Light travels in straight lines?a physical simulation of light propagation in a graded index optical fibre ~ M G Cornwall Largescale demonstrations of the propagation of a HeNe laser beam in graded Index opHcal medla have been developed, using interditfusing liquid isyarn. The sinuous paths Of a ray in a GRIN optical fibre and lenses have been simulated. of physical demonstrations of light propagation in graded index media. The physical simulations are achieved by using various liquids to create different refractive index profiles within a rectangular tank. A concept once treated as an interesting but peripheral topic in introductory physics-total internal reflection-has become the basis of the fibre optic communications revolution. Whilst modern A-level and College physics textbooks now mention this technological application of an old principle, the simple idea of a light ray ‘bouncing’ along inside a light pipe is usually the limit of the explanation given. In fact, there are relatively few actual fibre optic applications for which this simple model of light propagation is correct. Furthermore few, if any, simple demonstrations of the way in which light actually travels in fibres seem to have been developed. The simulations described here are the result of a series of undergraduate projects carried out by second- and final-year students in the Department of Mathematical Sciences at Brighton Polytechnic over several years. Originally, computer programs using simple ray-tracing techniques were developed to demonstrate the properties of stepped and graded index fibres. However, whilst computer simulations are excellent ways in which to model the real world in order to illustrate principles and to try out ‘what if‘ experiments, they lack the credibility of a real demonstration or experiment. We therefore developed a complementary package Beam propagation in a graded index medium Malcolm Cornwall is Principal Lecturer in charge of Physics at the University of Brighton (formerly Brighton Polytechnic) and a member of the Editorial Board for this journal. He often addresses schools and colleges on optics education. holography, lasers and related subjects. M31~912019LX15m73t07 104.5001992 IOP Publtshmg Lld The first and simplest arrangement was intended to demonstrate the bending of a ray in a medium with a continuously varying index. A layer of sugar is placed at the bottom of a large rectangular glass or plastic tank, covered with water and allowed to diffuse for several days. If sufficient sugar is used, a solid layer will still be present at the bottom of the tank, with a gradual merger between solid and solution, creating a ‘graded index’ medium. Sugar is used because, besides being a cheap and completely benign ‘chemical’, it has an extremely large solubility in water and produces large changes in refractive index, n. Moreover, values of n as a function of sugar concentration are listed in standard physical data handbooks (e.g. The Handbook of Chemistry and Physics). When a 0.5-1.0 mW HeNe laser beam is directed into the tank so as to travel through a relatively strongly graded index region it is observed to bend in a very obvious way. Figure 1 is a multiple exposure photograph showing the laser beam paths for several incident angles on the left-hand side of the tank. We have found that even for experienced physicists this bending is quite startling at first sight, so familiar are we with the ‘fact’ that, on a macroscopic scale, light rays always travel in straight lines. The equation for the path of a light ray in an optically inhomogeneous medium with a refractive index gradient in they direction given by n(y) and a constant index in the x direction is d2y/dx2=[l/(2n: sinzO0)]dn2(y)/dx 273 Flgure 1. Propagation of a H e N e laser beam in a graded index medium. The photograph shows an approximately 50 cm long 'sugar tank' (seetext) with a laser beam launched from t h e left into t h e tank at different heights. (The bright area at the bottom of t h e t a n k is the undissolved sugar layer.) where n, is the refractive index of the medium at the point of entry of the ray, and 00 the corresponding angle of incidence relative to the vertical. Note that the curvature of the ray is proportional to the gradient of the square of the refractive index function (Ghatak and Thyagarajan 1978). The bending, being related to the gradient ofthe refractive index, is very much more P " m c e d near to the interface between the undissolved solid sugar and the solution. Indeed, significant bending can be observedclose to this interface just a day or so after the water layer is added. An initial problem with the sugar solution was that the region with the highest index gradient was also optically diffuse and caused unacceptably large scattering of the beam and hence loss of beam definition. The use of chemically pure sugar in the form of high purity sucrose, and distilled water, avoidd this problem, H ~the absence ~ of impurities in the solution then produ& the opposite problem of insufficient scattering centres to make the beam visible except in virtual darkness. A 'beam visualizing' additive is a small drop n4 of white emulsion paint, which contains Ludoxt, a major component of some types of paint. Very small samples of the solution were taken with a fine capillary tube at various depths in the tank, and the refractive index measured with an Abbe refractometer. The index profile was found to be exponential within the experimental error. (The solid-liquid diffusion in the tank would, theoretically, be expected to give a 'complementary error function' (erfc) dependence of index on distance from the interface (Jost 1960); this function differs little from an exponential). This simulation incidentally illustrates the principle of the mirage, as observed in desert conditions (or, more mundanely, above a road on a hot day). manufactured by DuPont, consists of ~t This material, ~ , a suspension of extremely small 1-12 nm) particle of amorphous silica, each having a negative charge, such that they are maintained in suspension in a suitably buffered me,jium,with an approximately unifom den. sity within the suspension. The 'sugar tank' is simple to set up but requires a large tank and considerable waiting time (although this can be reduced by using warm water initially). we have fOund that an equally effective graded index medium can be produced by carefully floatink? 2-3 Cm Of isopropyl alcohol (IPA) on water. Beam visibility is greatly enhanced if Ludox-doped water is used (Ludox flocculates in alcohol). Interdiffusion of the liquids is quite fast, and within a few minutes a beam launched into the region of the alcohol-water interface is strongly refracted upwards even in a relatively short tank about 10 cm in length (figure 2(d. As the refractive index is 1.38 and is greater than that of water (1.33) the index gradient has the opposite sense from that in the sugar tank; the beam is therefore bent upwards as in a mirage. (In general. rays bend towards the layer of higher refractive index.) The dependence of the beam Cw-"Ure on the position at which the beam enten the tank. is illustrated in figure 2@). This shows the appearance of a 'fan' of rays incident on the tank such that *e top ofthe fan travels through a region Of high index gradient near the liquid interface and has a high curvature, whilst the lower part of the fan travels through almost pure water and is fan was p r o d u d by virtually undeflected. pe reflec.ing the laser beam from a small mirror attached to a loudspeaker driven at a few tens of hertz ( ~ ~ and~williams ~ ~lggl).) 1 1 If the laser beam is launched into the IPA layer at a suitable angle immediately after it is poured Onto the water, the beam can be made to 'bounce' &tween the water-air and water-rpAinterfaces. The beam is simply totally internally reflected at the upper surface, but is much mOre gradually curved at the liquid interface, due to the initial index gradient that exists as the two liquids begin to interdiffuse. me gradual d i r e c t i o n of the km observed here is the basic mechanism by which a graded index optical fibre operateS. Several other liquids can be used instead of IPA, but the only other relatively benign solvent we have tried that produces interesting results is ethanol, which we use in the form of industrial methylated spirit (IMS).This also floats on water and gradually diffuses into it, but it has a refractive index of 1.325 which is slightly lower than that of water. As a result a laser beam bends downward when it is launched into the tank. Because of the small difference in the refractive indices and the smaller solubility of IMS in water, the bending is less dramatic and can be seen only by careful manipulation of the incident beam height and angle. Graded Index fibres Flgura 2. (a) Path of a HeNe laser beam launchedfrom the right in a 10 cm long tank in which a 3 cm layer of iSOprOpYl alcohol (IPA) has been floated onto water. ( b )A 'fan' of rays, producedas described in the text. enteringan IPA tank from the left. The uppermostpart of the fan is strongly refracted:the lowest part of the fan traVeiS through almost pure water in a straight line. The success of the sugar and IPA tanks suggested the desirability of producing a physical simulation of a complete graded fibre, by using three layers of suitable refractive index and of suitable density SO that they would float on one another and gradually interdiffuse. Within such a structure the sinuous path of a light ray in a graded index fibre could be shown. A literature search produced a reference in R W Wood's classic book Physical Optics (Wood 1935) to a much earlier paper by Wood (1899), in which he describes a simulation of a mirage, and of a light beam in a layered medium. We found, however, that the solutions used by Wood were not as effective as the following sequence of solutions: Bottom layer: concentrated potassium alum solution. Middle layer: a mixture of 30% glycerol in IPA. Top layer: a mixture of 30% water in IPA. 275 The relative thicknesses of the layers are not important, but the middle one should be about 3-4 cm thick for a tank of about 50 cm in length. The layers need to be poured very carefully on top of one another and then left for some time. The three layers diffuse slowly into one another and we have found that at room temperature the optimum conditions are produced after about 18 hours. The layers need to be poured very carefully on top of one another, to avoid premature mixing. We have used a simple ‘pourer’ in the form of a piece of paper bent into ‘U’ shape, and held so that the bottom of the ‘U’ just touches the surface of the liquid layers as they are added. Launching a laser beam horizontally into the tank at various heights produced the beam paths shown in figure 3. The beam paths are apparently sinusoidal in shape and all have about the same periodic length. This is the path produced by a so-called ‘parabolic’ index gradient, which is approximately the gradient created in graded index fibres. (In such a fibre, the propagation time for any ray path along the fibre is (very nearly) the same, so that there is little ‘intermodal‘ dispersion due to the different times taken by rays travelling along different paths, which occurs in a step index fibre. So in a graded index fibre, digital signal pulses are transmitted with little distortion over long distances.) If a beam is launched into the tank immediately after it is set up, the beam ‘bounces’ between the interfaces, simulating the propagation of a ray in a stepped index fibre. However, due to optical distortions near the interface between the liquids the beam looks sinusoidal. In addition, the beam visually appears to be trapped in the interface layer for a significant distance before being totally internally reflected. To minimize the visual distortions the beam should be launched very close and parallel to the side wall of the tank through which it is observed. Using the technique mentioned earlier, we have measured the value of n as a function of depth after 18 h, and have found a variation from 1.342 near the top of the layers to 1.359 near the bottom ofthe tank, through a smooth maximum of 1.384 in the middle of the central layer. It is perhaps surprising that such small variations in n are nevertheless capable of producing such very obvious undulations in the beam path. (In fact for actual fibres the variation can be well under I%.) Unfortunately, it is not possible to use Ludox to enhance the beam visibility, because the colloid comes out of suspension in the alum and produces a floccular precipitate. We have had some success with other combinations of liquids within which Ludox is stable, and also using other visibilityenhancing techniques. 276 An improved fibre simulation It would obviously be desirable to develop a ‘three-layer’ simulation, which has a much larger change in refractive index across the layers, and hence produces larger and more obvious undulations of the laser beam and, ideally, with a much shorter ‘waiting time’. The criteria for the choice of liquids is that, if their densities are d,, d2 and d3 and their indices n , , n2 and n3, then we require that d , > d 2 > d 3 and n l < n 2 > n , , with the difference between n2 and the other two indices being as large as possible. In addition the liquids should be mutually miscible, should be able to be handled without serious hazards and should ideally be inexpensive. These criteria proved to be remarkably difficult to satisfy. An extensive literature search and experimentation resulted in the following set of liquids, which most nearly match the requirements: Bottom layer: dichloromethane. Middle layer: cinnamaldehydet . ethanol (IMSis suitable). Top layer: A 5 cm layer of IMS, a 1-2 cm layer of cinnamaldehyde and a 5 cm layer of dichloromethane were successively poured into a tank about 15 cm in length using the ‘U’ shaped pourer. The width of the tank should be small-we used one about 2 cm wide-to reduce the required volumes of the liquids. To make the beam clearly visible a few crystals of rhodamine 6G, a fluorescent dye, were dissolved in the cinnamaldehyde before it was added to the tank. The laser beam is then visible as a bright pinkish red fluorescence. The three layers in this tank need to be allowed to interdiffuse for a few minutes for optimum performance. Measurement of the refractive index as a function of depth immediately after setting up the tank gave a value of n varying smoothly from 1.384 near the surface, through a maximum of 1.598 in the middle layer to 1.470 just above the bottom of the tank. The waiting time, the most effectivethickness of the middle layer, and the best angle and position of incidence of the laser beam are interdependent. Typical beam paths are shown in figure 4(a) and are easy to produce, but even more ‘wavelengths’ can be produced by optimization of the variables. Although fewer ‘waves’ are t Cinnamaldehyde has an extremely strong and linger- ing smell (of cinnamon) and is classifiedas an ‘irritant’. A few people may be allergic to the vapour, and it is desirable therefore to handle the liquid with care and in a fume cupboard. It therefore does not satisfy one of our criteria. However, all other high index candidates for the middle layer ofwhich we are aware are more hazardous. propagation of a laser beam in t h e three-iayer 35 cm long tank simulating a two-dimensional graded index fibre (seetext). This iS a multiple exposure photograph showing the paths for a horizontal beam incident from the leftat five different heights. ngure 3. The wave-like visible here than in figure 3, it should be noted that the tank used for figure 4 is very much shorter and the curvature of the laser beam much greater, due to the very large index gradients produced in the ‘cinnamaldehyde’tank. By scanning the incident beam as described earlier, a fan of rays can be launched into the simulated ‘fibre’ as shown in figure 4(b). This is a more realistic simulation of the actual state of affairsin a fibre. If the laser beam is directed into the top layer it behaves like a ‘cladding ray’ (i.e. a ray that initially travels in the outer, lower index layer in a fibre) and is not properly propagated in the ‘core’ layer. Figure 4(c) shows that such a ray in fact initially diverges as it enters the ‘core’ and then reconverges as it enters the bottom ‘cladding layer’. Deficiemies In the simulations The most obvious deficiency in the physical simulations is that they are two-dimensional representations of a three-dimensional process, and cannot represent the more complex ‘skew’ modes that can occur in real fibres; all the ray paths simulated are in effect meridional, i.e. they are restricted to a plane passing through the axis of the ‘fibre’. The simulations are based on a ‘ray’ model of light propagation. This means that they cannot represent mode selection, mode conversions and evanescent modes-phenomena that are important in fibres, and which are essentially a result of the interference properties of waves. In particular the model cannot simulate the characteristics of a single-mode fibre. A~ interesting theoretical COntraStexists between a computer simulation and the physical simulations. The former is rigidly a ‘ray’-based model and illustrates directly the inadequacy of that model; a ray that is incident exactly parallel to the fibre axis will propagate along a path of constant index and will therefore not suffer any refraction. This is the caSe not only for a ray originally incident in an axial direction but alsp for any ray that is eventually refracted into an axial direction; such a ray will not experience any further ‘bending’ k a u s e it is then travelling in a constant index layer. This obviously fallacious conclusion from the ray model is demonstrated by the physical simulation, in which we are dealing with a real light beam with a finite width, which is of course refracted even when it is incident exactly normal to the index gradient. In a computer model the inadequacy of the ray model is disguised by the fact that the fibre is usually simulated by a set of finite layers of constant index; only if by chance a ray is incident at exactly the critical angle at a layer interface will it be refracted parallel to the layers. With the precision to which a computer operates, this is extremely unlikely and the ray is either completely reflected or refracted. The effect of the finite width of the laser beam is also obvious when the beam propagates through a region of sufficiently rapidly varying index to smear out the beam by bending different parts of it to varying extents (see figure 4(c)). Fu*rhvelopmmh We have investigated other combinations of liquids with larger index differences, but none has m Figure4. HeNe beam paths in the 'cinnamaldehyde' tank (seetext): the beam is visible as a resuit of t h e fluorescenceof Rhodamine 6G dissolvec in t h e cinnamaldehyde layer. ( a )Beam launched horizontally from t h e iefl into t h e middle layer. The tank length is onlyabout 15cm.comparedwithabout50cm in figure3 ( b )An approximately parallel fan of rays is incident from t h e left. The initially 'collimated' beam is 'focused' and 'defocused as it propagates. (c) The path of a beam launched into the upper 'cladding' layer. Note that it first diverges and then reconverges a s it enters and leaves the 'core' layer. been as effective as the system described above. We are also looking a t various possibilities for producing 'solid versions of the tanks, which would have the virtue of permanence and threedimensionality. A long solid Perspex (Plexiglass) tube ahout 2 cm in diameter can be used to show skew modes, but the beam is not easily visible. Much more effective is a long hollow glass Or Perspex tube, filled with Ludox-doped water. A suitably launched laser beam clearly shows the characteristic skew paths of non-meridional rays in a stepped fibre' In tbis caSe total internal reflection takes place a t the plastic (or dass)-air interface, and the weaker reflection a t the water278 solid interface is strong enough to cause a smeared appearance of the beam a s it propagates along the tube. A relatively short glass tube filled with cinnamaldehyde doped with Rhodamine 6G shows skewed modes much more dramatically. Because the liquid has a much higher refractive index than glass, not only does TIR takes place at the liquidsolid interface, resulting in a much 'cleaner' beam, but the beam is 'trapped' even for high incident angles (i.e. the simulated fibre has a very large numerical aperture). Wood (1935) described the in-diffusion of a sugar solution into gelatin a s a way of producing what we would now call a graded index (GRIN) lens. GRIN lenses-also called sewocs-usually consist of a short length of graded index rod, having a length of either 0.25 or 0.27 times the 'period' of the sinusoidal path (a few mm) within the fibre. These two lengths produce, respectively, a 'lens' that will collimate a diverging beam, or a 'lens' that focuses an initially parallel beam. In two dimensions we have simulated the operation of a GRIN lens by using a small rectangular glass 'cell' (as used for holding liquid samples in spectrometry) standing inside a longer tank. By filling the cell with Ludox-doped water and moving it to different positions along the beam path, the focusing or collimating action of a GRIN lens is graphically demonstrated (figure 5). The growing importance of fibre-optic-based communications and sensor technology, and the emergence of integrated optics and optical computing as the next innovative wave, suggest that our students should have more than a peripheral understanding of the optical principles involved. These simulations provide one very graphic means of demonstrating how light can be propagated in Figure 5. A demonstration of t h e operation of a collimating GRIN lens. The fan of rays is incident from t h e right. and a Small cell (filled with Ludox-doped water) is placed at a n appropriate position to show that a parallel beam can be produced from a point source, for a GRIN lens of length equal to0.25 'wavelengths'. The main tankcontainsa thre&ayer 'cinnamaldehyde' fibre simulation. real fibres and some of the characteristics of these fibres. There is undoubtedly considerable potential for further development of the ideas described here. In particular, it would be very desirable to find more pleasant and less expensive alternatives to cinnamaldehyde. Development of solid threedimensional simulations, perhaps along the lines suggested by Wood, would provide a more permanent and convenient teaching aid. Acknowledgments I am indebted to my students, notably Mark Greenway Stanley and Yanik Le ROY, and to the Technical Support team of my department for their invaluable practical contributions. Ghatak A K and Thyagarajan K 1978 Conremporury Oprics (New York: Plenum) Handbook of Chemisrry und Physics 54th edn (Boca Raton, FL: CRC) Jost W 1960 Diffusion 3rd edn (New York: Academic) Wood R W 1935 PhysicdOptics (London: Macmillan) -1899 Phil. Mug. Further reading Halley P 1987 Fibre Optic Sysfems(NewYork: Wiley) lizuka K 1986 Engineering Oprics 2nd edn (Berlin: Springer) Jones K A I987 lnrroducrion IO Opricol E/ecrronics (New York: Wiley) ZangerHandZangerC 1991FibreOplic CommunicarionsundOrher Applicurions (NEW York Merril) References Cornwall M G and Williams G T 199I Laser I Applicuriom in Science Educarion: U Teacher's Hondbook (Brighton Polytechnic: LASE Publications) 279