Publication No: 1455Search all ORC publications    

Thermally bonded planar waveguide lasers

C.T.A.Brown, C.L.Bonner, T.J.Warburton, D.P.Shepherd, A.C.Tropper, D.C.Hanna
Optoelectronics Research Centre

Onyx Optics INC, 6551 Sierra Lane, Dublin, CA 94568 USA


A new technique for fabricating active planar waveguide devices is reported. This process, based on the thermal bonding of precision finished crystal or glass components, allows waveguides to be assembled from very dissimilar materials and could be applied to a wide range of solid state laser or other optical media. The waveguide propagation losses, inferred from the laser performance, are found to be 0.7dB/cm for Nd:Y3Al5O12 bonded to Y3Al5O12, Nd:Y3Al5O12 bonded to glass, and 0.4dB/cm for Nd:Gd3Ga5O12 bonded to Y3Al5O12 devices.

Rare earth doped planar crystal waveguides have the potential to yield efficient, compact diode-pumped lasers and optical amplifiers. The waveguide geometry is well matched to the output from diode bars and offers significant advantages for thermal management over similar bulk systems. Two methods for coupling diode bar outputs into planar waveguides can be envisaged; firstly, the focusing of the diode bar using appropriate optics to a beam waist which is compatible with the waveguide, and, secondly, direct butt-coupling of the waveguide to the diode bar. In principle, butt-coupling is simpler than using a focusing arrangement since one avoids a complex train of cylindrical focusing elements. On the other hand, butt-coupling requires the use of relatively thick waveguides (>10μm) with high numerical apertures (N.A) to accommodate the highly divergent light from the diode. Previously reported methods of fabricating crystalline planar waveguides have often relied on epitaxial growth (eg.[1]), which imposes limits on the choice of waveguide core and substrate material combination due to lattice matching considerations. Likewise, waveguide fabrication by the modification of a surface layer via ion-diffusion or exchange has so far been applied to only a few materials. On the other hand, guide fabrication by ion-implantation, which can be used for a wide range of materials, gives a rather limited index change and guide depth [2]. In general, large differences in refractive index between the waveguide and substrate are difficult to achieve with such methods, and as a result of the modification to the core material, the spectra of the rare earth ions can be significantly changed from those of bulk crystals [3]. Adverse spectral changes can partially offset the benefits offered by the waveguide geometry.

In this letter, we discuss the fabrication of waveguides by thermal bonding of the core and substrate layers. In this way very dissimilar materials can be bonded, even to the extent of a crystalline waveguiding layer bonded to a glassy substrate. Here, we exploit the technique for the fabrication of high N.A. (> 0.8) waveguides. The spectra of the bulk materials from which the guiding regions are fabricated are not affected by the bonding process and the propagation losses are low - less than 1db/cm even in guides of ~10μm width.

The thermal bonding technique involves the assembly by optical contacting of precision finished crystal or glass components and heat treating them to increase bonding strength. Repetition of this process can allow the fabrication of a composite of the given design. If heat treatment does not proceed up to temperatures where interdiffusion between adjacent optically polished surfaces occurs, a wide variety of dissimilar materials may be bonded without formation of significant stress. Van der Waals intermolecular attractive forces are thought to be responsible for holding the individual components together. The bond formed has sufficient strength to withstand all conventional finishing operations and deposition of optical coatings just like conventional single crystals or glasses. These composites are usually referred to as thermally or optically bonded. Since they do not require any bonding agents, reflections at interfaces are mainly due to differences in refractive indices between the components. Scattering loss is restricted to that due to imperfections in the optical surfaces of the individual components and, possibly, to subsurface damage. A phenomenological description of the bonding technique for two contacted ideal isotropic or nonisotropic solid elastic bodies in terms of force, stress distribution and the potential energy of the interface, relates their Poisson ratios and elastic modulii to their surface figures before forming the interface [4].

Surfaces with an ideal clean and smooth surface would form a strong bond at room temperature. Heat treatment of actual optically contacted composites is, however, usually required to remove absorbed gases and hydroxyl groups and so increase bond strength by decreasing the distance between adjacent surfaces. Conditions of heat treatment may vary widely for producing viable composites which are stable during subsequent device fabrication and operation. Their effects on device performance have not yet been fully explored. The results presented here represent an initial exploration of waveguide devices.

Table I: Details of the thermally bonded waveguides used in this study.
Samples 2 and 3 have no cladding layer.

SampleGuide MaterialGuide ThicknessGuideLengthSubstrate MaterialCladding MaterialN.A
11at% Nd:YAG n=1.816338μm0.4cm YAG n=1.81523YAG n=1.815230.06
21at% Nd:GGG n=1.9617μm1.9cmYAG n=1.81523-0.74
31at% Nd:YAG n=1.8163322μm1.4cmGlass n=1.62-0.82

Three different waveguide structures were investigated as shown in table 1. Sample 1 has a Nd:YAG core with substrate and cladding layers of undoped YAG, which is similar to structures which have previously been grown by liquid phase epitaxy [5]. The small index difference between the core and the substrate is provided by the 1at.% Nd doping leading to a low numerical aperture, N.A=(n2core-n2substrate)1/2 , of 0.06. Sample 2 has a Nd:Gd3Ga5O12 (Nd:GGG) core on a YAG substrate (no cladding layer), similar to waveguides previously fabricated by pulsed laser deposition [6]. This structure has a much higher NA of 0.74. Finally, sample 3 has a Nd:YAG core on a glass substrate (no cladding layer) giving an NA of 0.82. Nd:YAG is one of the most interesting materials for high power diode laser pumping and is, therefore, attractive for use as the waveguide core. However, in order to accept the NA of the emission from diode bars, the substrate material would then need to have an index of ~1.6. Thus the Nd:YAG on glass waveguide was chosen to satisfy this condition and to demonstrate that thermal bonding can work with these very dissimilar materials.

The purpose of the experiments described here is to characterise the optical properties of the waveguides, especially their propagation loss as this is crucial to efficient device operation. We have not yet tested their optimised lasing performance or under high power (multi-watt) diode pumping to see their response to the associated thermal load. However, it should be noted that the planar geometry is well suited to efficient one-dimensional heat removal.

A Ti:Al2O3 laser operating around 807nm was used to excite the Nd3+ ions from the 4I9/2 ground level to 4F3/2 excited state. Fluorescence and lasing behaviour were observed on the 4F3/2 4I11/2 transition around 1.06μm, using either a large area Si detector, with appropriate filters to block any residual pump light, or an OMA2000 diode array spectrometer for determination of the lasing wavelengths. The pump light was end-launched into the waveguide using a X5 microscope objective giving a pump spot size (1/e2 intensity radius) of ≈4μm. No attempt was made in these preliminary investigations to optimise laser performance by finding the best pumping spot size and shape (eg. elliptical) or the best waveguide length as these are not required for the method we have used to find the propagation losses. When needed, plane mirrors were attached to the end faces of the waveguide using the surface tension of a thin layer of fluorinated liquid. The mirror adhesion was, in some cases, further improved by gluing the edges of the mirrors to the waveguide substrates.

The threshold absorbed power needed to produce lasing, at 1.064μm for samples 1 and 3 and at 1.062μm for sample 2, using a variety of output couplings was then determined. The absorbed threshold power, Pthreshold, is given by :


where C is a constant incorporating material properties and the spatial overlap of the pump and signal radiation, l is the crystal length, α is the propagation loss coefficient and R1, R2 are the reflectivities of the cavity mirrors. Following the method of Findlay and Clay [7], a plot of Pthreshold/2l against -ln(R1R2)/2l gives a straight line whose intercept, Pthreshold=0, gives the value of the propagation loss coefficient. This method eliminates any dependence on the pump and laser geometry as long as they remain the same when output couplers are changed. The results obtained for the three samples are illustrated in figure 1. From these results we can deduce that samples 1 and 3 have propagation losses of 0.7dB/cm and sample 2 has a propagation loss of 0.4dB/cm. These propagation losses are comparable to that observed in a number of other methods of producing rare-earth doped waveguides and so, although it does not quite match the losses for YAG guides grown by liquid-phase-epitaxy (~0.1 dB/cm), this is, nevertheless, an encouraging result for possible devices. Measurements of the fluorescence spectra, absorption spectra and fluorescence lifetimes of these guides confirm, as expected, no observable difference from bulk material. In order to further test the laser behaviour, the slope efficiency was measured for each device using an output coupler with a nominal transmission of 3.5%. This transmission is affected by Fabry-Perot effects between the endface of the waveguide and the output coupler so that, depending on the spacing of these surfaces the transmission of the output coupler could be up to ~6.2% for samples 1 and 3 and ~6.7% for sample 2. The results obtained are shown in figure 2 and demonstrate slope efficiencies of 16%, 13% and 15% with respect to absorbed power for samples 1, 2 and 3 respectively. In the case of samples 1 and 2, these efficiencies are consistent with the higher transmissions indicated above. The slope efficiency obtained for sample 3 suggests, taking into account the higher mirror transmissions, a slightly lower propagation loss than that obtained using the Findlay-Clay method (~0.5dB/cm). With some mirror sets, multiple line lasing was observed from samples 2 and 3. This lasing appeared to be simultaneous to within the 30ms scanning time of the OMA2000. Sample 2 was observed to lase at various lines from 1.054μm, 1.059μm, 1.061μm, 1.062μm, 1.105μm, 1.107μm and 1.110μm depending on the pump light launch condition and power. Sample 3 was observed lasing at 1.052μm, 1.064μm, 1.074μm, 1.078μm and 1.116μm. This behaviour is not yet fully understood, although the Fabry-Perot effects discussed above may play a role. Care had to be taken, therefore, when obtaining the thresholds for the propagation loss analysis to use mirror sets which only permitted lasing on one line (1.062μm for sample 2 and 1.064μm for sample 3.)

Figure 1: Modified Findlay-Clay plot for samples examined. Position of the intercept on the x-axis gives the value of α, the propagation loss coefficient. This plot allows losses of 0.7dB/cm, 0.4dB/cm and 0.7dB/cm to be determined for samples 1, 2 and 3 respectively.

Figure 2: Output power at 1.06μm versus absorbed pump power for the three samples.

Despite the highly multi-mode nature of the high NA guides, the laser could be made to operate in a single mode in the guided direction for all of the samples depending on the pump launch condition. As expected, the mode was well confined within the waveguide core for the high NA guides, whereas for the low NA guide, a significant proportion of the mode propagated within the cladding and substrate layers. For all of these guides, multimode output was observed in the non-guided direction. Single mode operation in both directions may be obtainable by optimisation of the pump launch along with the use of shorter waveguides. In order to provide a check on the N.A. values of these guides, the guide transmission was measured versus the angle between the guide plane and the propagation axis of the launched beam. The results are shown in figure 3 and demonstrate the insensitivity of the high NA Nd:YAG on glass and Nd:GGG on YAG guides to this angle compared to the low NA Nd:YAG on YAG guide.

Figure 3: Transmitted Ti:Al2O3 power as a function of the angle of the end face of the waveguide relative to the propagation axis of the input light.

In summary, we have demonstrated, for the first time to our knowledge, laser action in waveguides fabricated by a thermal bonding process. The waveguide material is then identical to bulk material. This process produces waveguides with propagation losses which are typically <1dB/cm. The ability of the thermal bonding process to produce guides with high numerical apertures has been demonstrated as has the ability to bond a crystalline laser host matrix to a glassy substrate. The fabrication process appears to offer great promise in the fabrication of planar waveguide devices and it is hoped that it can allow the realisation of highly compact diode bar pumped lasers and amplifiers capable of multi-watt output powers. This technique may also find uses with non-linear materials where waveguiding can increase conversion efficiency for low pump powers.

The Optoelectronics Research Centre is an EPSRC interdisciplinary research centre.

1. B. Ferrand, D. Pelnec, I. Chartier and Ch. Wyon, Journal of Crystal Growth, 128, 966, (1993).
2. L. Zhang, P.J Chandler, D.C Hanna, S.J Field, D.P Shepherd, P.D Townsend and A.C. Tropper, J. Appl. Phys., 69, 3440, (1991).
3. D.C. Hanna, A.C Large, D.C Hanna, S.J Field, D.P Shepherd, I Chartier, B. Ferrand and D.Pelnec, Appl. Phys. Lett., 63, 7, (1993).
4. L.D Landau and E.M Lifshitz in Theory of Elasticity, 3rd Ed, (Pergammon, Oxford, 1986)I.
5. Chartier, B. Ferrand, D.Pelnec, S.J Field, D.C Hanna, A.C Large, D.P Shepherd and A.C Tropper, Opt. Lett., 17, 810, (1992).
6. C.L Bonner, A.A Anderson, R.W Eason, D.P Shepherd, D.S Gill, C. Grivas, N. Vainos, , accepted Opt Lett, (1997).
7. D. Findlay and R.A. Clay, Phys. Lett., 20, 277, (1966).

Applied Physics Letters (1997) Vol.71(9) pp.1139-1141

doi: 10.1063/1.119846

Southampton ePrint id: 77958




Copyright University of Southampton 2006