WO2008080428A1

WO2008080428A1 - Waveguide photodetector in germanium on silicon

Info

Publication number: WO2008080428A1
Application number: PCT/EP2006/012603
Authority: WO
Inventors: Lorenzo Colace; Gaetano Assanto; Luciano Socci; Marco Romagnoli; Lorenzo Bolla
Original assignee: Pgt Photonics S.P.A.
Priority date: 2006-12-29
Filing date: 2006-12-29
Publication date: 2008-07-10

Abstract

It is described a photodetector structure (1,1′,1'), comprising a silicon-based waveguide (2) in which optical signals to be detected travel in a given direction (X) and are confined therein; a germanium-based layer (4) disposed on a portion of the silicon-based waveguide (2), the germanium layer (4) including a first mesa (10) defining an active region having a length (L) along the signal propagating direction (X) and a width (W) in a direction (Z) substantially perpendicular to the propagating direction (X) so that an evanescent tail of the propagating optical signal in the waveguide (2) is coupled into the active region, and a second mesa (30; 40) separated by a distance d from the first mesa (10) in said direction (Z) substantially perpendicular to the signal propagating direction. The photodetector further includes a first (7) and a second metallic contact (9a,9b) having opposite polarities, the first metallic contact (7) being located on the active region of the first mesa (10) and the second metallic contact (9a, 9b) being located on said second mesa (30), said first and second contact being used to collect electrons generated by light absorption to obtain an output electric signal.

Description

Waveguide Photodetector in Germanium on Silicon

Technical field

The present invention relates to a photodetector realized in germanium on silicon, particularly a silicon waveguide integrated germanium photodetector. The selected geometries of the photodetector combine a good responsivity and excellent speed of the overall device with a relatively simple fabrication process. Technological background The introduction of fiber-based optical communication has brought a great increase of long haul telecommunications: the inherent low cost, wide bandwidth and small attenuation of fibers are key factors in prevailing over copper wire. In short haul access network, however, fiber qualities are superseded by the current high costs of optical transceivers. These components are usually hybrid assembly of III - V devices such as lasers, modulators, photodiodes and waveguide.

In order to complete the deployment of widespread fiber communications, novel approaches for transceiver fabrication are required. Among others, silicon-based opto-electronics is attractive because of its potential low cost, scalability and reliability and integration with the mature and unsurpassed silicon VLSI technology.

In addition, light detection in the near-infrared (NIR) region is of extreme importance in optical telecommunications, particularly when high bit rates and low power levels are involved. It becomes therefore crucial to employ NIR detectors that not only exhibit good sensitivity and speed in the spectral range on interest, but that can be closely interconnected to driving/biasing and amplifying electronic circuits. Since the most common platform for electronic processing of signals is based on silicon, the integration of NIR photodiodes on standard silicon platform has been pursued in the past two-decades as a viable low-cost and high-efficiency solution to the growing request for compact semiconductors microsystems for optical signal processing.

One of the most appealing attempts of designing NIR photodetectors that can be integrated with standard semiconductor technology is based on polycrystalline Ge mainly because of the relatively low thermal budget required in the device fabrication. Polycrystalline films are deposited at low temperatures which guarantee a good compatibility with standard CMOS processing. The deposited films exhibit absorption spectra similar to those of monocrystalline Ge, but mobility and lifetimes are reduced. In "Fabrication and characterization of low temperature (<450°C) grown p- Ge/ n-Si photodetectors for silicon based photonics" written by P. R. Bandaru et al., published in Material and Science Engineering B, 113 (2004) pages 79-84, p-n hetero-junctions were fabricated by depositing p-Ge thin films on n-Si substrates using molecular beam epitaxy and electron - beam evaporation, with processing temperature less than 45O⁰C, to be compatible with back end silicon processing. The surface preparation of the Si substrate prior to Ge deposition was found to significantly affect the crystallinity of the deposited Ge layers and, hence, the p-n photodetector diode characteristics. In US patent n. 6897498 in the name of SiOptical, Inc., a photodetector for use with relatively thin (i.e. submicron) silicon optical waveguides formed in a silicon-on-insulator structure is disclosed. The photodetector comprises a layer of poly-germanium disposed to couple at least a portion of the optical signal propagating along the silicon optical waveguide. The poly-germanium detector is formed to either cover a portion of the waveguide or be butt- coupled to an end portion of the waveguide. When covering a portion of the waveguide, poly-germanium detector may comprise a "wrap around" geometry to cover the side and top surfaces of the optical waveguide, with electrical contacts formed at opposing ends of the detector. PCT application WO2006/066611 in the name of the Applicants describes a photodetector structure comprising a silicon-based waveguide in which optical signals to be detected travel in a given direction and are confined therein and a germanium layer is disposed in contact with a portion of the silicon-based waveguide, so that an evanescent tail of the propagating optical signal is coupled into the germanium layer. In addition, the germanium layer includes a mesa having a length along the signal propagating direction and a width in a direction substantially perpendicular to the propagating direction, in which the width of said mesa is smaller than its length. The photodetector also comprises a first and a second metal contact, the first metallic contact being located on the germanium layer the said second metallic contact being located on the silicon-based waveguide, the first and second contact being used to collect electrons generated by light absorption to obtain an output electric signal. Summary of the invention The invention is relative to a waveguide photodetector structure comprising a germanium-based layer, preferably a polycrystalline germanium layer (in the following shortened in "poly-Ge" layer), formed on a silicon-based waveguide.

The photodetector of the invention comprises a hetero-junction between a layer germanium based layer and a silicon-based layer. Preferably, the hetero-junction is a p-n junction between a p-type poly-Ge layer and an n- type silicon layer.

The region in which the hetero-junction between the silicon-based layer and the germanium-based layer is present is called the active region of the photodetector. The photodetector described above is integrated with a waveguide, i.e. the signals to be detected by the device travel in a waveguide, which may have any geometry, and are vertically confined therein.

The dimensions of the waveguide are such that an optical mode propagating in the waveguide has an evanescent tail which extends beyond the waveguide layer and thus the mode itself is sensitive to the presence of additional layer(s) possibly located on a surface of the waveguide. Typical waveguide dimensions are for example in the range of 150-300 nm of core thickness. In the preferred embodiments, the silicon-based waveguide forms with the underlying layers a silicon-on-insulator (SOI) structure.

The above-mentioned poly-Ge layer is deposited directly over the waveguide, which is a silicon-based waveguide in order to form the aforementioned p-n junction: as the signal propagates along the waveguide, it is coupled and then absorbed into the poly-Ge layer, thereby creating electron-hole pairs. Optical absorption takes place along the light propagation path in the overlap region between the photodetector active region and the guided mode profile of the mode travelling in the waveguide. Preferably, the germanium layer is intrinsically p-doped, due to its relatively high defect density produced during its deposition. The silicon waveguide preferably is made of intrinsic (i.e., substantially undoped) silicon or of silicon with a relatively low doping level, i.e., lower than 10¹⁶/cm³ in order to minimize the free carrier absorption and consequently the waveguide loss. On the other hand, in the active region of the photodetector it is desirable to have a higher n-type or p-type doping level (i.e. higher than 10¹⁷/cm³) in order to have good electrical properties of the p-n junction and to minimize the series resistance. According to an embodiment of the invention, the silicon waveguide is selectively doped only in correspondence of the active region and left substantially undoped or at low doping level elsewhere. Applicants have noted that the photodetector described in US patent n. 6897498 has a poly-Ge layer in which a p-i-n structure is realized, including a p-doped poly-germanium layer, an intrinsic (i.e., undoped) layer and a n- doped layer. Applicants have observed that a p-i-n structure in poly-Ge is difficult to realize, especially with deposition technologies that employ relatively low temperatures. Additionally, doping of selected areas of the poly-Ge layer is necessary in order to form a lateral p-i-n junction. Applicants have noted that this doping requires high temperatures treatments. However, relatively low deposition temperatures, i.e., not larger than 350-400⁰C, are desired in order to preserve compatibility with standard silicon CMOS technology. The p-n junction of the photodetector of the invention is realized forming a rib structure, i.e., the poly-Ge layer comprises a mesa structure, called in the following simply "first mesa". In the first mesa an active region is defined, having a given length L, a width W and a thickness T. It is to be noted that within this context with mesa structure or mesa it is not necessarily meant a planar top surface of poly-Ge (although preferred). For example, a rib-shaped or ridge-shaped poly-Ge layer could be envisaged. This mesa is in contact with the waveguide surface, preferably its top surface, and the active region of the mesa coincides with the active region of the photodetector, wherein the active region is the region in which light is absorbed. The active region is located so that the length L is the length along the signal propagating direction in the waveguide, while W is the width of the active region in the direction substantially perpendicular to L and to the mesa thickness. Light in the absorption region is confined both vertically and laterally.

According to the invention, in addition to the first mesa above described, a second germanium-based mesa, preferably in poly-Ge, is realized on the silicon-based layer. Preferably, the second mesa is located at a given lateral distance (i.e., parallel to W) from the first mesa and more preferably extends substantially parallel to the length L of the first mesa.

According to a preferred embodiment of the invention, in addition to the first mesa, two other mesas, a second and a third mesa, are preferably disposed symmetrically with respect to the central first mesa along the signal propagating direction in the waveguide. The second mesa and the third mesa, if present, do not form an active region because they are only marginally affected by the propagating optical mode in the waveguide, said mode being substantially confined within the active region. Both second and third mesas are deposited on top of the silicon layer but at a given distance from the mode location.

The electron-hole pairs generated in the active region of the first mesa due to the absorption of light can be efficiently collected using suitable metal contact structures. According to the invention, the photodetector comprises a first and a second metal contact, the first metal contact (anode) is located above the poly-Ge first mesa defining the active region and the second metal contact (cathode) is located above the poly-Ge second mesa. Preferably, the first metal contact has a length and a width equal to the length L and width W, respectively, of the active area of the photodetector, i.e. it covers the active region completely. Even more preferably, the first metal contact covers the whole first mesa.

In addition, the second metallic contact preferably covers completely the second (and third, if present) poly-Ge mesa.

Due to the location of these contacts, i.e., two poly-Ge mesas on which two metallic contacts are formed, laterally separated by a gap where only the silicon layer is present, the photogenerated carriers move towards the contacts within the silicon layer which is integral part of the p-n junction. According to a second embodiment of the present invention, which differs from the first preferred embodiment only in the shape of the first metallic contact, this latter comprises two metallic strips located on top of the active region of the first poly-Ge mesa. Preferably, the two strips are located on the first mesa symmetrically with respect to its longitudinal axis. More preferably, the two metallic strips are arranged along the two lateral edges of the top surface of the first mesa.

The width of the two metallic strips is preferably minimized in order to minimize the metallic losses.

In a third embodiment of the invention, the first poly-Ge mesa, in the region of the defined active area, is not covered by the first metallic contact: this latter is deposited over a portion of poly-Ge layer which extends from the active area along the longitudinal axis, i.e., substantially along the propagation direction. All the photogenerated charges are thus collected at the end of the active region.

The two latter embodiments, even if they show reduced losses due to the geometry of the metallic contacts which cover a reduced area with respect to the first embodiment, involve fabrication steps for the photodetector realization additional to those necessary for the fabrication of the photodetector according to the first embodiment, as it will explained more in detail in the following.

The fact that in the first embodiment of the photodetector of the invention the poly-Ge mesas and the metallic contacts cover the same areas, allows a single lithography process to be performed to create a structure comprising at least two mesas, which results to be a self-aligned structure. According to an additional embodiment of the invention, between the poly- Ge layer and the metallic layer, an additional poly-silicon ("poly-Si" in short) layer is formed, which acts as a mode constraint, lowering metallic losses due to the direct contact between the poly-Ge and the metal layers. Preferably, the poly-Si layer is patterned during the same lithographic process described above to realize the desired poly-Ge and contact layers configuration.

Applicants have noted that the geometry of the photodetector of the present invention reduces the number of fabrication steps required for its realization with respect of the photodetector described in WO 2006/066611. In addition, in the present invention, being both metallic contacts of the two mesas deposited over the poly-Ge layers, or over the poly-Si layers contacting in turn the poly-Ge layer, they act as good contacts when directly biased, thereby simplifying the polarization of the p-n junction. The first p-n junction corresponding to the poly-Ge/Si contact in the active region is an inverse biased junction, while the p-n junction formed between the second (and third) mesa and the underlying Si waveguide is a forward biased junction, wherein the photogenerated charges are easily collected. These forward biased junctions (corresponding to the second and, if present, third mesa) allow a good bias of the p-n junction defining the active region since ohmic contacts can be simply obtained by direct metal deposition without any thermal treatment. On the other hand, when the metal is directly deposited over the silicon layer, it is often necessary to perform a thermal treatment which generally requires annealing temperatures not compatible with standard CMOS technology. In all preferred embodiments, the thickness T of the poly-Ge layer forming the first mesa (and the thickness of the metal contacts) is chosen in such a way that the overlap of the optical mode travelling in the silicon waveguide with the active region is maximized in order to minimize the fraction of optical power absorbed in the metal contact which does not contribute to the photocurrent.

The preferred poly-Ge thickness T is comprised between 40 nm and 200 nm, more preferably between 90nm and 130nm, while the preferred thickness of the metal contact is comprised between 100 nm and 1000 nm. Referring now to the length of the active area, the length L is preferably kept as small as possible in order to achieve low capacitance and, therefore, preserve bandwidth. On the other hand, the length L is preferably longer than the absorption length so that nearly complete absorption is obtained. The preferred range of the length of the active region defined in the first poly-Ge mesa in the photodetector of the present invention is 10 μm < L < 1000 μm, more preferably L is comprised between 50 μm and 500 μm. Simulations performed by the Applicants have shown that the collection efficiency of the photodetector strongly decays when the width W of the active region in the poly-Ge first mesa exceeds the value of about 5 μm, having a value greater of 90% in the range 1-5 μm. On the other hand, calculations show that it is quite important to keep the optical mode in the waveguide at the photodetector input (i.e. at the beginning of the active area) as wide as possible in order to minimize the power loss due to mask misalignments during the fabrication process.

In detail, in the lithography process of the first poly-Ge mesa on top of the silicon waveguide, misalignments can be formed both in the longitudinal direction (i.e., along the signal propagation direction) and in the lateral direction (i.e., orthogonal to the longitudinal direction). Applicants have shown, that misalignments in the lateral direction are more critical than the ones in the longitudinal direction. Therefore, preferably the active region's width, W is comprised between 2 μm and 15 μm, more preferably between 5 and 10 μm. The latter values represent a good compromise between the maximization of the collection efficiency and the minimization of the sensitivity due to mask misalignments.

Generally, typical waveguide widths are smaller than the preferred values of active region widths in the photodetector fabrication. Therefore, preferably in the photodetector of the invention, a waveguide portion is tapered up to the required width and at the end of the taper the active region is formed. Brief description of the drawings

Further features and advantages of a waveguide photodetector in germanium on silicon according to the present invention will become more clear from the following detailed description thereof, given with reference to the accompanying drawings, where: - fig. 1 is a schematic perspective view of a first embodiment of the photodetector of the present invention;

- fig. 2 is a top view of the photodetector of fig. 1;

- fig. 2a is a top view of a second embodiment of the photodetector of the invention; - fig. 3 is a top view of a third embodiment of the photodetector of the invention;

- fig. 4 is a top view of a fourth embodiment of the photodetector of the invention;

- fig. 5 is a longitudinal cross section of the photodetector of figs. 1 and 2 along the Y axis of fig. 2 showing the mode profile of the mode traveling in the waveguide and its overlap with the depletion layer;

- fig. 6 is a lateral cross section of the photodetector of figs. 1 and 2 along the Z axis;

- fig. 7 is a graph showing the calculated absorption efficiency (left) and the absorption length (right) versus the polycrystalline germanium layer thickness for a W=IO μm of the active region of the poly-Ge first mesa in the photodetector of fig. 1 on a silicon waveguide 220 nm-thick. Contacts are in Au;

- fig. 8 is a graph showing the calculated collection efficiency versus the width of the poly-Ge first mesa, with L= 100 μm and T=120 nm;

- figs. 9a-9d are schematic perspective views of several steps of the process to realize the photodetector of fig. 1;

- fig. 10 is an enlarged top view of a detail of the top view of the photodetector of fig. 2; - fig. 11 is a graph showing the transmitted power from the Silicon waveguide in the active region of the photodetector of the invention of figs. 1 and 2 in the case of 2 μm vertical misalignments as in fig. 10 (top curve) and the power lost reflected and scattered (bottom curve), in the case of the same vertical misalignments. - fig. 12 is a SEM photograph of an embodiment of the photodetector of the invention;

- figs. 13a-13c are schematic perspective views of several steps of the process to realize the waveguide of the photodetector of fig. 1;

- fig. 14 is a top view of a different embodiment of the photodetector of the present invention; - figs.l5a-15c show schematic diagrams of applications of the photodetector of the invention. Preferred embodiments of the invention

With initial reference to figs. 1 and 2, 1 indicates a photodetector structure realized according to a preferred embodiment of the present invention.

The photodetector structure 1 comprises a silicon-based waveguide 2 in which optical signals travel along a given direction X, called the propagating direction, and are confined therein. In the present context, with the term "silicon-based" waveguide, it is meant a waveguide realized in silicon, preferably n-type silicon material. Preferably, the waveguide dimensions are such that the mode traveling therein is not fully confined vertically (i.e. along the Y axis) in the waveguide itself, but an evanescent tail of the mode extends outside the waveguide, so that the mode may be influenced by the location of additional layers on a waveguide's surface. Additionally, the waveguide geometry is arbitrary, i.e. a slab, rib or ridge waveguide, may be alternatively used depending on the final desired application of the photodetector structure 1. Preferably, the waveguide 2 is realized on a first cladding layer 8, which has a refractive index lower than that of the waveguide. Preferably, layer 8 is a SiO₂ layer. More preferably, the SiO₂ layer 8 is realized on a substrate 3. Substrate 3, cladding layer 8 and waveguide 2 form a silicon-on-insulator (SOI) structure, in which preferably the substrate 3 is made of silicon. Preferably, the thickness t of the waveguide 2 is comprised between 150 nm and 300 nm, more preferably between 200-250 nm. The waveguide 2 is realized starting from an initial silicon layer of the SOI wafer through, for example, two consecutive processes of e-beam or photo lithography and wet and/or dry etching of the silicon waveguide layer laying over the silicon insulating layer 8. The width of the waveguide is preferably comprised between 300 nm and 1 μm, which is the standard width of bus silicon waveguides.

The waveguide 2, as clearly seen from fig. 2, extends along the X direction and it comprises an input portion 2a, and an enlarged portion such as a platform 25 extending from the input portion 2a. More in detail, preferably during a single etching phase of the silicon layer, the waveguide input portion 2a, together with the platform 25, and preferably a waveguide taper portion 26 connecting the input portion 2a to the platform 25 and better defined below (see fig. 13a in which the etched silicon layer is shown forming the waveguide on top of the first cladding layer 8), are realized. The platform 25 comprises the area on which the layers forming the photodetector 1 are formed. The taper 26 can be seen as a continuation of the input portion 2a of waveguide 2, to join the input portion 2a to the platform. Additionally, the waveguide 2 may also extend beyond the platform 25. Preferably, waveguide 2, in particular its input portion 2a, is a single mode waveguide, while the taper portion 26 and the platform 25 are multi-mode waveguide.

Preferably, the input portion 2a of the waveguide 2 is made of substantially undoped silicon or silicon with low concentration of dopants (preferably n- type dopants), i.e. the doping level is lower than 10¹⁶/cm³. Preferably, the portion of the waveguide 2 in which the photodetector 1 is realized, i.e. in correspondence of the platform 25, is locally n-doped with a doping level higher than 10¹⁷/cm³, by using a suitable ion implantation or ion diffusion process step.

On top of the so-etched Silicon layer 2, a second cladding layer 23 is preferably deposited (step depicted in fig. 13b in which the platform 25 and the waveguide 2 are still visible under the cladding layer 23). Preferably, the thickness of the second cladding layer 23 is comprised between 100 nm- 5 μm, more preferably is of the order of 1 μm, and it is for example realized in tetraethylorthosilicate (TEOS). The second cladding layer 23 is thus removed form the photodetector area, using suitable techniques, such as wet and/or dry etching, creating substantially a "box" 25' into the second cladding layer 23 having as a bottom surface the underlying silicon waveguide 2 (see fig. 13c in which the configuration is shown after the box 25' creation). The input portion 2a, the taper 26 and the first cladding layer 8 - with the exception of the portion covered by the platform 25 -, even if still visible in figure 13c for clarity purposes, are covered by the second cladding layer 23. The box 25' in the second cladding layer 23 may correspond to the platform area 25, as shown in fig. 13c, i.e., the bottom surface of the box 25' corresponds to the top surface of the platform 25, or may be angled with respect to it, as depicted in fig. 14. The reason of this different box geometry will become clearer in the following.

In fig. 9a, a perspective view corresponding to the top view of fig. 13c is shown in a simplified manner: for the sake of convenience, only the region corresponding to the photodetector 1 (i.e., the region corresponding to the platform 25) is represented with only the SiO₂ layer 8 and the waveguide layer 2 depicted (i.e., no silicon substrate 3 is shown in the figures). It is therefore to be understood that the area of the waveguide layer 2 shown in figs. 9a-9d is the bottom surface of box 25' in the cladding layer 23 (not visible in figs. 9a-9d).

A polycrystalline Ge layer 4 is formed grown on top of the silicon-based waveguide 2 in correspondence of the platform area 25: this step is schematically depicted in fig. 9b. Suitable techniques for the poly-Ge layer formation might be infusion doping, thermal evaporation, sputtering and chemical vapour deposition, being the first three preferred because of their inherent compatibility with CMOS technology. Preferably, the poly-Ge is deposited by evaporation, sputtering or infusion doping and the temperature of the substrate is kept above 25O⁰C (for example around 300- 35O⁰C) so that the deposited germanium layer has a polycrystalline structure.

Preferably, before the poly-Ge deposition, the top surface of the silicon platform 25 is pre-treated (step not shown), for example through a short etching or a plasma treatment, to remove the native oxide which tends to grow on top of the Si layer. The etching is performed for example using a buffered HF. The presence of the native oxide is undesirable since it could prevent the p-n junction (described below) formed between the silicon layer and the poly-Ge layer from working properly.

Although the preferred embodiments of the present invention, which will be described in details below, refer to polycrystalline germanium, a photodetector device having a germanium layer with a (mono)crystalline structure, i.e. c-Ge, or a crystalline layer of SiGe with germanium concentration not smaller than 30% can be also included within the scope of the invention. Crystalline germanium or SiGe could be grown epitaxially on silicon. After poly-Ge deposition, metal evaporation is performed in order to form a metallic layer 9 on top of the poly-Ge layer 4 (step schematically illustrated in figure 9c). A suitable technique is metal evaporation and a preferred metal is gold due to its low optical losses. Preferred values of the metal layer thickness are comprised between 100 nm and 1000 nm.

According to an additional embodiment of the invention, not shown, between the metallic layer 9 and the poly-Ge layer 4, an additional poly- silicon layer is deposited, for the reasons illustrated below. The poly-Ge layer and metallic layer form a stack into the box 25' realized in the cladding layer 23, not shown in figs. 9a-9d. This stack is then etched (see fig. 9d), in order to obtain the desired geometry of the poly-Ge and metallic layers, as described in the following. The geometry realized in fig. 9d substantially corresponds to the embodiment of figs. 1 and 2. The metallic layer 9 and the poly-Ge layer 4 are etched so as to form a first mesa 10 defining an active region having length L in the direction of signal propagation X in the waveguide and a width W in a direction Z substantially perpendicular to the propagation direction, and a second mesa 30, laterally separated with respect to the first one by a distance equal to d. With the term "laterally" it is intended in the present context to define the direction substantially perpendicular to the direction of signal propagation in the waveguide 2, i.e. the direction along the Z axis. Preferably, the second mesa 30 has a width larger than 5 μm. According to a preferred embodiment of the invention depicted in figs. 1 and 2, the first mesa 10 is symmetrically flanked by two additional mesas, the second 30 and a third 40 mesa, extending parallel to the length L of the active region of the first mesa 10. Preferably the distance d between the first and the second mesa is comprised between 1 μm < d < 20 μm, more preferably between 3 μm < d < 10 μm. In addition, preferably, the distance d' between the first and the third mesas is comprised between 1 μm < d' < 20 μm, more preferably between 3 μm < d' < 10 μm.

The active region formed in the first mesa 10 is located in front of the input waveguide portion 2a, i.e. it is substantially the geometrical continuation along the X propagating direction of the input waveguide portion 2a at a different height, i.e. the first mesa lies on a plane parallel to the plane in which the waveguide 2 is defined. Preferably, the axis of symmetry of the input waveguide portion (2a) and the longitudinal axis of the first mesa (10) coincide.

From the above described geometry, it is identified a "detector region", or "active region" of the photodetector 1 given by L x W in which light absorption takes place in the first mesa 10. The active region starts at the interface between the input waveguide portion (2a) with the platform (25) and it extends for a length L

Additionally, the thickness of the first poly-Ge mesa 10 is identified in the following with T, which also preferably corresponds to the thickness of the second 30 (and third 40) mesa, being all deposited and etched from the same poly-Ge layer 4 in the same process steps.

The combination of the silicon-based waveguide 2 in contact with the poly- Ge layer 4 of the first mesa 10 forms a p-n hetero-junction. Preferably, the poly-Ge layer is p-type and the Si-waveguide is n-type: a p- type germanium is preferred when thermal evaporation is employed for fabrication of the layer 4, because the grown Ge layer turns out to be p- type without the need of an additional step of doping (e.g., by diffusion or ion implantation). Preferably, the p-doping level of the poly-Ge layer is of the order of 10¹⁷- 10¹⁸/cm³.

The absorption of the propagating light signal in the waveguide takes place along the light propagating path where the active region realized in first mesa 10 is present. The situation is schematically depicted in the cross- section of fig. 5 in which the propagating mode of the waveguide 2 is shown. A depletion layer 6 is formed at the poly-Ge/Si interface: when an incident photon reaches this region, due to the fact that the evanescent tail of the propagating mode in the waveguide 2 couples into the poly-Ge layer 4, it creates electron-hole pairs. The electrons generated by the light absorption are collected through the metallic electrodes, or contacts, realized by etching the metallic layer 9 deposited over the poly-Ge layer 4.

In a first preferred embodiment of the invention, shown in figs. 1 and 2, a first metal contact 7 is placed on a top surface of the first poly-Ge mesa 10 corresponding to the active region. Preferably, the first contact 7 (anode) comprises a metal strip 15 having a length equal to the length L of the active region in the first poly-Ge mesa 10. In particular, according to this first preferred embodiment, the first metal contact 7 covers the active region completely, following the profile of its top surface and thus defining a metallic strip 15 of width W and length L. A second metal contact (cathode) is placed on the top surface of the second mesa 30. The first and second contacts collect the electrons generated by the photon absorption in the active region L x W. Preferably the second contact has the form of a strip. More preferably, the second contact comprises two metallic strips 9a, 9b located on the second 30 and third mesa 40 symmetrically with respect to the first poly-Ge mesa 10, i.e. they "embrace" the first poly-Ge mesa 10, in particular its active region, and lay parallel to the latter. The two symmetric metallic strips 9a, 9b are also connected to each other (see for example the top view of fig. 2). The connection 9c is preferably located within the box 25', as shown in fig. 2, or it may extend outside the same.

A p-n junction is formed also in correspondence of the second 30 and third mesa 40. These junctions, affected only marginally by the propagating mode in the waveguide 2, are good contacts that allow an easy polarization of the "main" p-n junction in the active area of the photodetector. The junction between the poly-Ge and the silicon waveguide in correspondence of the first mesa is an inverse biased junction, while the junction(s) between the poly-Ge and the silicon waveguide in correspondence of the second (and third) mesa is(are) a forward biased junction(s). Additionally, generally the contact resistance of a metal - poly-Ge - Si stack is reduced with respect to a Si-metal contact. According to a preferred embodiment of the invention, the first metal contact 7 comprises a metal pad 11 extending from the metallic strip 15 along the X axis (see for example figs. 1 and 2) to more easily connect the first metallic contact 7 to an external circuit (not shown) through wire bonding. Even more preferably, the poly-Ge layer 4 is etched in such a way that a poly-Ge contact region 12 is formed, from which the active region projects, on top of which the metal pad 11 is deposited. Preferably, the poly-Ge contact region 12 (and thus the metal pad 11) has a square shape and for example its dimensions are at least 60 X 60 μm² in order to use standard probes in the art.

However, the metal pad (and the underlying poly-Ge contact region) may have a different geometry, such as a rectangle joined to a taper portion, as in the embodiments illustrated in figs. 2a, 3 and 4. The distance d between the first metallic contact 7 over the first poly-Ge mesa 10 and the second metallic contact 9a or 9b over the second and third mesa 30, 40 is preferably minimized in order to minimize the series resistance of the photodiode 1, thus increasing its optical bandwidth. However, this distance d should preferably not be too small in order to minimize the possibility of the formation of micro short circuits. Thus, the preferred distance between the first and the second metal contact is 1 μm < d < 20 μm, more preferably 3 μm < d < 10 μm.

According to a preferred embodiment of the invention, best seen in figs. 1, 2 and 9c, the first and second metallic contacts 7, 9a, 9b cover completely the mesas 10, 30, 40 realized in the poly-Ge layer. Therefore, a single lithography process, for example e-beam lithography as well as optical lithography, followed by two etching steps is necessary for the realization of this geometry, simplifying the fabrication process.

According to an additional preferred embodiment of the invention, the top view of which is shown in fig. 2a, the poly-Ge second 30 and third mesas 40 flank the first mesa 10 and then embrace the same, connecting after the pad 11 along the X axis. In addition, with the exception of the gap realized between the first mesa and the second and third mesas (which is a single continuous "channel" 18 being the second and third mesas connected one to the other), the mesas 10, 30, 40 cover completely the platform 25. In this embodiment, the only uncovered portion of the silicon waveguide is therefore the "channel" 18 formed between the poly-Ge mesas. In fig. 13 a SEM photograph is attached portraying the photodetector 1 described above corresponding to the embodiment of fig. 2a. Additional different contacts geometries are envisaged in the photodetector of the invention, two embodiments l',l" of which are shown in figs. 3 and 4.

In these two geometries, the amount of metal covering the top surface of the active region of the first mesa 10 is decreased with respect to the embodiment of fig 2, in order to increase the performances of the device, lowering metal losses. In fig. 3, photodetector 1' comprises a first contact 7 which has the shape of two parallel strips 32a, 32b disposed on the top surface of the first poly-Ge mesa 10, in correspondence of its active region, parallel to the signal propagating direction X and extending along the length L of the first mesa 10, and it may be located in proximity of a border of the mesa upper surface, i.e. it may be in contact to the border of the rectangle W x L or it can be located at a given distance from that border.

Preferably, the width of each strip 32a, 32b is comprised between 0.5 μm and 5 μm.

In fig. 4, photodetector 1" comprises a first contact 7 which leaves the active region of the first mesa 10 uncovered and starts at the end of the same, extending in a tapered shape joining the pad 11.

These geometries of figs. 3 and 4 require an additional lithography step with respect to the geometry of fig. 2 (or 2a).

As already said, according to an additional embodiment of the invention, not shown, between the poly-Ge layer 4 and the metal layer 9, an extra layer, in particular a poly-silicon (poly-Si) layer, is formed. This poly-Si layer can be introduced in all the above described photodetector 1, V₁ 1" embodiments of figs. 2, 3, 4.

Preferably, the poly-silicon layer is p doped and even more preferably the doping is in the 10¹⁷-10¹⁹ cm^"3 range; its thickness being preferably comprised between 50 nm and 500 nm.

The poly-Si layer better confines the propagating mode traveling in the waveguide 2 within the poly-Ge layer 4.The modes propagating in the waveguide 2, while in the input portion 2a of the same, remain confined vertically and horizontally mainly within the silicon waveguide layer, while in the active region, due to the refractive index difference between the silicon layer and the poly-Ge layer, the modes are guided also in the poly-Ge active region. This "mode shift" may cause metallic losses, because, especially for small poly-Ge thicknesses, the modes may contact the metal layer 9.

A poly-Si layer disposed on top of the germanium layer, better confines the mode in the latter, thus reducing the metallic losses.

For the fabrication of this latter embodiment, during the single lithographic step required to define the geometry of the poly-Ge layer and the metal contact layer (i.e., the mesas), also the geometry of the poly-silicon layer is realized.

In particular, the poly-Si layer is etched in such a way that it covers the poly-Ge mesas completely and is covered completely by the metallic layer. Applicants have noted that the thickness of the poly-Ge layer 4 is an important parameter when it comes to the photodetector responsivity. The maximum thickness of the depletion layer 6 in the poly-Ge layer is of about

10-50 nm. Above this thickness, the absorbed light is lost (i.e. it does not contributed to detection). Indeed, electron-pairs that are generated far away from the depletion region travel primarily under the effect of diffusion and may recombine without giving rise to a current in the external circuit

(connected to the photodetector through the contacts 7, 9a, 9b). This reduces the efficiency of the photodetector 1.

Conversely, a too small value of the poly-Ge thickness results in increased losses in the top metal contact 7. Since losses are due to absorption in metal and in the poly-Ge layer above one-diffusion length, a trade-off is expected.

Simulations have been performed by Applicants on the device according to the embodiment of the invention illustrated in figs. 1 and 2, where the active region of the poly-Ge first mesa 10 has W=IO μm and an L longer than the absorption length. Results of simulations are shown in fig. 7. The right ordinate represents the absorption length, while the left ordinate the absorption efficiency. For each selected value of poly-Ge first mesa thickness T (which is the abscissa of the graph of fig. 7, the thickness is given in nm), the corresponding maximum absorption efficiency of the photodetector 1 is calculated and visualized as a square in the graph of fig. 7. For the same T-value, the absorption length (at 10 dB) is calculated (the calculated absorption length can be considered as the minimum length necessary to achieve nearly complete absorption) and it is marked as a dot in the aforementioned graph. It can be clearly seen that the smaller the poly-Ge mesa thickness T, the longer is the absorption length (in other words L is inversely proportional to T).

A too long absorption length is not desirable because the device becomes bulky and thus less suitable for small devices realization, as typically required in integrated circuit technology. L is also preferably kept small in order to have low capacitance and, therefore, preserve bandwidth. The proposed preferred length range 10 μm < L < 1000 μm, more preferably 50 μm < L < 500 μm, has been selected in order to be able to span up to total absorption (-10 dB) and up to a bandwidth of 10 GHz. The expected trade-off caused by the above mentioned losses associated to the presence of the metallic anode on top of the poly-Ge layer 4 can be seen in the efficiency vs T curve of fig. 7, which has a maximum at about 110 nm. More generally, after having selected a preferred value or range of values of efficiency, which depends on the specific application of the photodetector, a range of suitable thickness for the poly-Ge mesa(s) is available. According to simulations, preferably the selected poly-Ge thickness T is comprised between 40 and 200 nm and more preferably between 90 nm and 130 nm. Having selected L and T, the width W of the active region of the first mesa 10 is to be determined. The width of the active region is varied in order to maximize the collection efficiency, which determines the responsivity of the device.

As shown in fig. 8, the collection efficiency strongly decays when the active region width W exceeds the value of 5 μm, being of 90% in the range 1-5 μm. More particularly, the two curves of fig. 8 represent calculations of the collection efficiency performed by the Applicants for different width W of the active region in the first mesa 10, the other active region dimensions being T=120 nm and L= 100 μm, and with a reverse bias applied to the hetero- junction of 10 V (upper curve) or 0 V (bottom curve). The geometry of the contacts 7, 9a, 9b and of the mesas 10,30,40 of the photodetector 1, the calculations of fig. 8 refer to, is as shown in figs. 1 and 2, and the waveguide 2 has a thickness t of 220 nm.

On the other hand, it is important to keep the optical mode traveling in the waveguide 2 at the photodetector input, i.e. at the interface between the input portion 2a of the waveguide 2 and the active region in the direction X of mode-propagation, as wide as possible in order to minimize the power losses due to mask misalignments which are generally present. A relatively wide first mesa renders the misalignments less relevant. Reference is now made to fig. 10 in which an enlarged view of the portion of the photodetector 1 (illustrated in figs. 1 and 2) in correspondence of the interface between the input portion 2a of the waveguide 2 and the active region of the first mesa 10 is shown.

Calculations performed by the Applicants have shown that misalignments between the end of the input portion 2a, or between the taper 26 -if present -, and the beginning of the active region (i.e. of the first mesa 10) along the signal propagating direction (X), indicated with Δx in fig. 10, are less critical than those in the Z direction, indicated with Δz. Typically, misalignments in both directions are of the order of, or less than, 2 μm. It has been found by Applicants that a suitable compromise between the two opposite needs of maximizing the collection efficiency and minimizing the sensitivity to masks misalignments used in the lithographic processes for the photodetector realization is obtained with an active region width W between 5 and 10 μm, even more preferably of approximately 5 μm. However, also values comprised between 2 and 15 μm can be used.

Considering a misalignment Δz in the Z direction equal to 2 μm, in fig. 18 calculations are reported for a photodetector 1 having T= 120 nm, L= 100 μm and W=5μm on a Si waveguide of thickness = 220 nm, of the power transmitted from the waveguide 2 in the active region of the photodetector (top curve) and of the power lost, reflected and scattered (bottom curve), for different λ of the propagating signal in the waveguide 2. Being the preferred width W of the active region equal to about 5 μm and, on the other hand, being the width of standard silicon waveguides generally comprised between 300 nm and 1 μm, the taper 26 is preferably included in the waveguide 2. The taper 26 between the width of the input portion 2a of the waveguide 2, having a typical waveguide width preferably comprised between 300 and 600 nm, and the width of the active region of the first poly-Ge mesa 10 realized on the platform 25, is formed at the interface between the input portion 2a and the first mesa 10 of the photodetector 1. The taper 26 is designed to maximize the optical power transfer from the waveguide input portion 2a into the photodetector active region. The taper 26 therefore increases the width of the input waveguide portion 2a up to the active region width W, which is located in front of the taper itself. Preferably, in case of waveguides 2 wider than 10 μm, instead of the taper 26, the input portion 2a of the waveguide is directly butt-coupled into the photodetector active region because in this case the sensitivity to misalignment is lower. In addition to the taper 26, the box 25' in the cladding layer 23 itself may be angled with respect to the platform 25. In detail, as shown in fig. 14, one of the corner of the box 25', which needs not to be square shaped (as in the embodiment of figure 14), is positioned in correspondence of the taper 26. In this way the active region of the photodetector, i.e. the first mesa 10, is itself tapered when the box 25' is filled with the poly-Ge layer 4: this taper formed in the poly-Ge first mesa using the geometry of the box 25' makes the entire structure less sensitive to misalignments that may occur within process tolerance.

Applicants have noted that poly-Ge has a good corner-filling properties, so that "automatically" a poly-Ge taper is formed during the poly-Ge deposition at the box corner in front of the taper 26 (see fig. 14). This "angled" box 25' is preferred in case of very small waveguides, i.e. of about 1-2 μm at the end of the taper 26.

The value of the dark current of the photodiode 1 has important implications on the overall power dissipation and noise performance through the associated shot noise. Typical measured dark current/ reverse bias

(applied at the heterojunction) characteristics of the photodetector 1 having an active region in the first mesa of L= 100 μm, W= 5 μm and a first mesa thickness of T= 120 nm, is at 1 V reverse bias, 0.05 μA and responsivity of

0.22 A/Watt. Applicants have noted that a relatively large responsivity is present even at very low or even at zero applied reverse bias voltage.

Example 1

The photodetector 1, some performances of which are listed above, has been realized via the following processes. An optical photolithography has been performed to pattern the three poly-

Ge mesas and the metallic contacts.

A two step etching process follows: to etch the Au layer, KI (Potassium

Iodide) and I₂ (Iodine) in aqueous solution has been used, while to etch the poly-Ge layer a standard Microprosit chrome etch solvent 18 from Shipley Corporation has been employed.

The photodetectors 1,1', 1" of the invention may be used as discrete components of optical fiber communication receivers. They can be also used as part of various higher-level systems. The compatibility with standard VLSI silicon technology allows the fabrication of silicon based optoelectronic integrated circuits.

Figs.l5a-15c show different configurations which can utilize the photodetector of the invention. Fig. 15a depicts anoptical add-drop multiplexer (OADM) 100, realized on a SOI substrate. The OADM 100 comprises a first and a second waveguide, the first of which defines an input port IP and a through port TP, while the second defines a drop port DP. Additionally, the OADM 100 (fig. 15a) comprises a microring based filter 50, which drops signals at a given wavelength coming from the input port IP to drop port DP. One or more photodetector(s) 1 according to the invention are realized on the same substrate optically coupled with either the first or the second waveguide, which may be considered analogous to the waveguide 2 described above. Conveniently, the platform 25 on which the layers of the photodetector 1 are deposited is formed during the same photolithographic and etching processes used to define the first and second waveguides of the OADM. The photodetector(s) 1 is(are) used to monitor the power of the signal travelling in the first waveguide, preferably before and after the micro-ring filter 50, as shown in fig. 15a. An additional photodetector 1 can also be used to monitor the power of the dropped signal at the second waveguide. Preferably, a variable optical attenuator (VOA) 60 intercepting the incoming signals is positioned before each photodetector 1.

An additional photodetector 80 is located at the drop port, called signal photodetector. Fig. 15b shows an OADM 100' realized substantially in an analogous way as the OADM 100. The difference between the OADMs illustrated in figs 15a and 15b lies in the position of the signal photodetector 80: in the OADM 100' is located within the SOI substrate, while in the OADM 100 is located externally. Photodetector 80 may have the same structure of photodetector 1. Fig. 15c shows an OADM 100", analougous to the OADM 100 of fig. 15a, in which a general add-drop filter 70, instead of the micro-ring filter 50, is included, such as a grating-based filter.

In the mentioned figures, the photodetector is indicated with 1 for sake of conciseness, but either a photodetector according to the first embodiment 1 of the invention, or according to the second 1', 1" embodiment can be alternatively selected.

Claims

1. A photodetector structure (1,1',1"), comprising

- a silicon-based waveguide (2) in which optical signals to be detected travel in a given direction (X) and are confined therein;

- a germanium-based layer (4) disposed on a portion of said silicon-based waveguide (2), said germanium layer (4) including a first mesa (10) defining an active region having a length (L) along the signal propagating direction (X) and a width (W) in a direction (Z) substantially perpendicular to the propagating direction (X) so that an evanescent tail of the propagating optical signal in said waveguide (2) is coupled into said active region, and a second mesa (30; 40) separated by a distance d from said first mesa (10) in said direction (Z) substantially perpendicular to the signal propagating direction;

- a first (7) and a second metallic contact (9a, 9b) having opposite polarities, said first metallic contact (7) being located on said active region of said first mesa (10) and said second metallic contact (9a, 9b) being located on said second mesa (30), said first and second contact being used to collect electrons generated by light absorption to obtain an output electric signal.

2. The photodetector structure (1,1',1") of claim 1, wherein said germanium-based layer (4) is a germanium layer.

3. The photodetector structure (1,1',1") according to claim 2, wherein said germanium layer (4) is a polycrystalline germanium layer.

4. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said first metallic contact (7) comprises a first metallic strip (15) located on the active region of said first mesa

(10).

5. The photodetector structure (1,1',1") according to claim 4, wherein said metallic strip (15) of said first contact (7) is disposed parallel to said propagating direction (X) and has a length along said propagating direction substantially equal to the length (L) of said first mesa (10).

6. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said germanium-based layer (4) comprises a third mesa (40) separated by a distance (d') from said first mesa (10) in said direction (Z) substantially perpendicular to the signal propagating direction.

7. The photodetector structure (1,1, ',1") according to claim 6, wherein said second (30) and third mesa (40) are disposed both at a distance d from the first mesa (10). 8. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said second metallic contact (9a, 9b) comprises a second metallic strip (9a; 9b) located on said second mesa (30).

9. The photodetector structure (1,1',1") according to claim 8, wherein said second metallic contact comprises said second metallic strip (9a) located on said second mesa (30) and a third metallic strip (9b) located on said third mesa (40).

10. The photodetector structure (1,1',1") according to claim 8 or 9, wherein said second and third metallic strips (9a, 9b) are disposed symmetrically with respect to said first mesa (10).

11. The photodetector structure (1,1',1") according to any of the preceding claims, wherein the length (L) of said first mesa (10) is comprised between 10 μm and 1000 μm.

12.The photodetector structure (1,1',1") according to any of the preceding claims, further comprising a first cladding layer (8) and a substrate (3) and wherein said silicon-based waveguide (2) is located on top of said first cladding layer (8) which is realized on said substrate (3), said waveguide (2), said first cladding layer (8) and said substrate (3) forming a silicon-on-insulator (SOI) structure. 13.The photodetector structure (1,1',1") according to claim 12, wherein said first cladding layer (8) is a SiO₂ layer. 14.The photodetector structure(l,l',l") according to claim 12 or 13, wherein said substrate (3) comprises silicon.

15.The photodetector structure (1,1',1") according to any of the preceding claims, wherein said silicon-based waveguide (2) is n-type.

16. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said germanium-based layer (4) is p-type. 17. The photodetector structure (1,1',1") according to claim 16, wherein said p-doping of said germanium-based layer (4) is of the order of 10¹⁷-10¹⁸/cm³. lδ.The photodetector structure (1,1',1") according to any of the preceding claims, wherein said first metallic contact (7) is formed above said first mesa (10).

19. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said second metallic contact (9a, 9b) is formed above said second mesa (30). 2O.The photodetector structure (1,1',1") according to claim 6, wherein said second metallic contact (9a,9b) is formed above said third mesa

(40). 21.The photodetector structure (1,1',1") according to any of the preceding claims, wherein a poly-silicon layer is realized between said first mesa (10) and said first metallic contact (7). 22.The photodetector structure (1,1',1") according to claim 21, wherein said poly-silicon layer is realized between said second metallic contact (9a, 9b) and said second (30) and/or third mesa (40).

23. The photodetector structure (1,1',1") according to claim 21 or 22, wherein the thickness of said poly-silicon layer is comprised between

50nm and 500nm.

24. The photodetector structure (1,1',1") according to any of the preceding claims, wherein the thickness (t) of said silicon-based waveguide (2) is comprised between 150 nm and 300 nm. 25. The photodetector structure (1,1',1") according to any of the preceding claims, wherein the width (W) of said active region in said first mesa (10) is comprised between 5 μm < W < 10 μm.

26.The photodetector structure (1,1',1") according to any of the preceding claims, wherein the thickness of said first (7) and/or second metallic contact (9a, 9b) is comprised between lOOnm and

1000 nm. 27.The photodetector structure (1) according to any of claims 4 to 26, wherein said metallic strip (15) of said first contact (7) has a width substantially equal to the width (W) of said active region in said first mesa (10).

28.The photodetector structure (1) according to any of the preceding claims, wherein the thickness (T) of said first mesa (10) is comprised between 90 nm ≤ T ≤ 130 nm . 29.The photodetector structure (l',l") according to any of the preceding claims, wherein said first metallic contact (7) comprises a first and a second metallic strip (32a, 32b) disposed on said first mesa (10). 30. A photodetector structure (l',l") according to claim 29, wherein said first and second metallic strips (32a, 32b) are positioned one parallel to the other. 31. A photodetector structure (1,1',1") according to any of the preceding claims, wherein the distance d between said first (10) and said second mesa (30) is comprised between 1 μm < d < 20 μm.

32.The photodetector structure (1,1',1") according to any of the preceding claims, wherein the distance d between said first (7) and said second metal contact (9a, 9b) is comprised between 1 μm and 20 μm.

33. The photodetector structure (1,1',1") according to any of the preceding claims, wherein said germanium-based layer (4) comprises a contact region (12) from which said active region extends. 34.The photodetector structure (1,1',1") according to claim 33, wherein said first metallic contact (7) comprises a metal pad (11) is located above said contact region (12) realized in said germanium-based layer (4). 35.The photodetector structure (1,1',1") according to any of the preceding claims, wherein said first (7) and said second metallic contact (9a, 9b) comprises gold.

36.The photodetector structure (1,1',1") according to any of the preceding claims, wherein said silicon-based waveguide (2) comprises an input waveguide portion (2a) and an enlarged portion

(25) on which said germanium-based layer (4) is formed. 37. The photodetector structure (1,1',1") according to claim 36, wherein the width of said input waveguide portion (2a) is comprised between

300 nm and 1 μm. 38.A photodetector structure (1,1',1") according to claim 36 or 37, wherein said waveguide (2) comprises a taper region (26) at the interface between said input waveguide portion (2a) and said enlarged portion (25). 39.A photodetector structure (1,1',1") according to claims 36-38, wherein said enlarged portion (25) as a n-doping higher than

10¹⁷/cm³.

40. A photodetector structure (1,1'_/1") according to claims 36-39, wherein said active region of said first mesa (10) extends from substantially the end of said taper region (26) into said enlarged portion (25) substantially parallel to said input waveguide portion (2a).

41. A photodetector structure (1,1',1") according to claims 36-40, wherein said germanium-based layer (4) is formed on said enlarged portion (25).

42.An optical add-drop multiplexer (100, 100',100") formed on a substrate, comprising a first waveguide having an input portion and a through portion and a second waveguide comprising a drop portion, and at least a photodetector structure (1,1',1") realized according to any of claims 1-41 in correspondence of said input portion and /or drop portion and/or said through portion. 43.The optical add - drop multiplexer according to claim 42, wherein said substrate and the first oand the second waveguides forms a SOL. 44. A method to realize a photodetector structure (1,1',1"), comprising the steps of:

- etching a silicon-based layer in order to form a silicon-based waveguide (2) in which optical signals to be detected travel in a given direction (X) and are confined therein; forming on said silicon-based waveguide (2) a germanium- based layer (4);

- forming on said germanium-based layer (4) a metallic layer (9); - performing a first lithographic process of said metallic layer (9);

- etching said metallic layer in order to form a first (7) and a second metallic contact (9a,9b) having opposite polarities, - performing a second lithographic process of said germanium- based layer (4);

- etching said germanium-based layer said germanium layer (49 to form a first mesa (10) defining an active region having a length (L) along the signal propagating direction (X) and a width (W) in a direction (Z) substantially perpendicular to the propagating direction (X) so that an evanescent tail of the propagating optical signal in said waveguide (2) is coupled into said active region, and a second mesa (30; 40) separated by a distance d from said first mesa (10) in said direction (Z) substantially perpendicular to the signal propagating direction, wherein said first (10) and second mesa (3=) are located below said first (7) and second metallic contact (9a, 9b) respectively.

45.The method according to claim 44, wherein said steps of performing a first and a second lithographic process are a single lithographic process.

46.The method according to claim 45, further comprising, after the step of etching said silicon-based layer, the steps of:

- forming a second cladding layer (23.) on said etched silicon- based layer (2); etching a portion of said second cladding layer in order to uncover a portion of said silicon-based layer (2).