AS history

Handbook of Mathematical Functions The Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical T ables [1] was the culmination of a quarter century of NBS work on coremathematical tools. Evaluating commonly occurringmathematical functions has been a fundamental need aslong as mathematics has been applied to the solution ofpractical problems. In 1938, NBS initiated its Mathe-matical T ables Project to satisfy the increasing demandfor extensive and accurate tables of functions [2].Located in New Y ork and administered by the W orksProjects Administration, the project emp loyed not only mathematicians, but also a large number of additionalstaff who carried out hand computations necessary toproduce tables. From 1938 until 1946, 37 volumes of theNBS Math T ables Series were issued, containing tablesof trigonometric functions, the exponential function,natural logarithms, probability functions, and relatedinterpolation formulae. In 1947, the Math T ables Projectwas moved to W ashington to form the ComputationLaboratory of the new National Applied MathematicsLaboratories of NBS. Many more tables subsequentlywere published in the NBS Applied Mathematics Series;the first of these, containing tables of Bessel functions[3], appeared in 1948. On May 15, 1952, the NBS Applied Mathematics Division convened a Conference on T ables. MiltonAbramowitz of NBS, who had been a member of thetechnical planning staff for the Math T ables Project,described preliminary plans for a compendium ofmathematical tables and related material. Abramowitzindicated that the Bureau was in need of both technicaladvice and financial support to carry out the project.With the support of the National Science Foundation(NSF), a two-day Conference on T ables was held at theMassachusetts Institute of T echnology on September15-16, 1954, to discuss the prospects for such an under-taking. Twenty-eight persons attended, including bothtable producers and users from the science and engineer-ing community. The report of the conference concludedthat “an outstanding need is for a Handbook of T ables for the Occasional Computer, withtables of usually encountered functions and aset of formulas and tables for interpolationand other techniques useful to the occasionalcomputer. ”(Note that here the term computer refers to a person performing a calculation by hand.) The report recom-mended that NBS manage the production of the Hand book and that NSF provide financial assistance. The conference elected the following committee to carryout its recommendations: P.M. Morse (Chair),M. Abramowitz, J.H. Curtiss, R.W. Hamming, D. H.Lehmer, C.B. Tompkins, and J.W. Tukey. The commit-tee was successful in persuading both NBS and NSF tosupport the project, and it began officially in Decemberof 1956. The Mathematics Division of the National Research Council also had an interest in mathematical tables.Since 1943, they had been publishing a quarterlyjournal entitled Mathematical Tables and Other Aids to Computation ( today known as Mathematics of Compu- tation). To provide technical assistance to NBS, as well as independent oversight for NSF, the NRC establisheda Committee on Revision of Mathematical Tables. Itsmembers were P. M. Morse (Chair), A. Erde ´lyi, M. C. Fig. 1. Portrait of Milton Abramowitz. 135 Gray, N. C. Metropolis, J. B. Rosser, H. C. Thacher, Jr., John T odd, C. B. T ompkins, and J. W . Tukey. This groupof luminaries in the fields of applied mathematicsand physics provided guidance to NBS throughout theproject to produce the Handbook. Milton Abramowitz, who was then Chief of the Computation Laboratory of the NBS Applied Mathe-matics Division, led the project. Abramowitz was bornin Brooklyn, NY , in 1915. He received a B. A. fromBrooklyn College in 1937 and an M. A. in 1940.He joined the NBS Math T ables Project in 1938 and in1948 received a Ph.D. in Mathematics from New Y orkUniversity. Abramowitz ’dedication, enthusiasm, and boundless energy led to substantial progress in theproject during its first year. The proposed outline for theHandbook called for a series of some 20 chapters, each with a separate author. Authors were drawn fromNBS staff and guest researchers, as well as externalresearchers working under contract. Most chapterswould focus on a particular class of functions, providingformulas, graphs, and tables. Listed formulas wouldinclude differential equations, definite and indefiniteintegrals, inequalities, recurrence relations, powerseries, asymptotic expansions, and polynomial andrational approximations. Material would be carefullyselected in order to provide information most importantin applications, especially in physics. Consequently,the higher mathematical functions, such as Besselfunctions, hypergeometric functions, and ellipticfunctions, would form the core of the work. Additionalchapters would provide background on interpolation intables and related numerical methods for differentiationand quadrature. Philip J. Davis of NBS first prepared Chapter 6, on the gamma and related functions, to serve as a model forother authors. This chapter portrayed the telegraphicstyle that is a hallmark of the Handbook, i.e., thematerial is displayed with a minimum of textual descrip-tion. In the course of developing his chapter, Davisbecame interested in the history of the topic. This led toa historical profile published in 1959 [4], which wonthe prestigious Chauvenet Prize for distinguishedmathematical exposition from the MathematicalAssociation of America. The Handbook project occurred during the period when general-purpose electronic computing machinerywas first coming into use in government researchlaboratories. (Early computer development of SEAC atNBS is described elsewhere in this volume.) Never-theless, most of the tables in the Handbook were gener- ated by hand on desk calculators. However, even at thattime it was clear to the developers of the Handbookthat the need for tables themselves would eventually be superseded by computer programs which could evaluatefunctions for specified arguments on demand. By the summer of 1958, substantial work had been completed on the project. Twelve chapters had beencompleted, and the remaining ones were well underway.The project experienced a shocking setback one week-end in July 1958 when Abramowitz suffered a heartattack and died. Irene Stegun, who was Assistant Chiefof the Computation Laboratory, took over managementof the project. Stegun, who was born in Y onkers, NY in1919, had received an M. A. from Columbia Universityin 1941, and joined NBS in 1943. The exacting work ofassembling the many chapters, checking tables andformulas, and preparing the work for printing tookmuch longer than anticipated. Nevertheless, theHandbook of Mathematical Functions, with Formulas,Graphs, and Mathematical T ables was finally issued as Applied Mathematics Series Number 55 in June 1964[1]. The volume, which is still in print at the U.S.Government Printing Office and stocked by manybookstores and online booksellers, is 1046 pages inlength. The chapters and authors are as follows. Fig. 2. Portrait of Irene Stegun. 136 1.Mathematical Constants, D. S. Liepman. 2.Physical Constants and Conversion Factors, A. G. McNish. 3.Elementary Analytical Methods, M. Abramowitz. 4.Elementary Transcendental Functions, R. Zucker. 5.Exponential Integral and Related Functions, W. Gautschi (American University) and William F.Cahill. 6.Gamma Function and Related Functions, P. J . Davis. 7.Error Function and Fresnel Integrals, W . Gautschi (American University). 8.Legendre Functions, I. A. Stegun. 9.Bessel Functions of Integer Order, F. W . J. Olver. 10.Bessel Functions of Fractional Order, H. A. Antosiewicz. 11.Integrals of Bessel Functions, Y . L. Luke. 12.Struve Functions and Related Functions, M. Abramowitz. 13.Confluent Hypergeometric Functions, L. J. Slater (Cambridge University). 14.Coulomb W ave Functions, M. Abramowitz. 15.Hypergeometric Functions, F. Oberhettinger. 16.Jacobian Elliptic Functions and Theta Functions, L. M. Milne-Thomson (University of Arizona). 17.Elliptic Integrals, L. M. Milne-Thomson (Univer- sity of Arizona). 18.W eierstrass Elliptic and Related Functions, T. H. Southard. 19.P arabolic Cylinder Functions, J. C. P . Miller (Cam- bridge University). 20.Mathieu Functions, G. Blanch (Wright-Patterson Air Force Base). 21.Spheroidal W ave Functions, A. N. Lowan (Y eshiva University). 22.Orthogonal P olynomials, U. W . Hochstrasser (American University). 23.Bernoulli and Euler P olynomials —Riemann Zeta Function, E. V . Haynsworth and K. Goldberg. 24.Combinatorial Analysis, K. Goldberg, M. Newman, and E. Haynsworth. 25.Numerical Interpolation, Differentiation, and Inte- gration, P . J. Davis and I. Polonsky. 26.Probability Functions, M. Zelen and N. C. Severo 27.Miscellaneous Functions, I. A. Stegun. 28.Scales of Notation, S. Peavy and A. Schopf (American University). 29.Laplace Transforms. The public reaction to the publication of the Hand- book was overwhelmingly positive. In a preface to the ninth printing in November 1970, NBS Director LewisBranscomb wrote“The enthusiastic reception accorded the ‘Handbook of Mathematical Functions ’is little short of unprecedented in the longhistory of mathematical tables that beganwhen John Napier published his tables oflogar ithms in 1614. Only four and one-half years after the first copy came from the pressin 1964, Myron Tribus, the Assistant Secre-tary for Commerce for Science and T echnol-ogy, presented the 100,000th copy of theHandbook to Lee A. DuBridge, then Science Advisor to the President. ” TheHandbook has had enormous impact on science and engineering. Likely the most widely distributedNBS/NIST technical publication of all time, the govern-ment edition has never gone out of print, and it hasappeared as a Dover reprint since 1965. It has beenreprinted (in all or part) by other publishers, such asMoscow Nauka, V erlag Harri Deutsch, and Wiley Inter-science. Government sales exceed 150,000 copies, withcommercial sales estimated at three to six times thisnumber. The Handbook ’scitation record is also remark- able. More than 23,000 citations have been logged byScience Citation Index (SCI) since 1973. Remarkably,the number of citations to the Handbook continues to grow, not only in absolute numbers, but also as aFig. 3. Photograph of Handbook. 137 fraction of the total number of citations made in the sciences and engineering. During the mid-1990s, forexample, about once every 1.5 hours of each workingday some author, somewhere, made sufficient use of theHandbook to list it as a reference. The success of the Handbook was due to several factors. It collected in one place, and in a well-organized way, the most importantinformation needed to make use of mathematical func-tions in practical applications. It served to standardizenotations and normalizations for the special functions ofapplied mathematics, thus easing the communication ofscientific results. In 1965, Irene Stegun was awarded aGold Medal from the Department of Commerce for herefforts in completing the project. A number of difficult mathematical problems that emerged in the course of developing the Handbook engaged researchers in the NBS Applied MathematicsDivision for a number of years after its publication.Two of these are especially noteworthy, the first havingto do with stability of computations and the second withprecision.Mathematical functions often satisfy recurrence relations (difference equations) that have great potentialfor use in computations. However, if used improperly,recurrence relations can quickly lead to ruinous errors.This phenomenon, known as instability, has tripped upmany a computation that appeared, superficially, to bestraightforward. The errors are the result of subtleinteractions in the set of all possible solutions of thedifference equation. Frank Olver, who wrote theHandbook ’schapter on Bessel functions of integer order, studied this problem in great detail. In a paperpublished in 1967 [5], Olver provided the first (and only)stable algorithm for computing all types of solutions ofa difference equation with three different kinds ofbehavior: strongly growing, strongly decaying, andshowing moderate growth or decay. Part of the impact ofthis work is reflected today in the existence of robustsoftware for higher mathematical functions. Olverworked on such topics in the Mathematical AnalysisDivision of NBS, and this work provided the foundationfor his very influential later book on asymptotic analysis Fig. 4. Screen shot of the NIST Digital Library of Mathematical Functions. 138 and special functions [6]. This book has been cited more than 800 times, according to SCI. Another important problem in mathematical compu- tation is the catastrophic loss of significance caused bythe fixed length requirement for numbers stored incomputer memory. Morris Newman, who co-authoredthe Handbook ’schapter on combinatorial analysis, sought to remedy this situation. He proposed storingnumbers in a computer as integers and performing oper-ations on them exactly. This contrasts with the standardapproach in which rounding errors accumulate witheach arithmetic operation. Newman ’s approach had its roots in classical number theory: First perform thecomputations modulo a selected set of small primenumbers, where the number of primes required is deter-mined by the problem. These computations furnisha number of local solutions, done using computernumbers represented in the normal way. At the end, onlyone multilength computation is required to construct theglobal solution (the exact answer) by means of theChinese Remainder Theorem. This technique was firstdescribed in a paper by Newman in 1967 [7]; it wasemployed with great success in computing and checkingthe tables in Chapter 24 of the Handbook. T oday, this technique remains a standard method by which exactcomputations are performed. Newman ’s research on this and other topics, performed at NBS, formed the basisfor his 1972 book [8], which quickly became a standardreference in the applications of number theory tocomputation. Research into the functions of applied mathematics has continued actively in the 36 years since the Hand- book appeared. New functions have emerged in impor- tance, and new properties of well-known functions havebeen discovered. In spite of the fact that sophisticatednumerical methods have been embodied in well-designed commercial software for many functions, there continues to be a need for a compendium of informationon the properties of mathematical functions. T o addressthis need, NIST is currently developing a successor totheHandbook to be known as the Digital Library of Mathematical Functions (DLMF) [9]. Based upon a completely new survey of the literature, the DLMF willprovide reference data in the style of the Handbook in a freely available online format, with sophisticatedmathematical search facilities and interactive three-dimensional graphics. Prepared by Ronald F . Boisvert and Daniel W . Lozier. Bibliography [1] Milton Abramowitz and Irene A. Stegun, eds., Handbook of Math- ematical Functions With Formulas, Graphs, and MathematicalT ables, NBS Applied Mathematics Series 55, National Bureau of Standards, W ashington, DC (1964). [2] Arnold N. Lowan, The Computation Laboratory of the National Bureau of Standards, Scripta Math. 15, 33-63 (1949). [3]T ables of the Bessel Functions Y 0(x), Y 1(x), K 0(x), K 1(x) 0 <=x<=1, NBS Applied Mathematics Series 1, National Bureau of Stan-dards, W ashington, DC (1948). [4] Philip J. Davis, Leonhard Euler ’s Integral: A Historical Profile of the Gamma Function, Am. Math. Monthly 66, 849-869 (1959). [5] F. W . J. Olver, Numerical Solution of Second-Order Linear Differ- ence Equations, J. Res. Natl. Bur. Stand. 71B, 111-129 (1967). [6] F. W . J. Olver, Asymptotics and Special Functions, Academic Press, New Y ork (1974). Reprinted by A. K. Peters, W ellesley, MA(1997). [7] Morris Newman, Solving Equations Exactly, J. Res. Natl. Bur. Stand. 71B, 171-179 (1967). [8] Morris Newman, Integral Matrices, Academic Press, New Y ork (1972). [9] D. Lozier, F. W . J. Olver, C. Clark, and R. Boisvert (eds.), Digital Library of Mathematical Functions, (http://dlmf.nist.gov) National Institute of Standards and T echnology, . 139