Interposers with circuit modules encapsulated by moldable material in a cavity, and methods of fabrication

The present application claims priority of a U.S. provisional patent application No. 61/952,066, filed Mar. 12, 2014, incorporated herein by reference.

The present invention relates to semiconductor technology, and more particularly to use of auxiliary substrates for interconnection of semiconductor integrated circuits and other components.

Semiconductor integrated circuits (ICs) are miniature devices with tiny contact pads that must be connected to other IC or non-IC components. The connection to other components is facilitated by auxiliary substrates such as printed circuit boards (PCBs) or interposers. FIG. 1 illustrates ICs 110 connected to a PCB 114 through an interposer 120 . ICs 110 have contact pads 112 attached to contact pads 120 C.T on top of interposer 120 . The attachment is via connections 140 that can be solder, adhesive, or thermocompression for example. The interposer's bottom contact pads 120 C.B are attached to the PCB's contact pads 114 C (by suitable connections 140 ). Interconnect lines 120 L in interposer 120 provide desired connection between the interposer's contact pads 120 C (i.e. 120 C.T and 120 C.B). The PCB's interconnect lines 114 L provide desired connection between the PCB's contact pads 114 C. Other interposers, ICs, or other circuits can be connected to the PCB.

To provide small size, high operating speed, low power consumption, and low manufacturing cost, each component 110 , 114 , 120 may have densely packed circuitry with correspondingly densely packed contact pads 112 , 114 C, 120 C and preferably short interconnect lines (“wires”) 114 L, 120 L. In particular, the interposer 120 should be thin to shorten the vertical segments of lines 120 L. Thin interposers can be inexpensively fabricated from organic or ceramic materials. However, a thin interposer can be fragile and flexible and can be easily warped or broken by mechanical stresses, such as thermal stresses arising from thermal expansion. Thermal stresses are particularly damaging in the presence of materials with different coefficients of thermal expansion (CTEs). For example, silicon has a lower CTE than organic or ceramic materials commonly used in PCBs and interposers, and this is a serious problem for using organic or ceramic PCBs and interposers with silicon-based ICs. Thermal stresses can damage the connections 140 and make the assembly inoperable. Also, thermal stresses contribute to warpage which complicates the assembly process.

Therefore, it is desired to stiffen the interposer against warpage and also to use thermally conductive materials that spread locally generated heat and conduct such heat into the ambient.

FIG. 2 illustrates a known ceramic or organic interposer 120 with a cover 210 over a die 110 (a die is an IC that was manufactured as part of a semiconductor wafer and then separated from the wafer; the wafer may contain multiple dies fabricated at the same time). The interposer connects the die 110 to bottom contact pads 120 C.B; lines 120 L are not shown. Cover 210 is attached to interposer 120 by epoxy 220 . Cover 210 stiffens and flattens the interposer, and acts like a heat spreader: the heat generated by die 110 is conducted through thermal grease 230 to cover 210 and then to the ambient. See U.S. Pat. No. 7,061,102 issued Jun. 13, 2006 to Eghan et al.

In order to reduce thermal stresses at the interface between stiffener 210 and interposer 120 , the stiffener and the interposer can be made in a single substrate 120 S as shown in FIG. 3 and described in the Eghan patent. Substrate 120 S has a cavity 304 containing the die 110 ; the die is surrounded by the substrate's “perimeter wall” 304 P. The assembly can be further strengthened by underfill 310 between the die and the interposer: the underfill is made of epoxy that glues the die to the interposer and thus relieves the stress on connections 140 . Also, the underfill can be thermally conductive to spread the heat. In addition, thermally conductive encapsulant 320 fills the spaces between the die and the walls 304 P to protect the die and help in heat spreading. The encapsulant may or may not cover the die, and an additional heat spreader (not shown) may be provided on top.

FIG. 4 illustrates an organic (molded epoxy or plastic) interposer 120 supporting a stack of three dies 110 . 1 , 110 . 2 , 110 . 3 , as described in U.S. Pat. No. 6,492,726 (Quek et al., Dec. 10, 2002). Here the interposer cavity 304 has stepped sidewalls, with different dies 110 . 1 , 110 . 2 , 110 . 3 mounted on respective different steps 410 . 1 , 410 . 2 , 410 . 3 ( 410 . 1 is the cavity bottom). For each step, contact structures 420 extend from the respective “contact balls” 140 down through the interposer where they can be accessed from below the interposer. See also U.S. Pat. No. 7,977,579 (Bathan et al., Jul. 12, 2011).

The dies 110 (i.e. 110 . 1 , 110 . 2 , 110 . 3 ) of FIG. 4 can be underfilled and encapsulated with suitable dielectric materials (not shown). A heat spreader (not shown) can be provided on top.

This section summarizes some features of the invention. Other features may be described in the subsequent sections. The invention is defined by the appended claims, which are incorporated into this section by reference.

Some embodiments provide architectures and fabrication techniques suitable for organic, ceramic, silicon, or other types of interposers with cavities. An interposer structure may include multiple levels of dies or multi-chip modules (MCMs). Some embodiments provide simple manufacturing processes.

The invention is not limited to the features and advantages described above except as defined by the appended claims.

FIGS. 1, 2, 3, 4 show vertical cross sections of structures with interposers according to prior art.

FIGS. 5A, 5B . 1 show vertical cross sections of interposers in the process of fabrication according to some embodiments of the present invention.

FIG. 5B . 2 is a top view of an interposer in the process of fabrication according to some embodiments of the present invention.

FIGS. 5C, 5D, 5E, 5F, 5G, 5H, 5I, 5J, 5K, 5L, 6, 7, 8A show vertical cross sections of interposers in the process of fabrication according to some embodiments of the present invention.

FIG. 8B is a top view of an interposer in the process of fabrication according to some embodiments of the present invention.

The embodiments described in this section illustrate but do not limit the invention. The invention is defined by the appended claims.

In some embodiments of the present invention, stacked dies or multi-chip modules (MCM) can be placed in an interposer cavity that may have any shape, as in FIGS. 2-4 or other shapes. Some embodiments use vertical wires similar to 420 for connecting upper dies to circuitry below the interposer. However, in some embodiments, the wires go only through a small portion of the interposer thickness. For example, the wires may go through the cavity's bottom wall but not through any portion of the peripheral walls.

In some embodiments, the assembly is manufactured as a stack of levels inside the cavity; each level includes a respective die or dies 110 and encapsulant encircling the dies. A vertical wire may go through multiple encapsulant levels. A wire's segment through each level is made separately by depositing conductive material in a hole made in this level. This short segment is easier to fabricate than depositing conductive material into a deeper hole whose depth is the entire length of the wire. Since the long wires are easier to fabricate, higher stacking is possible, and each wire can be narrower to improve the wire density.

FIG. 5A illustrates the beginning stages of interposer fabrication according to some embodiments of the present invention. As shown in FIG. 5A , the interposer fabrication starts with a substrate 510 , e.g. a wafer made of silicon (monocrystalline or some other type) or another semiconductor or non-semiconductor material. We will focus on fabrication techniques suitable for wafers made of silicon, but similar techniques can be used for other semiconductor and non-semiconductor materials including metals, glass, ceramic, polymers, organic or composite materials, and including laminates.

Wafer 510 is sufficiently thick to provide desired stiffness, heat dissipation, and/or other desired properties. For example, some embodiments use a monocrystalline silicon wafer of 200 or 300 mm diameter and at least 650 micron thickness. In some embodiments, the length of vertical wires to be made in the structure need not be taken into account in selecting the wafer thickness because the wire length is not defined by the wafer thickness.

Cavity 304 is formed in the top surface of wafer 510 , by a masked etch for example. Multiple etches can be used, with different masks, to provide a stepped sidewall as in FIG. 4 if desired. An exemplary cavity depth is 400 microns or less. The vertical wires will pass through the cavity's bottom wall 304 B.

We will refer to the top side of wafer 510 as the “cavity side”, and the bottom side as the “anti-cavity side”. Also, we will refer to the cavity side as the “top side”, and will always show it on top, even though the wafer can be turned upside down or at any angle during fabrication or operation. Likewise, we will call the anti-cavity side the “bottom side”.

If multiple interposers are fabricated in the same wafer, then multiple cavities 304 can be formed at the same time. The multiple interposers can be made simultaneously, and the wafer can later be diced to separate the interposers. Only one interposer will be shown for simplicity.

The lateral extent of the cavity is defined by perimeter walls 304 P. The walls are sufficiently wide (horizontally) to provide desired stiffness and keep flat the cavity's bottom wall 304 B and the entire wafer 510 .

Then ( FIG. 5B . 1 ) holes 514 . 0 are made through the cavity's bottom wall 304 B to accommodate the bottom segments of the vertical wires. FIG. 5B . 2 is the top view of the structure, with an array of holes 514 . 0 . The holes may be in any number, and may have any geometry (circular as in FIG. 5B . 2 , square, or some other shape).

The holes can be made by a masked etch. The mask can be made on the anti-cavity side, and the etch can be conducted from the anti-cavity side. Alternatively, masking and etching can be done from the cavity side. Non-masking techniques can also be used, e.g. laser ablation or stamping from the cavity or anti-cavity side (stamping is appropriate if the interposer is made of a soft material, such as soft plastic). Other methods can also be used. For moldable materials (e.g. plastics), the structure of FIGS. 5B . 1 and 5 B. 2 can be made by molding without any etches or masks.

Holes 514 . 0 will contain conductive through-substrate vias (TSVs) for the vertical wires. Conventional formation of high density TSVs is a challenging process because on the one hand, the TSVs should be narrow for high density, and on the other hand the substrate has to be thick to provide mechanical strength. Making narrow TSVs through a thick substrate is difficult. However, the wafer of FIGS. 5B . 1 , 5 B. 2 is strengthened by peripheral walls 304 P, and the TSVs pass only through the bottom wall 304 B. If the bottom wall 304 B is thin, the TSVs can be narrow and therefore can be at higher density. The bottom wall's thickness is 20 microns or less in some silicon embodiments.

If desired, circuitry (not shown) is formed in wafer 510 , possibly including transistors, diodes, capacitors, and/or other circuit elements. The circuitry can be formed in the walls 304 P and/or 304 B before formation of cavities 304 and/or at any other fabrication stages.

We will refer to holes 514 . 0 as “level-zero holes” or “level L 0 holes”. Other holes will be made at higher levels as described below. (We use the concept of levels for ease of description; the invention pertains to structures and methods and not to any description techniques.)

If wafer 510 is made of a non-dielectric material (e.g. silicon), dielectric 518 ( FIG. 5C ) can be formed on the holes' sidewalls. The dielectric may cover all the wafer as in FIG. 5 C, or may cover just the holes' surfaces and possibly additional areas as needed. If the dielectric covers the whole wafer, no masking is needed. If the dielectric covers only the hole surfaces, the dielectric can be formed using the same mask (not shown) that was used to form the holes 514 . 0 (the dielectric can be patterned by a lift-off process for example).

The dielectric can be formed by sputtering (this is appropriate for silicon dioxide, silicon nitride, and some other dielectric materials), or thermal oxidation (if the wafer is made of silicon or some other oxidizable material whose oxide is dielectric), or by some other process. The above processes and dielectric materials are not limiting.

If desired, a barrier layer (not shown) can be formed over the dielectric 518 in the holes and elsewhere, to protect the wafer from contamination by subsequently formed materials, e.g. metals.

As shown in FIG. 5D , holes 514 . 0 are filled (or lined) with conductive material 520 . 0 to form TSVs that will serve as wire segments. This can be done by sputtering or other techniques, e.g. electroplating (which is relatively inexpensive). In the example of FIG. 5D , the electroplating uses a seed layer 520 S formed to cover the anti-cavity side. The seed layer may or may not enter the holes 514 . 0 . In an exemplary embodiment, the seed layer is copper formed by sputtering. To prevent copper formation on the top (cavity) side, the sputtering can be performed at an angle, i.e. the wafer can be inclined to prevent the copper molecules from reaching the cavity through the holes. Alternatively, the seed layer can be foil (e.g. copper foil) laminated on the anti-cavity side. Other techniques can also be used.

Then the structure is placed in a plating bath (not shown), and the seed layer 520 S is connected to a terminal 534 of an electric power source to serve as a cathode in the plating operation. As a result, the holes 514 . 0 are plated with conductor 520 . 0 , e.g. copper. Conductor 520 . 0 fills the holes 514 . 0 , and thus provides contact pads at the cavity bottom, i.e. at the top surface of bottom wall 304 B. The contact pads may be wider than the holes 514 . 0 .

Conductor 520 . 0 may include multiple layers made in different electroplating and/or electroless-plating steps.

As shown in FIG. 5D , if seed 5105 is not masked from the bottom, the conductor can also be plated on the bottom of the seed layer. The conductor plated on the seed layer's bottom is marked as 520 ′ in FIG. 5D , but is not marked separately in subsequent drawings. Rather, to simplify the drawings, we use numeral 520 S both for the seed layer and the conductor 520 ′, and we refer to both layers as “seed layer 520 S”.

Conductor 520 . 0 , 520 S may include multiple conductive layers. In some embodiments, the conductor top surfaces (inside the cavities) are usable as seed for electroplating of other wire segments as described below.

An optional protective layer 532 ( FIG. 5E ), e.g. polyimide or some other organic polymer, can be deposited on the bottom surface of layer 520 ′ to protect the structure and avoid electroplating on the bottom surface in subsequent plating steps.

One or more dies 110 are placed into the cavity onto bottom wall 304 B. Three dies 110 A, 110 B, 110 C are shown, and any number can be present. The dies' contact pads 112 are attached to respective contact pads 520 . 0 (by connections such as 140 described above, e.g. solder, adhesive, thermocompression, or possibly other types (possibly discrete wires); different connections can be used for different dies in the same structure). We will refer to these dies as level-one dies or level L 1 dies or L 1 dies.

The dies are underfilled and encapsulated by a suitable dielectric encapsulant 524 . 1 (level-one or L 1 “encapsulant” below). Underfilling can be done by capillary or no-flow techniques. Encapsulation can be done by spin-on deposition and curing of moldable material or possibly by other techniques. Suitable underfill and encapsulant materials can be conventional, including organic polymer resins—such as BCB (benzocyclobutene) or epoxy—with suitable fillers, or possibly other types. We will refer to the underfill and encapsulant as “encapsulant 524 . 1 ”.

In FIG. 5E , encapsulant 524 . 1 reaches as high as the top surfaces of dies 510 B and 510 C, thus encapsulating these dies laterally and below, but the encapsulant does not cover these dies. The encapsulant covers the die 110 A. The encapsulant can reach any level, and in particular can cover more than one dies, and/or on the contrary can terminate below the top surfaces of one or more of the dies. The top surface level can be adjusted by lowering the encapsulant level after deposition, e.g. using chemical etching or etching by abrasives (e.g. dry or wet blasting).

Some of contact pads 520 . 0 are not covered by any L 1 die 110 .

Holes 514 . 1 ( FIG. 5F ) are formed in the encapsulant to expose contact pads 520 . 0 not covered by L 1 dies 110 . We will refer to holes 514 . 1 as level− 1 or L 1 holes. A suitable process for making L 1 holes 514 . 1 depends on the encapsulant's material. For example, for resins, laser ablation can be used, which is a known technique in fabrication of through-mold vias (TMVs). If the encapsulant is photoimageable, then photolithographic techniques can be used, i.e. exposure to light through a glass-based mask followed by development in a developer solution. These examples are not limiting.

L 1 holes 514 . 1 expose contact pads 520 . 0 . The holes are filled or lined with conductor 520 . 1 ( FIG. 5G ) to form level-L 1 wire segments. Segments 520 . 1 can be formed by electroplating: the plating current can be provided from the anti-cavity side through seed layer 520 S and LO wire segments 520 . 0 , by connecting the seed layer to an electric power source's terminal 534 . Conductor 520 . 1 fills (or lines) L 1 holes 514 . 1 , and thus provides contact pads (“L 1 contact pads”) at the top surface of L 1 encapsulant 524 . 1 . The L 1 contact pads may be wider than the L 1 holes 514 . 1 .

This process can be continued any number of times, to build any number of levels. For example, FIGS. 5H-5L illustrate level L 2 formation. One or more dies 110 (“L 2 dies”), including the die 110 D, are attached on top of level L 1 ; the dies' contact pads are attached to L 1 contact pads 520 . 1 . In this embodiment, die 110 D has contact pads 112 both on top and bottom. The top contact pads 112 will be attached to overlying circuitry (not yet formed).

L 2 encapsulant 524 . 2 is formed to underfill and encapsulate the L 2 dies. L 2 holes 514 . 2 ( FIG. 5I ) are formed in L 2 encapsulant 524 . 2 to expose those L 1 contact pads 520 . 1 that are not covered by a die. L 2 holes 514 . 2 are filled or lined with conductor 520 . 2 ( FIG. 5I ) which provides L 2 wire segments. Conductor 520 . 2 can be formed by electroplating; the plating current can be provided from the bottom side through seed layer 520 S and lower-level wire segments 520 . 0 , 520 . 1 by connecting the seed layer 520 S to a terminal 534 of an electric power source. The L 2 wire segments 520 . 2 contact the underlying L 1 wire segments 520 . 1 and provide contact pads at the top of L 2 encapsulant 524 . 2 . The contact pads can be wider than the L 2 holes 514 . 2 . The L 2 features can be formed using the same processes as described above for level L 1 .

FIG. 5K shows a final structure with four levels and a heat spreader 540 on top. Level L 3 includes a die 110 E some of whose bottom contact pads are attached to the level L 2 contact pads 520 . 2 (not marked in this figure). Other bottom contact pads of die 110 E are attached to top contact pads 112 of L 2 die 110 D. L 3 die 110 E is underfilled and laterally encapsulated by encapsulant 524 . 3 made of moldable material. L 3 wire segments 520 . 3 are vertical segments passing through encapsulant 524 . 3 . L 4 die 110 F has bottom contact pads attached to top contact pads of L 3 die 110 E and to the wire segments 520 . 3 (to the top contact pads provided by these segments). L 3 wire segments 520 . 3 can be formed by electroplating; the plating current can be provided from the bottom side through lower-level wire segments 520 . 0 , 520 . 1 , 520 . 2 . L 4 die 110 F is underfilled and laterally encapsulated by encapsulant 524 . 4 made of moldable material. L 4 wire segments 520 . 4 are vertical segments passing through encapsulant 524 . 4 . External circuits (not shown), e.g. dies, MCMs, or discrete wires or other discrete components, can be attached to the top ends of segments 520 . 4 . The top ends can be formed to facilitate such attachment; for example, if the attachment will be by solder, then a barrier layer such as nickel can be electroplated or electroless-plated to protect underlying portions of segments 520 . 4 from solder contamination; and gold can be electroplated or electroless-plated on top of segments 520 . 4 to block oxidation during soldering.

The structure of FIG. 5K includes a vertical wire 520 T having no electrical functionality but provided for heat removal. This wire consists of vertical segments 520 of levels 0 through 4 (i.e. 520 . 0 , 520 . 1 , 520 . 2 , 520 . 3 , 520 . 4 ). This wire reaches heat spreader 540 . The top level segments 520 . 4 can be formed by electroplating; the plating current can be provided from the bottom side through lower-level wire segments 520 . 0 , 520 . 1 , 520 . 2 , 520 . 3 . Any number of heat-removal wires can be provided; the heat-removal wires may or may not reach the heat spreader.

The top level L 4 includes wire segments 520 . 4 that provide contact pads at the top of encapsulant 520 . 4 side-by-side with heat spreader 540 . These contact pads can be connected to external circuits as mentioned above.

In some embodiments, each encapsulant layer 524 is deposited and cured in a separate operation; the encapsulant layer does not have any separately-cured sub-layers. Therefore, the encapsulant layer does not have any internal surface boundaries that would be present between sub-layers. An internal surface boundary is characterized by surface states. A surface state is a state with molecules having higher energies than below the surface; surface states may be detectable visually (possibly using microscopy) or perhaps by ultrasound or other ways. The invention is not limited to such embodiments however.

In some embodiments, if a die or dies 110 of a level Li are underfilled in a separate operation than lateral encapsulation, and the underfill is separately cured, then the encapsulant layer may have a surface at the boundary between the underfill and laterally encapsulating portion of the encapsulant. Further, the laterally encapsulating portion can be made of a different material than the underfill.

After or at any point during fabrication of all the levels, the interposer's bottom can be processed to remove or pattern protective layer 532 and conductor 520 S in any desired manner. In FIG. 5L , layer 532 is removed, and the conductor is thinned and patterned to provide contact pads 120 C.B at the ends of wire segments 520 . 0 . These contact pads can be attached to other dies, MCMs, or packaging substrates such as PCBs or interposers. Alternatively, other circuitry can be formed on the bottom, e.g. as part of a redistribution layer (RDL) including multiple dielectric, conductive, and semiconductor layers that provide circuitry coupled to wire segments 520 . 0 and accessible from below the interposer. The redistribution layer may have contact pads at the bottom that can be coupled to other dies, MCMs, or packaging substrates such as PCBs or interposers.

FIG. 6 shows additional features that may be available in some embodiments. Here level L 3 includes a redistribution layer 610 . 3 which may have a number of conductive layers separated by dielectric layers. The conductive layers provide interconnects 620 . 3 which extend both vertically and laterally (possibly horizontally). Interconnects 620 . 3 terminate in contact pads 630 . 3 at the top of the RDL. Interconnects 620 . 3 interconnect the underlying wire segments 520 . 2 and the overlying wire segments 520 . 3 as desired. In some embodiments, the dielectric layers of RDL 610 . 3 are formed of the same or similar materials as encapsulant layers 524 (e.g. moldable organic polymer deposited by a spin-on process, or from other materials described above), and are patterned by the same techniques (e.g. photolithography or laser ablation). The conductive layers can also be formed by other deposition and patterning techniques, e.g. using photolithography or by additive manufacturing (printing).

In the example of FIG. 6 , some RDL lines 620 . 3 connect L 2 wire segments 520 . 2 to L 3 wire segments 520 . 3 . A line 620 . 3 connects a top contact pad 112 of L 2 die 110 E to the top of an L 2 wire segment 520 . 2 . Another line 620 . 3 connects an L 2 wire segment 520 . 2 to a bottom contact pad of L 3 die 110 F.

Contact pads 630 . 3 at the ends of lines 620 . 3 can provide an area array pattern or other patterns as needed for connection to L 3 dies and wire segments.

In some embodiments, level L 2 includes a die that has no bottom contact pads, but has only top contact pads connected to other circuitry by RDL 610 . 3 . Any level may include a die having no bottom contact pads and having top contact pads attached to other circuits within or outside the cavity. Also a die may have no top contact pads. In the example of FIG. 7 , die 110 E has no top contact pads, and further L 2 encapsulant 524 . 2 covers the die, and RDL 610 . 3 covers the encapsulant.

Level L 3 wire segments 520 . 3 ( FIGS. 6, 7 ) can be formed at the top of RDL 610 . 3 by electroplating; the plating current is conducted from the bottom layer 520 S to the top of RDL 610 . 3 through wire segments 620 . 3 .

L 4 wire segments 520 . 4 and their respective underlying L 3 segments 520 . 3 are for heat removal.

Above, RDL 610 . 3 is associated with level L 3 , but this is simply a matter of terminology: we could say that RDL 610 . 3 is part of level L 2 . An RDL can be provided at any level. Also, multiple alternating RDLs and encapsulant layers can be provided in a single level, i.e. for a single set of dies 110 attached to the preceding level.

Substrate 510 can be assembled from different substrates, and an example is shown in FIGS. 8A (vertical cross section) and 8 B (top view on a smaller scale than FIG. 8A ). In this example, substrate 510 includes a planar substrates 510 . 1 and a substrate 510 . 2 attached to planar substrate 510 . 1 . The attachment is by adhesive 810 , but direct bonding can also be used. In FIG. 8B , multiple substrates 510 . 2 (called “frames” below) are attached to a single planar substrate 510 . 1 , and multiple structures are formed of the type of FIG. 7 or other types discussed above on the same planar substrate 510 . 1 . The planar substrate 510 . 1 can be diced along lines (not shown) passing between the frames 510 . 2 at the end of fabrication or at any other fabrication stage.

Planar substrate 510 . 1 and frames 510 . 2 can be made of the same or different materials, including any materials mentioned above for wafer 510 . Adhesive 810 , if present, can be any suitable adhesive, e.g. plastic deposited in a flowable form and then cured using heat, pressure, radiation, and/or other means. Non-flowable adhesives can also be used. Substrate 510 may include a stack of more than two substrates, e.g. each frame 510 . 2 or planar substrate 510 . 1 may be a laminate of multiple substrates.

Regardless of whether or not the substrate 510 is an assembly of multiple substrates, substrate 510 (and each constituent substrate if present) may be a uniform or non-uniform medium, i.e. may or may not have the same physical properties at each internal point. In some embodiments, substrate 510 is uniform with respect to any one or more of the following properties: chemical composition, density, elasticity modulus, conductivity, dielectric constant, ultrasound (US) propagation speed for one or more wavelengths (possibly all US wavelengths), light propagation speed for one or more wavelengths (e.g. wavelengths in the infrared range and below). For example, uniform substrate 510 can be a single crystal (e.g. monocrystalline silicon), and can be organic or inorganic, and can be a composite material or other materials described above. These examples are not limiting. If the substrate 510 is an assembly of multiple pieces such as 510 . 1 and 510 . 2 , then in some embodiments an interface region of the adjacent pieces (the region including adjacent parts of the two pieces) is not uniform with respect to at least one of the properties mentioned above, i.e. chemical composition, density, elasticity modulus, conductivity, dielectric constant, ultrasound propagation speed for one or more wavelengths, light propagation speed for one or more wavelengths. For example, in FIG. 7A , the interface region can be defined as including adhesive 810 and adjacent surfaces of substrates 510 . 1 and 510 . 2 , and the interface is not uniform if the adhesive has a different chemical composition than the adjacent surfaces of substrate 510 . 1 or 510 . 2 . If the attachment is by direct bonding of substrate 510 . 1 to 510 . 2 and the substrates are made of the same material (e.g. silicon), the interface can be uniform, and the substrate 510 can be indistinguishable from a substrate made initially from a single piece. However, the interface can also be non-uniform with respect to light or sound propagation speed, e.g. if the interface includes bubbles or other states detectable by US or spectral imaging.

The invention is not limited to the embodiments described above. Other embodiments and variations are within the scope of the invention, as defined by the appended claims.