Both the classic and optosource lamphouses use ellipsoidal reflectors, but the optical arrangements are significantly different. In the Classic lamphouse, the lamp is aligned along the optical axis, with the arc at the first focus of course. Ours is a standard implementation of a very commonly-used arrangement, which is shown in Figure 1.



To appreciate why it is so popular, you just have to be aware of the spatial characteristics of the output from an arc lamp. The arc emits light radially in all directions, but the electrodes begin to obstruct the light at angles of less than about 30 degrees away from the optical axis. This means that the Figure 1 arrangement collects and focusses most of the available light from the arc, even though the reflector is an incomplete optical surface (since it necessarily has openings at both ends!), so it is clearly unbeatable in this respect.

However, there are two disadvantages, one theoretical, and the other practical. The practical one is easier to appreciate, which is that for the usual requirement of a horizontal output beam, the lamp must also be mounted horizontally, whereas most xenon arc lamps are specified only for vertical operation. However, some lamps such as the UXL75 series from Ushio which we routinely supply, are also specified for horizontal operation, and we're pleased to confirm that they do so without any problems.

The theoretical disadvantage may not be immediately obvious, but the form of the out-of-focus image gives a clue. Instead of a diffuse beam, one sees a very characteristic "doughnut" shape, because as Figure 1 shows, there is no light converging at angles close to the optical axis. Because of this central dark region, the light intensity at the second focus is reduced in comparison with an unobstructed optical system, where light would be converging at all angles within the beam limits. This may sound paradoxical, because in principle most of the available light is being collected, but the reason is that this configuration has a relatively high magnification. In other words, the high light collection efficiency gives an image which is bigger rather than brighter. Specifically, the output aperture is f/2.5 and the magnification is approximately 24, which is a relatively high figure as we shall see.

For highest intensity at focus, lens-based systems are normally chosen, since they do not suffer from this obscuration problem. The lamp is usually mounted vertically for convenience, and a condenser lens is used to collect light over (usually) a wide angle, to produce a collimated output beam.



This can then be brought to a focus by a second lens, as shown in Figure 2. Of course, the two lenses could be combined into a single one, or they could actually be a multilens assembly, but the Figure 2 arrangement is often adopted in practice because it is so convenient. The output f number of the system is set by the second lens, and the magnification is set by the ratio of the focal length of this lens relative to that of the condenser lens. For highest point intensity at focus, the second lens should operate at as low an f number as possible, although in practice a limit is usually set by the subsequent optics, which would need to work at a similarly low f number in order to utilise all the focussed light.

Given this limit, why is the first (condenser) lens such a critical component? The reason is that for a given output f number, it sets the magnification of the system. This is a very important consideration, as the brightest part of a xenon arc is relatively small. For a 75W xenon lamp, the arc length is typically around 1mm, but most of the light is generated within about 0.1mm of the cathode, and to a first approximation the arc can be considered to be a sphere of about 0.1mm diameter. For a given diameter, the focal length of the condenser lens should be as short as possible in order to achieve the largest possible spot size at the output focus, up to the point at which it completely fills any aperture located there - in our case the monochromator input slit. In practice, the maximum magnification that can be achieved with an f/2.5 output from a condenser lens system is about fourfold (using an f/0.6 condenser). This means we're still unlikely to overfill the input slit of the monochromator, which is typically around 1mm wide for instruments such as our Optoscan. On the other hand, the 24fold magnification of the Classic light source is rather too high by this criterion.

But why not just use a more powerful xenon lamp instead? More powerful lamps have bigger arcs, and hence we wouldn't need so much magnification to fill the input slit of a monochromator. However, it turns out that the point intensity within the arc is lower for more powerful arc lamps, so for a given output f number and focussed spot size, it is better to use a less powerful lamp at a higher magnification. One trick which is often used to increase the effective magnification is to place a spherical reflector behind the lamp, to produce an image of the arc next to the arc itself. In theory this doubles the optical output, but in a practical system the improvement is nearer 50%, because the image light has to be reflected AND must pass throught the lamp envelope twice more. But whether or not a back reflector is used, it is best to use the fastest possible condenser lens, which in principle is no problem as such lenses are available at apertures up to about f/0.6. The only problem is that spherical aberration is an enormous problem at such large apertures, which necessitates using an aspheric lens here. Aspheric condenser lenses for use with visible light are readily available, but silica aspherics for use in the UV are prohibitively expensive, and in a wideband optical system there is also the problem of chromatic aberration to contend with.

Mirrors suffer from neither of these problems, so we were keen to stick with them if at all possible. The first alternative we investigated was to use an ellipsoidal mirror OFF-AXIS, as shown in Figure 3. By using an appropriately-shaped mirror, it is possible to collect at least as much light as a condenser system, but without the lamp obstructing the output beam. Furthermore, it is possible to use a spherical back reflector with this configuration as well, and the lamp can also be mounted vertically



This arrangement works well, and we are using it very successfully in a custom application using a deuterium lamp for UV illumination around 200nm (although use of a back reflector is not appropriate here). However, we weren't entirely happy with its performance when coupled to the monochromator. The asymmetric optical pathway gives the out-of-focus beam a rather disconcerting appearance, and the back reflector gives little if any useful improvement in this application (trying to locate the secondary arc image close enough to the primary one seems to cause a further intensity loss). It is also not a very easy system to set up, but nevertheless it IS an improvement over the Classic light source. So where next?

The answer turned out to be very simple! For a symmetrical optical pathway, we need to use an on-axis system, in which case the lamp will obstruct it again. However, we reasoned that if we used a relatively larger ellipsoidal mirror, sited relatively further away from the lamp, the obstruction would be relatively less. Figure 4 should make this clear.



In practice we lose only a few percent of the light in our chosen configuration, and this is more than compensated by the high light collection efficiency. The mirror in the Optosource collects light over a range of +/- 70 degrees, which works out to be f/0.18, or a numerical aperture (given by the sine of this angle, and which is a more commonly quoted figure of merit for very fast optical systems) of 0.94. This gives an output beam of f/2, which we chose because the monochromator can just about accept it if we fit a slightly larger diffraction grating. For comparison with the Classic light source, if we notionally reduce the output aperture to f/2.5, this reduces the light collection range to +/- 59 degrees, equivalent to f/0.3 or a numerical aperture of 0.86. These figures would be difficult if not impossible to match with a condenser lens system, especially in the UV.

Although the Optosource does not collect as much light as the Classic arrangement - we provisionally estimate that it collects about 1/6 as much (a more definitive measurement will be provided soon) - the intensity at focus is higher because the magnification is less. Whereas the Classic light source has a magnification of 24, the Optosource has a magnification of only 6, which means that it concentrates its light into only 1/16 of the area, so the intensity per unit area at focus is more than twice as high when the Optosource is compared at f/2.5, with a further improvement of about 50% if we use the full output aperture of f/2. Under conditions where the focussed image completely fills the input slit of the monochromator, the Optosource therefore gives a significantly higher optical throughput. However, this comparison is complicated by the fact that our Optoscan monochromator has variable slit widths to allow users to set the optical bandwidth, so we cannot quote a single figure here. What we can say though is that for optical bandwidths of up to 20nm, the Optosource will give at least twice as much throughput.. At higher bandwidths, the input slit width becomes greater than the image of the (brightest part) of the arc in the Optosource, so the improvement is not quite so great, but it is still useful.

Finally though,.we'd like to put things into perspective! Even with the Classic light source, light intensities at the experimental sample can easily be high enough to cause photobleaching and other damage within only a few minutes, so neutral density filtering is often needed to reduce such problems. Clearly there is no point in using the Optosource if this is already a problem. However, for high-speed imaging applications, where the sample is likely to be illuminted only for relatively short periods, the Optosource is definitely the light source of choice.