Sunteți pe pagina 1din 3
Exam talkback Data analysis ne of things that teachers and examiners of AS and [A-level physics have in mind is that it should give students a taste of what it is like to be a physicist. There are many reasons why this can never be a wholly successful exercise — the main one being that it is necessary to understand some basic material before you can go on to be creative in the subject. Creativity may not be one of the things most readily identified with Aclevel physics, but itis important yma aspect of the fulfilment that physi- 499 cists get from their subject. Some A-level specifications address this by including extended practicals or investigations, in ‘which the students have to choose 99, a topic and design their own method, apparatus and procedure. Often these are among the most popular parts of the course. Such investigations demand the ability 80 to analyse data and present them in an accessible manner. Indeed, this is arguably one of the most important skills that physicists possess — both for dealing with 70 the subject itself and for presenting to a potential employer. Physicists have good problem-solving and analytical skills, both highly’ mar- etable, 60. Data analysis in examinations the general approach that a candidate should take is nearly always the same. ‘The following question is taken from an Edexcel synop- tic paper. The question is reproduced by permission of Edexcel. The answers that follow are the responsibility of Paysics Review, and have been neither provided nor approved by the board. For AS and A-level students, learn- 50: ing how to analyse data most often takes place through the day-to-day experimental work that is carried out. However, some specifications test this skill specifically by includ 40 ing data analysis questions on ‘examination papers. Such ques- tions present the candidates with a description of an experiment Gometimes in an area of physics 39. unfamiliar to them) and some experimental results (in the form of a table or a graph) and then ask them to work through the data to verify (or not) the theory that is in 20 question, While the context of the © 2 questions can vary quite widely, Figure 4 24 _ 40 60 80 100 120 us Physics Review Question In an experiment to measure the half-life of radon-220, an tion chamber was interfaced with a computer and the printout, the jagged line shown in Figure 1, was obtained for the ionisation current fsa function ofthe time f. The smooth curve was then added. {@) Explain why the print-out is not a smooth curve. Use the smooth curve to estimate the half-life of radon-220, Explain how you obtained your answer and state any assumption you made. marks) (b) The decay process is represented by Ihe here Zs the decay constant. ‘The following values have been read from the graph and In(limA) calculated for each. tis 20 80100 39532 373s Plot a graph of In(l/mA) against t/s and hence find a value fori. (8 marks) (©) Use your value of / to obtain another value for the half- life. Discuss briefly whether this value for the half-life is likely to be more or less reliable than the value you obtained in a). (G marks) Answers Unfortunately, thi to the question, ‘When I set this question to my own students the major- ity of them suggested that the jagged curve was caused by some fault in the apparatus (or in the person using it). In fact, the most likely explanation for the variability in the data is the random nature of radioactive decay. Sometimes ‘more of the radon nuclei decay in a given time period than ‘one might expect, and sometimes fewer of them. I did give some credit to the couple of candidates who pointed out that the ionisation chamber did not react directly to the alpha particles produced by the radon. The alphas trigger a sequence of ionisations in the chamber and the resulting current pulse is detected. This process has an inherent degree of randomness in it, so one cannot expect a direct relationship between the size of the current pulse and the number of alphas detected. This is a plausible argument, but ionisation chambers are quite carefully designed and this is a much smaller effect than the random nature of the decay itself, Joshua then approached calculating the halflife in a rea~ sonable way, but let himself down by misreading the is not one of the best possible answers September 2003 graph. The smooth curve does not touch the yraxis at all on. the scale that has been used. A better move would have been to start with the 90mA data point which falls on exactly 105, Joshua then halved the (incorrect) current and read the time off the graph as 59s, This proved to be another very ‘common mistake made by my students — he did not read the scale correctly. The reading is one small square before the 60s mark, but this is 58s (each small square = 2s, not 1). The horrid expression ‘S0mA = 59 seconds’ is a misuse of the = sign, but examiners are likely to interpret this as, te student's carelessness and not penalise them for failing to express themselves clearly. A rather more serious flaw in Joshua's answer is failing to take two estimates from which he can take an average. Amodel answer would read When = 105,1=90mA When l= 45mA, t= 69s First estimate of half-life = 69 When I= 70mA, t= 30s and I= 35mA gives! = 94s, so the second estimate is 645 “Average estimation of halflife = (69 + 68)/2 = 66.55 Moving on to parts (b) and (c) of the question, one of my more able physicists produced the following answer. Carolyn’s answer (o) m4 60 0 add us halflife = Ini2)/0.02 = 69.38 | think that the first method is more reliable as it uses more of the data This is nearly a good answer However, there are some flaws to it First, Carolyn has produced a neat graph with labelled axes that have been plotted the correct way round. (Many students fal to learn the simple mantra: if the question asks you to plot one thing against another thing, then the first thing goes on the y-axis.) However, she has not chosen a 28 sensible scale. The data do not go below 0.3 so there is no need to start the y-scale at zero. It would have been better to expand the y-scale, giving a much steeper best-fit line, from which it would be easier to read the gradient. The only exception to this hint would have been if the question implied that the graph should pass through the origin — in which case zero should be included on the scale. Watch for cases when the examiners present an equation which ought to be a straight line passing through the origin (y = mx goes through the origin, y = mre +c does not) The rather Compressed scale contributes to a more seri- ous problem with Carolyn’s answer. In calculating the gradient, she used two points on the graph that are closer together than one might wish. The examiners generally have a mark reserved for an adequately large distance between the points used to calculate the gradient. The issue is one of precision. When a value is read from a graph, the best job that anyone can do is to read the scale to within half of one of the smallest squares on the graph paper (in other words, you can tell which side of a small square itis nearer to). You don’t generally quote the value to less than this smallest scale division - nothing better than that is pos- sible. In this case, Carolyn chose tivo points which, on the vertical scale, have only about one large square between them, Half a small square is rather a large percentage error in her values, Better results would be obtained by using points as far apart as possible, so the uncertainty in the scale reading is a much smaller fraction of the whole. Examiners prefer to see the triangle used to obtain the ‘gradient drawn on the graph but, at a push, will accept a calculation such as the one Carolyn shows, as itis possible to deduce the size of the triangle. The marks scheme for this question would have included a minimum acceptable value of dy and 8x for a gradient calculation. An issue with units While we are on the subject of gradients, strictly speaking, 2 gradient does not have any units. The graph that Carolyn plotted measures time on the x-axis in seconds. The argu ment goes that you can’t plot a number of seconds on a ‘graph, you can only plot a raw number (we are translating a time in seconds to a number of squares). This is why we ‘write t/s on the axis, which supposedly indicates that we have divided # (which is a number of seconds) by seconds to leave us with a raw number which we can plot. This is a rather technical argument, but it is more important on the jraxis of this graph. The logarithm of a number is a well- defined mathematical quantity, but it is not possible to make any sense of the logarithm of a unit — such as ma — In(mA) has no meaning. To avoid this, we have to divide out the unit before we take the log, which is why we write In(I/mA). Often students regard this as rather pedantic fussiness — and I have some sympathy with this view — but, strictly speaking, anything else is wrong and can poten- tially be marked down because of that. Anyway, as far as gradients are concerned, if both axes are unitless, the gradi- ‘ent must be unitless as well. Carolyn should have written: = 0.01, therefore A= gradient = 0.0157 The calculation of the half-life, using the formula (given on the formula sheet, is correct: 26 ‘The final part of the question asks the candidate to assess which of the two ways of obtaining the halflife is likely «0 be the mote reliable. Unfortunately, Carolyn’s answer is both wrong and too brief ‘The first method for calculating the half-life involves reading two values from a curved graph. There are two problems with this, First, itis inherently more difficult to read from a curve than itis from a straight ine. Second, the estimation relies on the value of the current at two distinet times. The second method obtains 2 from the gradient of a straight line which is drawn to bestft the whole of the data and so is less likely to be skewed by an individual rogue result. It is quite hard to tell how well the smooth line ‘its the jagged data on the first graph, but one can clearly see how well the straight line fits to the data on the log graph. This is illustrated by the two different half-lives that I obtained from different parts of the graph. With sufficient patience one might be able to get a decent result from the ‘curve by taking many half-lives to average over, but the log, ‘graph does this automatically when you take the gradient Conclusion ‘There are some rules of thumb to follow when analysing, and presenting data. Not all of these points apply to the particular question we have been looking at, but following, them will help you get full marks for data analysis ques- tions in general ‘¢ Always state clearly the units of measurements. * Use neat data tables with units at the top of the columns. ‘¢ Label axes of graphs so that the units and quantities are easily seen. ‘* Choose a sensible scale — let the data points spread across as much of the area of the graph as possible so that you can easily see how they line up. * When measuring the gradient of a straight-line graph, choose two points on the graph that are a good distance apart. + Hold the graph paper up and squint along it before drawing any best-fit lines — it is surprising how much more clearly you can see the way the points are arranged this way. It certainly helps avoid drawing a straight line through a set of data that is clearly curved. ‘¢ Try to plot a graph that will result in a straight line. If the relationship that you are testing is something like y = const/x or y = const/x, plot y against 1/x or y against 1/2° It is easier to see how well the data fit theory if you can make them compare to a straight line. One curve looks much like any other. Sometimes you ‘may have to plot log graphs, as in the question shown here + Never quote your final answer to a greater degree of precision than the data, Jonathan Allday teaches physics and is Head of Science at the King’s School, Canterbury, Physics Review

S-ar putea să vă placă și