Patent Application
20210354977
This relates to microelectromechanical devices that include a released mechanical structure positioned over an undercut in a silicon nitride layer.
Microelectromechanical (MEM) relays can play an important role as a device for adding functionality and decreasing the power consumption for various applications such as sensors and consumable devices for the Internet of things (IoT) and wearables. One type of MEMS device is a mechanical relay. These devices have the capability of a quasi-ideal switching behavior with a very abrupt on-off switching, and zero current leakage during the OFF-state. Multi-terminal operation of relays can also save energy. See, for example, Martin Riverolo, et al, “High Performance Seesaw Torsional CMOS-MEMS Relay Using Tungsten VIA Layer,” 2018. A complementary metal oxide semiconductor (CMOS) platform can be used for the monolithic fabrication of such MEMS relays in a combination with classical CMOS devices.
CMOS MEMS is a technology where Al (aluminum) metallization and chemical vapor deposition (CVD) of tungsten (W) in VIA masks are used to create MEMS structures. One characteristic of this approach is that silicon dioxide (SiO2) between metal layers is used as the removable spacer. The SiO2 is typically removed using vapor hydrogen fluoride (HF) or liquid HF. Some CMOS MEMS devices use silicon (Si) (single crystal or polycrystal) as the MEMS removable layer. The Si can be etched with plasma fluorine (F) process or xenon difluoride (XeF2).
In described examples, a microelectromechanical system (MEMS) is located on a substrate. There is a silicon nitride (SiN) layer on a portion of the substrate. A mechanical structure has first and second ends. The first end is embedded in the SiN layer, and the second end is cantilevered from the SiN layer.
In the drawings, like elements are denoted by like reference numerals for consistency.
CMOS MEMS has several characteristics that makes it attractive. CMOS is quite mature, and analog or other circuits can be incorporated on the same wafer as a MEMS device(s). CMOS wafers are relatively inexpensive and can be fabricated with a wide variety of known process technologies.
An issue with fabricating MEMS using an example CMOS process is the silicon dioxide (SiO2) undercut process. Usually, a dielectric layer of SiO2 is formed and then a metallic structure is formed on the layer of SiO2. A wet hydrofluoric (HF) etch that is typically is used to undercut a portion of the SiO2 to release a portion of the metallic structure to form a MEMS device creates a high stress that is limiting for MEMS structure. The HF etch also attacks some of the other materials, like titanium (Ti), that are typically used within CMOS structures. Vapor HF is even more reactive and attacks SiN, which is typically used as a dielectric material. This makes it difficult to create dielectric elements that are part of a MEMS structure.
An example CMOS process also has a limitation in that it may not include materials that are needed for MEMS devices such as relays. Conductive materials that are used in a CMOS process such as W and titanium nitride (TiN) do not create good contacts for MEMS devices, such as relays.
In described examples, CMOS metals (such as: aluminum (Al), Ti, TiN, titanium tungsten alloy (TiW), W, copper (Cu), tantalum (Ta), tantalum nitride (TaN)) may be combined with alternative materials such: as tantalum pentoxide (Ta2O5), titanium dioxide (TiO2), aluminum oxide (Al2O3), titanium aluminum nitride (TiAlN), chromium nitride (CrN), titanium aluminum oxide nitride (TiAlON), molybdenum (Mo), aluminum nitride (AlN), aluminum scandium nitride (AlScN), hafnium zirconium oxide (HfZrOx), platinum (Pt), iridium (Ir), Iridium oxide (IrOx), lead zirconate titanate (Pb(Zr,Ti)O3), lead (Pd), lead oxide (PdO), gold (Au), silver (Ag), nickel iron alloy (NiFe), iron (Fe), cobalt (Co), nickel (Ni), cobalt nickel iron alloys (CoNiFe), ruthenium (Ru), ruthenium oxide (RuO2), etc. which may be useful for creating piezoelectric actuation and/or relay contacts with undercut SiN dielectric.
These alternative materials can be included in a CMOS process by using SiN dielectric layers that are flattened using chemical mechanical polishing (CMP) steps during the fabrication process. This makes it possible to create CMOS using SiN as the dielectric between layers. A plasma process using carbon, fluorine, and oxygen (CxFy+O2) is used to provide selective gas removal of the SiN and yet not attack most of these metals and dielectrics. For example, a plasma carbon tetrafluoride plus oxygen (CF4+O2) provides a strong etch rate for SiN while only weakly etching SiO2. This selectivity applies to most of the other materials although there is some attack of W, Mo, and Ru or RuO2. A downstream plasma with a low substrate bias is used. This process is similar to that used to ash resist. This plasma process etches the SiN in a mostly isotropic fashion. This is a useful characteristic for undercut etch of MEMS structures. This is a completely different etch process compared to VIA etch where a directional etch is needed to produce vias.
In described examples, SiO2 and many other dielectrics such as Al2O3, Ta2O5, TiO2 can be used to create a dielectric feature that is not strongly attacked by the plasma etch undercut process.
While this figure shows the creation of a cantilever, in practice multiple patterned and un-patterned layers can be below or above the MEMS layers that have been created using the undercut process. These additional layers may be used to create a wide variety of MEMS devices that can be created using this approach.
In this example, a layer of SiN 104 is formed over a substrate 102, which is this example is silicon (Si). For simplicity, only a small upper portion of substrate 102 is illustrated. As is known, a CMOS process typically fabricates active devices in a thin epitaxial layer of silicon that is formed on top of a bulk wafer of silicon. Also, for simplicity, this example is not drawn to scale. Beam 120 is significantly longer than it is thick. Typical VIA thicknesses 121 are those used in CMOS devices which are typically between 0.1 um to approximately Sum. The total beam thickness in this example is VIA thickness plus any additional layers above layer 132 or below layer 120 (not present in this figure). The released beam length 124 is typically much longer than it is wide with a typical aspect ratio (length to height) of 5 to 1 or even much larger. For example, if the beam is 1 um thick 121, then the released portion 124 of beam 120 is typically much longer than Sum. In other examples, the released portion of such a beam can be longer than 20 um and potentially longer than even 100 um, depending on material and cross-section design. The total length 122 of cantilevered beam feature 120 is always longer than the released beam length 124 so that there is a reasonable length 125 still embedded in remaining SiN 104.
In this example, beam 120 is fabricated from tungsten (W) with a thin liner of TiN. The tungsten and TiN are deposited using a known chemical vapor deposition (CVD) technique in a damascene style process and CMP of the W and TiN are used to remove unwanted metal. The TiN also acts as diffusion barrier and also as an etch protection layer for the W after the SiN has been removed by the undercut etch process. Not shown is a thin layer of Ti which is typically used as an adhesion layer and also used to create a lower resistance electrical connection to metals below this structure. The Ti is typically deposited with directional sputter deposition using ionized metal plasma in order to achieve thin layer of metal on the bottom of the VIA type feature. Damascene is the art of encrusting gold, silver, or copper wire on the surface of iron, steel, bronze, or brass. A narrow undercut is made in the surface of the metal with a chisel and the wire forced into the undercut by means of a hammer. In this example CMOS process, a CMP process is used to remove unwanted tungsten after the CVD process, as described in more detail herein below. This requires a flat surface, so either the surface is un-patterned, or CMP has been used to make it flat prior to the W pattern step. While this is similar to what is done in a CMOS process, in this case the W is surrounded on the bottom and sides by a SiN layer 103 rather than SiO2.
In a damascene process, a dielectric layer is first deposited onto the substrate. The dielectric layer is then patterned and filled by metal deposition. A dual-damascene process is characterized by patterning vias and trenches in such a way that the metal deposition fills both at the same time while leaving interstitial regions between the vias and trenches. The damascene process for beam 120 uses existent interlayer dielectrics in which the vias and trenches for conduction paths are etched. In this case, the dielectric layer 103 is SiN and the metal for beam 120 is tungsten. In another example, the metal for beam 120 may be selected from other metals, such as TiW.
In this example, W or TiW is used to create the MEMS beam structure 120 and provides a low creep rate. Creep in this instance is the change in deflection of the beam after being subjected to time and temperature and possibly added stress. Typical product times/temperatures are 10 yr, 85C, 105C, 125C or 150C. Some products require longer times or even higher temperatures. The stress depends on the application. There is always built in internal stress that might cause the beam to change position even without the added external stress. Many devices require that the beam position (lateral and vertical) not change when not under outside stress. Of course, beams always have a spring constant and do bend under stress. The thickness of the W is selected to provide greater than approximately 75% of the stiffness for beam 120 so a slow creep of other materials does not degrade the properties of the overall MEMS structure. In the example shown, a thin layer 105 of TiN surrounds beam 120, and an aluminum conductor 132 is patterned on top of beam 120. Aluminum is known to have a high creep rate but the overall beam or MEMS structure will not move much if the W has little relative creep and is a dominant fraction of the beams stiffness. In this example, the Al layer can be used not only to protect the top of the W in the beam from the SiN undercut etch but also as an electrical contact to the beam. The Al layer can also be on the bottom of the beam and act as an etch stop for the beam. In various examples, this Al layer also typically includes other materials like Ti, TiAl, TiN. In various examples, other materials may be present as needed for specific functionality.
A sacrificial SiN layer 134 is formed on over SiN layer 104 and various other elements such as Al conductor 132. An SiO2 dielectric layer 136 is formed on top of the SiN layer 134. Opening 138 is patterned in dielectric layer 136 to guide an undercut process. The layer 136 can be multiple layers or even other materials than SiO2, such as AlOx, SiON.
Undercut region 140 is formed using carbon tetrafluoride plus oxygen (CF4+O2) to provide a strong etch rate for SiN while only weakly etching SiO2 etch stop layer 103 and SiO2 top dielectric layer 136. In practice this process needs plasma activated fluorine plus oxygen. There are multiple options for F such as SF6 or other fluorocarbons, plus F2, NF3, HF, etc. For O2 sources there are also multipole options such as H2O, O3, NO2, CO2, etc. The listing of CF4+O2 is a common process that has demonstrated good results but other processes with these alternate chemistries do exist. As illustrated in
At
One or more deposition steps may be required to achieve a sufficient thickness of SiN layer 304. The generic term “SiN” for silicon nitride is used herein to refer to any of the various forms of silicon nitride, such as Si3N4, Si(x)N(y)H(z), etc. In this example, the SiN layer has a composition of SiOxNyCz where O is less than 0.1 and C is less than 0.3 and the N makes up the remaining fraction of the material excluding the Si. The hydrogen (H) symbol is typically omitted from the chemical formulas, such as SiN, but is frequently present in many of these materials including metals.
In this example, SiN layer 104 is deposited using a chemical vapor deposition (CVD) process or plasma enhanced chemical vapor deposition (PECVD). Chemical vapor deposition is a coating process that uses thermally induced chemical reactions at the surface of a heated substrate, with reagents supplied in gaseous form. The most common CVD or PECVD silicon nitride typically contains up to 8% hydrogen. Other methods to deposit SiN is sputter deposition or electron evaporation, but this is less common.
After deposition of SiN layer 304 is complete, the wafer surface is typically flattened using a chemical mechanical polishing (CMP) process. This is necessary if there are layers underneath that have introduced topography. The CMP process uses an abrasive and corrosive chemical slurry (commonly a colloid) in conjunction with a polishing pad and retaining ring, typically of a greater diameter than the wafer. The pad and wafer are pressed together by a dynamic polishing head and held in place by a retaining ring. The dynamic polishing head is rotated with different axes of rotation (i.e., not concentric). This removes material and tends to even out any irregular topography, making the top surface 3041 of the wafer flat, also referred to as “planar”.
At
At
At
TiN layer 330 is deposited over the surface of wafer 300. A layer of aluminum is then deposited over the surface of wafer 300. A sacrificial layer (not shown) is then deposited, patterned, and etched to form aluminum conductor 332 that forms a contact for MEMS relay device 100 (
Another layer 334 of SiN is then deposited over the surface of wafer 300, followed by a CMP process to planarize the surface 3042 of wafer 300.
At
At
While not described herein, various CMOS transistors may also be fabricated on wafer 300 using known or later developed integrated circuit fabrication techniques. Upon completion, wafer 300 is sawn, or otherwise separated, into individual chips, also known as die. The separate die are then attached to a lead frame and encapsulated using a known or later developed IC packaging technique, such as molding with a mold compound, to provide a packaged MEMS device integrated with CMOS circuitry.
MEMS device 400 is fabricated in a similar manner as shown in
In this case, plasma etch 450 removes SiN from interstitial space 426 and then diffuses through interstitial space 426 to form portions of undercut region 441. In this manner, an extensive undercut region 441 can be formed under beams that are larger than a what can be formed under a solid beam, such as beam 120 (
In this example, undercut region 540 is fabricated using a plasma of carbon, fluorine, and oxygen (CxFy+O2) to provide selective gas removal of the SiN layers 503, 534 through an opening in dielectric SiO2 layer 536 and yet not attack the Ir layer 560 or etch stop SiO2 layer 503.
IC 671 is attached to a lead frame 670 that includes contacts 674. Bond wires 675 connect bond pads on IC 671 to contacts 674 using a known or later developed wire bonding technique.
A mold compound 676 encapsulates IC 671 using a known or later developed encapsulation technique. In this example, completed MEMS device 600 is packaged as a surface mount device.
In described examples, CVD tungsten protected by CVD TiN on 3 sides is used to form a beam structure within SiN. The top side can be protected by another patterned and etched layer or by a self-aligned process like recess the W followed by barrier metal like TiN and more CMP. In other examples, physical vapor deposition (PVD) of titanium, Ta, TiW, TiN, TaN may be used to form a beam structure within SiN, for example. In each case, the use of a carbon, fluorine, and oxygen plasma to undercut the SiN has less impact on other materials. SiN is stronger and has a higher thermal conductivity than SiO2.
SiN undercut etch results in a clean surface where residual carbon or fluorine can be removed using plasma or vapor cleaning process with H2, H2O, O2, N2, NH3, NO, etc.
In described examples, W is used for the low creep material. However, there are other low creep materials, examples of which are included in Table 1. All the materials in Table 1 have extremely high melting temperatures above 1500 C, with most of them having a melting temperature above 2000 C. Materials that are typically used in semiconductor processing that qualify as low creep materials are C, Ta, Mo, Ir, Ru, Ti, and Pd. In addition, there are compounds like TiN or TaN that also qualify as high melting point and low creep materials. Alloys including the commonly used W alloys like TiW and NiW are also high melting temperature and low creep. With a few exceptions, most of the materials listed in Table Tare not typically used in CMOS processing and therefore do not have a well-established materials processing infrastructure of deposition, etch and clean. Most of these materials may be deposited by sputter deposition and hence need to be patterned by pattern and etch process and not the damascene process discussed in described examples. Some of these materials will be etched by an SiN undercut etch process and therefore will need to be protected. This can be done using protective layers at the top and bottom of the stack such as Ta or even a thin layer of SiO2, TiN, Al or AlOx. If necessary, the sides can also be protected by depositing the protective material (CVD TiN, AlOx) followed by etch back process to remove materials on the planar exposed surfaces. One advantage of an etch process to create the beam instead of the damascene process is that a solid beam can be created.
In described examples, a simple beam structure that can be used as a relay is described. In described examples, a beam that has a generally rectangular shape is described. In other examples, more complex structures may be formed within SiN using the plasma etch process described herein. These structures can be used in an extremely wide variety of different MEMS structures. A wide variety of materials that are compatible with the plasma etch undercut process can be used to create many types of possible devices. These can be simple structures to create actuators using electrostatic, magnetic, piezoelectric, thermal. The high temperature metals such as Pt, Ir, W, Ru, Mo, Ti can be used to create high temperature heaters which have numerous applications such as gas flow sensors, IR sources. MEMS structures with these can be used to create resonators, IR detectors, heat detectors. Electrical structures such as relays or RF switches with reliable contact materials are possible. Variable capacitor devices can be made using a flexible beam as described herein.
In described examples, a portion of a beam is released from a SiN layer, while another portion remains embedded in the SiN layer. In other examples, a released mechanical structure that has no portion remaining in the SiN layer may be fabricated. In some examples, a released mechanical structure may be supported by a torsion bar or similar support mechanism that is attached to the SiN layer. As used herein, the term “mechanical structure” refers to both fully and partially released structures of various shapes and sizes.
In described examples, a cantilevered beam is positioned within a cavity in a SiN layer. In other examples, the SiN layer may be configured so that it does not completely enclose the cantilevered mechanical structure. For example, there may not be a top layer over the mechanical structure. In another example, a large portion of the SiN layer may be removed, in which case the cantilevered mechanical structure may project from an edge of the SiN layer into essentially open space.
In described examples, the finished packaged devices are surface mount devices with multiple contacts on a bottom side of the package. However, in other examples, the IC package may have any number of known or later developed configurations, and may have various form(s), material(s), shape(s), dimension(s), number(s) of contacts, shape(s) of contacts, etc. Moreover, the MEMS resonator(s) and/or any other components may be packaged, mounted, etc. in the IC package in various configurations. Other examples of IC packages include a wafer-level package and a die-level package.
Many devices are encapsulated with an epoxy plastic that adequately protects the semiconductor devices and has mechanical strength to support the leads and handling of the package. Some integrated circuits have no-lead packages, such as quad-flat no-leads (QFN) and dual-flat no-leads (DFN) devices that physically and electrically couple integrated circuits to printed circuit boards. Flat no-lead devices, also known as micro lead frame (MLF) and small outline no-leads (SON) devices, are based on a surface-mount technology that connects integrated circuits to the surfaces of printed circuit boards without through-holes in the printed circuit boards. Perimeter lands on the package provide electrical coupling to the printed circuit board. Another example may include packages that are entirely encased in mold compound, such as a dual inline package (DIP).
In this description, the term “couple” and derivatives thereof mean an indirect, direct, optical, and/or wireless electrical connection. Thus, if a first device couples to a second device, that connection may be through a direct electrical connection, through an indirect electrical connection via other devices and connections, through an optical electrical connection, and/or through a wireless electrical connection.
Modifications are possible in the described embodiments, and other embodiments are possible, within the scope of the claims.
This application claims priority to U.S. Provisional Patent Application No. 63/024,850 filed May 14, 2020, the entirety of which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 63024850 | May 2020 | US |
