Method of making resonant tunneling diodes and CMOS backend-process-compatible three dimensional (3-D) integration

This application is related to previously filed application Ser. No. 10/767,275 Filing Date: Jan. 29, 2004 that is assigned to the same assignee as the present application.

1. Field of the Invention

This invention relates to solid state electronics, in particular to a novel high frequency silicon based resonant tunnel diode with negative differential resistance.

2. Description of the Related Art

The tunnel diode formed by a heavily doped p-n junction was invented by Esaki in 1958. This diode operated on the basis of interband tunneling, wherein charge carriers moved between valence and conduction bands by tunneling through an intervening potential barrier. Subsequently, in 1974, Esaki and co-workers demonstrated a resonant tunneling diode (RTD) consisting of two potential barriers separated by a potential well using a III–V compound semiconductor (L. L Chang, L. Esaki, and R. Tsu, “Resonant tunneling in the semiconductor double barriers,” Appl. Phys. Lett., Vol. 24, pp. 593–595, June 1974). In this device, the tunneling was intraband, between conduction and conduction or valence and valence bands, through an intermediate quantum well whose bound state energies corresponded to those energies of injected electrons which would have the maximum probability for tunneling.

Over the past three decades, RTDs exhibiting negative differential resistance (NDR) have received a great deal of attention due to their potential for application in electronics. Since the RTD offers the capability of operation as an oscillator, an amplifier and a mixer at extremely high frequency and with high resonant current density and very low noise, its implementation in integrated circuits would minimize the total device counts, and standby current. Indeed, Noble (U.S. Pat. No. 6,208,555) provides an SRAM memory cell that includes two tunnel diodes coupled in series and a MOSFET. RTDs with good I–V characteristics have been demonstrated in heteroeptaxial systems such as GaAs/AlGaAs/GaAs (Dong-Joon Kim, Yong-Tae Moon, Keun-Man Song and Seong-Ju Park, “Effect of barrier thickness on the interface and optical properties of InGaN/GaN multiple quantum wells,” Jpn. J. Appl. Phys., Part 1, 40, 3085 (2001)) and SiGe/Si (U.S. Published Patent Application No. 2003/0049894) and will be briefly discussed below. In addition, Bate et al. (European Published Application No. 94107763.8, Publication No. 0 668 618 A2) discloses a resonant tunneling device in which a silicon well is sandwiched between epitaxially grown layers of CaF 2 .

Although RTDs have been known and used in demonstrating the operation of an oscillator, an amplifier and a mixer at extremely high frequencies and with high resonant current density and very low noise, while maintaining the minimum total device counts and standby current desirable in integrated circuit (IC) technology, they have been difficult to integrate into mainstream Si CMOS (Complementary Metal Oxide Semiconductor) IC technology.

In the RTD structure, the silicon film is sandwiched on each side by a SiO 2 dielectric layer. The quantum barrier is made from this dielectric film, which has a relatively larger band gap than silicon. SiO 2 is not the only material suitable for the barrier layer that has a wider band gap than silicon. The difference in the band gap between the silicon and its surrounding barrier layers results in a positive conduction band-offset (difference between the conduction band edge and barrier height) with respect to the smaller band gap of Si. The silicon layer between the two barriers, that has a width close to the electron's deBroglie wavelength, forms a quantum well that supports a band containing several discrete electron energy levels that may be broadened by various processes. The electron transport across the barrier occurs by means of this energy band (or bands), which, by its presence, promotes the tunneling of injected electrons and produces a corresponding tunneling current when an appropriate bias voltage is applied. When the band energy of the well is close to the conduction electron energy of the emitter electrode (the “resonance” referred to in the device name), the maximum tunneling current is produced. This current decreases as the conduction electron energy departs from the energy in the band due to the applied bias. This reduction in current as the voltage is increased gives rise to what is called the negative differential resistance (NDR) behavior in the I–V characteristics of the tunnel diode.

Although the SiO 2 double barrier structure surrounding a silicon well was reported in H. Ikeda, M. Iwasaki, Y. Ishikawa, and M. Tabe, “Resonant tunneling characteristics in SiO2/Si double barrier structure in a wide range of applied voltage,” Applied Physics Letters, vol. 83, pp. 1456–1458, 2003, it remains a challenge for SiO 2 /Si/SiO 2 type RTDs to find their way into applications. This is a result of their poor performance, which is due mainly to the large band offset between SiO 2 and Si and the excessive thickness of the SiO 2 of the buried oxide layer in a silicon-on-insulator (SOI) substrate. Okuno, in both (U.S. Pat. No. 5,466,949) and (U.S. Pat. No. 5,616,515) discloses a resonant tunneling diode formed by layering silicon dioxide barrier layers on either side of a germanium well, but, as already noted, this device structure is not compatible with silicon processing schemes.

Recently, new forms of bonding have provided a possible approach to the integration of RTDs with the backend copper interconnect technology commonly used in IC fabrications. This integration involves low temperature bonding techniques such as Cu—Cu thermocompression, and plasma bonding. The efficacy and detailed description of these and similar methods is described by Rafael Reif, Andy Fan, Kuan-Neng Chen and Shamik Das in “Fabrication Technologies for Three-Dimensional Integrated Circuits,” International Symposium on Quality Electronic Design, Mar. 18–21, 2002, San Jose, Calif., pp. 33–37. The type of integration demonstrated therein leads, in the present invention, to 3-D device and circuit architectures in which CMOS devices lie on the silicon plane, while RTD devices are placed at the interconnect level. As is pointed out by K. C. Saraswat, K. Banerjee, A. Joshi, P. Kalavade, S. J. Souri and V. Subramanian, in “3-D ICs with multiple Si layers: performance analysis and technology,” 197 th Meeting of The Electrochemical Society, Toronto, May 2000, these techniques offer an effective use of the device area, while the replacement of lengthy horizontal interconnects by much shorter and more direct vertical interconnects leads to lower interconnect delays in overall circuit performance. Moreover, the 3-D integration with CMOS/BJT/SiGe (BJT being bi-polar junction technology) devices can lead to innovative circuit designs for future analog and digital circuit technologies, especially those involving SRAM and DRAM. By combining the advantages of 3-D architecture with the inherent fast tunneling properties of RTDs, an excellent emerging candidate for the next generation of nanoelectronic devices and circuits is produced.

A first object of the present invention is to provide a method of forming an RTD device that is compatible with mainstream back end and front end CMOS/BJT/SiGe technologies.

A second object of the present invention is to provide a method of forming an RTD device that is compatible with front end CMOS/BJT/SiGe technologies and whose integration is compatible with back end copper interconnect technologies.

A third object of the present invention is to provide a method of forming an RTD device that allows flexibility in choosing the mixed and asymmetric barrier materials and electrodes so as to enhance the RTD performance while maintaining process and integration requirements and ease of fabrication.

A fourth object of the present invention is to provide a method of forming such an RTD device wherein front and back end integration includes simple integration with both active and passive interconnect-level devices such as MIM (metal-insulator-metal) capacitors, interconnect inductors, phase shift memories and random access memories.

A fifth object of the invention is to provide a method of forming such an integrated RTD device wherein a 3-D architecture relative to the silicon plane is achieved.

A sixth object of the present invention is to provide a method whereby RTD devices can be formed first and then subsequently integrated with planar CMOS/BJT/SiGe devices, circuits and structures using conventional low temperature bonding processes such as anodic bonding, thermal compression, eutectic bonding, plasma bonding, silicon-to-silicon bonding, silicon dioxide bonding, silicon nitride bonding and polymer bonding.

A seventh object of the present invention is to provide such a formation and integration method whereby a required device area can be selected so as to meet the desired performance parameters of connected devices and circuits.

An eighth object of the present invention is to provide such a method of forming integrated RTD devices, wherein the devices so formed have good I–V characteristics, such as high peak-to-valley ratio (PVR).

A ninth object of the present invention is to provide a method of fabricating a plurality of integrated RTD devices so that the devices so formed have a high device density across a wafer substrate.

A tenth object of the present invention is to fabricate both active and passive devices using a method and structure similar to that applied to the formation of the RTD devices.

An eleventh object of the present invention is to provide a method of fabricating integrated RTD circuits at the interconnect level.

A twelfth object of the present invention is to provide a method of producing integrated RTD circuits that have small interconnect delays when the RTDs and other associated circuit devices are bonded and electrically connected using Cu—Cu pads.

A thirteenth object of the present invention is to provide the integrated RTD devices and circuits so formed.

The objects of the present invention will be achieved by a process in which a thin silicon film (the quantum well), of thickness less than 10 nanometers (nm) is sandwiched on each side by a dielectric layer (the barrier layer) formed of low band offset, high-K (high dielectric constant) dielectric materials such as SiO 2 , Si 3 N 4 , Al 2 O 3 , Y 2 O 3 , Ta 2 O 5 , TiO 2 , HfO 2 , Pr 2 O 3 , ZrO 2 , or their alloys and laminates. These quantum barrier materials are characterized by a larger bandgap than silicon, the material forming the well. The properties of these dielectric barrier materials and their role in producing a tunneling diode structure of superior characteristics are fully set forth in related patent application Ser. No. 10/767,275, filing date: Jan. 29, 2004, which is fully incorporated herein by reference.

A schematic cross-sectional view of a typical double barrier (one barrier layer on each side of the well layer) RTD structure meeting the objects of the present invention is shown in FIG. 1 . The RTD device layers are formed in the x-y planar direction, which is the planar direction of the upper surface ( 10 ) of the “handling” substrate, the term used to describe a finished processed wafer that is held in place so that it can be bonded with the RTD device structure in this invention. The handling substrate, in the present invention, includes the requisite CMOS (complementary metal-oxide semiconductor), BJT (bipolar junction technology) and SiGe (silicon-germanium) circuit elements with which RTD integration is to be achieved.

The RTD device ( 1000 ) includes a Si quantum well layer ( 50 ) less than 10 nm in thickness, that is formed between an upper ( 60 ) and lower ( 40 ) dielectric barrier layer having a high dielectric constant, each barrier layer being formed to a thickness less than 3 nm. It is noted that a Ge or SiGe well layer is an equivalent embodiment. An electron injection electrode ( 30 ) is formed beneath and on the lower barrier layer, the injection layer being preferably formed of semiconductor materials such as polysilicon, a-Si, SiGe, or Ge, and preferably formed to a thickness less than 3 nm. A collection electrode ( 70 ) is formed above and on the upper dielectric barrier layer, the collection electrode being formed of a metal or a semiconductor material such as that used in forming the injection electrode. It is noted that the dimensions of the well and barrier layers produce the energy levels required to create the resonant tunneling probabilities and superior negative differential resistance characteristics. A thermocompressionally bonded Cu—Cu layer ( 200 ), which is two Cu bonding pad layers (which can be pads or films), thermocompressionally bonded to each other, surface-to-surface, and illustrated as a single layer, joins the RTD device to the handling substrate ( 10 ). When aligned and joined by the Cu—Cu bonding pads, vias and associated interlayer interconnects (not shown) as well as the pads themselves, provide electrical connections between the RTD and the circuitry within the handling substrate. The lateral edges of the RTD device are contacted by dielectric isolation layers ( 75 ) of SiN to isolate the device from other nearby devices. The process steps leading to the formation of this structure will be described in detail below with reference to FIGS. 2 a–i and FIGS. 3 a–f.

The objects, features, and advantages of the present invention are understood within the context of the Description of the Preferred Embodiment as set forth below. The Description of the Preferred Embodiment is understood within the context of the accompanying figures, wherein:

FIG. 1 is a schematic view of the cross-section of the integrated double barrier resonant tunnel diode (RTD) of the present invention.

FIGS. 2 a–i are a series of schematic illustrations of the process flow of the fabrication of the double barrier resonant tunnel diode of FIG. 1 integrated with a Cu—Cu thermocompressionally bonded back end.

FIGS. 3 a–g are a series of schematic illustrations of the process flow in forming a double barrier resonant tunnel diode using the “smart-cut” technique, a particular fabrication technique appropriate to certain forms of Si substrate.

FIG. 4 is a schematic cross-sectional illustration showing an exemplary integration of three double barrier resonant tunnel diode with CMOS/BJT/SiGe circuits and/or devices, using Cu—Cu bonding pads and conducting interconnects.

The preferred embodiments of the present invention each teach a method of fabricating, within a SOI substrate, a double barrier resonant tunnel diode (RTD) of the cross-sectional form schematically illustrated in FIG. 1 and described above, and of integrating that RTD with an additional level of conventional CMOS/BTJ/SiGe devices, circuits and structures formed within a separate handling substrate. The integration is achieved using processes such as Cu—Cu thermocompression or plasma bonding to physically bond and electrically connect the RTD to the handling substrate to achieve the integration. The method produces a three-dimensional (3-D) integrated architecture relative to the handling substrate that contains the CMOS/BJT/SiGe devices, structures and circuits. The 3-D architecture provides operating advantages, such as reduced interconnect delay, and when those advantages are combined with the inherent operational speed and negative differential resistance (NDR) characteristics of the RTD, a basis for ultra fast circuits having very low noise is formed.

In the case of an RTD having a silicon layer as the quantum well layer, such as the RTD depicted in FIG. 1 , the fabrication process of the RTD begins most advantageously with a silicon-on insulator (SOI) wafer as a first substrate. The SOI wafer is a substrate of choice in many IC fabrication processes other than RTD fabrication, so its present use makes the invention described herein compatible with broader IC fabrication processes. However, the method of the present invention can also be applied advantageously to RTD structures having Ge quantum wells or SiGe quantum wells, in which cases the first substrate of choice would be a Ge-on-insulator (GOI) substrate or a SiGe-on-insulator substrate. It is also envisioned that other semiconductor materials could be formed into quantum well structures, in which case other substrates could be employed. Although the fabrication processes presented herein are applied in an exemplary manner to the formation of a Si quantum well RTD, and although Si is most probably the most common semiconductor material being employed in semiconductor fabrications, the generality of this method should be noted.

Referring to FIG. 2 a , there is shown an initial process step in a first embodiment that utilizes a semiconductor-on-insulator wafer as the substrate for the RTD fabrication. Within the method of the invention, the wafer can be a Si on insulator (SOI), a Ge on insulator (GOI) or a SiGe on insulator (SiGe—OI) wafer, depending on the nature of the quantum well that is desired. Within the present embodiment a SOi wafer is used as a substrate, but the choice should not be construed as a limitation on the generality of the invention.

As is well known in the art, the SOI substrate ( 20 ) is a tri-layer including a lower Si layer ( 100 ), on which is formed a layer of insulating material ( 110 ), typically SiO 2 , which is also called a buried oxide or “BOX” layer, on which is formed an upper layer ( 50 ) of monocrystalline silicon (the SOI layer). The upper layer, which will become the quantum well layer, has been reduced in thickness, by an etching process such as oxidation or chemical mechanical polishing, to a final thickness of less than 10 nm. It is understood that the thickness of the well in this embodiment is sufficient to form a plurality of electron bound states, with at least one bound state being required, and associated energy levels in order to provide the required resonant tunneling. It is also understood that electron scattering within the well can be reduced by reducing the thickness of the well and that such reductions can be used to optimally tune the performance characteristics of the RTD device. Apart from this, an n-type or a p-type doping with doping level between approximately 10 16 to 10 19 cm −3 is used to further tune the performance characteristics of the RTD device.

Referring next to FIG. 2 b , there is shown the formation of a thin layer ( 40 ) of low band offset, high-K dielectric material, such as SiO 2 , Si 3 N 4 , Al 2 O 3 , Y 2 O 3 , Ta 2 O 5 , TiO 2 , HfO 2 , Pr 2 O 3 , ZrO 2 , or their alloys and laminates as discussed earlier. This layer is deposited on the thin silicon layer ( 50 ) by a method such as chemical-vapor deposition (CVD), atomic-layer deposition (ALD), or sputtering, to a thickness between approximately 0.5 nm. and 5.0 nm. This layer will serve as the lower tunneling barrier layer. Immediately following this deposition, a layer ( 30 ) of n+ polysilicon (polysilicon, heavily n-doped in situ) is deposited on the barrier layer to a thickness approximately 0.5 μm (microns), using methods such as e-beam evaporation, CVD, ALD (atomic layer deposition), or sputtering. It is noted that all layer depositions in the fabrication are accomplished using methods known in the art, such as those cited above and also including LPCVD (low pressure chemical vapor deposition), APCVD (atmospheric pressure chemical vapor deposition) and MOCVD. The polysilicon layer will ultimately serve as the injection electrode of the RTD. A first metal barrier layer ( 25 ), preferably a layer of Ti whose purpose is to prevent Cu diffusion of a subsequently deposited Cu layer, is deposited to a thickness between approximately 2 nm and 200 nm on the polysilicon layer.

A first metal bonding layer, which can be formed as a thin film or patterned to form a bonding pad, which is formed of Cu ( 200 a ) in this embodiment, but which can be Au or other suitable metal, is formed on the Ti barrier layer to a thickness between approximately 2 nm and 200 nm. This bonding layer (film or bonding pad), which has an exposed surface ( 201 a ) will ultimately be thermocompressionally bonded, at that exposed surface, to a corresponding exposed surface of a second, identical Cu (or Au) bonding layer (or bonding pad), to be formed on a handling wafer (the wafer having the already formed required circuitry) as described below.

Referring next to FIG. 2 c , there is shown a second substrate ( 10 ), denoted a handling wafer or handling substrate, already including CMOS/BTJ/SiGe circuitry (not shown) and associated vias and interconnects (not shown) by which electrical connections and integration will be achieved. In accord with the objects of the invention, this handling wafer will be bonded to the SOI wafer already described above. To prepare for the bonding, a second metal barrier layer is formed on the handling wafer, preferably a Ti barrier layer ( 250 ), whose dimensions, method of formation and purpose is identical to layer ( 25 ) in FIG. 2 b . A second Cu bonding layer ( 200 b ) is formed on the Ti barrier layer, to the same thickness as first Cu layer ( 200 a ) of FIG. 2 b . This second Cu layer ( 200 b ) will be subsequently thermocompressionally bonded at its exposed surface ( 201 b ) to the exposed surface ( 201 a ) of first Cu layer. As already noted, the handling wafer ( 10 ) will typically contain the CMOS/BTJ/SiGe devices and circuitry to which the RTD will be physically and electrically interconnected by means of the bonding layer and with which it will thereby be integrated. Physical and electrical interconnections between the RTD and the circuitry within the handling wafer will be made as a result of the joining of the wafers by the Cu-to-Cu bonding layers (films or pads) ( 200 a ) and ( 200 b ) and electrical connections are also made by interconnects passing through vias (not shown) that pass between the various layers of the fabrication and are aligned (using conventional alignment tools) during the bonding process. Depending on the nature and location of the circuitry in the handling wafer, the Cu bonding pads may be patterned to allow access to vias and interconnects or the vias and interconnects may be formed peripherally to the bonding pads. As an alternative to the Cu bonding pads and layers, connecting layers and pads of other appropriate metals can be used, in which case the invention would include metal “A” to metal “B” bonding pads.

Referring next to FIG. 2 d , there is shown the thermocompressional bonding of the first and second Cu layers at their exposed surfaces, to join, thereby, the fabrication of FIG. 2 c with the fabrication of FIG. 2 b . A detailed description of the prior art method of thermocompressional bonding is to be found in Rafael Reif et al., cited above. It is noted that the bonding process requires alignment of the substrates so that conducting interconnects pass correctly through corresponding vias insuring proper electrical connections between the circuitry. The alignment can be achieved using conventional alignment tools.

The bonded, combined ( 200 a )–( 200 b ) Cu layers will be collectively denoted as bonded layer ( 200 ). It is noted that other appropriate bonding methods can also be used to satisfy the objects of the invention, namely: anodic bonding, eutectic bonding, plasma bonding, Si—Si bonding or polymer bonding. Several of these bonding methods can be implemented to bond the two substrates (SOI substrate and handling substrate) without the use of the metal bonding layers ( 200 a ) and ( 200 b ). The remaining layers formed above bonded layer ( 200 ) are the layers of FIG. 2 b , now inverted because of the face-to-face bonding that has occurred.

Referring next to FIG. 2 e , there is shown the fabrication of FIG. 2 d wherein the lower Si layer ( 100 ) of the SOI first substrate (now shown as an upper layer in the fabrication) has been reduced in thickness to a thickness of approximately 40 microns by an etching process such as grinding.

Referring next to FIG. 2 f , there is shown the fabrication of FIG. 2 e after a continuation of the etch of layer ( 100 ), this time using an XeF 2 /SF 6 plasma etch, has removed the remaining Si layer ( 100 ), wherein the SiO 2 BOX layer ( 110 ) serves as an etch-stop layer.

Referring next to FIG. 2 g , there is shown the removal of BOX layer (( 110 ) in 2 f ) using a buffered oxide etch, hydrophilic acid or other oxide etches well known in the art. In this etch, the Si quantum well layer ( 50 ) serves as an etch-stop layer.

Referring next to FIG. 2 h , there is shown the formation of an upper dielectric barrier layer ( 60 ) on the quantum well layer ( 50 ). The upper dielectric barrier layer is formed of the same materials, in the same manner and to the same thickness as the lower barrier layer ( 40 ). A metal collecting electrode ( 70 ), preferably a layer of Ti, formed to a thickness between approximately 2 nm and 200 nm is formed on the upper dielectric barrier layer.

Referring finally to FIG. 2 i , there is shown the results of a sequence of patterning and device isolation steps and contact formations to produce the final RTD device configuration. First, a desired RTD device area is defined by a patterning process that etches away peripheral portions of RTD layers ( 70 ), ( 60 ), ( 50 ) and ( 40 ), stopping at the upper surface of the injection electrode ( 30 ), leaving substantially vertical edges that define a regular horizontal cross-sectional area. Then SiN isolation layers ( 75 ) are laterally disposed against the edges of the patterned RTD device. A contact ( 77 ) is made to the collection electrode ( 70 ) and a contact ( 79 ) is also made to the injection electrode ( 30 ). It is through these contacts that electrical connections are made to the RTD. Connections, both physical and electrical, between the RTD and whatever embedded CMOS/BTJ/SiGe devices and circuits are contained (not shown) within the handling wafer ( 10 ) are made by means of the Cu bonding pads ( 200 ) and by interconnecting vias (not shown) that pass through the various layer formations.

A second embodiment of the present invention is fabricated using a SOI wafer as in the previous embodiment and applying what is denoted the “smart-cut” technique (a cleavage technique known in the art) to laterally cleave its upper Si layer, then thin it further so that it can be used as a Si well.

Referring to FIG. 3 a , there is shown the first step in this fabrication method in which an upper Si layer ( 50 ) of a SOI wafer ( 20 ) is implanted through an upper surface ( 55 ) with hydrogen to a depth that is less than 100 nm. The implant is represented as shaded region ( 57 ). It is noted that the implanted hydrogen, when combined with appropriate subsequent thermal processing of the wafer, produces a lateral cleavage of the upper Si layer ( 50 ) (hence the term, “smart-cut”) allowing the lower portion of the layer and the remainder of the SOI substrate beneath it to be removed. It is not the purpose of this invention to describe the smart-cut process, only to indicate that it is a known method of producing a lateral cleavage of the upper monocrystalline Si layer in an SOI wafer. To achieve the objects of the present invention, the smart-cut process is applied to a SOI wafer already bonded to another Si substrate, whereby the already bonded portion of the cleaved Si layer can be thinned to quantum well proportions while the remainder of the SOI substrate is removed.

Referring next to FIG. 3 b , there is shown a second substrate ( 300 ), which is a doped Si substrate that will ultimately serve as an injection electrode. A layer of high-K dielectric material ( 40 ), such as SiO 2 , Si 3 N 4 , Al 2 O 3 , Y 2 O 3 , Ta 2 O 5 , TiO 2 , HfO 2 , Pr 2 O 3 , ZrO 2 , or their alloys and laminates, is deposited on the second substrate and formed to a thickness less than 3 nm. The layer has an upper surface denoted ( 45 ). This layer will serve as the lower dielectric barrier layer for the finally formed RTD.

Referring next to FIG. 3 c , there is shown the second substrate ( 300 ) with the upper surface ( 45 ) of its barrier layer ( 40 ) bonded to the upper surface ( 55 ) of the hydrogen implanted Si layer ( 50 ) of the SOI wafer. The surface-to-surface bonding is achieved by low temperature plasma bonding or fusion bonding.

Referring next to FIG. 3 d , there is shown the SOI wafer, subsequent to thermal processing in accord with the smart-cut process, now cleaved parallel to the surface plane ( 55 ) of its hydrogen implanted Si layer ( 57 ), leaving the hydrogen implanted layer adhering to dielectric layer ( 40 ). The remainder of the SOI wafer, now denoted ( 25 ) is shown separated from the cleaved portion by double arrows. It will be discarded.

Referring to FIG. 3 e , there is shown the fabrication of FIG. 3 d wherein the cleaved-away portion of the SOI wafer has been removed and the remaining cleaved portion ( 57 ) of the implanted layer, bonded to the barrier layer ( 40 ), has been smoothed and thinned by a method such as CMP to the quantum well thickness of less than 10 nm.

Referring next to FIG. 3 f , there is shown the formation, on the quantum well layer ( 57 ), of an upper dielectric barrier layer ( 60 ), formed of the same high-K dielectric material and to the same thickness as the lower barrier layer ( 40 ). A metal collection electrode ( 70 ) is then formed on the upper barrier layer, the electrode being formed of a metal like Ti and to a thickness between 2 nm and 200 nm.

Referring finally to FIG. 3 g , there is shown the results of a sequence of patterning and device isolation steps and contact formations to produce the final RTD device configuration in a manner substantially identical to that described in FIG. 2 i . First, a desired RTD device area is defined by a patterning process that etches away peripheral portions of the RTD layers ( 70 ), ( 60 ), ( 57 ) and ( 40 ), stopping at injection electrode layer ( 300 ). Then SiN isolation layers ( 75 ) are laterally disposed about the edges of the patterned RTD device. A contact ( 77 ) is made to the collection electrode ( 70 ) and a contact ( 79 ) is also made to the injection electrode ( 30 ).

The lower surface ( 310 ) of the substrate ( 300 ) serving as an injection electrode can now be conventionally bonded to a handling substrate (not shown) that contains embedded CMOS/BTJ/SiGe devices and circuits. This bonding does not require metal—metal thermocompressional bonding, but can be done using such conventional methods as Si—Si bonding, polymer bonding, eutectic bonding, plasma bonding, and other methods known in the art. Conducting interconnects, such as those shown schematically in FIG. 4 below, aligned during the bonding process, can provide the electrical connections between the RTD device and the circuitry level in the additional substrate.

Referring finally to FIG. 4 , there is shown a schematic diagram illustrating the integration of three RTD devices ( 400 ), ( 500 ) and ( 600 ) with a handling substrate ( 10 ) on which several levels of CMOS/BJT/SiGe devices and circuits are formed. Small squares ( 600 ) symbolically indicate possible circuit elements at different levels. A conducting interconnect ( 700 ) runs from RTD ( 300 ) to a bonding pad ( 350 ). Another conducting interconnect ( 710 ) connects RTD's ( 400 ) and ( 500 ), through a via ( 800 ), to a bonding pad ( 900 ) connecting to a still lower level of circuitry. The Cu interconnects ( 700 ) and ( 710 ) are shown for example purposes only and the formation of such interconnects is known in the art and not described herein. As is noted above, the interconnections shown in this figure can be used whether the RTD is formed in accord with the method outlined in FIGS. 2 a–i or in FIGS. 3 a–g.

As is understood by a person skilled in the art, the preferred embodiment of the present invention is illustrative of the present invention rather than being limiting of the present invention. Revisions and modifications may be made to methods, processes, materials, structures, and dimensions through which is formed a double barrier RTD device that is integrated with CMOS/BTJ/SiGe devices and circuits through the thermocompressional bonding of Cu—Cu bonding pads, or other metal-to-metal bonding and conducting interconnects, to form a three-dimensional integrated architecture, while still providing such a three-dimensionally integrated and bonded RTD device, formed in accord with the present invention as defined by the appended claims.