Electro-optical device
The present invention relates to an electro-optical device, in particular to an electro-optical device sensitive to radiation.
A wide variety of electro-optical devices are currently produced using semiconductor materials, including for example silicon charge-coupled device (CCD) imagers and photodiodes. Many applications now require image detectors which have fully integrated enhanced amplification and readout functionality, which is predominantly realised using silicon technology. Unfortunately the photon absorption characteristics of silicon largely precludes its use where detection of infra-red or filtered ultra-violet radiation is required. Therefore there is a need to integrate silicon with a range of semiconductor materials which have spectral responses over chosen wavelength ranges.
According to a first aspect of the present invention, there is provided a monolithic electro- optical device comprising vertically integrated active semiconductor and radiation sensitive semiconductor layers, with an insulating layer interposed therebetween.
Preferably, the radiation sensitive layer comprises doped junction regions, most preferably photodiodes, which detect radiation having a wavelength falling within a predetermined wavelength range.
The doped junction regions are interconnected with the active semiconductor layer.
Preferably, the active semiconductor layer and the radiation sensitive semiconductor layer are each formed from different materials each having different energy bandgaps .
Preferably, the active semiconductor layer, insulating layer and radiation sensitive semiconductor layer are formed on a handle layer.
Preferably, the handle layer can be selectively etched, most preferably to remove or thin at least one portion corresponding to a region of a pixel to be formed in the radiation sensitive layer.
The region of a pixel is defined as the pitch between successive doped junction regions in the radiation sensitive layer.
Preferably, the handle layer is transparent to the radiation having a wavelength falling within the predetermined wavelength range.
Preferably, the handle layer is formed from the same base material as the radiation sensitive layer.
Preferably, the handle layer may comprise at least one further insulating layer.
Preferably, the insulating layer can be employed as an etch-stop layer in a thinning process.
Preferably, the device comprises at least one further radiation sensitive layer.
Preferably, the device includes an electrically conductive shield layer above, below or within the insulating layer(s) . Preferably, said electrically conductive shield layer is formed from a refractory metal or a suicide thereof.
Preferably, said electrically conductive shield layer is formed from tungsten or tungsten suicide.
Preferably, the active semiconductor layer comprises for example CCD, MOSFET, CMOS or MESFET pixel circuitry, most preferably active pixel circuitry.
Preferably, the radiation sensitive semiconductor layer comprises a germanium layer.
The use of narrow bandgap germanium means that the electro-optical device has an enhanced response to incident radiation in the infra-red or near infra- red spectrum.
Optionally, the radiation sensitive semiconductor layer comprises a silicon carbide layer.
The use of wide bandgap silicon carbide means that the electro-optical device has an enhanced response to incident radiation in the ultra-violet spectrum.
According to a second aspect of the present invention, there is provided an image sensor comprising the electro-optical device of the first aspect.
Preferably, the image sensor is arranged such that, in use, incident radiation impinges on a rear surface of the radiation sensitive layer.
According to a third aspect of the present invention, there is provided a method of making a monolithic electro-optical device, comprising the steps of:
combining a handle wafer and a wafer comprising a radiation sensitive semiconductor to form a substrate; and combining the substrate with an active semiconductor wafer, said active semiconductor wafer comprising an insulating layer,- such that the active semiconductor wafer and radiation sensitive semiconductor are vertically integrated, the insulating layer being interposed therebetween.
Each wafer then becomes a corresponding layer of the electro-optical device.
Preferably, the step of combining the handle wafer and the wafer comprising a radiation sensitive semiconductor to form a substrate comprises a first wafer bonding and ion splitting process
Preferably, the step of combining the substrate with the active semiconductor wafer comprises a second wafer bonding and ion splitting process.
Preferably, the method further comprises polishing an upper surface of the structure obtained after the first and/or the second combination steps.
The method may further comprise forming pixel circuitry in one or more portions of the active semiconductor wafer, most preferably employing standard semiconductor device fabrication techniques.
The method may also further comprise forming doped junction regions, most preferably photodiode regions, in one or more portions of the radiation sensitive wafer, most preferably by ion implantation or diffusion techniques.
The doped junction regions are interconnected with the active semiconductor layer.
Preferably, the method also comprises the step of selectively etching the handle wafer, most preferably to remove or thin at least one portion corresponding to a region of a pixel to be formed in the radiation sensitive layer.
By "selective" etching, it is meant that an etchant is used which has suitable properties to erode the handle layer without destroying the insulating layer.
Preferably, the method further comprises the step of forming at least one further insulating layer within the handle wafer.
Preferably, the further insulating layer acts as an etch stop in a back-thinning process.
Preferably, the method further comprises the step of forming at least one further radiation sensitive layer.
The present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:
Fig. 1 shows an electro-optical device according to a first embodiment;
Fig. 2 shows an electro-optical device according to a second embodiment; and
Figs. 3 and 4 illustrate process steps in the fabrication of the electro-optical devices shown in Figs. 1 and 2.
Fig. 1 shows a cross section of a first embodiment of a multi-layer electro-optical device. A radiation sensitive semiconductor layer 12 is provided below an insulating layer 14, on which an active semiconductor layer 16 is formed. Pixel circuitry 18, for example active pixel circuitry, is formed in the active layer 16 and insulating layer 14. A metal interconnect 15 is provided to connect a doped junction region 26 to the pixel circuitry 18. The radiation sensitive semiconductor layer 12 interacts with electromagnetic radiation 24, for example by absorbing it. A handle layer 10 is optionally provided to provide mechanical support for the device. The handle layer 10 may be made from a semiconductor material.
A rear surface 25 of the device is defined as the bottom surface as illustrated in the drawings, i.e. the surface opposed to the surface of the pixel circuitry 18. Radiation 24 incident on the rear surface 25 of the structure is transmitted through the handle layer 10 and absorbed by the radiation sensitive semiconductor layer 12, creating electron hole pairs which are separated and swept to the appropriate p+ or n+ wells 26 by the electric field in a reverse bias depletion region of the radiation sensitive semiconductor layer 12. The device has on its rearward surface metal contacts 20 to reverse bias the photodiode junctions.
Additional vertically integrated circuit and/or sensor layers may be added to this structure, separated by insulator layers as required.
For example, an electrically conductive shield layer can be included above, below or within the insulating layer. The shield layer can suitably be formed from any refractory metal or a suicide thereof.
In particular, Silicon-on-insulator (SOI) substrates incorporating low resistivity silicon-rich tungsten suicide (WSix) buried ground plane layers have demonstrated the lowest reported crosstalk figure of merit for application in mixed signal integrated circuits. To form such a shield layer amorphous, silicon-rich WSix (W:Si, 1:2.6) is deposited using a LPCVD process employing SiH4 and WF6 source gases, at
a temperature of 3700C. Grain growth during a 10000C anneal causes the suicide layer to become polycrystalline, and the resistivity reduces to less than 50 μΩ-cm.
The incorporation of a shield layer protects the vertically integrated readout electronics from possible radiation-induced damage, and has application for back-illuminated detectors used in high-energy physics applications.
The structure illustrated in Fig. 1 can have a variety of purposes and can be made in a number of ways. In particular, the materials chosen to form the layers 10, 12, 14, 16 may be varied and selected as appropriate for the required use of the electro- optical device. In one particular embodiment, an imaging device is constructed that comprises an electro-optical device having an enhanced response in the near infra-red spectrum (0.77μm to 1.5μm); this will hereinafter be referred to as an infra-red imager. Such an infra-red imager is required for a number of applications, including Raman spectroscopy and astronomy.
The energy bandgap of silicon (1.12eV) results in a photon absorption cutoff wavelength of l.lμm, and it is thus an inappropriate material for manufacturing an infra-red imager. Germanium, however, has an energy bandgap of 0.66eV, and thus exhibits a cutoff wavelength of 1.88μm. For example, infra-red photons of 1.5μm wavelength are effectively
completely absorbed by a 3μm thick germanium layer, and a 0.7μm germanium layer fully absorbs 1.lμm photons, where a 525μm thick moderately doped silicon wafer is essentially transparent.
Therefore, an imaging device built on a silicon substrate, but containing a germanium layer ranging in thickness between approximately 0.7μm to 3μm, would offer useful absorption properties in the near infra-red region. However, germanium is generally less suitable than silicon as a material for fabricating metal-oxide-semiconductor field effect transistors, whether CCD imagers or readout registers, as it is difficult to produce a reliable gate dielectric on germanium. It is therefore desirable to vertically integrate the infra-red absorption properties of germanium with silicon MOSFET readout circuitry.
Fig. 2 shows a cross-section of a monolithic vertically integrated active pixel infra-red imager. In standard silicon-on-insulator (SOI) technology, a substrate acts only as a mechanical support for the 'active' silicon layer on which circuits are fabricated. However, in the illustrated infra-red imager a region of the substrate forms an integral part of the imager.
A substrate of the imager comprises a lightly doped n or p type buried germanium layer 30 bonded to a silicon 'handle' wafer 32 to provide mechanical support. The handle wafer 32 should be of the same
doping type as the germanium layer 30, but may be doped at a different level. The buried germanium layer 30 may be between approximately 0.7μm and 3μm thick.
It will be appreciated that the germanium layer could be of other thicknesses, depending on the precise wavelength of the radiation to be detected. Furthermore, the range given above is an approximate range, and the device may also be able to function at values outside this range.
Detector photodiodes are formed by a pixel matrix of doped junction regions 34 (i.e. wells which can be formed by diffusion or ion implantation) in the surface of the germanium layer 30, and these p-n junctions are sufficiently reverse biased to ensure that their depletion regions extend to the interface between the germanium layer 30 and the silicon handle layer 32. The structure may be fabricated using either p+-n~ or n+-p" germanium photodiodes, and the doping properties of the buried germanium layer 30 and silicon handle wafer 32 are chosen accordingly.
Infra-red photons 40 having a wavelength greater than 1.lμm impinging on the back surface 42 of this structure are transmitted through the silicon substrate 32 and are absorbed by the buried germanium layer 30. This creates electron-hole pairs which are then separated and swept to the
appropriate p+ and n+ regions 34 by the electric field in the germanium depletion region.
To avoid possible carrier absorption of the infra- red photons, the back surface 42 of the silicon substrate 32 may be selectively etched away, in particular in the pixel regions. For applications which also require detection of shorter wavelength photons which would be absorbed by silicon, all or part of the silicon substrate in the pixel regions may also be selectively removed, provided adequate mechanical support is provided for the device.
For either embodiment, the buried layer itself may act as an etch stop. This makes better use of material and leads to a thinner overall sensor. Alternatively, the handle layer can comprise a further buried layer which can act as an etch stop.
Causing radiation to impinge upon the rear surface of the imager is desirable in order to prevent photon collection by the readout circuitry, and thus achieve a high fill factor. However such back illuminated devices will require thinned substrates to yield acceptable quantum efficiency and spatial resolution. It is therefore desirable to incorporate a thin buried layer of radiation sensitive semiconductor material within the detector substrate, and to vertically integrate this material with a silicon layer.
Amplification and readout circuitry is also required to provide spatially resolved detection of incident photons 40. This has previously either been situated remotely from the sensor and interfaced to it using high density wire bond interconnect technology, or custom readout chips have been bump- bonded to the sensor. However, both of these techniques suffer from yield and reliability problems, and add significant parasitic elements to the circuitry.
The device described here avoids these difficulties because pixel circuitry 18 comprising (for example) an amplifier and readout circuit for each pixel is fabricated in the active silicon layer 38, and is vertically integrated with the buried germanium photodiode sensor matrix 34, thus forming a monolithic active pixel infra-red image detector. Each active pixel contains a MOSFET-based circuit to realise (for example) transfer gate, reset, source follower, buffer amplifier and column select functions. Heavily doped semiconductor contact regions 22 are also provided.
In a further embodiment, a detector with a■ wide bandgap, such as silicon carbide, is used in order to fabricate an electro-optical device having an enhanced response in the ultra-violet spectrum. For example, in applications such as astronomy, a detector responsive to ultra-violet radiation, but which has reduced sensitivity in the visible or infra-red spectrum at wavelengths greater than
approximately 390nπi is required. The most common polytype of silicon carbide (hexagonal 6H) has a bandgap of 3eV, resulting in a photon absorption cutoff wavelength of 410nm. Therefore, an imager can be provided in which the radiation sensitive semiconductor layer comprises a silicon carbide layer. This imager is similar in all other aspects, and can comprise the same features as the abovedescribed infra-red imager.
Silicon carbide based photodiodes have the further advantage that they may be used without substrate cooling, since their wide bandgap yields very low detector dark current.
The structures shown in Figs. 1 and 2 are formed using a novel combination of existing semiconductor wafer direct bonding and ion splitting techniques to form the buried germanium layer and to enable vertical integration of the buried germanium layer with the active silicon layer. Such techniques are for example described in US patent 5,374,564 to Bruel.
The first stage of the device fabrication process is void-free direct bonding of a silicon handle wafer 50 to a germanium wafer 52, as shown in Fig. 3.
When mirror-polished, flat and clean wafers are brought into intimate contact, they are known to bond to each other at room temperature in ambient air without the use of adhesives or significant
external forces. This process is termed "direct wafer bonding", and may occur between suitably prepared hydrophillic or hydrophobic wafer surfaces.
Wafer surfaces may be successfully bonded provided the root mean square (RMS) surface roughness is no greater than 0.5nm, which may be readily achieved using existing chemical-mechanical polish (CMP) technology. Bonding may be initialised in a small area of the wafer by applying slight pressure to expel any trapped air. Thereafter, the bonded region propagates over the entire wafer within a few seconds .
A germanium-on-silicon substrate 54 is then formed by ion splitting the germanium wafer 52. Ion implantation of hydrogen or helium into a semiconductor substrate forms a plane of weakness which can be used to delaminate a thin film (the active layer) and transfer it to another substrate (the handle wafer)
In this device, the germanium wafer 52 contains a hydrogen or helium implant with an energy designed to produce an ion peak 53 at a chosen depth within the range of approximately 0.7μm to 3μm below the germanium bonding surface. The silicon handle wafer 50 acts both as a stiffener and also as a support for the transferred layer. Annealing, typically at 5000C, propagates microcracks parallel to the implanted wafer surface. Provided the low temperature bonding energy of the wafer pair is
sufficiently high, complete wafer splitting and layer transfer occurs. The germanium layer is thus transferred to the silicon handle wafer. The germanium-on-silicon wafer 54 is then annealed in N2 at around 8000C to strengthen the interface bond. The RMS roughness of such an ion split surfaces- is of the order of IOnm, and so touch polishing of the transferred germanium layer is required to complete the germanium-on-silicon substrate 54.
Alternatively, the germanium layer may be formed by epitaxial growth of germanium on a silicon substrate. As a further alternative, defects due to the lattice mismatch between silicon and epitaxial germanium may be avoided using ion splitting to transfer a germanium layer approximately 0.5μm thick onto a silicon handle wafer, followed by epitaxial growth of germanium. As a further alternative, the required germanium layer thickness may be obtained by precision grinding and polishing the germanium wafer after bonding.
The next stage of the fabrication process (shown in Fig. 4) is to bond and ion split an oxidised silicon wafer 56 onto the germanium-on-silicon substrate 54, in a manner similar to that described above, thus forming a buried germanium silicon-on-insulator substrate 62. This substrate 62 comprises an' active silicon layer 58 and an insulating silicon dioxide layer 64 disposed on the germanium-on-silicon substrate 54. As an alternative, the required active silicon layer thickness may be obtained by
precision grinding and polishing the oxidised silicon wafer after bonding.
The doping properties of the silicon wafer 56 are chosen to be suitable for the fabrication of MOSFET- based active pixel circuitry. In order to obtain a final active silicon layer 58 thickness of approximately 0.5μm the oxidised silicon wafer 56 includes a hydrogen or helium implant with an ion peak 60 approximately 0.5μm below the silicon dioxide-silicon interface. After elevated temperature bond annealing and touch polishing, the buried germanium silicon-on-insulator substrate 62 is complete.
Fabrication of an imaging device can then proceed by etching photolithographically patterned windows in the active silicon layer 58 and buried silicon dioxide layer 64 to define the germanium photodiode pixel regions 34 (see Fig. 2), which are then doped by ion implantation or diffusion. Alternatively, these doped junction regions 34 may be formed in the germanium layer prior to the stage of bonding and ion splitting the oxidised silicon wafer 56 onto the germanium-on-silicon substrate 54.
Active pixel silicon MOSFET fabrication is desirably carried out at a temperatures no greater than 8000C in order to preserve the integrity of the buried germanium layer, and can be achieved using technology routinely employed for the manufacture of thin film transistor (TFT) devices on glass
substrates. As a possible alternative, the silicon active layer may be replaced by a gallium arsenide active layer. The use of such a material enables the active pixel circuitry to be implemented using metal-semiconductor field effect transistor (MESFET) integrated circuit technology which does not require high temperature processing.
Therefore, it can be seen that an electro-optical device can be constructed in which the photon absorption and electronic readout device properties of different semiconducting materials can be successfully combined, which gives flexibility in the design of an electro-optical device. Furthermore, the described electro-optical device can be constructed using standard silicon process technology.
Improvements and modifications can be made to the above without departing from the scope of protection. In particular, the buried radiation sensitive semiconductor layer may be made from any suitable material, and any suitable form of pixel circuitry can be formed on the active layer. Furthermore, the active layer does not have to be silicon; other materials such as gallium arsenide can be used.