IELTS Academic Reading Practice Test 155

The magnetic resonance imaging (MRI) scanner has become a familiar sight in hospitals: a large white cylinder into which the patient slides on a motorised bed. It produces pictures of soft tissue — brain, muscle and ligament — that X-rays cannot show clearly, and it does so without exposing the patient to ionising radiation. Yet the technique grew out of an experiment in physics that had nothing to do with medicine, and its development was driven by a handful of researchers whose training was in physics rather than in clinical practice.

In 1946 two American physicists, Felix Bloch at Stanford and Edward Purcell at Harvard, independently detected a phenomenon that became known as nuclear magnetic resonance. Certain atomic nuclei, including the single proton that forms the nucleus of ordinary hydrogen, behave like tiny magnets. Placed in a strong magnetic field, they line up with it; struck by a pulse of radio waves at exactly the right frequency, they absorb energy and then release it as they relax. The frequency depends on the strength of the field and on the element, which makes the response a kind of signature. Bloch and Purcell shared the Nobel Prize in Physics in 1952.

Chemists quickly adopted the technique for identifying molecules, because the precise frequencies at which nuclei responded revealed the structure of a compound. The other useful piece of information was timing. Once a nucleus had been excited, the period it took to relax depended on its surroundings, so the same nucleus could give different signals in different chemical environments. It was this observation that eventually suggested medical uses.

In 1971 Raymond Damadian, an American physician, reported that the relaxation times of hydrogen nuclei in certain rat tumours differed from those in healthy tissue. He proposed that the technique might be used to detect cancer and filed patents covering the idea of scanning the human body with magnetic resonance. His work convinced many people that a medical application was possible, but it did not show how to produce an image. A signal gathered from a whole region of tissue gives a single number, not a picture.

The crucial step was taken by Paul Lauterbur, a chemist then working at the State University of New York at Stony Brook. In 1973 he published a method in which the magnetic field was deliberately made uneven, with a gradient from one side of the sample to the other. Because the resonant frequency depends on field strength, nuclei at different positions then responded at different frequencies. A spectrum could therefore be decoded to give the distribution of hydrogen along the gradient, and by taking such measurements from several directions Lauterbur built up an image. He called the result zeugmatography, a term that never caught on.

Peter Mansfield, a physicist at the University of Nottingham, developed the mathematics further and found ways of making the process much faster. His work on switching gradients rapidly while the signal was being recorded turned a slow laboratory procedure into something that could be completed in a practical scanning time. Speed mattered because a scan of a moving patient is blurred, and a patient who can hold still for only a few seconds can still be examined. Mansfield also described echo-planar imaging, which in principle captures an entire picture in a fraction of a second and later made it possible to watch the brain at work.

The first image of a human being was produced in 1977 by Damadian's team, using a machine he had built himself and named Indomitable. The patient was placed in a large superconducting magnet, and the resulting picture of the chest took hours to acquire. Commercial scanners followed in the early 1980s. Early models were marketed as NMR machines, but the word 'nuclear' alarmed patients, who confused it with radioactivity. The technique was renamed magnetic resonance imaging.

Modern scanners work at magnetic fields of 1.5 or 3 tesla, tens of thousands of times stronger than the Earth's field, and the magnet is kept at superconducting temperatures by liquid helium. Three sets of gradient coils allow position to be encoded in three dimensions, while radio-frequency coils both transmit the pulse and listen for the reply. Because different tissues return their signal at different rates, the operator can weight an image towards fat or towards water, so that the same anatomy can be shown in several distinct ways. Software then assembles these measurements into pictures of astonishing detail.

The scanner has hazards as well as limitations. The magnetic field is always on and is strong enough to pull a steel oxygen cylinder across a room, which is why departments control access so carefully. Patients with certain pacemakers or metal fragments cannot be scanned at all, and the gradients switch on and off noisily, producing bangs loud enough to require ear protection. Movement spoils the images, so children and anxious adults may need sedation. Claustrophobia is common, although wider and shorter tunnels have made the experience less alarming than it once was.

In 2003 Lauterbur and Mansfield received the Nobel Prize in Physiology or Medicine. Damadian, whose earlier work had pointed the way, was not included, a decision that provoked considerable argument at the time. MRI is now one of the most versatile tools in medicine, used to examine the brain, the joints and the heart, and, in functional imaging, to map activity while a person is thinking or feeling.

A robot vacuum cleaner has to answer two questions at once: where am I, and what does this room look like? The second question is harder than it appears. A home is a shifting landscape of chairs, cables, pets and closed doors, and the machine has no map to begin with. The earliest mass-market models, which appeared in the early 2000s, dodged the problem entirely. They moved in a mixture of spirals and straight lines until a bump sensor told them that they had hit something, then turned at a random angle and continued. Such robots cleaned adequately over time but could never say where they had been.

Even without a map, a robot needs to know what is happening to it. Wheel encoders count how far each drive wheel has turned, and a gyroscope measures how fast the body is rotating. Combining the two gives an estimate of the robot's position, a technique known as dead reckoning. The estimate drifts, however, because small errors accumulate: a machine that believes it has travelled ten metres may in reality have covered nine and a half. Two other sensors guard against disaster. Bump sensors detect a collision, and infrared cliff sensors, mounted underneath, measure how much light bounces back from the floor and stop the robot falling down stairs.

The instrument that transformed the market was the laser rangefinder, or lidar. A small turret on top of the robot spins several times a second. Inside it, a laser diode fires short pulses of invisible infrared light and a detector records how long each pulse takes to return; since light travels about thirty centimetres in a nanosecond, the timing gives the distance to whatever the beam struck. A single rotation produces a ring of several hundred measurements, and the machine repeats this a few times every second. The result is a cloud of points describing every wall, table leg and doorway within range, usually several metres.

From this point cloud the robot builds a map. The floor plan is divided into a grid of small squares, and each square is marked as occupied, free or unknown, so that a wall gradually appears as a line of occupied cells. The harder task is working out where the robot itself is. The standard solution is known as simultaneous localisation and mapping, or SLAM. The machine compares the pattern of distances it has just measured with the map it has already built, looking for the position and orientation that make the two fit best; it then updates both its estimated position and the map. Done repeatedly, this corrects the drift that dead reckoning alone cannot.

Not every robot carries a laser. A cheaper alternative is a camera, sometimes two, combined with software that tracks visual features — the corner of a skirting board, the edge of a rug — from one frame to the next. This approach, called visual SLAM, costs less and can recognise objects, but it struggles in a dark room or against a blank white wall, and it demands far more computing power than a laser rangefinder does. A third option uses structured light: a projector throws an invisible pattern of dots onto the scene, and a camera judges distance from the way the pattern distorts.

All of these systems have blind spots. Infrared cliff sensors can be fooled by a black carpet, which absorbs the light, and by a highly reflective surface; some robots have therefore been fitted with additional sensors or simply programmed to avoid dark flooring. Thin cables, socks and the trailing cords of phone chargers are notoriously difficult to see, and a machine that tries to swallow one may stop dead until its owner rescues it. Because the lidar turret adds height, a robot carrying one cannot pass under a low sofa. Nor can a laser see a mirror, which returns the beam so directly that it appears, confusingly, to be open space.

Once a robot knows the layout of a room, the practical benefits follow. It can clean in neat parallel strips rather than wandering, cover every part of the floor and return to its dock when the battery runs low. Owners can mark areas to be avoided on a phone, such as a pet's bowl or a child's play mat, without laying down physical barriers. Maps of several rooms can be stored, and the machine can be told to clean the kitchen only. Most models also divide a large open space into sections, cleaning one before moving to the next, which is more efficient than treating the whole area as a single task.

Current research aims at something more ambitious than geometry. A map that records only where walls are cannot distinguish a chair from a waste bin, and a robot that recognised objects could decide to clean around them rather than bumping into them. Neural networks trained on images can now label furniture on the map, and some machines send their maps to cloud servers to be processed and stored. That raises a question of privacy, since a floor plan reveals a good deal about a household. Researchers are also trying to shift the computation onto the robot itself, so that less data leaves the home. Battery life remains a constraint: a machine that runs for ninety minutes may need to recharge before finishing a large flat, though it can usually resume where it stopped.

Few technologies provoke as sharp a disagreement as nuclear power. To its supporters it is the only dependable source of low-carbon electricity on a scale that matches demand; to its critics it is an expensive and dangerous distraction from cheaper alternatives. My own view is that nuclear fission deserves a place in the mix, but a narrower one than its enthusiasts generally claim. The argument for it is strongest when it is made in comparative terms, and weakest when it is presented as a matter of principle.

A reactor generates heat by splitting the nuclei of uranium-235, an isotope of a common element. The heat boils water, the steam drives a turbine, and electricity follows. No carbon dioxide is produced at the plant itself, and when the mining, the enrichment and the construction of a station are taken into account, its emissions over a lifetime are of the same order as those of wind power and far below those of gas. Nuclear stations supply roughly a tenth of the world's electricity, and a larger share of its low-carbon electricity.

Their great practical advantage is consistency. A well-run reactor produces at close to full power for more than ninety per cent of the hours in the year, while a wind farm in a mediocre location typically manages between twenty-five and forty-five per cent. Because output does not depend on the weather, nuclear plants complement wind and solar rather than competing with them: they keep the grid stable when the wind drops. They also occupy very little land for the electricity they deliver.

The difficulty is cost. Reactors are enormous construction projects, and the recent record in Europe is poor. Work on Olkiluoto 3 in Finland began in 2005 with a planned start of 2009; the plant did not enter regular production until 2023. A similar French project at Flamanville suffered comparable delays. In Britain, the Hinkley Point C station was agreed in 2016 at a guaranteed price for its electricity that looked expensive then and looks worse now that wind and solar have become so much cheaper. Each of these projects has been financed at rates of interest that punish delay, and each has taken so long that the technology has been overtaken while it was being built.

Safety is where the argument becomes least rational, in both directions. The two worst accidents, at Chernobyl in 1986 and at Fukushima Daiichi in 2011, were genuinely terrible. Chernobyl killed workers and firefighters within weeks, forced the permanent evacuation of a wide area and scattered radioactive material across Europe. At Fukushima the tsunami that overwhelmed the plant drowned some eighteen thousand people along the coast, and the accident itself forced more than a hundred thousand residents from their homes; the United Nations scientific committee has concluded that no discernible rise in cancer has been observed among the public exposed to the radiation released. Against this must be set the World Health Organization's estimate that outdoor air pollution, much of it from burning fossil fuels, kills millions of people every year.

Waste is the objection I take most seriously, because it is a question of time rather than of probability. The volume of highly radioactive spent fuel produced by a large reactor each year is modest, but some of it will remain hazardous for tens of thousands of years, and no country yet operates a permanent geological repository. Finland has excavated one at Onkalo, designed to bury spent fuel deep in stable rock and the most advanced in the world, but it is still being commissioned. Until such facilities exist, the cost is carried by generations who had no say in producing the waste.

New designs, particularly small modular reactors, are often offered as the answer to the cost problem. Built in factories and assembled on site, they promise shorter construction times, and several such designs have reached the stage of regulatory review in Europe and North America. A few have been built elsewhere, though none has yet been ordered in large numbers by a Western utility. Whether the savings will materialise is another matter: the first of any new design is always the most expensive, and the promised economies depend on producing the same unit many times over.

I am also uneasy about the way the debate is conducted. Supporters talk about capacity factors and dismiss opponents as emotional; opponents talk about Chernobyl and dismiss supporters as lobbyists. Both habitually compare the best version of their own technology with the worst version of the other. Both also tend to treat a single figure, whether a death toll or a cost per megawatt hour, as though it settled the question. A fair comparison would price the carbon emissions of every source, count the full cost of managing waste and decommissioning old plants, and include the cost of the storage needed to make intermittent renewables reliable. Almost nobody campaigns in that way.

So what should governments do? My conclusion is modest. Existing reactors that are safe and economical should be allowed to keep running, since closing them early is expensive and usually means burning fossil fuel instead. New stations should be judged project by project, on the same terms as any other low-carbon source, without subsidy that is hidden from the customer. Nuclear power is neither the devil nor the answer to our prayers. It is a tool, and it should be used where it works.