WO2008098404A2

WO2008098404A2 - Method for manufacturing a single-crystal film, and integrated optical device comprising such a single-crystal film

Info

Publication number: WO2008098404A2
Application number: PCT/CH2008/000065
Authority: WO
Inventors: Gorazd Poberaj; Andrea Guarino; Peter Günter
Original assignee: ETH Zürich
Priority date: 2007-02-16
Filing date: 2008-02-14
Publication date: 2008-08-21
Also published as: WO2008098404A3

Abstract

The invention relates to a method for manufacturing a single-crystal film, e.g. a metal oxide or organic crystal film, based on the known ion slicing technology. In a first step, ions are implanted into a donor crystal structure to form a damage layer within the crystal structure at an implantation depth below a top surface of the crystal structure, the top surface and said damage layer defining at least in part the single- crystal film to be detached from the crystal structure. Thereafter, the crystal structure is indirectly bonded to a substrate by a bonding layer between the crystal structure and the substrate, this bonding layer comprising a polymer adhesive. After curing the polymer, the laminate is exposed to a temperature increase to effect detachment of the single-crystal film from the crystal structure. The invention has the advantage of less stringent requirements for surface smoothness and flatness. Therefore, it enables a very reproducible fabrication of high-quality and large area thin films, e.g. of a metal oxide (e.g. ferroelectric) or organic crystal.

Description

METHOD FOR MANUFACTURING A SINGLE-CRYSTAL

FILM, AND INTEGRATED OPTICAL DEVICE COMPRISING

SUCH A SINGLE-CRYSTAL FILM

FIELD OF THE INVENTION

This invention is related in general to the field of manufacturing integrated optical devices utilizing single-crystal films. More particularly, the invention is related to a method for detaching micron-thin single-crystal films from donor crystal structures for bonding onto substrates.

BACKGROUND OF THE INVENTION

Crystal Ion Slicing (CIS) is a well-known technique to fabricate mesoscopically thin films. By means of ion implantation, a narrow damage layer is introduced into a donor crystal structure, for example an epilayer/substrate crystal structure or a bulk crystal structure. After ion bombardment, the crystal structure is bonded to a. substrate with the surface of the ion bombardment (top surface) facing the substrate. Chemical or thermal treatment, e.g. etching or rapid temperature increase, can slice a thin film from the crystal structure. The film remains bonded to the substrate. Its thickness corresponds to the depth of the damage layer, which is controlled by the energy of the implanted ions. CIS has been used in particular in the field of microelectronics for the fabrication of thin structures on a silicone wafer.

Direct wafer bonding of the donor crystal structure to the substrate requires very clean, flat and smooth surfaces (tolerable roughness typically around 1 ran rms). This means that the surfaces of commercially available wafers have to be treated before bonding. However, even with such special treatment of the surfaces it is very difficult to obtain thin films which have a larger area than several mm² because larger films tend to split off the substrate because of material tensions.

US 6,120,597 proposes to use a bonding layer in between the donor and the substrate. However, this bonding layer is supposed to be a low-temperature melting material which melts at temperatures below the typical temperatures necessary for detaching the thin film. Consequently, the thin film is not securely bonded to the substrate after the slicing/detachment process.

SUMMARY OF THE INVENTION

It is thus an object of the present invention to provide a method which allows to manufacture a thin single-crystalline film bonded to a substrate in an as easy manufacturing process as possible. In particular, it is desired to reduce the requirements for surface cleanliness and smoothness as compared to prior art processes.

It is a further object of the invention to provide a large area thin single-crystalline film securely bonded to a substrate, and a corresponding production process. It is yet another object of the invention to provide an integrated optical device wherein light is at least in parts guided by a thin single-crystalline film, or a specific structure therein.

These and other objects are solved by a method for manufacturing a single-crystal film as claimed in claim 1, an integrated optical device as claimed in claim 21. Beneficial embodiments are described in the dependent claims, the following description and the figures.

The present application describes a method for the fabrication of large area (several cm²) single-crystalline thin films, in particular metal oxide (preferred ferroelectric) films or films of an organic crystal, in particular a crystal with optical non-linear properties, combining the crystal ion slicing technique and adhesive bonding using a polymer adhesive. This polymer adhesive is preferably benzocyclobutene (BCB) from the Dow Chemical Company, but other materials could be used as well. Both, the crystal ion slicing technique and adhesive bonding have already been used for several applications, however, they have not yet been combined and used simultaneously for fabrication of thin films, in particular metal oxide (ferroelectric) or organic films. Comparing to the direct bonding, the adhesive bonding using a polymer adhesive like BCB or the like has a big advantage of less stringent requirements for surface smoothness, flatness, and cleanliness. Therefore, it enables a very reproducible fabrication of high-quality and large area thin (e.g. metal oxide, ferroelectric, organic) films. In addition, the adhesive bonding of ion-sliced thin ferroelectric or organic films with optical nonlinear properties using a polymer adhesive like BCB opens also new possibilities in connection with photonic devices, in particular for optimizing the waveguide design and the electrode configuration in low-voltage electro-optically tunable photonic devices. - A -

The method according to the invention comprises the following steps:

Implanting ions, in particular He+ and H+ ions, into a donor crystal structure to form a damage layer within the crystal structure at an implantation depth below a top surface of the crystal structure, the top surface and said damage layer defining at least in part the single-crystal film to be detached from the crystal structure;

bonding the donor crystal structure to a substrate by applying a bonding layer comprising a polymer adhesive between the crystal structure and the substrate;

curing the polymer;

- exposing the laminate to a temperature increase to effect detachment of the single-crystal film from the crystal structure.

The polymer adhesive used for indirectly bonding the donor crystal structure to the substrate is preferably a thermosetting plastic, in particular a heat-curable or UV- curable polymer. Such a material cures upon deposition of energy, e.g. by increasing the ambient temperature or irradiation, and can thus be exposed to the temperatures necessary to effect detachment of the film from the crystal structure without losing its bonding strength, in particular without melting away. The material properties of the polymer are chosen such that curing temperature is below a typical detachment temperature for the single-crystal film. A first heating phase takes place within a first temperature range which is adapted to the curing temperature of the polymer. Preferably, the temperature increase within this heating phase is comparatively slow to avoid a temperature shock that may lead to premature detachment of the film. After curing, the temperature is increased in a second heating phase to effect detachment of the film. The temperature increase may be comparatively fast. A third heating phase may be added to anneal implantation-induced crystal defects in the single-crystal metal oxide film. In this heating phase, the temperature may be increased even further, but preferably below the glass transition temperature of the polymer.

As will be shown below in more detail, the combination of Lithium Niobate (LiNbO₃) or related crystals (e.g. MgOiLiNbOs, EnLiNbO₃, Er:MgO:LiNbOs, Nd:MgO:LiNbθ3, LiTaOs, KNbOs, BaTiOs, KTaOs, KNbi-xTaxOs, Bai-xSrxNb∑Oβ, Ba.NaNb5θi5) as a donor and/or a substrate with a benzocyclobutene (BCB) as heat- curable polymer adhesive has significant advantages for the fabrication of in particular integrated optical devices, as the typical curing temperature of BCB (around 170°C) is lower than the typical detachment temperature (around 220- 250°C) which is again lower than the typical glass transition temperature of BCB (around 350°C), such that an annealing step can be performed at temperatures lower than 350°C. Furthermore, BCB is transparent and allows for light propagation. The difference in the refractive indices of BCB and LiNbO₃ opens up new design possibilities for mesoscopic optical elements attached to the LiNbO₃ substrate indirectly via the BCB layer, like waveguides (e.g. ridge/slab waveguides), photonic crystal structures, resonators (e.g. ring resonators), filters and the like, where the BCB serves as "cladding" for the optical structure and provides a refractive index contrast to the waveguide. Such elements can be manufactured by selectively removing material of the film by known processes like, for example, photolithography and etching, laser ablation.

Similar effects can be achieved by using organic crystals as a donor crystal. Organic crystalline materials can be, e.g., stilbazolium salts, such as 4-N,N-dimethylamino- 4'-N'-methyl-stilbazolium tosylate (DAST), 4-N,N-dimethylamino-4'-N'-methyl- stilbazolium 2,4,6-trimethylbenzenesulfonate (DSTMS) or molecular crystals, such as 2-(5-methyl-3-(4-(pyrrolidin- 1 -yl)styryl)cyclohex-2-enylidene)malononitrile

(MH2) or 2-(3-(4-hydroxystyryl)-5,5-dimethylcyclohex-2-enylidene)malononitrile (OHl). These materials have good electro-optic properties and a high potential for different applications in integrated optics.

Other polymer adhesives that are transparent in the desired part of the electromagnetic spectrum can be used as well. Examples are methylsilsesquioxane (MSSQ), Flare™, and Parylene-N.

Furthermore, the method according to the invention allows for the introduction of additional conductive structures in between or on top of the crystal film-polymer— substrate laminate, e.g. in the shape of conductive layers. Such conductive structures may serve as electrodes to influence the optical properties of the (ferroelectric, organic) film and thus to modulate an optical signal guided in the device. Such layers are a conductive coating with Chromium (Cr) or Indium Tin Oxide (ITO), for example. The latter has the advantage that it is transparent and can be arranged in between the thin film and the polymer layer by coating the donor crystal prior to the bonding step. The conductive layer may also be structured by known methods.

Further dielectric layers may be applied to the laminate after detachment of the single-crystal film from the crystal structure by a suitable material deposition process. They may also serve as cladding in an integrated optical device.

Though use of Lithium Niobate and related materials as donor and substrate is preferred, the method according to the invention is equally suited for other materials susceptible to ion slicing, if a suitable polymer is used and the characteristics of the heating steps are adjusted to the material properties of the polymer, the donor and the substrate. Preferably, donor and substrate are of the same material to avoid material tensions due to different thermal expansion.

The integrated device according to the invention comprises a substrate, a thin single- crystal film, and a polymer layer in between. The film and the substrate are preferably a ferroelectric or optical nonlinear organic material, in order to produce an integrated optical device suitable for light propagation and preferably electro-optic modulation.

Advantageously, the method of the present invention allows to produce a plurality of integrated devices on wafer scale, and thus in a mass production process.

Preferred embodiments of the invention are schematically shown in the drawings and are described below.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 shows different steps of the crystal ion slicing and wafer bonding technique according to the invention;

Fig. 2 shows a stopping range of implanted He⁺ ions in LiNbθ3 calculated as a function of their energies; Fig. 3 shows a photograph of a 600-nm thick LiNbCb film with an area of 12 x 10 mm² bonded by benzocyclobutene (BCB) on a Cr-coated LiNbCb substrate;

Fig. 4 shows an alternative manufacturing process according to the invention;

Fig. 5 shows a simplified scheme of a proposed active photonic crystal (PhC) device based on a structured LiNbCb thin film;

Fig. 6a-c show cross-sections of a lithium niobate microring resonator structure;

Fig. 7 shows a structured lithium niobate microring resonator;

Fig. 8 shows a transmission spectrum of a 100 μ m-radius ring resonator;

Fig. 9. Electro-optic shift of the resonance curve upon application of a voltage;

Fig. 10 shows the Q-factor of a microring resonator as a function of the refractive index contrast Δn = n∞re - risubstrate for different ring radii;

Fig. 11 shows the effective index of the first guided TE (straight line) and TM (dashed line) modes in a lithium niobate planar waveguide as a function of the film thickness; Fig. 12 shows the effective index of the first guided TE (straight line) and TM (dashed line) modes in a lithium niobate channel waveguide as a function of the width w for a film thickness h = 0.6 μ m;

Fig. 13 shows the effective (N) and group effective (Ng) index as a function of the wavelength for the first TE and TM guided mode in a lithium niobate microring resonator with radius R = 100 μ m.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

1. Method for fabricating thin ferroelectric films

In the following the disclosed fabrication method is described for the case of one particular lithium niobate (LiNbθ3) crystal. Four steps of the fabrication procedure of sub-micrometer-thick single crystalline LiNbθ3 films are schematically shown in Fig. 1:

Step (a): High dose He⁺ implantation (D=4-10¹⁶/cm²) of LiNbθ3 donor wafer 1; step (b) Preparation of LiNbθ3 substrate 3: Cr-electrode 5 deposition (50 nm) and spin- coating of benzocyclobutene BCB layer 4 (1-2 μm); (c) Wafer bonding; (d) Thermal treatment (up to 330°C) increases the bonding strength, induces detachment (split- off) of donor wafer 1, and partially recovers implantation-induced crystal defects in LiNbOathin film β. A crystal structure 1, here a LiNbCb wafer with a diameter of 3" and a thickness of 0.5 mm is implanted with He⁺ ions at an energy of 195 keV, a fluence of ca. 4-10¹⁶/cm², and at an angle of 7° to avoid ion channeling effects. A damage layer 2 in about 670 nm depth is formed. The implanted wafer 1 is then cut into smaller rectangular platelets, which are then bonded to a substrate 3, here LiNbCb substrates having a Cr-coating 5 on their front surface, by means of polymer adhesive layer 4, here a benzocyclobutene (BCB, Cyclotene, Commercial Product of the Dow Chemical Company) with a refractive index of 1.55. Compared to the direct bonding technique reported in the past, indirect bonding using a thin polymer adhesive layer 4 - here BCB - has the advantage of substantially lower requirements for smoothness and cleanliness of the bonded wafer surfaces. After bonding, the samples are heated in an oven at a temperature of 250 - 330°C for 5 - 15 hours.

The thermal treatment has a threefold role: (1) during the slow-ramp heating phase the bonding strength is increased (BCB hardens at around 170°C); (2) afterwards, at around 220-250°C, the implanted layer is split-off from the donor wafer and remains attached to the BCB layer as a single-crystalline film; (3) during the last (longest) heating phase, implantation-induced crystal defects in the LiNbCb thin film are partially recovered. Annealing process can be performed at temperatures up to 350°C (glass-transition temperature of the BCB). In this way, high quality LiNbCb thin films 6 with a thickness of 670 nm and a surface area of > 1.5 cm² have been routinely produced.

The refractive index contrast of LiNbCb films 6 with respect to the BCB layer 4 is around 0.7, which makes these films very suitable as a platform for very large scale integrated (VLSI) photonics devices. The described fabrication method does not depend on the size and thickness of chosen LiNbCb wafers 1 (Fig. 1, LiNbCb (a) ). High-quality, optically polished LiNbCb crystal wafers with a diameter of up to 4" are commercially available and can be implanted with appropriate implantation facilities. The thickness of fabricated films 6 depends on the energy of implanted He⁺ ions. Several different metals (including Cr as example in Fig. 1) can be used as electrodes by depositing a thin film thereof on either the substrate or the donor surface prior to the bonding step. The thickness of the BCB layer 4 can be controlled (1-20 μm), by choosing an appropriate type of the BCB and an appropriate spinning speed during the coating process.

LiNbCb film 6 thickness depends on the stopping range, and therefore, the energy of the implanted He⁺ ions, and is in a good agreement with the SRIM (The Stopping and Range of Ions in Matter) calculation. For example, a stopping range of the implanted He⁺ ions in LiNbCb as a function of their energies up to 800 keV is shown in Fig. 2. For an ion implantation energy of 195 keV, the thickness of LiNbCb film is 670 run. Since the ions range distribution has a finite width (see bars in Fig. 2 showing ion straggling), the upper film surface (splitoff side) exhibit a roughness of 6 nm rms. This value, however, is much smaller than the ion straggling.

As a simple and effective method to reduce the surface roughness (if required) we propose the Ar-ion sputtering. For example, sputtering away of a 70-nm thick uppermost layer reduces the surface roughness to 4 nm rms.

Fig. 3 shows a photograph of a 600-nm thick LiNbCb film with an area of 12 x 10 mm² bonded by benzocyclobutene (BCB) on a LiNbCb substrate. The adhesive

(indirect) bonding step using the BCB does not require a surface roughness of the bonded wafers to be smaller than lnm rms, which is a typical value required for a successful direct bonding of two wafers. The whole area of the thin film is free of defects. The substrate is Cr-coated.

It is possible to insert also layers of different materials between the polymer adhesive 4, e.g. BCB, and the thin metal oxide film 6, e.g. LiNbCb film. For example, in order to achieve a high electric field in LiNbCb optical waveguides structured in thin films (important for electro-optical devices), a transparent Indium Tin Oxide (ITO) layer 5', later acting as electrode, can be deposited directly on the implanted surface of the donor LiNbθ3 wafer 1, which is then bonded to the LiNbθ3 substrate 3 by means of the BCB 4.

Such a procedure, similar to that described above with respect to Fig. 1, is shown in Fig. 4 which shows a modified fabrication technique for LiNbθ3 thin film with a transparent Indium Tin Oxide (ITO) electrode attached directly to the LiNbθ3 thin film: Step (a) shows high dose He+ implantation (D=4-10¹⁶/cm²) of LiNbθ3 donor wafer 1, leading to a damage layer 2; step (b) shows deposition of Indium Tin Oxide 5' (ITO, 50nm) on the implanted donor wafer at a temperature below 220°C; step (c) shows the preparation of LiNbθ3 substrate : spincoating of benzocyclobutene BCB layer 4 (1-2 μm); step (d) Wafer bonding of wafers 1 and 3; step (e) Thermal treatment (up to 330⁰C) increases the bonding strength, induces detachment (split- off) of the donor wafer 1, and partially recovers implantation-induced crystal defects in LiNbO₃ thin film 6.

A high-quality ITO layer 5' can be deposited on an implanted LiNbθ3 donor wafer 1 at temperatures well below the critical temperature of 220°C at which the detachment (split-off) of the implanted LiNbθ3 layer 6 takes place. Furthermore, the method according to the invention enables very reproducible fabrication of thin films. Due to its relaxed requirements for a surface roughness as compared to the ion-slicing technique combined with the direct bonding technique, it is possible to transfer defect-free films of much larger areas. In addition, several other ferroelectric and non-ferroelectric materials, which do not show strong exfoliation upon high-dose ion implantation, might also be transferred successfully. The fabrication method described above can be also used for other ferroelectric materials such as: MgO:LiNbθ3, EnLiNbOs, Er:MgO:LiNbθ3, Nd:MgO:LiNbOs, LiTaOs, KNbOs, BaTiOs, KTaOs, KNbi-xTaxOs, Bai-xSrxNb2Oe, Ba-NaNb₅Ou, etc.

2. Structuring methods of thin films fabricated by the disclosed method

Several methods used for structuring of integrated optics devices can be applied for structuring of thin films produced by the disclosed method:

_: Standard mask lithography using UV lamp + (reactive) ion etching .- Laser lithography + (reactive) ion etching .- Direct laser writing of optical waveguides in LiNbθ3 .- Femtosecond laser ablation micromachining : e-beam lithography + (reactive) ion etching - Focused ion beam milling

A thin film 6 structured in a direction parallel to the top surface of the substrate is schematically shown in Fig. 6a+b. After structuring, the thin film 6, e.g. the LiNbθ3 film, can be covered by various dielectric layers 7 serving as a waveguide cladding.

Deposition of metallic electrodes 8 on the top of the laminate (in combination with the bottom electrode) enables exploitation of the electro-optic effect in ferroelectric thin films.

Also, transparent conducting electrodes, such as ITO, can be deposited directly on the structured thin film (without any dielectric buffer layer) to achieve the maximum electric field in the thin film for a given electrode voltage. Due to the use of transparent conducting electrodes, low-loss optical waveguiding is preserved.

3. Applications of thin ferroelectric films fabricated using the disclosed method

a) Conventional integrated optics devices based on:

Ridge or channel optical waveguides; - Curved waveguides with curvature radii as small as a few micrometers;

Modulators and switches based on microring resonators and Mach-Zehnder interferometers;

Large area thin films (> lcm²) enable fabrication of complex optical devices on a single chip; - Periodically poled ferroelectric thin films;

b) Photonic crystal slabs

The disclosed method enables fabricating high-index-contrast LiNbCb planar waveguides (slabs), which are very suitable as a platform for photonic crystal structures. An example thereof is shown in Fig. 5. Fig. 5 shows a simplified scheme of a proposed active photonic crystal (PhC) device based on a structured LiNbθ3 thin film. The film 6 is structured such that it comprises a bar-like first waveguide structure 11 (ridge waveguide) and a slab-like second waveguide structure 12 with a PhC channel waveguide in the middle. The second structure 12 comprises a periodic pattern 13 (PhC structure) of high and low dielectric constant, that affects the propagation of light through the second structure 12. A BCB layer 4, acting as low index cladding, is arranged underneath/behind the film 6 and connects it to the substrate 3 with a conductive layer 5, acting as bottom electrode, in between. The upper cladding and the upper electrode are not shown. The photonic bandgap in the photonic crystal 12/ 13 can be tuned by applying a voltage on the electrodes, exploiting the electrooptic effect in LiNbCh.

c) Combination of conventional integrated optics building blocks and photonic crystal structures;

d) Thin film based ferroelectric memory chips;

e) Thin film based pyroelectric detectors (e.g. LiNbCb, LiTaCb, Ki-xNaxNbi-yTayCb).

4. Further investigations

Optical microresonators have attracted a growing attention in the photonics community over the last decade. By confining circulating light in small volumes, their applications range from quantum electro-dynamics to sensors and filtering devices for optical telecommunication systems, where they are likely to become an essential building block. The integration of nonlinear and electro-optical properties in the resonators represents a very stimulating challenge, as it would incorporate new and more advanced functionality. Lithium niobate is an excellent candidate material, being an established choice for electro-optic and nonlinear applications.

Optical microring resonators in submicrometric thin films of lithium niobate have first been realized by means of the method according to the invention. The high index contrast films are produced by the improved crystal ion slicing and bonding technique using benzocyclobutene, according to the invention. The rings have radius

R = 100 μ m and their transmission spectrum has been tuned using the electro-optic effect. These results open new perspectives for the use of lithium niobate in chip- scale integrated optical devices and nonlinear optical microcavities.

The established use of wavelength division multiplexed (WDM) systems has raised the demand for new filtering and switching functions. In order to integrate these devices on a wafer scale, whispering gallery mode microresonators represent the most compact and efficient solution. They consist of a bus waveguide evanescently coupled to a micrometer-size ring resonator; the characteristic size-dependent frequency spectrum of the ring allows only selected wavelength channels to be transmitted or shifted to another waveguide. Small radii allow a large free spectral range - i.e. large separation between the filtered channels - but increase the propagation bending losses, which can compromise the quality factor Q - i.e. the wavelength selectivity - of the device. To overcome this limitation, high refractive index contrast between the ring core and the surrounding materials is mandatory.

A second requirement relates to the tunability. The possibility to electrically control the transmission spectrum, via electro-optic effect, would allow extremely compact and ultrafast modulation and switching. By integrating arrays of microring resonators on a single optical chip, the realization of complex functions would be feasible. Besides, large-Q resonators based on non centro-symmetric materials would exploit the high amount of stored energy for enhancing the efficiency of nonlinear optical phenomena.

Several examples of microring resonators have been proposed and successfully realized in the last years in a variety of materials like semiconductors, silica and polymers. The advanced structuring technology in semiconductor materials enables the realization of very high-Q resonators 1 even for radii as small as 10 μ m. Silicon- based resonators can be tuned by electrically-driven carriers injection in the core, but do not own truly nonlinear optical properties and their application is limited to infrared wavelengths. Polymers represent a very flexible solution in terms of processing and structuring, but the minimum resonator dimensions (and therefore the maximum achievable free spectral range) are limited by the low refractive index of the material. Silica rings, finally, do not provide any fast nonlinear or electro-optical property.

Here, it is proposed to use lithium niobate as a very attractive new choice for microresonating devices. It has the potential for ultrafast modulation since it has large electro-optic coefficients (m = 31 pm/V, rπ = 8 pm/V ), large transparency

range (0.4 - 5 μ m) and a wide intrinsic bandwidth. It is well known in existing electro-optic and nonlinear optical applications, and large dimension wafers of crystalline quality are available. The new technique according to the invention, based on crystal ion slicing and wafer bonding, produces sub-micrometric thin films of single-crystalline quality; it provides much higher refractive index contrast than the standard waveguide production methods in lithium niobate. This is an essential asset for the fabrication of small radius ring resonators, as is shown below in the Supplementary Information, Section 5.1. An electro-optic modulator has been demonstrated by using lithium niobate films bonded to Siθ2 as substrate. However, the direct bonding method does not provide large area films and lacks of sufficient reproducibility, due to the severe requirements on the surface roughness and imperfections. Bonding of lithium niobate films to other substrates (for instance, semiconductors) has also been reported, but suffers of film cracking due to the large mismatch between the thermal expansion coefficients of films and substrates and does not provide the optical contrast needed for the realization of optical microresonators.

We have improved the lithium niobate thin film fabrication technique by introducing the use of benzocyclobutene (BCB), a well known adhesive polymer for the realization of 3D semi- conductor devices, to successfully and reproducibly bond large area (> 1.5 cm²) submicrometric films. Full details of the fabrication procedure are presented below in the Methods section. The films are realized by implanting z- cut lithium niobate wafers with He⁺ ions which accumulate below the surface. The ion energy (E = 195 keV in our experiments) determines the position of their density

peak (here, 0.67 μ m). Subsequently, a sample of the implanted wafer is cut and bonded to another lithium niobate wafer, covered by a metallic electrode and a BCB layer (approximately 2.5 μ m). The bonded pair is thermally treated for several hours; this heating step, on one hand, strengthens the bonding by curing the polymer, on the other it causes helium bubbles to aggregate and leads to splitting of the film.

Finally, it also provides partial annealing of the defects introduced by ion implantation. The use of BCB offers several advantages: its planarization and adhesion properties reduce the role of surface defects and greatly enhance the reproducibility and the size of the trans- ferred films; optically, BCB has excellent transparency in the visible and infrared region, and as a substrate provides a suitable optical confinement due to its low refractive index (n about 1.55). After the splitting, the film thickness is reduced by Ar⁺ ion-etching of a sacrificial layer of approximately 60 nm. This step reduces the surface roughness inherently induced by the straggling of the implanted ions. The waveguides and the rings are structured by photo-lithographic techniques explained below; the ridge height is 0.4

μ m, as a compromise between a low surface scattering from the lateral walls and the

need for a suitable lateral confinement. Finally, a 0.85 μ m-thick covering SiCh layer reduces the scattering losses and ensures optical insulation between the core and the upper chromium electrode. We emphasize that the geometry chosen allows the applied electric field to be along the z-axis of lithium niobate and therefore to exploit

the electro-optic coefficient m = 31 pm/V.

A schematic representation of this device can be seen in Fig. 6a-b. Fig. 6c is a scanning electron microscopy (SEM) image of one end face of the structure cleaved before the deposition of the oxide and upper electrode layers.

Fig. 6a-c show a cross-section of a lithium niobate microring resonator structure. Fig. 6a-b show the schematic layout and cross section of a microring resonator 14 and coupling waveguide 15. The waveguide 15 and ring core 14 consist of structured lithium niobate thin film 6, bonded using BCB polymer 4 to a lithium niobate wafer 3 and covered by a SiCh layer 7. The upper and lower electrodes 5, 8 enable the application of an electric field along the z-axis of lithium niobate.

Fig. 6c shows a scanning electron microscopy image, viewed at an angle, of a cleaved end before the deposition of the oxide and upper electrode layers. The high-refractive index contrast structures produced with this technique (Δn about 0.65) are ideally suited for the realization of microresonators: the numerical calculations presented in the Supplementary Information, Section 5.1, show the bending losses are negligible even for ring radii of 10 μ m. The high contrast also implies stringent conditions on the waveguide dimensions to obtain single-mode operation (see Supplementary Information, Section 5.2 for details). The submicrometric thickness of our films support only one guided mode in the vertical direction. Single-mode opera- tion in the horizontal direction requires a waveguide width of approximately w = 1 μ m, which is too narrow for standard lithographic

techniques. Our waveguides have a width of approximately W = 4 μ m, hence they are multi-mode. However, the results demonstrate that in our structures the contribution of higher-order modes is nearly negligible, because these modes have higher propagation losses. More sophisticated structuring techniques (laser or electron-beam lithography) could potentially achieve true single mode operation without excessive scattering losses.

Another critical issue in the structuring of microresonators relates to the coupling coefficient between the waveguide and the resonator. To maximise the light extinction at the resonant wave- length, the coupling should be equal to the total propagation loss per resonator round trip. The horizontal coupling geometry requires a very accurate separation between the ring and the waveguide. To achieve a sub- micrometer gap, we lithographically define the waveguides and the rings in two steps, using a negative-tone photoresist. In the first step the straight waveguides are created in the photoresist using mask photolithography and hardening. Subsequently, the rings are formed on a second photoresist layer with the same procedure' and positioned using a standard mask-aligner. The two-step technique reduces the diffraction effects that would inhibit the formation of the narrow gap if a single-step illumination was used. The structures are then transferred into lithium niobate by Ar⁺ ion etching.

The scanning electron micrographs of Fig. 7 show a structured micro-ring resonator 14 in lithium niobate with radius R = 100 μ m (a) and a sub-micrometer gap 16 (b) between the waveguide 15 and the ring 14 obtained by this technique. The gap size is approximately 0.2 μ m.

The measured transmission spectrum of a coupled ring resonator around λ = 1.55

μ m is presented in Fig. 8. The measured normalized transmitted light at the through

port for both TE (left) and TM (right) modes using a tunable source in the λ =

1.55-1.57 μ m region is shown. The free spectral range is 1.66 ran and the finesse 5. The modulation depth is approximately 7 dB.

Both TE (electric field direction mainly parallel to the film) and TM (perpendicular to the film) polarisations of the waveguide bus can be coupled into the cavity and show the distinctive features of a microresonator. The extinction ratio at the resonant wavelengths is approximately 7 dB. The free spectral range of the resonator is about

Δλ FSR as 1.66 run, as predicted by the calculations presented in the Supplementary Information, Section 5.3, which account for the modal dispersion of the structure. The resonator finesse is approximately F = Δλ FSR/6A FWHM « 5 and the

corresponding Q value is Q = 4 ^χ 10³. This value is probably limited by implantation-induced defects and scattering losses. The electro-optic properties of lithium niobate microrings have been tested by shifting the transmission spectrum applying a static electric field to the device electrodes. In the Supplementary Information, Section 5.4, it is shown in detail how the induced refractive index change affects the resonance condition for both TE and TM modes.

Fig. 9 shows the electro-optic shift of the resonance curve at a wavelength around 1.555 μ m (left) and the corresponding electro-optically shifted curve (right) by

applying a voltage V = 100V to the device electrodes. The shift corresponds to an

approximate tunability of 0.14GHz/V

The resonance of a TM mode displayed in Fig. 9 shows a Δλ = 105 pm-shift in

response to an applied voltage of ΔV = 100V. This wavelength shift corresponds to

frequency tunability of 0.14GHz/V. This value indicates a reduction of the electro- optic activity of our structure by approximately 50% compared to the bulk material. A partial decrease of the electro and nonlinear optical properties in lithium niobate thin films due to implantation-induced defects has already been reported in a previous work. This reduction could be restored by optimizing the annealing conditions.

Two strategies can be implemented to reduce the switching voltage for a specific wavelength channel. First, an optimization of the polymer and oxide thickness would increase the electric field in the lithium niobate film, whose value is only 1/10 of the field in the underlying polymer, due to the large lithium niobate dielectric constant (s33 = 28). Second, the required wavelength shift is inversely proportional to the Q- factor of the cavity, therefore it could be dramatically decreased by reducing the propagation losses with advanced lithographic techniques. This would also allow the fabrication of a smaller resonator radius.

In summary, the first microring resonator based on sub-micrometric thin films of lithium niobate, produced using BCB-assisted bonding, have been realized. The resonance condition could be tuned using the electro-optic activity of the material. The size of the device is 30 times smaller than previously demonstrated resonators in lithium niobate. This work unveils the potential of electro-optically tunable optical microring resonators based on lithium niobate for telecommunication applications. The availability of optical microresonators in lithium niobate will lead to a variety of other experiments and applications, including nonlinear optical generation and amplification in the microcavities. This may be the start into a new direction in the realization of highly integrated nonlinear photonic devices.

Methods:

Here in detail the device fabrication is described. The implanted wafer is a pure congruent lithium niobate z-cut wafer (Crystal Technology, Inc.). The He⁺ ions had energy E = 195 keV. The implantation fluence was 4 * 10¹⁶ ions/cm² and the sample

holder was heated to T = 100°C during the process. The implanted wafer was cut in 12^χ 14mm² pieces and cleaned using standard RCAl solution. The substrate consists of another pure congruent z-cut lithium niobate wafer. The bottom electrode was formed by deposition of a 50 nm-thick chromium layer. BCB, under its commercial name of Cyclotene 3022-46 (Dow Chemical) was spun at 4000 rpm, after the use of the Adhesion Promoter AP3000. The polymer thickness was approximately 2.5 μ m. Thermal treating of the bonded pair was performed at T = 290°C in vacuum conditions (to avoid BCB oxidation) for several hours. No bonding pressure was applied during this step. The splitted film was subsequently smoothed by sputtering ofAr⁺ ions for 50 minutes (200W), which removed approximately 60 nm of material. Atomic force microscope measurements demonstrate that the RMS surface roughness was reduced by 40% to 4 nm by this process.

The photolithographic structuring of waveguides and rings, the negative tone photoresist SU-8 was used in two steps. In each step, the photoresist layer was 1.4

μ m-thick and after illumination and development the structures are hard baked at 120°C. The positioning of the samples was performed using a Karl-Suss MJB3UV3OO mask-aligner.

The ridges and rings were transferred into lithium niobate after 320 minutes of 200W etching using Ar⁺ ions. After removing the remaining SU-8, the sample was covered by a PECVD-layer of SiCh of approximately 0.85 μ m. The upper electrode was deposited with the same parameters as the bottom electrode.

Finally the sample was sawed and the sides were polished to ensure efficient end-fire coupling. Typical sample length is 3mm.

Experiments: The microrings were tested using a tunable laser diode Santech TSL-220. The tuning range is 1.530-1.610 μ m and the spectral width is IMHz. The light was spatially filtered using a single mode fiber and end-fire coupled into the waveguide using a 10Ox microscope objective with NA=O.9. The transmitted light was collected using a 4Ox (NA=0.45) long working-distance microscope objective.

5. Supplementary Information

5.1. Bending losses and Q-factors of high-index contrast resonators

In this section it is shown how a high refractive index contrast between the lithium niobate and the surrounding material minimise the bending losses and therefore can lead to an improved Q-factor of these microring resonators. Decreasing the microring radius is attracting, because the free spectral range (FSR), defined as the separation between two adjacent resonant wavelengths, increases and may become even larger than the wavelength range used for WDM applications. For example, a lithium niobate microring resonator with radius R = 10 μ m will have a FSR of approximatively 16 run. However, one of the most critical factor limiting the minimum useful ring radius are the bending losses. These can be qualitatively understood by describing the bend as a straight waveguide, where the effective index is a decreasing function in the radial direction. This implies that at a certain distance from the waveguide core, the solution of the Maxwell equations becomes a radiating field; this radiation is a loss source, as in a leaky waveguide. The larger the refractive index difference, the smaller will be the leakage. The losses essentially determine the Q-factor of the optical cavity: if only propagation losses α are present, the Q factor can be calculated using:

wN

Q (1)

where N is the effective index and λ is the vacuum wavelength. A bending loss of

α = 10 dB/cm would already imply that Q « 4000 without any other loss source.

Using a commercial software, we have calculated the bending losses for a single mode waveguide as a function of the refractive index difference between the core and the surrounding material, assuming a TE guided mode and a light wavelength of λ = 1.55 μ m.

The results are shown in Fig. 10: This figure shows the Q-factor of a microring resonator (due to bending losses only) as a function of the refractive index contrast

Δn = ncore - nsubstrate for different ring radii. The calculations refer to a TM mode at

λ = 1.55 μ m. The core dimensions are W = 1 μ m and h = 0.6 μ m for _n = 0.65

and are scaled to be single mode for every value of Δn. The grey line shows that for lithium niobate microring resonators, surrounded by BCB polymer, small ring radii are possible because Δn « 0.6.

It is evident that lithium niobate thin films, bonded using BCB polymer, have a large advantage over other nonlinear or electro-optic materials like polymers, whose typical contrast is about 0.1 - 0.2. The refractive index contrast of Δn » 0.6 enables

the fabrication of rings having a radius as small as R = 10 μ m, if other loss sources can be neglected.

5.2. Conditions for single mode operation in lithium niobate waveguides

The high refractive index difference between the lithium niobate (no = 2.21 and rie =

2.13 at λ = 1.55 μ m) and the BCB polymer (n « 1.55) requires the film thickness

and the waveguide width to be smaller than a limiting value (about 1 μ m) to achieve

single-mode operation. The modal curve at λ = 1.55 μ m for planar lithium niobate waveguides surrounded by Siθ2 and bonded using BCB has been calculated.

The effective index dependence on the film thickness is presented in Fig. 11. This figure shows the effective index of the first guided TE (straight line) and TM (dashed line) modes in a lithium niobate waveguide as a function of the film thickness. The waveguide is surrounded by SiCh (top) and bonded using BCB (bottom); the modes are calculated at λ = 1.55 μ m. In case of a thickness h = 0.6 μ m, the confinement of any higher order mode is either not existing or very weak.

The maximum film thickness to achieve single mode operation is given by h = 0.7

μ m. With the technique explained above, the film thickness is determined by the

energy of the implanted ions. The present films, being created using E = 195 keV and subsequently Ar⁺ sputtered, have a thickness h = 0.6 μ m and therefore support only one guided mode.

The dependence of the effective index on the waveguide width, assuming a waveguide height h = 0.6 μ m is shown in Fig. 12. This figure shows the effective index of the first guided TE (straight line) and TM (dashed line) modes in a lithium niobate waveguide as a function of the width W for a film thickness h = 0.6 μ m. The waveguide is surrounded by S1O2 (top and aside) and bonded using BCB (bottom); the modes are calculated at λ = 1.55 μ m. Single mode operation can be obtained

only with structures narrower than 1 μ m.

As it can be seen, a single-mode operation requires waveguide structures narrower than 1 μ m. This condition is not met in the microrings presently investigated, which

have w w 4 μ m and therefore support several modes. However, single-mode operation could be reached by using more advanced structuring techniques, like laser or electron beam lithography. Nevertheless, in the present experiments higher order modes have limited impact, since they have higher propagation losses.

5.3. Group effective index and calculation of the free spectral range

To properly analyse the transmission properties of the microring resonators in thin films of lithium niobate, the dispersion properties of the guided modes have to be considered. In this section the modal dispersion of the present structure, i.e. the sensitivity of the effective index of the guided modes to wavelength changes, is calculated. The resonance condition of the microring resonator is given by:

2τnV(λ)L

= 2πm or N(X)L = mλ (2)

where N is the effective index of the guided mode, L is the resonator's length, λ is

the light wavelength and m is an integer number. The effective index depends on the wavelength by two distinct mechanisms. First, the refractive index of the materials are wavelength dependent (material dispersion); second, the guiding properties of the structure (i.e. the solutions of the propagation equation) depend on the wavelength (modal dispersion). The second mechanism is specially relevant, due to the tiny vertical dimension of the core (the film thickness) with respect to the wavelength. The free spectral range (FSR) is defined as the difference between two adjacent resonant wavelengths Δλ FSR = λ m - λ m+i; it can be approximately calculated by differentiating Equation (2):

X² X²

^AXFSR ~ r(\τ \M\ ⁼ TW ^

where Ng := N — λ dWdK has been defined as the group effective index. This

quantity contains the material and modal dispersion. Ng for the first guided mode of both TE and TM polarisations in the straight waveguide and in the microring resonator has been calculated, considering both modal and material dispersion, using a mode solver software.

The results are shown in Fig. 13. This figure shows the effective (N) and group

effective (Ng) index as a function of the wavelength for the first TE and TM guided

mode in a lithium niobate microring resonator with radius R = 100 μ m. Waveguide

width and height are W = 4 μ m and h = 0.6 μ m, respectively. The waveguide core

is surrounded by BCB polymer (n « 1.55) and Siθ2 (n » 1.45) as explained above.

The discrepancy between N and Ng increases with higher wavelength due to reduced guiding effect.

The difference between the effective index N and the group effective index Ng is large (more than 25%), therefore the role of dispersion in lithium niobate films shall not be overlooked for a proper calculation of the free spectral range. In the visible region of the spectrum the mode is confined and the main contribution to dispersion is given by the material dispersion. In the infrared region of interest, the main contribution is given by the modal dispersion, since the film thickness is smaller than the wavelength.

To estimate the free spectral range around λ = 1.55 μ m we therefore use Ng₁TE «

2.323 and Ng₁TM « 2.286, which in (3) yield a free spectral range of 1.65 nm and 1.67 nm, respectively. This is well in agreement with the spacing experimentally measured (1.66 run) as presented above.

5.4. Electro-optic effect in lithium niobate microresonators

In this section more details about the electro-optic tuning effect induced in lithium niobate microresonators are presented. Applying an electric field between the upper and lower electrode of the structure shown in Fig. 6, the refractive indices of lithium niobate are changed according to its electro-optic tensor njk. This change modifies the

effective index N of the microring resonator, hence its round-trip phase and its resonance condition. Since the applied electric field is directed along the z-axis of lithium niobate, the only relevant electro-optic coefficients are rn and H3, which are responsible for the change of the ordinary and extraordinary refractive index, respectively. For a small electric field δE, the effective index N varies according to:

The derivatives of the effective index N with respect to the ordinary and extraordinary refractive indices can be determined by using a developed mode solver routine in MATLAB which accounts for the anisotropy of the structure. For the given dimensions (h = 0.6 μ m, W = 4 μ m) and the wavelength λ = 1.55 μ m we calculated the following values result for the first guided TE and TM mode:

The TE mode is sensitive to the ordinary refractive index only, while the TM mode is sensitive to both. The values indicate that, due to the small core dimensions, the change of the effective index is smaller than the change of the material refractive index. To estimate δN, the electric field in the lithium niobate core must be determined. Due to the large difference between the dielectric constant of the film (s33 = 28) and of the other materials (SBCB = 2.6 and εsiθ2= 3.8) the field in the film is considerably smaller than in the adjacent layers. The field in the i-th layer can be calculated from the applied voltage ΔV using the continuity of the vertical

component of the electric displacement field Di:

i (6)

D_f = £ø&j.Ej = const

Therefore, the field in the lithium niobate layer is given by: AV AV

ELN =

CLN Σ O,⁴7 d_eff (7)

The quantity deff is an effective thickness. It is a practical quantity which accounts for every layer thickness and dielectric constant if the field in the lithium niobate layer has to be determined. The effective thickness which corresponds to the structure presented in the text is deff « 34 μ m. The effective index change can be finally expressed as

δ_{N (8)}

The shift of the resonance wavelength, due to a perturbation of the effective index, can be derived by differentiating Equation (2) for a fixed integer value m. The effect of the dispersion is taken into account by using the definition of group effective index given in the previous section:

The tunability of the microring resonator does not depend on its length. Substituting in the last expression the result of (8) we obtain:

Ng

dn_β 2 ) d_eff ^{K J} The tunability T is often expressed in terms of frequency shift per voltage δv /δV :

Using the bulk lithium niobate values for the electro-optic coefficients and the previously determined deff, T and the expected wavelength shift can be calculated if a

voltage δV = 100V is applied.

mode tunability |(r Shift. SX [pm] for SV = 100 V [GHz/V] expected measured

TM 0.28 224 105 ± 10 TE 0.10 80 30 ± 8

The measured values are approximately 40 - 50% of the expected values. To confirm the results, alternative measurements have been performed by applying a small

(10.5V amplitude) oscillating voltage and detecting the oscillating light intensity with a lock-in amplifier. The wavelength has been chosen off-resonance, to ensure the maximum response. The experiments confirmed the values determined by the resonance shift. This decrease can be attributed to the ion implantation induced crystal defects, since the samples could not be annealed at a temperature higher than 300 °C.

Claims

WHAT IS CLAIMED IS:

1. A method for manufacturing a single-crystal film, in particular a metal oxide film or an organic crystal film, comprising the steps of: implanting ions into a donor crystal structure to form a damage layer within the crystal structure at an implantation depth below a top surface of the crystal structure, the top surface and said damage layer defining at least in part the single-crystal film to be detached from the crystal structure; bonding the donor crystal structure to a substrate by applying a bonding layer between the crystal structure and the substrate; and exposing the laminate to a temperature increase to effect detachment of the single-crystal film from the donor crystal structure, wherein the bonding step comprises providing a polymer adhesive, and curing the polymer adhesive prior to detachment of the single-crystal film from the donor crystal structure.

2. The method according to claim 1, wherein the polymer adhesive is a thermosetting plastic, in particular, a heat-curable or UV-curable polymer.

3. The method according to claim 1 or 2, wherein the polymer adhesive is a heat- curable polymer with a curing temperature that is below a typical detachment temperature for the single-crystal film.

4. The method according to one of the preceding claims, wherein the polymer adhesive is benzocyclobutene (BCB), methylsilsesquioxane (MSSQ), Flare™, or Parylene-N.

5. The method according to one of the preceding claims, wherein the crystal structure is a metal oxide crystal structure, in particular a ferroelectric crystal structure, or an organic crystal structure, in particular a nonlinear optical crystal structure.

6. The method according to claim 5, wherein the metal oxide crystal structure is at least one of LiNbOs, MgC-LiNbCb, EnLiNbO₃, Er:MgO:LiNbOs, Nd:MgO:LiNbθ3, LiTaO₃, KNbO₃, BaTiO₃, KTaO₃, KNbi-χTaxO₃,

Bai-xSrxNb2θ6, Ba-NaNbsOis.

7. The method according to claim 5, wherein the organic crystal structure a stilbazolium salt, for example 4-N,N-dimethylamino-4'- N'-methyl- stilbazolium tosylate (DAST), 4-N,N-dimethylamino-4'-N'-methyl-stilbazolium 2,4,6-trimethylbenzenesulfonate (DSTMS), or a molecular crystal, e.g. 2-(5- methyl-3-(4-(pyrrolidin-l-yl)styryl)cyclohex-2-enylidene)malononitrile (MH2), 2-(3-(4-hydroxystyryl)-5,5-dimethylcyclohex-2- enylidene)malononitrile (OHl).

8. The method according to one of the preceding claims, wherein the substrate has an equal or similar thermal expansion coefficient as the donor crystal structure, in particular is a LiNbO₃ wafer.

9. The method according to one of the preceding claims, wherein the bonding step comprises a first heating phase within a first temperature range, which is preferably performed for 0.5-2h.

10. The method according to claim 9, wherein the maximum temperature in the first temperature range is between 20°C and 200°C, preferably approximately 150-200°C, in particular approximately 170°C.

11. The method according to one of claims 9-10, wherein the exposing step comprises a second heating phase wherein the temperature is increased to a value above the first temperature range.

12. The method according to claim 11, wherein the temperature is increased to approximately 200-250°C, preferably approximately 220⁰C.

13. The method according to one of claims 9-12, wherein the exposing step comprises increasing the temperature of the laminate from a first temperature to a higher second temperature.

14. The method according to claim 13, wherein the temperature increasing step is performed at a rate of 2-5°C/min.

15. The method according to one of the preceding claims, comprising, after detachment of the single-crystal film from the crystal structure, a further heating phase to perform annealing of implantation-induced crystal defects in the single-crystal film.

16. The method according to one of the preceding claims, comprising the step of applying an electrically conductive coating to at least one of the top surface of the crystal structure and the substrate, preferably at least one of a Chromium (Cr) or Indium Tin Oxide (ITO) layer.

17. The method according to one of the preceding claims, wherein the implanting step comprises implanting at least one of He⁺ and H⁺ ions.

18. The method according to one of the preceding claims, comprising the step of smoothening of the donor crystal structure surface and/or edges prior to the bonding step, in particular by Ar+ sputtering or chemo mechanical polishing.

19. The method according to one of the preceding claims, further comprising structuring the single-crystal film by selectively removing material of the film, in particular to form an optical waveguide and/or photonic crystal structure.

20. The method according to one of the preceding claims, further comprising applying at least one dielectric layer to the laminate after detachment of the single-crystal film from the crystal structure.

21. Integrated device, comprising a single-crystal film attached to a substrate with a polymer adhesive layer in between, as produced by the method according to one of the preceding claims.

22. Integrated device according to claim 21, wherein the polymer adhesive layer comprises a heat-curable polymer, in particular benzocyclobutene (BCB), methylsilsesquioxane (MSSQ), Flare™, or Parylene-N.

23. Integrated device according to claim 21 or 22, wherein the single-crystal film is a metal oxide film, in particular a ferroelectric crystal structure, preferably at least one of LiNbOs, MgOrLiNbOs, EnLiNbOs, EnMgOrLiNbOs,

Nd:MgO:LiNbOs, LiTaOs, KNbOs, BaTiOs, KTaOs, KNbi-xTaxOs,

Bai-xSrxNbaOδ, Ba2NaNbsOi5.

24. Integrated device according to one of claims 21 to 23, wherein the substrate has an equal or similar thermal expansion coefficient as the donor crystal structure, in particular is a LiNbOs wafer.

25. Integrated device according to one of claims 21 to 24, wherein the single- crystal metal oxide film is structured as seen in a direction parallel to the film, in particular to form optical waveguides and/or photonic crystal structures.

26. Integrated device according to one of claims 21 to 25, further comprising at least one electrically conductive coating acting as electrode.