Article comprising a planar optical waveguide with optically non-linear core

[0002] This application pertains to articles (e.g., an all-optical fiber switch for an optical fiber communication system) that comprise an optical waveguide (typically a planar optical waveguide) with optically non-linear core.

[0003] Switching of optical pulses is an essential functionality of substantially any practical optical fiber communication system. Current technology limits practical electronic switching speeds to about 40 Gbit/s. Thus, in order to attain single channel bit rates higher than about 40 Gbit/s, it is necessary to develop the ability to switch and process optical pulses at speeds substantially higher than 40 Gbit/s.

[0004] This application discloses means for ultra fast switching and/or processing (collectively “switching”) of optical pulses, exemplarily permitting sub-picosecond switching times.

[0005] Japanese Kokei Patent Application No. Hei 3 [1991]-21934 discloses nonlinear optical devices comprising Ge, As, S and Se-containing chalcogenide nonlinear material.

[0006] All cited references are incorporated herein by reference.

[0007] All-optical switching (i.e., switching of an optical signal pulse with an optical control pulse) in principle can be made ultra-fast.

[0008] Typically, signal and control light propagate together through an optical waveguide comprising an optically nonlinear core material. Of particular interest herein are planar optical waveguides.

[0009] The refractive index n of a nonlinear material is conventionally expressed as n₀+n₂I, wherein I is the optical intensity (in W/cm²), n₂is the nonlinear refractive index (in cm²/W), and n₀is the conventional linear refractive index. Similarly, the nonlinear attenuation α of a material is expressed as α₀+α₂I, wherein α₂is the nonlinear attenuation (in cm/W), and α₀is the conventional linear attenuation (in cm⁻¹). It will be understood that typically the above parameters are wavelength-dependent.

[0010] Using the fact that an all-optical switching device of the type that is of interest herein typically utilizes a phase shift of order π, it can be deduced that such a device requires a waveguide of length L equal to about λ/2n₂I, where λ is the optical wavelength in vacuum. In fiber devices L can be quite long, but in planar waveguide devices L desirably should be no more than about 1 cm. Furthermore, devices of interest herein will typically operate at a wavelength at or near 1.55 μm, exemplarily in the range 1.45-1.65 μm. For the sake of concreteness, in the discussion below it will be assumed that λ=1.55 μm. Still furthermore, a reasonable light intensity I is about 1 GW/cm², and an acceptable attenuation ao is less than about L⁻¹.

[0011] For an operating wavelength at which the nonlinear attenuation of a given waveguide core material dominates, the above requirements lead to the inequality (α₂λ/2n₂)<1, and for a wavelength at which the linear attenuation dominates, the requirements lead to the inequality (α₀λ/2n₂I)<1. In both cases it is clearly desirable if n₂is relatively large, typically more than several 100 times larger than the nonlinear refractive index of silica, designated n₂(SiO₂), which is about 2.8×10⁻¹⁶cm²/W at 1.55 μm. Furthermore, α desirably is not more than about 1 cm⁻¹.

[0012] We have carried out a theoretical analysis of the optical properties of some chalcogenide glasses and have determined that some such glass can meet the above recited requirements, and thus can be used as nonlinear core material in a planar waveguide for all-optical switching.

[0013] More specifically, the invention is embodied in an article that comprises a planar waveguide adapted for guiding light of vacuum wavelength λ in the range 1.45-1.65 μm. The planar waveguide comprises a nonlinear core material having at λ a refractive index n=n₀+n₂I, where n₀is the linear refractive index, n₂is the nonlinear refractive index, and I is the intensity of guided light in the waveguide.

[0014] Significantly, the nonlinear core material is a chalcogenide glass comprising selenium and a member of the group consisting of arsenic, germanium, and arsenic and germanium. Associated with the chalcogenide glass is an optical energy gap E_g. See, for instance, R. Zallen, “The Physics of Amorphous Solids”, John Wiley and Sons, New York, 1983, especially p. 266. The chalcogenide glass further comprises one or more dopant elements selected from the group of elements whose presence in the chalcogenide glass results in a change in E_g. The amount of the one or more dopant elements in the doped chalcogenide glass is selected such that E_gis substantially equal to a two-photon energy 2hν, where h is Planck's constant and ν=c/λ, where c is the speed of light in vacuum, and the amount is furthermore selected such that the doped chalcogenide glass has a nonlinear refractive index n₂greater than 200n₂(SiO₂), preferably greater than 400n₂(SiO₂), where n₂(SiO₂) is the nonlinear refractive index of vitreous silica at about 1.55 μm.

[0015] Exemplary dopant elements are tellurium, antimony, thallium, copper, silver and sulfur. Of the recited exemplary dopant elements, all but S reduce E_g.

[0016] It should be understood that the prior art contains data on some of the optical properties of substantially any chalcogenide glass of interest herein, including data on some effects of doping. However, to the best of our knowledge, nothing in the prior art suggests a technique for evaluating the nonlinearity of a given chalcogenide glass and, significantly, nothing suggests the use of chalcogenide glass of sufficiently large nonlinearity as core material of a planar optical waveguide, the planar waveguide for instance being used for all-optical pulse switching or processing.

[0017] FIG. 1 shows the frequency dependence of the nonlinear refractive index n₂as a function of the ratio of optical energy to bandgap;

[0018] FIG. 2 shows the functional dependence of n₂(0) on n₀;

[0019] FIGS. 3a-3dshow approximate materials parameters as a function of composition for As_ySe_1−y;

[0020] FIGS. 4A-4dshow approximate materials parameters as a function of composition for Ge_ySe_1−y;

[0021] FIG. 5 schematically depicts an exemplary optical fiber communication system comprising all-optical switches according to this invention;

[0022] FIG. 6 schematically shows an exemplary planar optical waveguide according to the invention; and

[0023] FIG. 7 schematically illustrates an optical time division multiplexer according to the invention;

[0024] FIGS. 5, 6 and 7 are not intended to be to scale or in proportion.

[0025] We have carried out an extensive theoretical analysis of the optical properties of chalcogenide glasses and have concluded that at least some chalcogenide glasses can be advantageously used in all-optical switches. In particular, we have found that there are two material parameters that are of particular importance in a chalcogenide glass for all-optical switching. These are the optical energy gap E_gof the chalcogenide glass (also referred to as the Tauc energy gap) and the linear refractive index no of the glass. Appropriate choice of these (and optionally other) material parameters can result in highly nonlinear glass with relatively small attenuation for light of predetermined wavelength λ, typically about 1.55 μm. Such material can be advantageously used as the nonlinear core material in a planar waveguide for all-optical switching.

[0026] Our theoretical analysis inter alia yielded the following expression for the nonlinear refractive index n₂(x):

[0027] n₂(x)=0.6(n₀²+2)³(n₀²−1)(d/n₀E_s)²F(x),  1)

[0028] in units of n₂(SiO₂). In this expression, x=hν/E_g, hν is the photon energy, n₀is the linear refractive index as x goes to zero, E_s(in eV) is the Sellmeier energy gap, d(in μm) is the mean bond length of the bonds that are primarily responsible for the dielectric response of the material, and the function F(x) is shown in FIG. 1.

[0029] From the expression

[0030] n₀²(ν)−1=E_dE_s/(E_s²−h²ν²)  2)

[0031] the Sellmeier energy gap can be determined experimentally from the long wavelength frequency dependence of the electronic linear response, while the energy parameter E_d(in eV) can be determined by extrapolation to zero frequency, yielding

[0032] E_d=E_s(n₀²−1).  3)

[0033] As can be seen from FIG. 1, F(x) (and thus n₂(x)) increases rapidly as x approaches 0.5 from below. Values of x greater than 0.5 permit 2-photon absorption and are therefore undesirable in the context of high intensity devices, due to the resulting large attenuation.

[0034] A principal result of our theoretical analysis is the recognition that chalcogenide glass having an optical gap E_gat or near the two-photon energy 2hν of the operating light (typically about 1.55 μm) can have a large nonlinearity and is a candidate for all-optical switching, provided the attenuation is small enough to meet practical requirements.

[0035] Equation (1) also shows that n₂increases with increasing bond length d. Thus, it is desirable to maximize d.

[0036] Equation (1) above shows that the value of the optical gap is an important parameter for choice of a nonlinear material, but it also shows that other parameters typically are also important. The Sellmeier gap E_sdesirably is small, since both n₂and n₀increase with decreasing E_s. However, E_sscales to a significant degree with E_g. This conclusion follows from the recognition that E_smeasures the mean gap between the dielectrically active energy levels near the top of the valence band and those near the bottom of the conduction band, while E_gis a measure of the minimum gap. For instance, in chalcogenides, E_s/E_gtypically is confined to a range a little above 2, exemplarily 2.2-2.6.

[0037] From equation (1) it also follows that n₂(0) is proportional to (n₀²+2)³(n₀²−1)/n₀², and thus is a sensitive function of the linear refractive index n₀, as can be seen from FIG. 2. Desirably, n₀is as large as possible, commensurate with the selected E_g.

[0038] From equation (2) one can see that n₀is also significantly influenced by the energy parameter E_d. It has been shown that E_dincreases with the number of (electronically active) valence electrons per anion, the bond coordination number of the anion, and the degree of covalence of the bonds involved in the electronic response. See, for instance, S. H. Wemple, J. Chemical Physics,Vol. 67, page 2151 (1977).

[0039] In summary, our analysis shows that a chalcogenide glass for all-optical switching of light of wavelength λ desirably is selected to have an optical gap (Tauc gap) E_gat or just below the 2-photon energy 2hν. For λ=1.55 μm, the 2-photon energy is about 1.6 eV.

[0040] The analysis also shows some desirable optional conditions, namely, a relatively high anion coordination number; a relatively large number of valence electrons per anion; a relatively high degree of covalency; and relatively long (covalent) bonds.

[0041] Furthermore, it is important that the chalcogenide glass be substantially free of any tendency towards actual or incipient phase separation, exhibit substantially no microcrystallinity, and have a glass transition temperature sufficiently above ambient temperature for the glass to be stable under anticipated operating conditions.

[0042] The above discussion of the theoretical analysis is provided for tutorial purposes only and is not meant to limit the scope of the claims.

[0043] Of the four principal binary chalcogenide glass-forming compositions (namely As—S, As—Se, Ge—Se and P—Se), only two (namely As—Se and Ge—Se) have E_gsmall enough (less than 2 eV) to make them potentially useful for optical switching.

[0044] Our calculations show that for bulk As_ySe_1−ythe maximum value of n₂at λ=1.55 μm is about 430 times n₂(SiO₂) at As₂Se₃, with E_g=1.74 eV. However, such a high value of n₂may not be attainable in the thin-film context, since glassy thin film As_ySe_1−yis expected to exhibit decreased packing density (as expressed by E_d), as compared to bulk material.

[0045] Similarly, our calculations show that for bulk Ge_ySe_1−ythe maximum value of n₂at λ=1.55 μm is about 160 times n₂(SiO₂) at about Ge₂Se₃, with E_g=1.9 eV. Pseudo-binary chalcogenide (As,Ge)_ySe_1−yis expected to have intermediate E_g.

[0046] The above remarks indicate that the chalcogenide glasses As—Se, Ge—Se and (As,Ge)—Se have relatively high nonlinearity. They also have relatively low values of attenuation.

[0047] The above recited numerical values of n₂pertain to bulk material, and n₂is expected to be somewhat less for most of the respective chalcogenide glasses in thin film form. Nevertheless, the above mentioned chalcogenide glasses may have utility as nonlinear core material in planar waveguides for all-optical switching, but are not preferred.

[0048] It would be desirable to have available chalcogenide glass compositions that have even higher non-linearity, together with relatively low attenuation, even in thin film form. We have found that this is possible by appropriate doping of the binary (or pseudo-binary) chalcogenide glass base composition, and preferred embodiments of our invention comprise such doped chalcogenide glass.

[0049] More specifically, we have found that doping can be used to tune E_gto a value at, or close to, the 2-photon energy.

[0050] Furthermore, doping can optionally be used to enhance E_d, either via improved packing efficiency or by increasing the mean number of electronically active valence electrons per anion. Doping can optionally also be used to increase the mean bond length of the strongly electronically active bonds and to substantially maintain or improve the homogeneous glass-forming capability of the glass.

[0051] It will be understood that a given dopant will not necessarily provide all of the above recited improvements. In view of the strong dependence of n₂on E_gnear the 2-photon energy, the first of the above condition will generally be the most important one.

[0052] Tellurium can be incrementally substituted for Se in both As—Se and Ge—Se, and is currently the preferred E_g-decreasing anionic dopant. Exemplarily, the composition As₂(Se_1−zTe_z)₃can be tuned to E_g˜1.6 (eV) for z approximately equal to 0.15. For this composition the following approximate parameter values pertain:

[0053] n₀=2.90

[0054] E_s=3.6

[0055] E_g=1.6

[0056] d=2.46

[0057] x=0.5

[0058] F(x)=3.4,

[0059] yielding n₂of about 950n₂(SiO₂). Other compositions As_ySe_1−y, Te_zin the compositional vicinity of y˜0.4 and z˜0.09 can also have large values of n₂, as can Te-doped Ge—Se compositions.

[0060] Exemplary cationic dopants for E_g-reduction in As—Se and Ge—Se chalcogenide glasses are Sb, Bi, Sn, Pb, In, Tl, Cu and Ag. However, cationic dopants other than Sb, Tl, Cu and Ag rapidly destabilize the selenide glasses with increasing dopant concentration, and thus are not preferred.

[0061] Exemplarily, Sb-doped chalcogenide glass of composition (1−y) As₂Se₃.ySb₂Se₃can form homogeneous glasses for y<0.45, with E_gbeing about 1.6 eV at y˜0.1. Antimony-doping of As—Se glass is expected to yield a maximum value of n₂of about 650n₂(SiO₂). Similar behavior (but with somewhat smaller maximum n₂) is expected for Sb-doped Ge—Se.

[0062] As—Se glass can be tuned to E_g=1.6 eV by addition of only a few atomic percent of Tl. The doped glass is expected to have n₂˜700 n₂(SiO₂). Ge—Se requires considerably more Tl than As—Se. maximum n₂is expected to exist close to GeTlSe₂, and to have a value of about 600n₂(SiO₂).

[0063] Both copper and silver reduce E_gin As—Se glass and can be used to tune E_gto 1.6 eV, but Cu reduces E_gmore efficiently than Ag. The amount of Cu will typically be in the range 5-15 atomic percent. A maximum value of n₂of about 700n₂(SiO₂) is expected. Similar results are expected for copper or silver-doped Ge—Se, but a larger dopant concentration is typically required.

[0064] The above-discussed dopants reduce E_gand thus can be used to achieve E_g=1.6 eV in chalcogenide glass having E_g>1.6 eV. On the other hand, partial substitution of sulfur for selenium increases E_g, and thus can be used to achieve E_g=1.6 eV in chalcogenide glass having E_g<1.6 eV.

[0065] FIGS. 3a-3dshow approximate values of some materials parameters (n₀, E_d, E_sand E_g) as a function of composition for the binary chalcogenide glass As_ySe_1−y, and FIGS. 4a-4dshow approximate values of the same parameters for the binary chalcogenide glass Ge_ySe_1−y. The data presented in the above recited figures will be useful in the practice of the inventions since it can facilitate choice of an appropriate base material for doping. Once an appropriate base material is selected, a minor amount of routine experimentation will generally suffice to find doped material according to the invention.

[0066] FIG. 5 schematically depicts an exemplary optical time division multiplexed (OTDM) fiber communication system 50 that comprises all-optical switches. Numerals 511-51n refer to transmitters operating at electronic bit rates, e.g., 10 Gb/s, with n being the number of channels, exemplarily 10. Optical signals from the transmitters are provided to optical multiplexer 52 and are interleaved in time. The interleaved optical data stream (exemplarily 100 Gb/s) is coupled into optical fiber 53 and transmitted therethrough to optical splitter 54, where it is split into n optical data streams. Each of the data streams is guided to one of optical demultiplexers 551-55n, where the data stream is optically demultiplexed. It will be understood that the demultiplexers 551-55n are nonlinear optical switches according to the invention. The demultiplexed optical data streams (exemplarily 10 Gb/s) are then provided to conventional n-channel receiver 56.

[0067] FIG. 7 schematically illustrates an optical switch 70 (more specifically an optical time division demultiplexer) according to the invention. Splitter 71 splits the OTDM data stream into two substantially identical data streams that propagate through a Mach-Zehnder-type device to output port 72. In the absence of a control pulse, arms 731 and 732 are of essentially the same optical length. Numeral 75 refers to a length of nonlinear waveguide according to the invention. In the presence of a control pulse the two arms of the Mach-Zehnder-type device are not of the same optical path length, but exemplarily exhibit a π phase difference, resulting in destructive interference at the output. Alternatively, the device could be designed with a built-in phase difference of π, with the presence of a control pulse resulting in a change in the optical path length of arm 731 such that the pulses constructively interfere at the output. For more detail on optical demultiplexing see, for instance, R. Ramaswami et al., “Optical Networks-A Practical Perspective”, Morgan Kanfman Publishers, Inc., San Francisco, Calif., 1998, especially pp. 515-531.

[0068] FIG. 6 schematically shows an exemplary planar waveguide 60 according to the invention. Numerals 61-63 respectively refer to the substrate (exemplarily Si), the nonlinear chalcogenide glass core, and the upper cladding (exemplarily a lower refractive index chalcogenide glass). Such a waveguide structure can be readily made by known methods, exemplarily including sputter deposition of a doped chalcogenide layer, patterning of the layer by photolithography and etching, and deposition of the upper cladding layer by sputtering, optionally followed by patterning by means of photolithography and etching.