Modern telecommunications rely on the transmission of light signals along fibre-optic cables — fast, but how to minimize signal degradation over large distances? The secret is to eliminate contaminants that cause unwanted absorption. These intruders tend to creep into the fibres during the manufacturing stages. Elsewhere in this issue ( Nature 404, 262–264; 2000), Gordon Thomas and colleagues reveal how the main culprit, water, sneaks in. Putting this knowledge into practice allows the manufacture of dry fibres with near maximal transparency and improved bandwidth.

The problem with water, in particular the hydroxyl (OH) group, is that it gets excited when irradiated by certain infrared wavelengths. As this band is used for telecommunications, precious signal power can be lost in vibration of the water molecules.

So where does the water come from? One process for making optical fibres involves heating silica rods with ultra-pure glass cores to very high temperatures. Once softened, the rods can be drawn out into fibres some tens of kilometres long. To achieve the necessary temperatures — in excess of 2,000 °C — it is common to use torches that burn hydrogen and oxygen: the perfect recipe for water.

But the incorporation of water into the silica rods is a more subtle matter. To investigate this process, Thomas and colleagues measured the transmission of infrared light through a small section of rod that had been cooled before drawing out, thereby ‘freezing in’ the water. They found that the absorption, and hence the hydroxyl concentration, is much stronger in the outer layers of the rod. This can be seen in the figure, which indicates the varying hydroxyl concentrations through a cross-section of the rod. (The colour scale is logarithmic, blue representing the highest concentrations.) Clearly, the water diffuses from the outside in.

Given that the signals in optical fibres are confined to a narrow core region, the distribution of water in the rod might appear to be good news. But Thomas and colleagues found that the drawing process — which increases the aspect ratio by a factor of about a hundred million — lets the water in much further. They confirmed this by calculating the diffusion coefficient, a quantity that describes the flow of both water and glass during the contamination process. This parameter was much higher than expected from low-temperature diffusion, probably reflecting an increased mobility of hydroxyl groups in the hot, molten state.

There is no doubt that absorption by hydroxyl groups contributes significantly to transparency loss in optical fibres. Uncovering the physical origin of the contamination points to an obvious solution: pick a water-free heat source.