existing conventional structures which suffer from CTE mismatch among components not incorporating the improvements of the present invention, but demonstrating the need for such improvements. In particular, FIG. 1 illustrates a semiconductor wafer 100 comprising two adjacent dice 102, 104 with electrodes 106 mounted on an upper surface of the wafer 100. Often additional structures are subsequently provided on the upper surface of the wafer 100, and those additional structures can comprise materials which have a CTE significantly different from that of the wafer 100.
For example, FIGS. 2 and 3 illustrate a spacer 200 provided on the upper surface of the wafer 100 through which conductive vias 202 are formed. The vias 202 may be filled with the metal, such as by a plating process, to provide electrical connection to the electrodes 106 encased within the spacer 200. Additional electronic components 300, such as a layer of passives, may be provided on top of the spacer 200, and placed in electrical communication with the metal filled vias 202. The passives 300 may include capacitors, resistors, inductors, impedance matching circuits, or may be many layers of patterned metals such as copper, dielectrics, and air or void has been described in the art of PolyStrata® processing as practiced by Nuvotronics, Inc. and described in Nuvotronics' patents.
The spacer 200 can be a low loss dielectric material which, besides improving the Q-factor, also alleviates/decreases the sources of interference between the passives 300 and active circuitry of the semiconductor wafer 100, as well as the ground plane.
Whether the spacer 200 is built sequentially or in parallel and attached to the semiconductor wafer 100 later as shown in FIG. 4, the CTE mismatch between the wafer 100 and the material of the spacer 200 is always a significant problem. (More particularly, FIG. 4 illustrates the situation where the passives 300 and interconnect 210 to the wafer 100 are formed independently and then bonded to the wafer 100 in sequential steps using any suitable methods such as Au-Au compression, solder, epoxy, and so on. Handle wafers that may be used in this process are not shown for simplicity as many variations are known and practiced in the art of wafer level packaging.)
The layers of interconnects 210, passives 300, and dielectric spacer 200 typically have a CTE that does not match the CTE of semiconductor or ceramic materials which may comprise the wafer 100. The metals and dielectrics present in the spacer 200 and passives 300 are usually several to 10 times or more the CTE of the wafer 100. This CTE mismatch can cause many problems including wafer bow or wafer breakage and fundamentally limits the thickness and reliability over temperature of the spacer 200 and passives 300 when formed or joined with or on the wafer 100. Such CTE mismatch also limits the size of the chip or wafer 100 that can be formed for any such contiguous structures. Thus, the CTE mismatch between wafer 100 and spacer 200 presents a serious problem that could cause the entire wafer 100 to bulge, curl or even break.
Applicant has recognized that there is a serious need for structures and methods to decouple the stress and strain in such devices. Such a solution is important to both wafer level packaging and chip scale and wafer scale solutions, e.g., for applications such as phased arrays, which often need complex antenna solutions and distribution networks that are interconnected down to pads and/or active devices formed in or on the semiconductor wafer 100.
As a result, in one of its aspects, FIGS. 5-12 schematically illustrate exemplary approaches in accordance with the present invention to address the CTE mismatch problems between semiconductor wafer and the interconnect and passive structures. FIG. 5 illustrates a wafer 100 comprising two adjacent dice 102, 104 along with electrodes 106 disposed on an upper surface 115 of the wafer 100. A portion of the wafer 100 is removed in a trench region 108 between the dice 102, 104 to assist with future separation of the dice 102, 104 in a manner that ameliorates CTE mismatch in accordance with the present invention. The trench region 108 may be provided in the form of a trench having sidewalls 116 that may be formed either chemically or by dry etching, and may desirably be placed around saw lines of every die, or clusters of dice, on the wafer 100. The trench region 108 of removed wafer material may extend upwardly proximate to the upper surface 115 of the wafer 100 but not reach the surface 115, to keep the wafer 100 in one piece, FIG. 5. To assist in this regard, etching of the trench region 108 may be stopped by a deposited stop layer 110 of oxide, nitride, metal or polymer deposited on the upper surface 115 of the wafer 100 in the vicinity of the upper extent of the trench region 108. The stop layer 110 may comprise a similar or different material from that of the electrodes 106. In the case where the removal of wafer material in the trench region 108 extends all the way to the upper surface 115, it may be desirable that the stop layer 110 be sufficiently mechanically strong to hold the two dice 102, 104 together while maintaining the upper surfaces 115 of each dice 102, 104 in the same plane.
Once the trench region 108 of removed wafer material is formed, a stretcher material 400, such as Parylene, may be disposed on the back surface 117 of the wafer 100, FIG. 6. In one exemplary approach in accordance with the present invention, selective adhesion of the stretcher material 400 may be used. Selective adhesion could be achieved, for example, by depositing an adhesion promoter to the back surface 117 of the wafer 100, while preventing the deposition of the adhesion promoter to the sidewalls 116, such as by masking the sidewalls 116. Alternatively, the sidewalls 116 may be treated or formed in a way that lessens adhesion of the stretcher material 400 to the sidewalls 116.
In preparation for separation of the adjacent dice 102, 104 from one another, the stop layer 110 may be removed, and the two dice 102, 104 may be maintained in the original planar relationship relative to one another by the stretcher material 400 and/or a portion of the wafer 100 that remains intact above the trench region 108. To separate the adjacent dice 102, 104, a portion 112 of the wafer 100, located proximate the trench region 108 of removed wafer material, may be removed using either chemical or dry etch, a laser, or dicing removal, for example, FIG. 7. Thereafter, the wafer 100 may be stretched using die matrix expansion with slight force to break the poor adhesion between the stretcher material 400 and the sidewalls 116, thus separating each die 102, 104 on the wafer 100, while the stretcher material 400 still preserves the separated dice 102, 104 as a planar structure similar in planarity to that of the original wafer 100, FIG. 8. The expanded wafer 100 from this point on can still be handled as a wafer with precise X-Y-Z coordinates for each die 102, 104. Depending of semiconductor substrate thickness, the wall angle of the trench region 108, and the stretching capabilities of the material 400, one can increase the overall size of the semiconductor wafer 100 to provide additional real estate over the semiconductor dice, e.g., for building more passives. Some stretcher materials such as Parylene C can stretch up to 200% of their length. Additionally even without any stretching of the stretcher material 400, delaminating the stretcher material 400 from the sloped walls 116 will extend the x-y dimensions between each dice by 2× length of a single sloped wall 116.
Further processing can include depositing a spacer material 500 in the region 109 between the two dice 102, 104 as well as on the upper surface 115 of the wafer 100. The CTE of the spacer material 500 may closely match the CTE of the stretcher material 400 and/or have a CTE larger than that of the wafer 100. In such a configuration, the CTE mismatch may still be present, but in this case the CTE mismatch can propagate only over the area of each single die 102, 104, which even at 200° C. with a CTE difference of 50 PPM/° C. between the wafer 100 and spacer material 500, is only 2 microns over the areas of a 1 cm×1 cm die, in X-Y dimensions. As a result, the wafer 100 can remain flat and free of stress throughout any of the subsequent process steps.
For example, subsequent process steps can include providing vias 602 through the spacer material 500 to electrically communicate with respective ones of the contacts 106, FIG. 10. A further layer 600 may also be provided on the spacer material 500 and may include electronic components, such as passives, e.g., resistors, capacitors, inductors, in electrical communication with the vias 602. Alternatively or additionally, the layer 600 may include three-dimensional metal/dielectric structures, such as coaxial waveguides, antennas, Wilkinson or Gysel combiner/dividers, and filters, formed by the Poly Strata® fabrication processing (Nuvotronics, Inc., Virginia, USA). In a further exemplary configuration for lowering CTE while maintaining wafer-level planarity between adjacent dice 142, 144 in accordance with the present invention, FIGS. 11, 12 schematically illustrate two adjacent dice 142/144, 152/154 of a wafer 100 with a 3D spring membrane 700 disposed around the periphery of each dice 142/144, 152/154 on the wafer 100. The membrane 700 may act as a spring or CTE cushion to absorb the CTE mismatch that may occur as additional materials are added, such as spacer materials 200, 500 or additional layers 300, 600 like those illustrated and discussed in connection with FIGS. 1-10. Placement of the membrane 700 at the periphery of each die allows any CTE mismatch to propagate along each die 142, 144, 152, 154, instead of having the CTE mismatch propagate over the entire wafer 100, thereby stressing, bulging, and/or cracking the wafer 100. The membrane 700 can be a metal such as copper, a dielectric such as a low stress silicon nitride, or a polymer such as Parylene, benzocyclobutene (BCB) or polyimides, which can keep the wafer 100 in one piece but move and give way under CTE induced expansion.
The structure of FIG. 11 may be made by a process flow as shown in FIGS. 11A-11F to create a 3-D membrane 700 connecting each die 142, 144. FIG. 11A shows a beginning substrate is used, such as a silicon wafer 100. FIG. 11B shows that a hard mask 162 is provided having a hole 163 through which the wafer 100 may be crystallographically etched <100> using crystallographic silicon planes to create an etched hole 165, FIG. 11C. The mask 162 is stripped, as shown in FIG. 11D. Metallization is placed on the wafer 100, including in regions of the etched hole 165, which form a membrane 700 and optionally contacts 106, FIG. 11E. The membrane 700 may be thick copper (10 or more microns) or a polymer or any suitable material. The back side of the wafer 100 may then be processed using (for example) dry etching from the backside to singulate the dice 142, 144, leaving them connected by the membrane 700, FIG. 11F. A similar process may be used to create the structure of FIG. 12 in which steps 11A-11E are repeated, but wet etching on the crystallographic planes is used to singulate the dice 152, 154 instead of the dry etching of FIG. 11F.
In yet another of the aspects of the present invention, the structures of FIG. 8 may be leveraged to utilize the aforementioned increased real estate created by the separation of dice on the stretcher material 400. For example, the increased real estate affords a number of possibilities for making chip scale packages at the wafer level.
Chip scale packages are usually processed in strip or lead frame format using singulated dice, since the size of a chip scale package is slightly larger than the die size, due to industry definitions and the fact that the wire bond locations on the tape have to extend beyond the actual size of the die. FIG. 13 shows a typical existing Chip Scale Package (“CSP”) including a semiconductor chip 3001, wire bonds 3009, traces 3005, and solder ball contacts 3010. Using the concept of expanding the wafer size, shown in FIGS. 5-8, one can further process the deposited material 400 at the back 117 of the wafer 100 to create metallization, bonding pads and solder pads.
For instance, FIG. 14 illustrates an example of a CSP 800 in accordance with the present invention formed at the wafer or grid level. The wafer level CSP 800 includes electrodes 106, dice 102, 104, and a stretcher material 400 configured in a similar way as previously discussed with regard to FIGS. 5-8. In addition, an opening 3008 may be provided in the stretcher material 400 to permit wire bonds to electrically communicate between electrodes 106 and traces 3005 disposed on a dielectric substrate 3004. Solder pads 3010 may also be provided in electrical communication with the traces 3005 to complete the electrically conductive pathway from the top side of the CSP 800 and electrodes 106 to the bottom side of the CSP 800 and the solder pads 3010.
FIGS. 14A-14F represent exemplary process flow steps by which the CSP 800 may be fabricated. Starting with FIG. 14A, a stretched wafer 3000 is provided having a configuration similar to that shown in FIG. 8. The stretched wafer 3000 includes two adjacent dice 102, 104 adjoined and supported in a wafer-level planar fashion by a stretcher material 400. One or more electrodes 106 may be provided on each of the dice 102, 104. In this particular configuration the stretcher material 400 may be a dielectric, selected to have nonconductive properties to be compatible with the remainder of the structure in the CSP 800. FIG. 14B shows a separate base substrate 3000B with metallic traces 3005 and pads 3011 disposed on a dielectric material 3004, which may be supported on a dimensionally stable substrate carrier 3007. A sacrificial layer 3006 may be provided between the substrate carrier 3007 and dielectric material 3004 to facilitate removal of the carrier 3007 later.
The stretched wafer 3000 and the base substrate 3000B may be attached to each other using the dielectric stretcher material 400 or by adding an adhesive there between, FIG. 14C. A portion of the dielectric stretcher material 400 may be removed in the region 3008, such as by plasma etching, laser removal, or other suitable means for example, to expose end portions of the traces 3005, FIG. 14D. Then, wirebonds 3009 may be provided between the electrodes 106 and traces 3005 to provide electrical communication therebetween, FIG. 14E. FIG. 14F shows the removal of the dimensionally stable substrate carrier 3007, and FIG. 14G shows that solder balls 3010 may be placed at the wafer-level on the backside of the wafer assembly shown in FIG. 14F. It is further possible to test these packaged assemblies at the wafer-level prior to dicing or singulation of the chip-scale packaged devices. Also, lids may be applied to the wafer-level packaged devices.
In addition variations of the above structures shown in FIGS. 5-8 may be provided in accordance with the present invention, where a material can be applied at the side of the wafer 100 where the trench region 108 is located. The trench region 108 may be formed either from the front side or back side of the wafer 100, and then traces can be formed on the sloped sidewalls 116, where instead of wire bonds one can directly extend conductive trace metallization from one side of the wafer 100 to the other side, or onto a substrate on either side of the wafer 100, thus scaling the X-Y dimensions of the wafer 100, while still keeping the process at wafer level, for parallel and cost effective processes. This is merely one example in accordance with the present invention of how the stretching mechanism could be used for electronic device fabrication and packaging.
The approach of creating a spring membrane 700 between cells or dice to form a structure that can be reliably bonded to another structure that has a different CTE can be applied to various problems. For example in the art of PolyStrata technology (typically copper structures), it may be desirable to connect an antenna and/or feed and filter and micro-coaxial device structures to non-CTE matched circuits such as semiconductor wafers or large die or ceramic modules. When forming antenna arrays having a continuous ground plane for the antenna array; however, it may be required to create a larger area that may be formed in a CTE mismatched material, such as copper, to be joined for example to a silicon wafer using silicon germanium (SiGe) as the active semiconductor material, for high frequency applications.
FIGS. 15-17 schematically illustrate another exemplary structure in accordance with the present invention—an array 1100 of ground plane tiles 1102 forming an electrically continuous ground plane with signal conductors 1104 that pass through each ground plane tile 1102 and one or more ground conductors 1106 mounted on, and electrically connected to, the ground plane tiles 1102. The signal conductors 1104 may be suspended using dielectric supports 1108 which may be a thermally conductive ceramic and may be embedded into the ground plane tiles 1102, FIG. 17. The dielectric supports 1108 may be strips partially embedded in the signal conductor 1104, but could be membranes or shaped differently. FIG. 16 shows a bottom isometric view of the array 1100 of ground plane tiles 1102 with the ground connections 1106 for connection to a semiconductor chip (e.g., a chip having a significantly lower CTE than copper). There could be as many as eight or more ground connections 1106 around each signal conductor 1104, or there could be as few as one or two, depending on the required impedance matching, design rules and other possible factors. In addition, the unit cells 1102 could house other circuits (whether passive or active) therein.
The ground plane tiles 1102 (collectively the ground plane) of the three-by-three array 1100 may be electrically connected by CTE-decoupling membranes 1112, in this case formed as U-shaped springs 1112. The gap between the ground plane tiles 1102 and signal line feedthrough, as well as the dimensions of the conductors 1104, can be electromagnetically designed, simulated and optimized with software such as HFSS™ or CST™ to ensure the signals or RF power may move from structures above the ground plane tiles 1102, such as antennas, to structures below the ground plane tiles 1102, such as semiconductor devices on or in or even below a wafer/chip region connected to these structures. Creating designs with such software can minimize impedance mismatch causing reflections and power loss. The ground plane tiles 1102 may be formed by PolyStrata® technology, comprising beam forming passives, that may include couplers, baluns, filters, splitters, combiners, and so on.
Although one signal conductor 1104 is shown centered in each ground plane tile 1102, multiple signal conductors 1104 could pass through each ground plane tile 1102, such as a pair of differential lines for a single polarization antenna or four feeds for a pair of differentially fed orthogonal polarizations. While the U-shaped spring regions 1112 cause local deviations from the flat conductive ground plane, by minimizing their electrical length, the antenna ground plane can look like an electrically-continuous ground plane operating to high frequencies. In addition to U-shaped spring regions 1112 extending into the positive Z direction, the spring regions 1112 could extend into the negative Z direction. Also, the spring regions 1112 could be ‘V’ shaped, or repeat in sequence, or use different suitable shapes. Thus, the present invention allows one to join two otherwise-CTE-mismatched structures, because at least one of the structures is comprised of unit cells 1102 that are connected by regions 1112 that behave as springs. If the thickness of the metal membranes 1112 is chosen properly, the membranes 1112 will contract and expand against the thermal expansion of copper, thus decoupling all or most of the thermal expansion induced stress from adjacent terminals, e.g., signal conductor 1104 and ground connections 1106, on each array element 1102. That is, the CTE mismatched structure(s) (for example a semiconductor wafer and PolyStrata® passives) mounted on the array of unit cells 1102 no longer behave as rigid bulk slab(s) of different CTE being bonded together.
An exemplary use of the ground plane array 1100 of the present invention is a wafer-scale phased array. FIG. 18 schematically illustrates a cross-sectional view of an antenna array 1800 comprising the ground plane array 1100 of FIG. 15 with the addition of antenna emitters 1110.
In general, wafer-scale phased arrays include electronic circuitry behind each antenna element to provide beam steering and signal amplification and are of roughly the same size as the unit cell of the antenna element, which is generally approximately half of a wavelength at the upper end of the frequency of operation, but may be less or more depending on the electronic scanning requirements of the particular application. Such an architecture eliminates some levels of packaging and testing and may be a best method of making phased arrays at upper millimeter-wave frequencies. For wafer-scale arrays, precision definition of metal features, including an electrically continuous ground plane across an area of a few wavelengths or more on a side at the top end of the frequency of operation may be required.
Copper is an excellent electrical conductor at the frequencies of interest for wafer-scale phased array antennas, such as in the millimeter-wave frequency range, but copper has a significantly different CTE compared to that of a semiconductor wafer or reticle to which the copper is connected. Because of the size of the interconnected metal and semiconductor materials, the size of the ground plane required makes it difficult to have a metal with a thickness greater than 25 microns bonded to the semiconductor wafer or reticle for structures more than a few millimeters. As such, a method to decouple the mechanical interface between the two materials will reduce the CTE mismatch stress induced as the assembly is subjected to a variety of temperatures.
Because of electrical skin depth limiting the penetration of electric fields into the conductors at higher frequencies, copper thicker than 25 microns may not be required to pass RF signals, but other factors (such as required height above a ground plane of the radiating structure of an antenna for radiation efficiency or bandwidth considerations, or protection of circuits from damage caused by ionizing radiation in space based applications) may mean that thicker metal may be more interesting than common thin-film or printed circuit board metal thickness values. In such cases, there is a desire to make an antenna that provides radiation shielding to an integrated circuit, while also limiting the front-end loss of the system. This is used to limit single-event upset-type failures from occurring in a space-based electronic system.
The PolyStrata® process may be a good way of creating such an antenna because copper is a high-density material; however, copper may not be the best way of doing this because it has a coefficient of thermal expansion (CTE) that is different than the bulk CTE of most integrated circuits. The idea of using tungsten, molybdenum or other metals that have a low coefficient of thermal expansion to create these antennas may solve the thermal expansion problem, but these metals may not be as easily formed as copper by a process such as the PolyStrata® process due to, for example, limitations of electroplating such metals. One advantage of including tungsten is that it has a density twice that of copper, which means that it can roughly provide the same shielding properties in half the thickness. This can decrease the overall height of an antenna assembly on an integrated circuit. Three-dimensional copper spring structures made using PolyStrata® fabrication technology can address the CTE problems present in applications ranging from wafer level packaging to wafer level phased array constructions.
If the passive PolyStrata® ground plane circuitry of the antenna array behaves mechanically as a continuous copper slab, the thermal stress induced by different rates of expansion between the semiconductor chip or wafer and PolyStrata® circuitry may cause the terminal connections to either fail or cause the semiconductor wafer to crack due to thermal expansion mismatch between the metal and semiconductor. The proposed disclosure attempts to solve this problem by mechanically isolating each elements of the array with membranes (or springs) 1112, FIGS. 15-17, while keeping the electrical and high frequency performance intact. A typical two dimensional spring will result in RF scattering and unwanted antenna behavior. Other fields of use for this technique and such structures may include any focal plane array where a two-dimensional pattern of pixels with a transducer and processing electronics at every pixel is required (at microwave or millimeter-wave frequencies these transducers are antennas, but these could be ultrasound transducers or infrared detectors, among other things). Although not a phased array, passive millimeter-wave imaging arrays also have repeating unit cells with electronics for each pixel and an antenna as a transducer. Alternatively, the CTE-compensating features in accordance with the present invention may be used to conform the array across a surface that has some amount of non planarity where electrical connections must be maintained across a wide area. Hence the spring-like features joining unit cells or die can operate not only in one plane, but compensate for non planarity or curvature in the axis orthogonal to the plane in which they are formed.
FIG. 18 shows a completed antenna array with antenna radiators 1110 connected to the signal conductors 1104 of the ground plane array 1100 of FIG. 15. Although the signal conductors 1104 are shown to connect to the center of the antenna radiators 1110, the signal conductors 1104 could be located in different positions or multiple signal conductors could be attached to the antenna radiators 1110. The antenna unit cells can extend in X and Y beyond what is shown, even if they are connected to something with a dissimilar CTE because of the CTE decoupling provided by the membranes or springs 1112.
An exemplary process flow for fabricating the ground plane array 1100 shown in FIGS. 15-17 is provided in a step-by-step fashion in FIGS. 18A-18L, and an exemplary process flow for fabricating an antenna array 1800 is shown in FIGS. 18M-18O. The build steps are illustrated in an X-Z plane cross section. FIGS. 18A-18B show a photoimageable photoresist 2002 either laminated or spin coated over a temporary carrier substrate 2001. The photoresist 2002 is exposed to UV light, and a pattern is developed/formed by removing regions of photoresist using a photo-developer liquid. Electrode terminals 2003 are plated in locations where the photoresist 2002 has been removed. FIGS. 18C and 18D depict similar process steps to deposit or plate more metal and routing circuits. In FIG. 18C, photoresist is deposited in a relatively thicker layer, patterned, and non-desired portions removed to provide cavities defined by the remaining photoresist 2004. Metal 2005 is then deposited in the regions between the remaining photoresist 2004 to provide a ground plane structure, FIG. 18D. Dielectric supports 2006 are deposited to support the signal conductor on the ground plane metal 2004 in anticipation of future removal of the remaining photoresist 2005, FIG. 18E. In the Poly Strata® process, the supports 2006 are photopatternable.
FIGS. 18F-18G depict similar process steps of photoresist application, UV exposure, photo imaging, removal of non-desired portions of the photoresist to provide the remaining photoresist 2007, FIG. 18F, with the spaces in between the remaining photoresist 2004, 2007 plated with a metal 2008, FIG. 18G. Using similar techniques of photoresist application, UV exposure, photo imaging, and selective removal of photoresist, a ridge of photo resist 2009 is formed over remaining photoresist 2004, FIG. 18H. A conformal conductive seed layer 2010 is applied over the upper surface as shown in FIG. 18I in preparation for the deposition of additional metal. Additional photoresist 2011 may be provided in areas where metal plating is not desired by selectively depositing photoresist application, UV exposure, photo imaging, and selective removal, FIG. 18J. A conformal, thin membrane 2012 of copper may be applied in the areas where the seed layer 2010 is exposed and not covered by masking photoresist 2011, FIG. 18K. The thin copper membrane 2012 can provide a flexible three-dimensional membrane spring once the underlying photoresist 2009, 2004 is removed. Were the remaining photoresist 2004, 2007, 2009 and carrier substrate 2001 removed from the structure shown in FIG. 18K, the resulting structure would resemble ground plane array 1100 of FIGS. 15-17. However, the exemplary process of FIGS. 18A-18O continues in FIGS. 18L-18O to create the antenna array.
Specifically, FIGS. 18L-18N show a further set of photoresist coating, UV exposure, resist removal to provide photoresist 2013, 2014, after which a metal 2015 is provided in the gaps therebetween, where the shape of the metal 2015 is provided in the shape of antenna radiators 1110, FIG. 18O. All of the remaining photoresist 2004, 2007, 2009, 2013, 2014 and carrier substrate 2001 may be removed to provide the antenna array structure with ground plane array 2100 as shown in FIG. 18O. Instances of membranes or springs 2012 are shown as being flush with the top surface of the ground plane tiles 1102, but this is not necessary for all embodiments of the invention.
The antenna array 1800 may also be attached to a semiconductor chip or wafer 1900 using epoxy, solder or other suitable means, FIG. 19. The semiconductor chip or wafer 1900 could also be a ceramic or glass substrate, printed circuit board or other medium with an effective coefficient of thermal expansion that may be different from that of the antenna array 1800. In addition, although the steps shown in FIGS. 18A-18O indicate that the temporary carrier substrate 2001 may be removed to produce the antenna array 1800, by making the carrier substrate 2001 permanent, the full device shown in FIG. 19 may be created monolithically. In this case, semiconductor chip or wafer 1900 may be an active wafer, reticle or chip with integrated circuits or it may be a handle that provides rigidity and a known coefficient of thermal expansion that is set by the semiconductor chip or wafer 1900. If electrical through vias are put into the semiconductor chip or wafer 1900, the device of FIG. 19 could then be assembled into a larger system with antenna-element-level electrical connections to amplifiers or other electronics in an antenna system, whether an electronically-scanned phased array or otherwise. In addition, the semiconductor chip or wafer 1900 could serve as a barrier to prevent underfill material from penetrating the rest of the device or to provide a solder wick stop if the apparatus of FIG. 19 is bonded into a higher level assembly, to deter the flow of solder or other materials that may flow due to wicking back toward the antenna leads or connection points. An underfill material may be used to improve the reliability of the connections from the device of FIG. 19 to another package, but such materials may flow into regions they are not intended to penetrate.
Additionally, using multiple layers for the springs 2012 may suppress unwanted electromagnetic modes that may be supported in regions of air or vacuum between the membranes springs 2012 and semiconductor chip or wafer 1900. These modes may come from energy that leaks through the membranes or springs 2012, as it may not be physically continuous, although it is designed to be electrically continuous. The effect of these undesirable modes may appear or disappear depending on the electronic scanning conditions of the antenna array. This can effectively increase the frequency at which these electromagnetic modes are supported beyond the frequency band of interest for a given application.
In addition, the three-dimensional membrane 2012 formation, comprising the steps of FIGS. 18H-18L for example, two-dimensional springs 1112-2 may be provided between ground plane tiles 1102-2 with center conductors 1104-2, which springs 1112-2 may not be as efficient either mechanically or electrically, but nevertheless represents another embodiment of the present invention for providing mechanical decoupling while maintaining electrical connections, FIG. 20. The 2-dimensional springs 1112-2 having a serpentine shape and can provide similar methods of CTE decoupling can be provided as has been described in FIGS. 15-19; however, the aspect ratio of the metal and gap definition attainable using a particular fabrication process may constrain the ability of the designer to create springs 1112-2 with minimal impact on the overall performance of the antenna array or circuit. This is due to the additional inductance inherent in the springs 1112-2 compared to the springs 1112.
In addition, the ground plane tiles 1102-2 and springs 1112-2 can serve a dual purpose. First, the ground plane tiles 1102-2 and springs 1112-2 may provide an essentially electrically continuous ground plane over the frequencies of interest. The features of the springs 1112-2 and the gaps may be sized to be electrically small compared to the operating wavelength, so the ground plane of tiles 1102-2 appears continuous to the antenna elements. An electrically continuous ground plane can be important to the antenna's electrical performance. Second, the spring features allow for thermal expansion mismatch between a mainly-copper part (e.g., PolyStrata® part) and a substrate with significantly different thermal expansion, e.g., silicon, SiC, alumina, Low Temperature Cofired Ceramic (LTCC), etc. The springs 1112-2 may allow a part and/or a substrate adjoined thereto to thermally expand in-plane with different rates. The springs 1112-2 can flex to allow the individual ground plane tiles 1102-2 to move relative to one another without putting undue stress either within the ground plane itself. Additionally, the springs 1112-2 can flex to deter stress between the ground plane and attached structures, including, for example: a metal part (e.g., Poly Strata® part); the bonded electrical junction between the part and a substrate below, e.g., a wafer; and/or, an antenna radiator structure. The effect of the springs 1112-2 on a bonded electrical joint, such as that to a semiconductor wafer, can have a negligible effect on the joint's fatigue life, and can survive the required number of thermal cycles that would be expected during qualification testing and on-orbit lifetime for a typical spacecraft mission.
FIG. 21 schematically illustrates a modification to the apparatus shown in FIG. 18 to improve radiation shielding across the entire device surface. By making a jog 1120 along the path of the center conductor 1104, the regions of lowest line-of-sight radiation shielding have higher shielding levels. Item 1122 is a shield staggering feature in the CTE decoupling structure region. This lowers the effective shielding in those areas where the staggering occurs, but if the gaps between material of adjacent unit cells can be minimized, the reduction of shielding is minimized. With reference to FIG. 19, if the radiation shielding capability of the antenna array for the circuits on 1900 is important, it is possible to ensure those circuits most sensitive to ionizing radiation damage are not placed directly under the U-shaped spring regions 1112 or other areas with less shielding.
In yet a further aspect of the present invention, an array of antenna radiators may be provided using a singulation technique as illustrated in FIGS. 22A, 22B. A continuous array 2200 of antenna radiators may be provided that may include a bottom-side metal pattern 2220, a top-side metal pattern 2230, and circuit board 2240, FIG. 22A. The bottom-side and top-side metal patterns 2220, 2230 may be generally periodic and repeat on the antenna unit cell pitch. It is possible to not have a pattern on one or both sides of the board 2240, and there could be multiple pads on one side of the board (for instance in a case where two pads from the bottom-side metal pattern are capacitively coupling through the board 2240 to the topside metal pattern 2230). The continuous radiator array 2200 may be bonded to the ground plane array 1100 shown in FIG. 15 or the ground plane 1810 of FIG. 18, FIG. 22B. The individual unit cells may be separated from each other by laser ablation or using a dicing saw to decrease the amount of continuous circuit board material to reduce stress induced by CTE differences over the entire length of the continuation array 2200, as the antenna circuit board material is broken into unit-cell-sized pieces, FIG. 22B. By bonding a single continuous radiator array 2200 to the rest of the structure before singulation of the unit-cell-size pieces, there are fewer assembly steps than if each unit-cell-sized piece were assembled individually. As the size of the gap between adjacent unit cells can be small (on the order of 25 microns), the overall electrical effect of making the circuit board material discontinuous can be minimal.
These and other advantages of the present invention will be apparent to those skilled in the art from the foregoing specification. Accordingly, it will be recognized by those skilled in the art that changes or modifications may be made to the above-described embodiments without departing from the broad inventive concepts of the invention. For example, while fabrication via PolyStrata® technology has been mentioned, other methods of photolithography-based fabrication, 3d printing or other fabrication techniques could be employed in accordance with the present invention. Alternatively, the previous discussion has focused on the spring-like features being included in the material that has the higher CTE; however, through bulk micromachining (or other methods), the spring features could be integrated into the lower-CTE device bonded (such as silicon). It should therefore be understood that this invention is not limited to the particular embodiments described herein, but is intended to include all changes and modifications that are within the scope and spirit of the invention as set forth in the claims.
Patent Application
20170330836
