MEI Conference Friday, 3rd July 2008 Mathematics: an historical perspective Notes Stephen Lee A remarkable formula: summation being equal to an Integral (Bernoulli). Historical Snippets Stephen Lee MEI/FM Network Images from www.mathematik.ch and wikipedia Sources of information • Professional Bodies ‘free’ Publications – Institute of Maths and its Applications • Mathematics Today – Maths Association • Mathematics in School • Mathematical Gazette • Mechanism for gaining students interest in mathematics Historical perspective 1 of 11 MEI conference 2008 A remarkable formula (or two) • Mathematical Gazette – March 2008 – Vol 92 – No. 523 – Pgs 94-96 Discovered by Bernoulli in 1697: ∞ 1 1 1 = dx ∑ n ∫ 0 xx n n =1 Investigate ∫ 1 0 1 dx x x = = 1.29129 (to 5 d.p. using CAS system) Historical perspective 2 of 11 MEI conference 2008 Investigate n ∞ 1 ∑ n n =1 n 1 1 1 1 = 1 + 2 + 3 + 4 + ... 1 2 3 4 Summation 1 1 2 1.25 3 1.28703703703704 4 1.29094328703704 5 1.29126328703704 6 1.29128472050754 7 1.29128593477322 8 1.29128599437787 9 1.29128599695904 10 1.29128599705904 11 1.29128599706255 12 1.29128599706266 13 1.29128599706266 14 1.29128599706266 Proof (Dealing with the interchange of summation and Integration) ∞ 1 ∑n n =1 n =∫ 1 0 1 dx xx (1) n ∞ 1 n ( x ln x ) exp( x ln x ) ( 1) = − = − ∑ xx n! n =0 (2) The convergence of this series being uniform on [0,1] we get that ∫ 1 0 ∞ 1 ( −1) n dx = ∑ xx n! n =0 1 ∫ ( x ln x) n dx 0 (3) Proof (cont.) To evaluate the integral of ( x ln x) it is more convenient, but not essential to look at the more general function, x n (ln x) m where m, n are positive integers, since it is easy to get a general recurrence formula: n I m,n = ∫ 1 0 x n (ln x) m dx = − m I m −1,n n +1 m! m! I 0,n = (−1) m (n + 1) m (n + 1) m +1 n! = (−1) n (n + 1) n +1 I m,n = (−1) m I n,n Historical perspective 3 of 11 MEI conference 2008 Proof (cont.) I n ,n = (−1) n n! (n + 1) n +1 Substituting this result in (3) ∫ 1 0 ∞ 1 (−1) n dx = ∑ xx n! n =0 ∞ Gives (1) 1 ∑n n =1 n 1 ∫ ( x ln x) 0 =∫ 1 0 1 dx xx n (3) dx (1) 1 (−1) n = ∫ x x dx n 0 n =1 n ∞ ∑ Same argument Proof Same argument proves the equally interesting formula: 1 (−1) n = x x dx ∑ n ∫ 0 n =1 n ∞ Proof…is irrational Common knowledge of proof that using proof by contradiction. 2 is irrational, Equally interesting, but perhaps slightly less common is proof by contradiction that e is irrational. If you search for it there are several webpages that show this, including for example the excellent meikleriggs webpages (meikleriggs.org.uk) Historical perspective 4 of 11 MEI conference 2008 Andrew Rogers Frank Nelson Cole and his 1903 lecture, “On the Factorisation of Large Numbers”, in which he proved that 2^67 – 1 was composite. “On the Factorisation of Large Numbers” (MEI Conference, 2008) Starting point Core 3, January 2008, Question 5 These are Mersenne primes. More on Mersenne in a minute. If I remember correctly from awarding, in part (i) candidates identified the primes p less than 11 correctly, and produced 3, 7, 31 and 127 which are all prime. In part (ii), they could calculate 23 × 89 correctly (!) but misunderstood the last part, often attempting to evaluate 22047 − 1. Anyway, 211 − 1 is composite. Fr Marin Mersenne, 1588-1648 He was a French monk, philosopher and mathematician. He was friendly with Fermat, Pascal and Galileo. There were no mathematical journals in his time and he acted as a clearing house for mathematical ideas: people sent him details of their discoveries and he passed them on. But he had many ideas of his own. One of them concerned numbers of the form 2n − 1, and if any such numbers were prime. It is easy to see that 2n − 1 is not prime if n is not prime, e.g. by using the binary representation (although I doubt if Mersenne did this!). So consider the case where n is prime. Mersenne, in a letter to Frenicle de Bessy, proposed that 2n − 1 was prime for a list of prime values of n, not including 11 (see C3 W08) but including 31, 67, 127 and 257. The smallest of these numbers has 10 digits and it was beyond the capacity of anybody at the time to prove or disprove this assertion. In 1750 Euler succeeded to prove that 231 − 1 was prime. Edouard Lucas, 1842-1891 He devised a simple test for checking the “primeness” of these Mersenne numbers, and showed that 2127 − 1 was prime (it was the largest known prime for more than 70 years after this discovery) and that 267 − 1 was not prime. However, his method did not exhibit the factors. He does not seem to have said anything about 2257 − 1. Subsequent searching showed that Mersenne had missed a few: 2n − 1 is prime for n = 61, 89 and 107. Frank Nelson Cole, 1861-1926 Cole was born in Massachusetts and educated at Harvard. In 1903 he gave a presentation to the American Mathematical Society called “On the Factorisation of Large Numbers”. Obviously he had no modern technology to help him deliver his lecture. Without saying a word, he wrote up on the blackboard the calculation for raising 2 to the 67th power, then carefully subtracted 1. (More carefully than Wikipedia or the BBC web site, which have the answer wrong.) On the other side of the board, he then wrote out 193,707,721 × 761,838,257,287 and worked out the multiplication by hand. The two answers agreed. “Cole returned to his seat, not having uttered a word during the hour-long presentation.” (Wikipedia). He received a standing ovation. Historical perspective 5 of 11 MEI conference 2008 Three years of Sundays How had Cole found the factors? He admitted that it had taken him “three years of Sundays”. But these days the problem, and its history, can be presented to pupils and the answer worked out before their eyes. Do you know the program Derive? It can work out calculations to integer accuracy, and the most truculent Year 7 class can be entertained by asking Derive to work out 100000! and then asking them to see if they can see their phone number in the sea of digits. It disposes of the primeness of 267 − 1 in 0.04 seconds. The Texas TI-92 has a version of the Derive software on it, and can evaluate the expression to integer accuracy, but cannot do the factorisation. Derive on a PC can factorise 2257 − 1, and it takes about one minute. The smallest prime factor has 15 digits. Frank Nelson Cole He lectured at Harvard, the University of Michigan and Columbia University, was the secretary of the American Mathematical Society and editor of its Bulletin. There is a Frank Nelson Cole prize for work in number theory. He died suddenly of heart failure in New York City in 1926. A Google Image Search produced the following photograph (Nat King Cole and Frank Sinatra). Frank Nelson Cole “On the Factorisation of Large Numbers” Andrew Rogers King Edward VI Camp Hill Boys’ School Core 3, January 2008 Historical perspective 6 of 11 MEI conference 2008 Fr Marin Mersenne, 1588-1648 2n – 1 prime for n = 2, 3, 5, 7, 13, 17, 19, 31, 67, 127, 257 Edouard Lucas, 1842-1891 Proved that 2127 – 1 is prime (largest known prime for over 70 years) Showed 267 – 1 wasn’t, but couldn’t find the factors Frank Nelson Cole, 1861-1926 Advertised a lecture called “On the Factorisation of Large Numbers” (1903) He had a blackboard He showed that 267 – 1 = 147,573,952,589,676,412,927 (spot the error on Wikipedia) Then he worked out 193,707,721 × 761,838,257,287 The two answers agreed Three years of Sundays Children like large numbers Use Derive to work out 267 – 1 and factorise the result It takes 0.04 seconds The TI-92 can’t do it I tried 2257 – 1 which has 78 digits Derive took 1 minute to factorise it The smallest prime factor has 15 digits Historical perspective 7 of 11 MEI conference 2008 Google Image Search Historical perspective 8 of 11 MEI conference 2008 Peter Mitchell The Königsberg Bridge Problem. Reference http://www.contracosta.edu/math/konig.htm Gerald Goodall Experimental design in Wales. The 1920s and 1930s must have been an incredibly exciting time for applied statisticians. Wonderful new advances in the design and analysis of experiments came out all the time. Nowadays we teach this as classical material - not least, in MEI Statistics 4 and in the AS Statistics - and it is in absolutely standard practical use in all application areas everywhere all round the world. Including in Wales. References http://www.york.ac.uk/depts/maths/histstat/images/latin_square.gif A key to the layout of this Latin square is at http://www.york.ac.uk/depts/maths/histstat/images/latin_square_key.gif. Some information is available by Googling for something like "beddgelert forest latin square". Note the correct spelling of "beddgelert" rather than the incorrect "bettgelert" which is often quoted for this material. Note for example that it appears in the key given above. Googling using "bettgelert" will obtain some of the information but not all of it. Several of the internet items that can be found make interesting comparisons between Latin Squares and Sudoku puzzles. Historical perspective 9 of 11 MEI conference 2008 Bernard Murphy How mathematicians before Newton and Leibniz attempted to find the gradients of tangents to various curves. Alternative methods for finding the gradient of a tangent Mathematicians before Newton and Leibniz attempted to find the gradients of tangents to various curves. Here are two methods, using the curve y = x 2 as an illustration. Fermat’s method y y = x2 ( x + e) 2 x2 x s e By similar triangles, x2 ( x + e) ≈ s s+e 2 Making s the subject: x2 s≈ 2x + e So as e → 0 , s → x2 x = 2x 2 Therefore, the slope of the tangent approaches Historical perspective 10 of 11 x2 2 x2 = = 2x s x MEI conference 2008 y Descartes’ method y = x2 The circle, centre ( v, 0 ) , radius r , touches the curve y = x 2 at the point (α , α ) . Therefore the tangent to the 2 circle is also the tangent to the curve at this point. r x α v The equation of the circle is ( x − v ) + y 2 = r 2 . The circle touches, rather than crosses, 2 the curve y = x 2 . Therefore we have a double root of x = α when we try to solve the equations simultaneously. One way to solve the equations simultaneously is to substitute y = x 2 into ( x − v) 2 + y 2 = r 2 . This gives ( x − v) 2 + ( x2 ) = r 2 2 We need values of v and r so that this equation has a double root of x = α . In other words, we need ( x − v ) + x 4 − r 2 to be of the form ( x − α ) ( x 2 + px + q ) 2 2 Comparing coefficients (in the expanded forms) gives: p − 2α = 0 α − 2α p + q = 1 2 α 2 p − 2α q = −2v α 2q = v2 − r 2 These simplify to p = 2α , q = 1 + 3α 2 , v = α + 2α 3 α2 and the gradient of the v −α 3 v − α (α + 2α ) − α = 2 = = 2α α α2 Therefore, the gradient of the radius shown is − −1 (perpendicular) tangent is − Historical perspective α2 v −α 11 of 11 MEI conference 2008