arXiv:1403.5164v1 [physics.hist-ph] 20 Mar 2014 n h nwcmeca)GUSA ut fprograms of suite us- GAUSSIAN (Crawford 90% commercial) performed (now that are the calculations estimated ing chemistry is quantum it improvements all functionality, of of and decades accuracy after in Today, Hartree-Fock the approximation. in molecules the solved for which his Schr¨odinger equation of program release computer A free GAUSSIAN 1970 1990). group’s the Schaefer was contribution and proper- early (Handy notable and systems structure molecular the of predict ties career- to first- methods and Pople’s semi-empirical principles honored developing to had commitment Years long Ten “Forty Chemistry”, [1] one. conference, Quantum no international of surprised an Pople earlier, to (No- award years development The chemistry” his quantum 2013). in for bel shared methods Pople was computational density John of of prize and development his theory The 2000, for functional Kohn Schaefer “Walter by 2012). and Sim˜oes equally ad- (Barden and and molecules problems Gavroglu study chemical to used dress is where mechanics enterprise quantum theoretical a chemistry, quantum of field ∗ lcrncaddress: Electronic ycnrs,tefc htWle onere share a earned Kohn Walter that fact the contrast, By h 98NblPiei hmsr eonzdthe recognized Chemistry in Prize Nobel 1998 The tal. et yms cetsst aclt h hsclpoete fm of properties physical approximate the its o of calculate that one to thought (in scientists descr was DFT most complete it introduction, by a its for DFT, after needed Before years was breakthroug function a solids. wave was many-electron and This pr nuclei liquids, system. ex and that gases, electrons its of of In system description any (DFT). th complete of to theory density approach charge functional an total of density creation called mid-1960’s mechanics his for Chemistry in n icsino h ewsuuulywl-utdt creat to well-suited unusually “their” was award he to why came of papers chemists discussion foundational the a two how and the of analyze explanation an and S with DFT, California, to of University directly th the led to of move agenda his physics discuss solid-state sections theoretical visitor the c frequent his of a of context and years Technology ten of first Institute graduat the Carnegie during the his Theoreti made with for he choices continues Institute research Bohr’s the and Niels Kohn’s II and with War University begins World Harvard account during My experie Canada DFT. educational and of Kohn’s creation of the essay including biographical a present h hoeia oi-tt hscs atrKh a awar was Kohn Walter physicist solid-state theoretical The .INTRODUCTION I. 2001). [email protected] n h raino est ucinltheory functional density of creation the and h dcto fWle Kohn Walter of education The eri nttt fTechnology of Institute Georgia nrwZangwill Andrew tat,G 30332 GA Atlanta, colo Physics of School hti h ou ftaiinlqatmchemistry. quantum traditional function molecule, of wave focus atom, many-electron the an the fo- is on in that theory than electrons rather This of solid density or committee. the Nobel theory on the functional cuses by density cited period the (DFT) the The created in conducted which chemistry? Kohn quantum 1963-1965 research the with in do do lies what to answer have ask: things reasonably 2013a). these could (APS chemist surfaces” card-carrying at A phenomena applica- electronic its to and “under- tion gas “the and electron solids” respectively, inhomogeneous of the theory to, standing electron contributions the of his foundations for Phys- Society American and the (1961) ical of Prize (1977) Buckley Prize E. Davisson-Germer previ- Oliver the had the work him of theoretical earned body a ously his as Indeed, gained physicist. was solid-state reputation international his a xso hsgah(e i.1 niae h number the indicates 1) Fig. (see horizon- electrons graph The and of this atomic of 1965). on axis (Pople symposium tal theory 1965 quantum a of molecular proceedings the the published from in Pople which understood chemistry” quantum be of can ‘hyperbola laureates Nobel Chemistry fteNblPiein Prize Nobel the of re ficesn cuay sdt eemn h sys- the determine structure’, to ‘electronic used (in tem’s accuracy) methods increasing mechanical quantum of of order sequence a lists axis l,adohrpoete aclbefo hs.Pople these. from calculable properties lev- other energy and electron density, els, charge electronic function, h oncinbtentewr fteto1998 two the of work the between connection The ae15’ n al 90s Subsequent 1960’s. early and 1950’s late e nDeo dniytersac su which issue research the identify Diego, an oteBl eehn aoaois nthe in Laboratories) Telephone Bell the to a hsc nCpnae.Ite study then I Copenhagen. in Physics cal ∗ csadpoesoa aeru oand to up career professional and nces re we ewsafclymme at member faculty a was he (when areer vdsalteifrainnee o a for needed information the all ovides traso l id.I hsppr I paper, this In kinds. all of aterials h hoyi h rtplace. first the in theory the e e n-afo h 98NblPrize Nobel 1998 the of one-half ded aypril rbe nquantum in problem many-particle e c om F salse htthe that establishes DFT form, act o h td faos molecules, atoms, of study the for h om)i h ehdo hieused choice of method the is forms) pino uhsses oa,fifty Today, systems. such of iption fteter.Tepprconcludes paper The theory. the of tdn er nAsra England, Austria, in years student oe rz otepyiitKohn physicist the to Prize Nobel l h atymr complicated more vastly the nly N n otgaut riigat training post-graduate and e ntesse fitrs.Tevertical The interest. of system the in chemistry i.e. t ayeeto wave many-electron its , upie aybecause many surprised 2 suggested that the activities of most quantum chemists II. AN UNSENTIMENTAL EDUCATION tended to cluster around the extremities of the hyper- bolic solid line. Those interested in the highest accu- In 1933, ten-year-old Walther Kohn began the eight- racy were forced by computational constraints to focus year course of study at the Akademische Gymnasium, the on small molecules (small N) while those interested in oldest and one of the best secondary schools in his home large molecules (large N) were forced by computational city of Vienna. [2] In doing so, he was not unlike the chil- constraints to use methods that were capable of only low dren of many middle-class Jewish parents who were ac- accuracy. He noted that progress would occur by moving tively engaged in the intellectual and artistic life of their off the hyperbola either horizontally from left to right or city. [3] His father Salomon owned an art publishing com- vertically from bottom to top. pany that specialized in the manufacture and distribution of high-quality art postcards. Despite a worldwide clien- tele, it was a struggle to keep the business going in the face of a global economic depression which hit Austria particulary hard. Nevertheless, there was a tacit under- standing that Walther would eventually run the family business. Kohn’s mother Gittel was a highly educated woman who spoke four languages and it was she who chose the humanistically oriented Akademische Gymna- sium to educate her son. Walther excelled at Latin and ancient Roman history but showed no apparent aptitude for mathematics. The only grade of C he ever received was in that subject (Kohn 1998). Besides marking the beginning of Kohn’s secondary school education, 1933 was also the year that Adolf Hitler and his Nazi party took power in neighboring Germany. The Nazi party was illegal in Austria, but its many Fig. 1. The applicability range of Kohn’s density sympathizers worked tirelessly to undermine the exist- functional theory placed on Pople’s hyperbola of ing democratically elected government. Finally, in May quantum chemistry. Adapted from Pople (1965). 1934, the country succumbed to a form of authoritar- The density functional theory created by Walter Kohn ian rule known to historians as Austrofascism (Berkley represented a dramatic move away from Pople’s hyper- 1988). Four years later, cheering crowds welcomed Hitler bola, particulary for systems with a large number of elec- when the German army crossed the border and marched trons (see the arrow in Fig. 1). Many chemists were wary into Vienna. The Anschluss (political union) of Germany (or dismissive) of this theory at first because it made no and Austria was a fait accompli. These events were vivid use of the N-electron wave function, a quantity thought in Kohn’s memory many years later (Kohn 1998, Kohn to be indispensable for a proper description of any atom 2013a): or molecule. Eventually, improvements to Kohn’s the- The [March 1938] Anschluss changed every- ory made by chemists themselves led to quantitative suc- thing. The family business was confiscated cesses that could not be denied. Today, it is the method but my father was required to continue its of choice used by most scientists who wish to calculate management without any compensation. . . the properties of real materials. . He wrote to a London art distributor and Unlike many other scientific achievements, the techni- business client named Charles Hauff (whom cal question which led Kohn to create DFT in the mid- he had never met) and asked if he and his wife 1960’s was not “in the air” among physicists, chemists, would temporarily accept me and my older or anyone else. It is entirely possible that the theory sister Minna into their home. Hauff replied would be unknown today if Kohn’s background, technical affirmatively and Minna emigrated to Eng- skills, and scientific experiences differed very much from land very soon thereafter. For reasons of their what they were. Accordingly, this article (i) recounts own, the Nazis made it much more difficult the unusual personal and intellectual journey which led for young Jewish boys to leave. I remained in to Walter Kohn’s success as a theoretical physicist; (ii) Vienna, but was expelled from my school. identifies the scientific issue which motivated the creation of DFT in the context of theoretical solid state physics Many expelled students never went to school again. research in the 1960’s; (iii) summarizes the key elements Kohn was lucky because he was permitted to finish the of the founding papers of DFT; (iv) argues that Kohn’s school year at a segregated high school for Jews. Then, in scientific background made him particularly well-suited August 1938, he was one of a few hundred high-achieving to create density functional theory; and (v) explains in Jewish students from the various Viennese secondary brief how Kohn came to win a share of the Nobel Prize schools who were given the opportunity to continue their in Chemistry for the creation of DFT. education at the Zwi Perez Chajes Gymnasium, a private 3 all-Jewish high school. [4] This experience was transfor- mative for Kohn because his interest in science was ig- nited by the physics teacher Emil Nohel and the mathe- matics teacher Victor Sabbath. [5] He later recalled that (Hollander 2000, Kohn 2004)

Nohel was a tall, quiet, noble man who de- voted himself to his students. It was really a combination of my admiration for this man as a person and his deep knowledge of physics that started me off . . . Though I was only fifteen going on sixteen years old, I al- ready understood—due to Nohel’s role model and by comparing myself to others—what it meant to really comprehend something in Emil Eliezer Nohel (ca. 1938) was a high school physics physics. This is one of the most important teacher who inspired Kohn to become a scientist. insights for a future theoretical physicist. Courtesy of the Yad Vashem Photo Archive. Sixteen-year old Walther was saved from probable ex- termination by the Nazis by the Kindertransport, an or- ganized effort to evacuate Jewish children under the age Sabbath was also a fantastic guy. The thing of 18 from Germany, Austria, and Czechoslovakia. Be- I remember about him is that while he was tween December 1 1938 and September 1 1939, this col- teaching us he told us about a new book laboration between British Quakers, World Jewish Re- he was reading by the great French physicist lief, the Refugee Children’s Movement (a British rescue [Louis] de Broglie called Matter and Light... organization), and the Jewish agencies in the affected Sabbath was a teacher with great enthusiasm countries made it possible for almost 10,000 unaccompa- and it was very exciting what he told us. nied children to board special trains bound for port cities in the Netherlands. Boats carried them from there to safety across the English Channel (Fast 2011). By agree- ment with the British government, the children were to be placed with sponsor families or volunteer foster par- Kohn’s inspirational teachers Emil Nohel and Victor Sab- ents for a period of up to two years, whereafter repatria- bath did not survive the Holocaust. tion to their home countries was anticipated. The Jewish community of Britain bore full financial responsibility for November 10 1938 was particulary memorable at the the transport. The host families were expected to arrange Chajes school because it was the day following the night educational opportunities for their guest children. of the infamous, state-sanctioned orgy of destructive vi- One quarter of the Kindertransport children came olence against Jewish homes, businesses and synagogues from Austria and the Jewish Community of Vienna [Is- known as Kristallnacht (the night of broken glass). The raelitsche Kultusgemeinde Wien (IKG)] was responsible principal dismissed the students early to avoid attract- for deciding which children would receive exit visas from ing attention, but Kohn and a classmate were arrested amongst the great many applications they received (Cu- on their way home. They were released after several ter- rio 2004). Initially, the IKG focused on urgent cases, in- rifying hours in the police station, but Kohn returned cluding stateless children under threat of expulsion and home to find “our apartment absolutely vandalized by a children in orphanages. Priority was given to children group of hooligans, including the person who had taken whose parents had been arrested or deported. Later, over my father’s business” (Hanta 1999). The classes preference was given to children with guaranteed spon- at Chajes got smaller and smaller in the months after sors in Britain, but this alone did not ensure selection. Kristallnacht. Emigration was on everyone’s mind, but The children were interviewed and the additional crite- it was difficult and expensive to make it happen (Ehrlich ria of good health, a pleasant personality, likely success 2003). Kohn was again lucky. He escaped from Aus- in school, and the ability to “fit in” were used to make tria to England just three weeks before Germany invaded the final choices. Kohn apparently met all these criteria Poland and World War II began. His parents were un- and thereby earned a seat on one of the last Kinder- able to leave. Both were eventually deported, first to the transport trains. He arrived in England safely with the Terezin concentration camp in Czechoslovakia and then pre-arranged plan to live with Charles and Eva Hauff, to Auschwitz, where they were murdered in 1944. the same couple who had previously welcomed his sister 4

Minna. In fact, Minna met Walther at London’s Victo- and algebraic reasoning and the reader is expected to ria station and they traveled together by train to West understand first derivatives, second derivatives, and the Sussex where the Hauff’s lived (Kohn 2013a). law of integration which connects them. This is mate- Before leaving Vienna, Kohn and his parents had rial found in present-day American textbooks intended agreed that he should learn to be a farmer in England for first-year college students. (Kohn 1998). They had seen too many unemployed in- Kohn’s life at East Grinstead was happy and peace- tellectuals in pre-war Vienna and farming seemed like an ful for five months. Then, on May 10 1940, Germany occupation that would make him less subject to economic invaded Holland, Belgium, and Luxembourg. Winston dislocations. [6] Moreover, Salomon Kohn was 66 in 1939 Churchill replaced Neville Chamberlain as Prime Min- and it was understood that Walther would soon be re- ister of Great Britain and the British newspapers be- sponsible for his parents’ financial security. All of this came filled with war hysteria and reports of ‘fifth colum- was communicated to the Hauffs ahead of time and they nists’. [7] The British War Office feared that an invasion were able to arrange for Kohn’s formal English education was imminent and recommended to the Home Office that to begin at a training farm in Sittingbourne, Kent, about the government “intern all enemy aliens in areas where one hour away from the Hauff’s home by car. There, German parachute troops are likely to land” (Gilman Kohn recalls (Hollander 2000), & Gilman 1987). Arrests begin immediately and, when Italy entered the war on June 10, Churchill demanded I pulled carrots and looked after piglets. . that police officials ‘collar the lot’. This terse order ex- . Unfortunately, on the farm, I contracted panded internment to all parts of the country and to all what turned out to be meningitis and so was enemy aliens, i.e., to all holders of passports from Italy very ill. Sulfa drugs had just been invented, or Nazi-occupied countries age 17 or older. It took Scot- so I pulled through, but it was touch and go. land Yard less than six weeks to intern 24,000 men and After that, I was very weak and going back 4000 women (Cesarini and Kushner 1993). Those ar- to the farm was out of the question. rested included not just foreign nationals of quite short The Hauffs dealt swiftly with this setback and, in Jan- residency in England like university students but also uary 1940, Walther entered the nearby East Grinstead highly trained scientists, engineers, doctors, artists, and County Grammar School. His limited English skills led musicians who had lived in England for many years. [8] the headmaster, Thomas W. Scott, to estimate that 2-3 Kohn had turned 17 on March 9 and thus was sub- years could be required for this new student to earn the ject to the order of internment. He was arrested and school certificate needed to enter college. Scott’s method shipped by train to a large camp that had been hastily to redress this situation was to enroll Kohn in the lower constructed in the town of Huyton on the outskirts of 6th form (to avoid extra English requirements) and in- Liverpool. It was at Huyton that Walther met and be- struct his teachers to overlook their new student’s de- gan a lifelong friendship with Josef Eisinger, a boy who ficiencies in English. He then created a daily German had been one year behind him at the Akademische Gym- class where Walther was the only student. The instruc- nasium in Vienna. [9] A week or so later, both boys were tor taught English to Kohn for half the class and Kohn transferred to an internment camp on the Isle of Man. taught advanced German to the instructor for the other The Hauffs arranged for Walther’s East Grinstead teach- half (Kohn 2000, Ford 2013). ers to send him his physics textbooks because they were Kohn’s preparation in math and science was equal to told he would return home soon (Hanta 1999). Instead, or exceeded that of his fellow 16-year-old English class- he spent a month on the Isle of Man where there was no mates (Kohn 2013a). The level of physics he was ex- work and little food. He lost 30 pounds. posed to at East Grinstead can be judged from two of his The burden of warehousing this great mass of civil- textbooks, both intended for students preparing for Uni- ian internees and the belief that many German pris- versity Scholarship Examinations. Heat (1939) by R.G. oners of war would arrive soon led Churchill’s govern- Mitton is a thorough introduction to the thermal proper- ment to seek help from the British Commonwealth na- ties of matter, the kinetic theory of gases, heat engines, tions. Accordingly, on July 4 1940, Kohn and his friend entropy, and the laws of thermodynamics. The prose Eisinger found themselves bound for Canada aboard the is brisk, yet clear, and the author assumes familiarity Sobieski, a Polish cruise ship that had been captured by with the laws of algebra and the geometrical meaning of the British and converted to a troop transport. The ship the derivative. The latter occurs in a section devoted to arrived in Quebec City two weeks later as part of a con- the Clausius-Clapeyron equation, an advanced topic not voy that zigzagged across the Atlantic to avoid German often found in present-day American high school text- submarines (Koch 1980, Auger 2005).[10] By the time books. The oddly-named Properties of Matter (1937) by the British government discontinued its policy of intern- D.N. Shorthose is a textbook of particle and continuum ments and deportations in early August 1940, nearly 4400 mechanics which includes chapters devoted to ballistic civilian internees and 1950 prisoners of war had been re- motion, circular motion, simple harmonic motion, rigid located to Canada. What the Canadians did not know— body motion, hydrostatics, friction, elasticity, and vis- because the British did not tell them—was that the ma- cosity. The exposition moves freely between geometrical jority of the civilian internees were not Nazi sympathizers 5 but Jewish refugees from Nazi barbarism. the construction of the camp was complete, each group Kohn spent the next eighteen months in four differ- elected a leader and a deputy leader, as well as chair- ent Canadian internment camps, a burden lessened only men for the various departments of the internee camp slightly by the Red Cross, which made it possible for him Cabinet. to exchange letters with his parents every two weeks or The camp’s Canadian officials had a clear idea of how so. He was also able to engage in learning much of the the able-bodied internees would spend most of their time: time, albeit not exactly of the sort he had experienced lumberjacking in the forest outside the camp for 20 cents before. For example, on the evening of his arrival in Que- a day. Kohn was not unhappy to do this because the bec City, he and 710 others internees were moved eighty sheer physical labor of chopping trees warded off the miles down the St. Lawrence river to the town of Trois Canadian winter cold. However, much more important Rivi`eres, where barbed wire had been strung around the to him were the many hours he spent at the ‘camp school’ perimeter of a local agricultural exhibition ground to cre- where internees with special technical or academic knowl- ate Camp T. Barracks were built in an arena designed to edge offered classes for the benefit of anyone who wanted house livestock (Jones 1988). However, a baseball field to learn. Each class met at a designated table in the adjacent to the arena was the home of the Trois Rivi`eres camp’s Recreation Hut and internees sought and received Renards of the Quebec Provincial Baseball League. By time off from their work details to attend. Appeals to standing on several tables stacked on top of one another humanitarian organizations and letters written by the to reach the windows, Walther and others could watch instructors to acquaintances around the world produced the games. They pieced together the rules of baseball donations of textbooks, writing materials, exercise books, during the month they were interned there (Koch 1980, blackboards, painting supplies, musical instruments, etc. Kohn 2013a). Two months after it opened in November 1940, the camp library boasted a collection of nearly 1000 volumes. The Camp B school was organized by the chair of the Education Department, Alfons Rosenberg, a 39-year-old former gymnasium teacher from Berlin who had taught briefly at the Cranbrook School in Kent, England before he was ‘collared’ and deported to Canada. His motiva- tion to create the camp school was simple: 60% of the internees were under 20 years old and in desperate need of organized instruction. The many Cambridge men in the camp offered their services and, over time, courses were given in accounting, acting, anthropology, architecture, art history, astronomy, biology, composition, chemistry, economics, engineering, English, French, geography, Ger- man, history, Latin, law, literature, mathematics, music theory, philosophy, physics, physiology, political theory, psychology, sex education, Spanish, and typing. In a brilliant stroke, Rosenberg arranged with the Walther Kohn in Canada at age 18 (1941). camp commandant for official examination booklets to Courtesy of Josef Eisinger. be brought into the camp so internees could sit for the nation-wide McGill University Junior Matriculation Camp T was closed on August 12, 1940 and its popu- Exams—a necessary step to enter a Canadian college. lation was transferred to Camp B, a 15 acre compound Thomas Cassirer, a future professor of French at the Uni- built in the heavily wooded Acadia forest twenty miles versity of Massachusetts, Amherst characterized Rosen- east of Fredericton, the capital of the province of New berg as (Jones, 1988) Brunswick.[11] Kohn spent nearly a full year at Camp B and a vivid portrait of the camp written by Ted a schoolmaster in the good sense: he could Jones (1988) makes it possible to reconstruct life there in listen to others, he had a wide outlook, and some detail. For our purposes, a salient fact is that the he gave suitable advice. . . . He was always internees organized themselves into a highly-structured at everybody’s disposal. At discussions, and mini-society composed of sixteen distinct groups based we had many with him, he would often solve on age, education, geography, past friendships, religious everything with just a short sentence. practices, and political views. Kohn belonged to a group of academically-minded boys between 16 and 20. A dif- Walther Kohn attended a daily physics class offered ferent group was populated largely by academics from by Kurt Guggenheimer, a physical chemist who antic- Cambridge University (graduate students, recent PhD’s, ipated the shell model of the nucleus by pointing out post-doctoral fellows, and junior faculty), another group similarities in the systematics of the binding energies of self-identified as communists, and another group was molecules and nuclei. [12] Kohn and future McGill Uni- composed entirely of ultra-orthodox Jews. Even before versity mathematics professor Joachim Lambek were the 6 only students in a set theory course taught by Fritz Roth- Kohn would have learned from his camp-mates that these berger.[13] According to Lambek (1980), ‘new examples’ were drawn from the demanding exam- inations used by Cambridge University to evaluate its Rothberger was an outstanding teacher with students who hoped to earn a BA degree in Mathemat- an inimitable style of lecturing. He gave un- ics. sparingly of his time and managed to bring The Slater book that Walther bought and read (doubt- the most abstract concepts down to earth. He less cover to cover) was very unusual for its time. The instilled a love of mathematics in numerous preface states that the author worked hard “to make it young people. intelligible to a reader with a knowledge of calculus and A similar impression was left on Kohn (Kohn 1998, Kohn differential equations, but unfamiliar with the more diffi- 2013a): cult branches of mathematical physics”. For that reason “the quantum theory used is of a very elementary sort Rothberger normally taught us out-of-doors . . . and it has seemed desirable to omit wave me- where he wore shorts and boots and nothing chanics.” On the other hand, the content is far from else. He used a stick and a sandy area as elementary. Slater notes that “it is customary to write a blackboard to teach us about the different books either on thermodynamics or on statistical me- types of infinities . . . . He was a most kind chanics; this one combines both.” Moreover, “atomic and and unassuming man whose love for the in- molecular structure are introduced, together with a dis- trinsic depth and beauty of mathematics was cussion of different types of substances, explaining their gradually absorbed by his students. interatomic forces from quantum theory and their ther- mal and elastic behavior from our thermodynamic and Walther “took the classes very seriously because I felt statistical methods.” The preface does not warn the a huge responsibility to support my parents after the reader about the author’s frequent use of kinetic the- war” (Kohn 2013a). The level of his commitment is ory, which is a non-trivial subject of its own. All told, not difficult to demonstrate. First, he earned all pass- the first 100 pages cover “Thermodynamics, Statistical ing marks when Rosenberg and his camp staff admin- Mechanics, and Kinetic Theory”, the second 200 pages istered the McGill Junior Matriculation Exam in June discuss “Gases, Liquids, and Solids”, and the final 200 of 1941 (Giannakis 2013).[14] Second, Kohn took the pages concern themselves with “Atoms, Molecules and 20 cents a day he earned lumberjacking and used the the Structure of Matter.” This material would be chal- “princely sum, carefully saved” to order two books which lenging for a good American college student. It must were sent to him at the camp: A Course in Pure Mathe- have been even more so for a 17-18 year old student with matics (1938) by G.H. Hardy and Introduction to Chem- a twice-interrupted high school career who was still learn- ical Physics (1939) by J.C. Slater (Kohn 1998). The ing English. [18] Cambridge-educated cohort at Camp B would have been The Canadian military dissolved Camp B in the spring very familiar with the Hardy book because its author and summer of 1941. Some Jewish internees were re- was a Cambridge mathematician and the book was in turned to England and released when the British govern- its seventh edition at the time. By contrast, a brand ment finally acknowledged that they posed no threat to new book by the American physicist Slater would have the war effort. [19] Kohn was not so lucky. At the end of been completely unknown to his fellow internees. Kohn July, he and Josef Eisinger were transferred to Camp A in purchased it entirely on the basis of a catalogue descrip- the southern Quebec town of Farnham, 60 km southeast tion (Kohn 2013b). In an uncanny way, the contents of of Montreal (Eisinger 2013). Kohn was assigned to the both books foreshadow the mathematical rigor, taste for knitting shop where he spent hundreds of hours making foundational issues, and deep interest in the properties of woolen socks and camouflage nets. On the other hand, matter that characterize much of this subsequent work. Farnham boasted a camp school every bit as good as the Hardy[15] and Slater[16] were both first-rate re- one at Fredericton and the many hours of quiet repetitive searchers with a strong interest in the pedagogy of their work allowed him to think deeply about his schoolwork fields. Hardy wrote the first edition of his book in (Kohn 2013a, Auger 2005). 1908 for the express purpose of making the teaching of The Camp A school was organized by William mathematics more rigorous at British universities. [17] It Heckscher, a non-Jewish native of Hamburg who was was aimed specifically at first-year University students working as an art historian in England when he was of ‘scholarship standard’, i.e., the top 10-20% in ability. interned (Sears 1990). Kohn singles out Heckscher for The preface to the first edition is explicit: this is “a book special praise in his autobiography (Kohn 1998) and in for mathematicians: I have nowhere made any attempt to the first history of the Canadian internment camps, Eric meet the needs of students of engineering or indeed any Koch (1980) reports that class of students whose interest are not primarily mathe- matical.” Thirty years later, in the preface to Kohn’s edi- William Heckscher was a remarkable figure. tion, Hardy remarks that “the general plan of the book is He was the ideal headmaster. The adjective unchanged” but “I have inserted a large number of new with which several ‘old boys’ of the Farnham examples from the papers of the Mathematical Tripos”. camp school described him was ‘elegant’. He 7

had grace, style, and patience. (Kohn 2013a). Mendel and Infeld arranged a meeting for Kohn (and five other camp boys with a similar prob- Heckscher told Koch that the Farnham camp comman- lem) with the Dean of the College of Arts & Science, dant, Major Eric D.B. Kippen, once said to him, “You Samuel Beatty, who was also the chair of the Mathe- know, Heckscher, I wish I could send my two sons to matics department. [25] Beatty was sympathetic, but he your school.” At the end of September 1941, Heckscher was unable to move the inflexible registrar. Therefore, escorted Kohn and a group of other internees when they he proposed to admit Walther and the others as ‘special traveled to Camp S on St. Helen’s Island in Montreal students’ who did not need the prerequisites. This artful to sit for McGill University’s Senior Matriculation Ex- maneuver required only the assent of the Department amination. The records show that Kohn did well in chairs whose departments were involved in the Math- all the subjects tested: algebra, geometry, trigonometry, Physics curriculum. This time, it was the chair of the physics, chemistry, and coordinate geometry (Giannakis Chemistry department, Frank Kenrick, who threw up a 2013). roadblock by refusing to allow a foreign national from Salvation for Kohn and Eisinger came in October 1941 any Triple Axis country to enter his chemistry building when they received a letter from the wife of a faculty where war research was being conducted.[26] Beatty ar- member at the University of Toronto. She had heard ranged for Kohn to plead his case in person, first with about them from a former internee and offered to spon- Kenrick (who refused to acknowledge that Kohn was a sor them to come live with her family after their release refugee) and then with the University Chancellor, Rev. (Eisinger 2011).[20] Letters were exchanged with Scot- Henry John Cody (who was unwilling to overrule one of land Yard and, by the end of January 1942, the boys his department chairs). found themselves sharing a comfortable attic space in the home of Hertha and Bruno Mendel. [21] The Mendels were refugees from the Nazis themselves who had earlier helped young Jewish couples escape Germany and start new lives in Canada (Feldberg 1960). [22] Their new mis- sion was to arrange educational opportunities for recently released Jewish internees. Within days, Eisinger was en- rolled in a Toronto high school. Kohn, who was one year older, needed help from two complete strangers to get started at the University of Toronto. [23] He first received some valuable advice (Kohn 2013a), Dr. Mendel worked at the university and his good friend, Leopold Infeld, came to their home very soon after I arrived. Infeld ques- tioned me about my plans and I told him I wanted to be an engineer (another practical profession). He asked “ Is that your main in- Dean Samuel Beatty (ca. 1953) bent the admission rules terest?” and I said no, it was mathemat- so Kohn could enter the University of Toronto. Courtesy of the University of Toronto Archives. ics and physics. He told me that engineer- ing at the university was good but that the In the end, the creative Beatty simple redefined the Math-Physics program was superb and that I meaning of ‘special student’ to constitute the normal should pursue a degree there. With the train- Math-Physics curriculum minus the usual chemistry re- ing I received, I could always do engineering. quirement. In this way, Kohn and his compatriots en- At that time, the University of Toronto had a Mathemat- tered the Math-Physics program several weeks after the ics department, a Physics department, and a small (five- beginning of the spring 1942 term. Relatively little time person) department of Applied Mathematics (Robinson was lost because Beatty had permitted the group to au- 1979). Leopold Infeld was a member of the Applied dit his gateway mathematics course the entire time the Mathematics faculty, having come to Toronto in 1938 af- admission negotiations were going on. Even after the for- ter working with Albert Einstein for two years at the In- malities were completed, Beatty tutored the group pri- stitute for Advanced Study in Princeton. [24] The Math- vately for a month to bring them up to speed with the Physics program recommended to Kohn by Infeld was an rest of the class. honors curriculum where all physics, mathematics, and Kohn’s undergraduate transcript shows that he com- applied mathematics students took the same courses for pleted the two-year Math-Physics “common core” in the first two years and then specialized in their final two three semesters. [27] Besides algebra, analytic geometry, years (Allin 1981). differential and integral calculus, differential equations, Walther attempted to enroll in Toronto’s Math-Physics physics laboratory, mechanics, properties of matter, dy- program, but was rebuffed by the University registrar namics, electricity & magnetism, and light & acoustics, because he lacked some of the mandatory prerequisites one finds two required courses in actuarial science, two 8 required courses in French and German replaced by En- Borden, Ontario. This substantial piece of work (Kohn glish courses, and two elected courses in ‘oriental litera- 1946) establishes bounds on the motion of a heavy spher- ture’ where texts in ancient Egyptian, Arabic, Hebrew, ical top using a contour integration method used by We- Persian, and Turkish were read in translation. Kohn took instein (1942) to study a spherical pendulum. The text twelve more advanced courses during the 1943-1944 aca- makes clear that Kohn had at least some familiarity with demic year, including algebraic geometry, differential ge- Uber¨ die Theorie des Kreisels (1898), the great treatise ometry, partial differential equations, the theory of func- on tops by Felix Klein and Arnold Sommerfeld. tions, group theory, thermodynamics, classical dynamics, In the 1945-1946 academic year (Kohn 1998), quantum mechanics, variational principles in physics, En- glish literature, and modern ethics. These two semesters after my discharge from the army, I took an turned out to be his last as an undergraduate because excellent crash masters program, including he was inducted into the Canadian army in September some senior courses I had missed, graduate 1944. [28] He served until the war ended in August 1945 courses, and a master’s thesis consisting of and was awarded his BA in Applied Mathematics at the my paper on tops and a paper on scaling summer 1945 convocation ceremony while still on active atomic wave functions. duty. [29] The atomic wave functions paper, “Two Applications of The bare list of courses Walter took during his five un- the Variational Method to Quantum Mechanics” (Kohn dergraduate semesters does not communicate the elite 1947) was the first of many papers to come (including the quality of the instructors who taught and mentored density functional papers) where Kohn exploits a varia- him. [30] Leopold Infeld, who lectured to upperclassmen tional principle. He first learned about such principles only, had been invited to join the Applied Mathematics from an advanced undergraduate course where Weinstein Department by its chair, the eminent Irish mathemati- discussed the Lagrangian and Hamiltonian formulations cian and theoretical physicist, J.L. Synge. [31] Synge and of classical mechanics. Weinstein regarded Kohn as a his Applied Mathematics colleague Bernard Griffith were potential PhD student and thus shared with him his re- the authors of Principles of Mechanics, the introductory cent work on variational methods to study the vibrations but quite sophisticated textbook used by all the students of clamped plates and membranes (Aronszajn & Wein- in the Math-Physics program. Synge was always eager stein 1941). A review paper by Weinstein (1941) summa- to add talent to his faculty and in 1941 he succeeded rized the original contributions to this subject by Lord to recruit the Russian mathematical physicist Alexander Rayleigh and Walter Ritz. Weinstein, a mature scientist with a strong reputation Notwithstanding the foregoing, it was Applied Math- for his work on free boundary problems and variational ematics Professor Arthur Stevenson who broadened principles. [32] Kohn’s perspective to include quantum problems and the It is significant to our story that variational methods use of variational methods to study them. He is thanked were something of a Toronto speciality at the time. Be- in Kohn (1947) “for his kind advice and interest”. [33] sides Weinstein, Synge, and Griffith, one should include Stevenson’s early research concerned quantum mechan- Gilbert Robinson (Mathematics) and Arthur Steven- ical methods to calculate the energy levels of electrons son (Applied Mathematics) because Cornelius Lanczos in atoms and he had performed variational calculations thanks them in the preface to his now-classic 1949 text for the helium atom in collaboration with a colleague in The Variational Principles of Mechanics because they to- the Toronto physics department (Stevenson and Craw- gether “revised the entire manuscript”. In later years, ford 1938). Kohn surely read this paper because the he- Kohn singled out the algebraist Richard Brauer and lium atom figured into his work also. the non-Euclidean geometer H.S.M. (Donald) Coxeter Kohn assumed that the readers of his paper were fa- as “luminous faculty members whom I recall with spe- miliar with the variational method to find approximate cial vividness” (Kohn 1998). He also recalled the first- solutions to the Schr¨odinger equation. Indeed, most text- year electricity and magnetism lectures given by Lachlan books of quantum mechanics written between 1930 and Gilchrist, a 1913 PhD student of Robert Millikan, be- 1945 devoted more than passing attention to this topic cause Gilchrist told his Toronto students that it was he because Egil Hylleraas (1929) had used it with spectac- who had purchased the oil used by Millikan in his famous ular success to calculate the ionization energy of the he- oil-drop experiment (Kohn 2003). lium atom. This provided the first convincing evidence Weinstein’s influence on the undergraduate Kohn is ap- that quantum mechanics could achieve quantitative suc- parent from Kohn’s first two published scientific papers. cess for a system of more than one electron. The first, submitted in July 1944, is a two-page report on For future reference, I sketch here a simple form of the an exact solution for the oscillations of a spherical gyro- Rayleigh-Ritz variational method appropriate to an N- scope which generalizes a method presented in the Synge electron system with ground state energy E0 and Hamil- and Griffith book but thanks Weinstein for “his advice tonian operator H. If r = (x,y,z), the starting point is a and criticism” (Kohn 1945). The second paper was com- trial wave function, ψ(r1, r2,..., rN ), which depends on pleted and submitted in November 1944 at a time when the Cartesian coordinates of all the electrons. One then Kohn was engaged in advanced basic training at Camp computes a 3N dimensional integral with the configura- 9 tion space volume element dτ = dr1 ··· drN and exploits trial function ψ as a linear combination of a set of basis the inequality, [34] functions and minimize the integral on the right side of Eq. (1) with respect to the expansion coefficients. In- ∗ ∗ creasing the number of basis functions generally lowers 0 ≤ E[ψ]= dτψ (H − E0)ψ = dτψ Hψ − E0. (1) Z Z the bound obtained for E0. An important feature of this procedure emerges if we consider a trial function of the 0 The exact ground state wave function ψ satisfies the form ψ = ψ0 + δψ where δψ is “small” by some mea- Schr¨odinger equation, Hψ0 = E0ψ0. Using the latter sure. Inserting this trial function into Eq. (1) gives the 0 in Eq. (1) shows that E[ψ] = 0 when ψ = ψ and sug- variation δE = E[ψ0 + δψ] as gests a strategy to find an upper bound to E0: write the

∗ ∗ 2 δE = dτ(ψ0 + δψ) (H − E0)(ψ0 + δψ)= dτδψ (H − E0)δψ = O(δψ ). (2) Z Z

One says that the energy functional E[ψ] is “station- the rest of the time was given over to (Anderson 1999) ary” in the sense that a trial function that differs from ψ0 essentially everything Schwinger knew about. by a small amount (first order in δψ) produces an energy All about Green functions, all about nuclear which differs from E0 by an amount that is very small physics and so on. All the numerical tricks (second order in δψ). For that reason, minimizing E[ψ] he had devised to solve quantum mechanical with respect to a set of variational parameters produces problems and nuclear physics problems. . . a much better estimate for the ground-state energy than there was a lot of physics and there were a one might have supposed. The elegance and generality lot of variational techniques, for example to of this technique must have appealed powerfully to the solve the deuteron. . . . He was also start- young Kohn, because he “read many of the old papers ing to build the machinery that was going to on the subject and variational methods became the first solve the problems of quantum electrodynam- tool in my theoretical physics toolbox” (Kohn 2013b). ics. We were treated to a lot of that machin- In the spring of 1946, Kohn completed his MS studies, ery. taught calculus and analytic geometry as an Instructor for the Mathematics Department, and applied to a dozen Kohn himself has given one of the best descriptions of or so PhD programs with the clear idea to study theo- Schwinger’s teaching style (Kohn 1996): retical physics. Offers of admission with financial sup- Attending one of his formal lectures was com- port came from Rudolf Peierls at Birmingham and Eu- parable to hearing a new major concert by a gene Wigner at Princeton, among others. Kohn accepted very great composer, flawlessly performed by the Birmingham offer on the advice of Infeld, who knew the composer himself. . . Old and new mate- Peierls personally (Kohn 2013b). One day later, an offer rial were treated from fresh points of view and arrived from Harvard which included a prestigious Arthur organized in magnificent overall structures. Lehman Fellowship. Kohn again consulted Infeld, who The delivery was magisterial, even, carefully without hesitation told him to communicate his regrets worded, irresistible like a mighty river. He to Peierls, accept the offer from Harvard, and try to work commanded the attention of his audience en- for the young physics superstar Julian Schwinger. [35] tirely from the content and form of his mate- Accordingly, a somewhat insecure twenty-three year rial, and by his personal mastery of it, with- old Walter Kohn arrived on the Harvard campus in the out a touch of dramatization. fall of 1946 as one of a group of about thirty first-year Quite early on, the chairman of the Harvard Physics graduate students. The twenty-eight year old Schwinger department, John Van Vleck, approached Kohn and had joined the Harvard faculty the previous spring and asked him whether he would like to work with him on immediately began teaching a three-semester sequence of a solid state physics problem. [36] Kohn was not inter- courses on special topics in theoretical physics (Schweber ested in solid state physics and instead presented him- 1994). Kohn and his cohort stepped into the middle of self to Schwinger as Infeld had urged him to do. [37] He this sequence as a supplement to their required courses in described to Schwinger his experiences in Toronto and classical mechanics, electrodynamics, quantum mechan- (Kohn 1998, 2001b) ics, and statistical mechanics. In principle, Schwinger’s course was devoted to nuclear physics. In practice, he Luckily for me, we shared a common inter- devoted part of the time to a highly personal exposition est in the variational methods of theoreti- of quantum mechanics in the style of Dirac (1935) and cal physics. . . . He accepted me within 10

minutes as one of his 10 PhD students. He suggested I should try to develop a Green function variational principle for three-body scattering problems, like low-energy neutron- deuteron scattering, while warning me omi- nously that he himself had tried and failed.

Schwinger was an acknowledged expert in the use of both variational principles and Green functions to solve a wide variety of problems. In Eq. (1), an energy functional E[ψ] with the stationary property δE = 0 made it possible to estimate the ground state energy of a bound-electron system like helium. For a scattering problem, the energy is known and one is led to seek stationary variational functionals for other quantities. An example is the Green function, an energy-dependent operator defined in terms of the system Hamiltonian by Julian Schwinger (ca. late 1940’s) was Kohn’s PhD supervisor at Harvard University. 1 G(E)= . (3) Source unknown. E − H Kohn did not have a close personal relationship with his advisor. [41] None of Schwinger’s students did, in part For two-body scattering, one writes H = Hf +V (r) where because it was notoriously difficult to schedule a per- Hf ϕ = Ef ϕ is the Schr¨odinger equation for a free par- ticle and V (r) is the potential responsible for the scat- sonal meeting with him. Kohn met with him only “a tering. Using a coordinate-space representation of the few times a year” (Kohn 1996) and according to John David Jackson, an MIT graduate student who listened to corresponding free-particle Green function Gf , the scat- tered wave function ψ satisfies the integral equation, [38] Schwinger’s lectures at Harvard, there was one occasion where “Kohn was miffed by Julian’s unavailability. He completed his thesis, wrote up a paper, and submitted it 3 ′ ′ ′ ′ ψ(r)= ϕ(r)+ d r Gf (r, r |Ef )V (r )ψ(r ). (4) to Physical Review without ever consulting him” (Mehra Z and Milton 2000). Such feelings must have passed quickly because, in a moving tribute at a memorial symposium Kohn worked on the three-body Green function prob- after Schwinger’s death, Kohn makes it clear that (Kohn lem for half a year before abandoning it. [39] Instead, 1996) he generalized Eq. (1) and developed a variational prin- It was during these meetings, sometimes more ciple for the two-body scattering phase shift, a quan- than 2 hours long, that I learned the most tity which characterizes the final state when two par- from him. . . . to dig for the essential; to ticles interact via a short-range potential. [40] He also pay attention to the experimental facts; to try derived a variational principle for the scattering ampli- to say something precise and operationally tude for two-particle scattering with an arbitrary inter- meaningful, even if one cannot calculate ev- action potential. For both cases, Kohn borrowed from erything a priori; not to be satisfied until one the Rayleigh-Ritz method and expanded the trial scat- has embedded his ideas in a coherent, logical, tering wave function in a set of basis functions with the and aesthetically satisfying structure. . . . I correct long-distance behavior. Finally, he derived a vari- cannot even imagine my subsequent scientific ational principle for the elements of the scattering matrix life without Julian’s example and teaching. for the special case of nuclear collisions where multiple disintegrations are energetically forbidden. It is inter- Besides Schwinger, Kohn benefitted from members of esting that Kohn made no use of Schwinger’s “beloved his graduate student cohort who either contributed ma- Green functions” (Kohn 1998) in his thesis, “Collisions terially to his education at Harvard or who played an of light nuclei”, or in the published version of his thesis important role subsequently. One group consisted of (Kohn 1948). He did, however, use his scattering ampli- fellow Schwinger students: Kenneth Case, Frederic de tude variational principle to rederive an alternative vari- Hoffmann, Roy Glauber, Julian Eisenstein, Ben Mottel- ational principle for the phase shift that Schwinger had son, and Fritz Rohrlich. Another group did their PhD derived in his spring 1947 theoretical physics class using work in other areas of physics: the theorists Thomas Eq. (3) and reported at a meeting of the American Phys- Kuhn, Rolf Landauer, and Philip Anderson, and the ex- ical Society (Schwinger 1947). The latter is commonly perimentalists Nicolaas Bloembergen, George Pake, and called “Schwinger’s variational principle for scattering” Charles Slichter. Joaquin Luttinger, another MIT grad- (Adhikari 1998, Nesbet 2003). uate student who made the short trip to Harvard to au- 11 dit Schwinger’s classes, later became a close friend and Eidgen¨ossische Technische Hochschule in Z¨urich. Never- a scientific collaborator. All these members of Kohn’s theless, he was painfully aware that he had so far made student network went on to have successful scientific ca- only a “very minor contribution” to field-theory research. reers.[42] Glauber, Mottelson, Anderson, and Bloember- This fact, and the stunning quality of the latest achieve- gen won Nobel Prizes themselves. ments by Schwinger and Richard Feynman, made it easy Walter’s life changed profoundly in two important for him when (Kohn 1998) ways when he accepted an offer by Schwinger to remain at Harvard as a post-doctoral fellow. First, the income Van Vleck explained to me that he was about from this job permitted him to bring to Boston and marry to take a leave of absence and ‘since you are Lois Mary Adams, a former nursing student he had met familiar with solid state physics’, he asked at the University of Toronto who had been working in me if I could teach a graduate course on this New York City while he finished his PhD (SDUT 2010). subject he had planned to offer [for the fall A baby daughter soon arrived and family responsibilities 1949 semester]. This time, frustrated with were added to the research and teaching responsibilities my work on quantum field theory, I agreed. that came with his position as Schwinger’s assistant. The Kohn not only taught the solid-state physics course, he research project he undertook was an investigation of the collaborated with Harvard graduate student Richard Al- electromagnetic properties of mesons done in collabora- lan Silverman to find approximate numerical solutions of tion with fellow Schwinger post-doc Sidney Borowitz. His the Schr¨odinger equation for the purpose of calculating teaching consisted of an introductory physics course in the cohesive energy of metallic lithium (Silverman and the summers of 1949-1950 and a junior/senior level clas- Kohn 1950). [46] Using the same numerical data, he and sical mechanics course in the summer of 1950. [43] Bloembergen estimated the Knight shift for lithium, the The second profound change in Kohn’s life occurred latter being a measure of the electron wave function am- through the good offices of John Van Vleck, the solid plitude at the atomic nucleus accessible to experimenters state theorist he had rebuffed as a thesis supervisor. Van using magnetic resonance techniques (Kohn and Bloem- Vleck re-enters the story because Kohn supplemented his bergen 1950). It is notable that Silverman and Kohn summer 1949 income by working for the Polaroid Corpo- conclude with the remark, “One of us (W. Kohn) is in- ration at their Cambridge, Massachusetts research labo- vestigating the cohesive energy by means of a variation it- ratory. His job was to discover the mechanism whereby eration procedure based on the integral equation [Eq. (4) high-energy charged particles produce an image when of this paper] and a Green function appropriate to a pe- they impinge on photographic plates. [44] This task re- riodic lattice.” quired a knowledge of solid state physics, which he ac- Kohn, a naturalized Canadian citizen since 1943, quired by reading Frederick Seitz’ The Modern Theory of looked everywhere in Canada and the United States for Solids (1940) and consulting with Van Vleck when nec- an entry-level academic position. Nothing turned up, essary.[45] but an early 1950 interview trip to the Westinghouse Re- search Laboratory in Pittsburgh, Pennsylvania bore fruit even though his foreign citizenship precluded a job offer from Westinghouse. In Pittsburgh, Kohn stayed at the home of Alfred Schild, a friend from Toronto who had found a job teaching mathematics at the Carnegie Insti- tute of Technology. Schild told him that the chairman of the physics department, Frederick Seitz, had just re- signed and was moving his solid state group to the Uni- versity of Illinois (Seitz 1994). Perhaps there was an op- portunity at Carnegie Tech itself. The new chair, nuclear physicist Edward Creutz, interviewed Kohn and offered him a job as an Assistant Professor 48 hours later.[47] It turned out that Creutz needed someone who could teach solid state physics and mentor a few graduate students who had lost their advisors when the Seitz group left (Kohn 1998). Kohn was delighted to accept. Harvard’s John Van Vleck facilitated Kohn’s transition from nuclear physics to solid state physics. Courtesy of the UW-Madison Archives. III. PORTRAIT OF THE PHYSICIST AS A YOUNG MAN By this time, Kohn’s first paper with Borowitz had ap- peared (Borowitz and Kohn 1949) and he had applied to Walter Kohn was thinking about the start of his aca- the National Research Council for a Fellowship to spend demic career at Carnegie Tech when, in March 1950, the the 1950-1951 academic year with Wolfgang Pauli at the National Research Council approved his application for a 12 fellowship to spend a year in Europe. Ed Creutz agreed ing (Kohn 1952a) and a non-Green function variational to a one-year leave of absence, but only if Walter agreed principle for electron waves in a periodic potential (Kohn to teach solid state physics for the fall 1950 semester. At 1952b). the same time, Kohn was having second thoughts about Back in Pittsburgh, the Physics Department had his original plan to work with Wolfgang Pauli. This led changed somewhat during Walter’s absence. The senior him to seek and secure the approvals needed to switch the experimentalist Immanuel Estermann had left to head venue for his fellowship year from Z¨urich to Niels Bohr’s the physics section of the Office of Naval Research (ONR) Institute for Theoretical Physics in Copenhagen. Accord- and his last PhD student, Simeon Friedberg, had taken ingly, Kohn moved to Pittsburgh, taught his course, and over his low-temperature physics laboratory. A senior left for Copenhagen at the end of the term. [48] When he theorist, Gian-Carlo Wick, had joined the faculty from arrived at the Institute in January 1951, Walter Kohn Berkeley and Roman Smoluchowski, an expert in the the- was an expert in scattering theory who had begun to ory of defects in solids, had transferred to the Physics De- think of himself as a solid state physicist. Unfortunately, partment from the Metallurgy Department. A young ex- “no one in Copenhagen, including Niels Bohr, had even perimenter, Jacob Goldman, and a young theorist, Paul heard the expression “solid state physics’” (Kohn 1998). Marcus, had joined the solid state physics group to com- Kohn managed to publish two papers that year, but plement the senior experimentalist Emerson Pugh. One more important to his future was the fact that the Insti- familiar face was Norman Rostoker, a Toronto native tute attracted a steady stream of short-term and long- who had graduated from his hometown university as a term visitors from around the world from whom he could physics major one year behind Kohn and then received learn new physics.[49] It was good luck for Kohn that his PhD at Carnegie Tech under Pugh’s supervision (Ros- post-war freedom of movement motivated Bohr to orga- toker 2013). Kohn and Rostoker had become friends dur- nize a meeting for all foreign physicists who had ever ing the fall 1950 semester and Norman was still working worked at the Institute (Rozental 1967). The resulting in the Physics Department as a post-graduate research Conference on Problems in Quantum Physics (July 6-10 scientist when Kohn returned from Europe. 1951) was attended by an outstanding collection of the- The fall 1952 semester found Kohn teaching thermo- oretical physicists, many of whom Walter was able to dynamics to undergraduates and nuclear physics to grad- incorporate into his expanding professional network. [50] uate students. He was also named as co-principal investi- A few weeks later, Kohn was tapped to lecture on solid gator with Jack Goldman on an ONR contract to conduct state physics for two weeks at the first Summer School of solid state research. [53] His main research project was to Theoretical Physics organized by C´ecile DeWitt at Les develop a Green function method to calculate the energy Houches, near Chamonix in the French Alps. [51] band structure for crystalline solids. In other words, he At the end of 1951, Bohr wrote a formal evaluation wanted to use Eq. (3) to solve the Schr¨odinger equation which concluded that (Bohr 1951) to find the energy eigenfunctions and eigenvalues for elec- Dr. Kohn has proved himself a highly qual- trons moving in a periodic potential. A distraction arose ified theoretical physicist with great knowl- in the spring 1953 semester when Carnegie Tech learned edge of a wide field of problems. His ability that Walter had received job offers from the Department to stimulate others in their work and his will- of Mathematics at McGill University in Montreal and the ingness to assist them with his knowledge has Physical Research Department at Bell Telephone Labo- been of great value to the many members of ratories in Murray Hill, New Jersey. Evidently, others our group. besides Niels Bohr had formed a very positive impres- sion of this new Assistant Professor. In the end, Carnegie This good opinion led Bohr to arrange a Rask Ørsted Tech retained his services by promoting him to Associate Foundation fellowship for Kohn so he could remain in Professor (WKP 1953a) after only three semesters of aca- Copenhagen through the summer of 1952. This was wel- demic service. come news because Walter and his family enjoyed living Kohn had brought to Carnegie Tech the germ of in Denmark. Moreover, he had just begun a scientific his Green function method to solve the electron band- collaboration with Res Jost, a Swiss mathematical physi- structure problem. He recruited Norman Rostoker to cist five years his senior who had lectured on quantum help with the numerical calculations and that activ- field theory at the Les Houches summer school. [52] Jost ity continued (part-time) while Walter was in Copen- was interested in scattering theory and his “predilection hagen.[54] The work accelerated when Kohn returned for mathematical rigor” (Enz 2002) struck a responsive to Pittsburgh and he reported their still-unpublished re- chord in Kohn. Together, the two theorists completed sults at two invited talks, one at the June 1953 Summer three papers (including an ‘inverse scattering problem’ Meeting of the American Physical Society and one at a where one deduces characteristics of the scattering poten- July 1953 Gordon Research Conference devoted to the tial from phase shift information) before Kohn returned Physics and Chemistry of Metals. The latter was a par- to Carnegie Tech to begin the 1952-1953 academic year ticularly prestigious venue and it is notable that of the (Jost and Kohn, 1952a, 1952b, 1953). On his own, Kohn seven theorists invited to speak, the three youngest (by studied the validity of the Born expansion for scatter- far) were Walter Kohn, Jacques Friedel, and David Pines 13

(WKP 1953b). Friedel, an expert on the theory of met- by Frederick Seitz and David Turnbull in 1954 to provide als and alloys, and Pines, an expert on electron-electron “broad surveys of fields of advanced research that serve interactions in solids, had both published half a dozen to inform and stimulate the more experienced investiga- papers in their fields by the time of the Gordon Con- tor” (Seitz and Turnbull 1955). The inaugural volume ference. It is an indication of Kohn’s rising reputation contained articles devoted to five issues: the band struc- that he had published only one full-length paper in solid ture problem, the properties of valence semiconductors, state physics by this time. Kohn’s Green function paper electron-electron interactions, cohesion in solids, and the (Kohn and Rostoker 1954) finally appeared in the June theory of order-disorder phase transitions. Kohn had 1 1954 issue of the journal Physical Review. Therein, he already made a significant contribution to band struc- and Rostoker (Kohn 1998) ture theory and he now added to his repertoire research projects devoted to semiconductors and to the effects of developed a theory for the energy band struc- the electron-electron interaction (soon relabeled many- ture of electrons in solids harking back to body physics). The total energy (cohesion) problem be- my earlier experience with scattering, Green came a central concern when he developed density func- functions, and variational methods. We tional theory a decade later. [59] showed how to determine the band structure Kohn became interested in semiconductors because his from a knowledge of purely geometric struc- flirtation with permanent employment at Bell Telephone ture constants and a small number ( ∼ 3) of Laboratories led to a summer consulting arrangement scattering phase shifts of the potential in a that lasted from 1953 to 1966. His first summer project, a single sphericalized cell. theoretical study of the damage done to germanium crys- It happens that the same basic idea had been published tals after bombardment by energetic electrons, was mo- several years earlier by the Dutch physicist Jan Korringa. tivated by experimental results obtained at Bell Labs by However, Korringa (1947) included no numerical appli- Walter Brown and Robert Fletcher (Brown et al. 1953). cations and his paper went largely unnoticed. [55] Kohn These experiments, in turn, were part of an enormous in- and Rostoker illustrated their method by calculating the house effort to investigate the properties of the elemental energy as a function of wave number for the 2s conduction semiconductors germanium and silicon after the 1947 in- band of lithium metal and comparing their results with vention of the transistor at Bell Labs by John Bardeen, previous calculations in the literature. It is entirely char- Walter Brattain, and William Shockley (Millman 1983). acteristic of Kohn that he did not take his band structure Kohn was happy to return to Bell Labs summer after formalism and begin applying it to one material after an- summer, both to gain access to exciting experimental other. [56] Instead, he made one use of his lithium results results and for the opportunity to interact with senior (Kohn 1954) and then proceeded to expand his personal theorists on the Bell Labs staff like Conyers Herring and research activities into other areas of solid state physics. Gregory Wannier and junior theorists closer to him in To understand the choices he made, we interrupt our nar- age like Peter Wolff and his Harvard classmate Philip rative briefly to survey the research agenda of solid state Anderson. He would later remark that “I owe this in- physics in the mid-1950’s. stitution my growing up from amateur to professional” Wartime developments in computers, instrumentation, (Kohn 1998). and materials processing had a profound effect on the is- Walter’s 1954 ‘summer vacation’ at Bell Labs was par- sues addressed by solid state physicists at the mid-point ticulary important because he began a long-lasting scien- of the twentieth century. [57] In June 1954, the National tific collaboration with Joaquin Luttinger, another con- Science Foundation and the American Society for Engi- sultant to the semiconductor group. [60] They worked to- neering Education sponsored a meeting at Carnegie Tech gether to create a rigorous “effective mass theory” for the attended by representatives from forty-five colleges and electronic energy levels produced when impurity atoms universities and several industrial and government lab- are purposely substituted for germanium or silicon atoms oratories. According to the conference co-chair, Profes- in pure crystals of the latter. The crucial importance of sor Jack Goldman, [58] the purpose of the meeting was these impurities and their quantum mechanical states to to “make more definitive the state of knowledge of solid the extraordinary electrical properties of semiconductors state physics and the levels at which various parts of it had been explained qualitatively by William Shockley in may be expected to be integrated into engineering educa- his seminal treatise, Electrons and Holes in Semiconduc- tion” (Goldman 1957). To this end, the conferees identi- tors (1950). However, careful electron spin resonance and fied six broad areas of active solid state physics research: cyclotron resonance experiments at Bell Labs and else- the structure of crystalline matter, metals and alloys, where demanded a quantitative theory. Not for the first surfaces, magnetism, semiconductors and dielectrics, and time and not for the last time, Kohn combined the cre- non-crystalline materials. ation of a novel and sophisticated theory with variational One needed to attend conferences reserved for special- calculations designed to produce numbers for comparison ists to learn the cutting-edge issues in each area. Happily, with measurements for specific material systems. That the same purpose was soon served by the articles pub- fall, Kohn and Luttinger completed three substantial lished in Solid State Physics, a series of volumes initiated papers in semiconductor physics (Luttinger and Kohn 14

1955, Kohn and Luttinger 1955a,b) and thereby finished theory ten years later: “how can you use a band model in a virtual dead heat with Berkeley solid state theo- when the potential felt by the electrons is not periodic” rist, Charles Kittel, who published similar work inde- (Arrott 2013)? pendently (Kittel and Mitchell 1954, Dresselhaus et al. As 1955 turned into 1956, Walter found himself think- 1955). [61] ing more and more about the effective mass equation he had derived with Luttinger for the energy levels of impu- rity states in silicon. Their “one-particle” method treated the impurity atom as unaware of its silicon host except for whatever influence could be captured by two num- bers: an effective mass m∗ which parameterized the ar- rangement of atoms in the silicon crystal and an effective dielectric constant κ∗ which parameterized the ability of the silicon conduction electrons to “screen” or “shield” the Coulomb potential produced by a positively charge impurity embedded in the semiconductor. [64] Why then did the energy levels calculated using the effective mass equation agree so very well with the energy levels mea- sured in the laboratory? Surely, he reasoned, it must be that “this equation can be derived from some very general properties of the entire many-electron wave func- tion without any recourse to the one-particle picture” Joaquin Luttinger was Kohn’s principal (Kohn 1957). For the first time, Kohn attacked the scientific collaborator in the 1950’s. quantum-mechanical “many-body problem” in solid state Courtesy of Walter Kohn. physics where the repulsive Coulomb interaction between all pairs of electrons is taken seriously. Working alone, he In the spring of 1955, Walter worked hard to convince managed to demonstrate his assertion for a hypothetical the British theoretical solid state physicist Harry Jones situation where the charge on the impurity nucleus ex- to accept an offer of a chaired position at Carnegie Tech. ceeds that of the other nuclei by an infinitesimal amount. Jones had co-authored (with Nevill Mott) the influen- He announced this result in a comprehensive review of tial book The Theory of the Properties of Metals and Al- the Kohn-Luttinger theory written for Seitz and Turn- loys (1934) and he had spent the spring 1954 semester as bull’s Solid State Physics series (Kohn 1957a). A full a visiting professor in Kohn’s department. Kohn wrote account appeared later (Kohn 1957b). to Jones and pointed out that “all of us in solid state Meanwhile, back at Bell Labs, the resident theoretical physics, as well as all the people in metallurgy, would physicists had successfully convinced the vice-president be delighted to see you come here. With your field of for research, William Oliver Baker, to create (Anderson interest, I honestly think that probably no other school 2011) in this country could offer you better opportunities for creative work along research and teaching lines” (WKP a separate ‘super-department’ for theorists . 1955). Jones ultimately declined for personal reasons. . . with post-doctoral fellows, a rotating boss The undeterred Kohn pursued the physics of semicon- on whose identity we were consulted, sabbat- ductors and metals simultaneously and submitted two icals, a travel budget under our control, and manuscripts to the Physical Review. The first paper, a spectacular summer visitor program. . . . co-authored with his first Carnegie Tech Ph.D. student, One of the reasons for our success with man- Daniel Schechter, reported calculations for the wave func- agement was the fact that for several years we tions and energy levels associated with shallow (weakly had had Walter Kohn and Quin Luttinger as bound) impurity states in germanium. [62] The second regular summer visitors and they had become paper reported a Knight shift calculation for metallic so useful that our bosses desperately wanted sodium with Terje Kjeldaas, a full-time employee of the to attract them permanently. Westinghouse Research Laboratories in East Pittsburgh (Kohn and Schechter 1955, Kjeldaas and Kohn 1956). [63] The advent of a ‘spectacular summer visitor program’ For later reference, it is important to note that Kohn meant that an unusually large number of theoretical acted as an informal consultant to the transition-metal physicists passed through and interacted with Kohn and magnetism groups of his faculty colleagues Emerson Luttinger during their 1956 summer stay at the Labs. [65] Pugh and Jack Goldman. Goldman’s PhD student An- A hot topic was the effect on the properties of semicon- thony Arrott recalls Kohn’s surprise when Arrott suc- ductors when one systematically increased the number of cessfully used a simple energy band model to analyze impurities present. When the impurity concentration is his magnetic data for concentrated Cu-Ni alloys. At Ar- low and the temperature is low, it was well known that rott’s oral thesis defense, Kohn asked him a question that electron scattering from impurities is the main source of foreshadowed his motivation to invent density functional a solid’s electrical resistance. What happens when the 15 concentration of impurities is high? Luttinger had been work. [69] Brueckner, in turn, led an effort by his Penn thinking about the general subject of electric current flow colleagues to hire Kohn away from Carnegie Tech. A in solids already in connection with his studies of the Hall similar effort was mounted by the Physics Department effect in ferromagnets and it was not difficult to convince of the University of Chicago (WKP 1957). Kohn took Kohn to work with him to produce as rigorous a the- these overtures seriously and Carnegie Tech responded ory of electrical conductivity as they could. [66] After all, by awarding him tenure and promoting him to the rank Luttinger’s understanding of the experimental facts for of Professor with a substantial increase in salary (WKP the Hall effect came directly from a review paper writ- 1958). Walter made his decision to return to Pittsburgh ten by Kohn’s Carnegie Tech colleagues Emerson Pugh while completing his sabbatical and spending the spring and Norman Rostoker (Pugh and Rostoker 1951). The 1958 semester with Harry Jones and his group at the fruits of that summer’s labors were two long papers on Department of Mathematics of the Imperial College of the quantum theory of electrical transport in solids [67] Science and Technology in London. (Kohn and Luttinger 1957, Luttinger and Kohn 1958). The many-body revolution introduced new ideas into In contrast to their effective mass theory work, which solid state physics like the quasi-particles of Lev Landau more reflected Kohn’s style to address important phys- (1956) and new objects for study like the one-particle and ical questions with intuition, a good idea, and mathe- two-particle Green functions exploited by Victor Galitskii matical elegance, the transport theory papers more re- and Arkday Migdal (1958). [70] The physical and math- flected Luttinger’s preference for general formalism and ematical elegance of this subject caused some solid state mathematical rigor. In this way, the two young theorists theorists to focus their attention exclusively on prob- enlarged each others’ perspectives of their craft. lems where many-body effects dominate the physics. [71] For Walter Kohn, professor of physics, the calendar Kohn did not follow this particular path because not ev- change from 1956 to 1957 meant little more than a change ery problem that interested him demanded a many-body in his teaching assignment from statistical mechanics for analysis. For example, a theorem due to Felix Bloch physics majors to classical physics for engineers. How- (1928) demonstrated that the eigenfunctions and energy ever, for Walter Kohn, solid-state physicist, the new eigenvalues of the Schr¨odinger band for a spatially peri- year saw changes in his field that had a profound effect odic system like a crystal have the form on his future research efforts. In the words of Cana- ψk(r) = exp(ik · r)uk(r) and E(k), (5) dian physicist Allan Griffin, 1957 was a “magic year” when “the way all theoretical physicists thought about where the three quantum numbers collected in the vector interacting many-body systems underwent a revolution” k = (kx, ky, kz) are real numbers confined to a finite vol- (Griffin 2007). The key event was the realization that ume of the three-dimensional k-space called the Brillouin the methods of quantum field theory could be applied zone, and the function uk(r) has the spatial periodicity of with equal success to study the many-electron problem the crystal. While at Imperial College, Kohn performed in solid-state physics. [68] In particular, diagrammatic an extensive study of the properties of the Bloch solutions methods like those invented by Richard Feynman (1949) when k is a complex-valued vector. This allowed him to to study quantum electrodynamics made it possible to make precise statements about the exponential decay of define a perturbation theory that remained consistent as a class of spatially localized wave functions first intro- the number of particles in a system increased. Feynman duced by Wannier (1937). In a separate project, Kohn diagrams posed no problem for a Harvard PhD like Kohn analyzed the motion of Bloch electrons in a magnetic field who was both well-trained in quantum mechanics and fa- with more rigor than had been done previously. He suc- miliar with quantum field theory from Julian Schwinger’s ceeded to show that an approximation for this problem lectures. He also had a ready-made problem: his own first made by Peierls (1933) had a much broader range desire to understand the success of the Kohn-Luttinger of validity than previously thought. effective mass equation from a many-body point of view. Kohn’s Imperial College projects in mathematical A breakthrough paper by Jeffrey Goldstone (1957) pro- physics did not disengage him from the more practical vided all the technical details he needed. aspects of solid state physics. [72] For example, he was Walter became a naturalized citizen of the United invited to the 1958 International Conference on Semi- States in 1957 and he had arranged a sabbatical leave conductors in Rochester, New York, to report his many- from Carnegie Tech for the 1957-1958 academic year. body analysis of the effective mass equation for shallow He spent the fall of 1957 at the Physics Department impurity states. While there, he attended a session de- of the University of Pennsylvania and it was there that voted to the calculation of energy bands and listened to he wrote up Kohn (1958), his first contribution to the a talk by fellow-theorist James Phillips. Kohn asked many-body revolution–a diagrammatic analysis of the Phillips whether he was doing “physics or magic” be- static dielectric constant of an insulator. At the end cause Phillips’ ‘pseudopotential method’ reproduced the of this paper, Kohn acknowledges “stimulating conversa- results of much more elaborate band structure calcula- tions” with Keith Brueckner, a senior member of Penn’s tions for silicon and germanium using only three param- Physics Department whose own thinking about the quan- eters for each (Bassani and Tosi 1988). Similarly, Kohn tum many-body problem had stimulated Goldstone’s paid careful attention when the phonon spectrum of a 16 crystal was measured for the first time (Brockhouse and as post-doctoral fellows in Kohn’s theoretical solid-state Stewart 1958) and also when the Fermi surface of a metal physics group. was measured for the first time (Pippard 1957, Gold, The problem Kohn set for Langer and Vosko was the 1958). [73] shielding of a positively charged impurity embedded in These experimental breakthroughs stimulated Walter’s a metal host. This is the analog of the problem Walter scientific imagination and he soon completed a simple had studied previously for the case of a non-metal host. and elegant analysis which predicted that the phonon Kohn’s many-body perturbation theory calculation for spectrum of a metal possesses observable “anomalies” the non-metal case confirmed the classical result that the which depend only on the existence and shape of the Coulomb potential q/r at distance r from a point charge Fermi surface. More precisely, the Fermi surface lo- q in vacuum changes to q/κr when the point charge is em- cates a singularity in a linear response function which bedded in an insulator with dielectric constant κ. Nevill describes the ability of the conduction electrons to screen Mott (1936) studied the screening of a point charge in the ions which move during a lattice vibration. The a metal in connection with a calculation of the electrical short communication which described what came to be resistivity of dilute metal alloys. He used semi-classical known as “Kohn anomalies” was one of four manuscripts Thomas-Fermi theory (March 1957) and showed that the he submitted for publication during the 1958-1959 aca- potential q/r in vacuum changes to (q/r)exp(−r/ℓ) in a demic year at Carnegie Tech (Kohn 1959a,b,c, Ambe- metal. The screening length ℓ depends on the density of gaokar and Kohn 1959). Two longer papers described conduction electrons and takes the value 1-2 A˚ in a typ- the results of the projects begun at Imperial College and ical metal. Jacques Friedel (1958) revisited this problem the paper co-authored by his PhD student Vinay Ambe- using a scattering theory method and found that the dis- gaokar reported a new sum rule for insulators. The Kohn turbance of the electronic charge density at a distance r anomaly and Ambegaokar papers appeared in the same from the impurity charge, δn(r), varied as issue of Physical Review Letters, a journal spun-off from cos(2k r + ∆) the Physical Review to provide “speedy publication” of δn(r) ∝ F , (6) “new discoveries of major importance and for significant r3 contributions to highly active and rapidly changing lines where k and ∆ are two constants. The 1/r3 decay of of research in basic physics” (Goudsmit and Trigg 1964). F this function falls off much more slowly with distance Ambegaokar (2004, 2013) recalls that than the exponential variation predicted by Mott’s the- Before going on leave [to Pennsylvania], Wal- ory and thus implies that the effect of isolated impu- ter advised me to take a second course in rities might not be completely screened at the position quantum mechanics taught by Gian-Carlo of nearby atoms. The oscillatory behavior in Eq. (6) is Wick even though I had not finished a first an intrinsically quantum effect. Kohn did not fully be- course. . . . Upon his return [from Eng- lieve Friedel’s result and therefore asked his post-docs land], he suggested a research project that to attack the problem themselves (Kohn 2012b). Much was very much to my taste. We met for at to his surprise, Langer and Vosko (1959) fully confirmed least an hour a week and his supervision was Friedel’s formula using diagrammatic many-body pertur- both precise and constructive. He thought bation theory. [75] Kohn promptly applied their results to hard during our meetings to keep the project a quantitative calculation of the magnitude of the nuclear moving along. . . . [Walter] could be formal magnetic resonance signal in copper metal when small as a person, but he opened up considerably amounts of impurity atoms are introduced (Kohn and with people he respected. He got me a sum- Vosko 1960). Confirmation of the Kohn-Vosko theory mer job at Bell Labs and we played tennis came from comparison with experiments performed by together there frequently. Theodore Rowland at the Union Carbide Metals Com- pany (Rowland 1960). Kohn knew Rowland from his Another student, James Langer, was an undergraduate Harvard days when Rowland was a PhD student of Nico- physics major at Carnegie Tech from 1951-1955. Langer laas Bloembergen. never took a formal course from Kohn, but in his se- The summer of 1959 reunited Walter and Quin Lut- nior year, “Walter somehow became my private instruc- tinger at Bell Laboratories. Once again, the pair pro- tor for a year-long supervised reading course. We went duced an interesting paper (Kohn and Luttinger 1960) through the first edition of Leonard Schiff’s classic text and once again, Kohn grappled with an offer from Keith Quantum Mechanics essentially cover to cover” (Langer Brueckner to leave Carnegie Tech. This time, however, 2003). Langer won a Marshall Scholarship to attend Brueckner was not soliciting on behalf of the University of graduate school in Great Britain and Kohn directed him Pennsylvania. He had resigned from Penn a few months to Rudolf Peierls at the University of Birmingham. [74] previously and his mission now was to convince Walter Langer earned his PhD for a problem in nuclear physics to help him create the Physics Department at the soon- and then returned to Carnegie Tech as an Instructor in to-open University of California at La Jolla (later San the fall of 1958. For the next year, he and Seymour Vosko Diego). Earlier in the year, Brueckner had flown Kohn (a recent PhD student of Gian-Carlo Wick) functioned to the beautiful site of the proposed campus to meet and 17 hear the vision of its principal advocate, Roger Revelle, old Harvard friend and magnetic resonance practitioner, the Director of the Scripps Institute for Oceanography. Charles Slichter, and the liquid helium experts William The salary was attractive and La Jolla seemed like an Vinen and Russell Donnelly. As it turned out, none of ideal place to relocate his wife and two elementary school- these people joined the UCSD faculty. [79] aged daughters.[76] Moreover, Ed Creutz, the man who Kohn’s arrival in San Diego in January 1960 coincided had hired Kohn at Carnegie Tech and who was now Vice- with his election as a Fellow of the American Physical President of Research at the General Atomics division of Society (APS), the professional organization of Ameri- the General Dynamics Corporation in San Diego, had can physicists. [80] Resettlement and administrative is- recently concluded a consulting contract with him. This sues dominated his time at first, so Kohn used his stu- time, the allure was too great and Walter agreed to sign dents, post-docs, and short-term visitors to pursue a on. His only condition was that Keith Brueckner must research agenda now focused primarily on the physics serve as chair of the new department (Brueckner 2013). of metals and alloys. [81] A new post-doctoral fellow, Walter’s research group had grown to include three Stephen Nettel, studied the spatial arrangement of elec- PhD students and four post-doctoral fellows by the fall of tron spins in the ground state of a homogeneous and 1959 when he submitted his resignation to the President non-interacting electron gas, a much-studied hypothet- of Carnegie Tech. His senior student, Vinay Ambegaokar, ical system composed of a collection of mobile electrons was one semester from graduation. His junior students, (with their mutual Coulomb interaction turned off) dis- Larry Glasser and Edwin Woll, Jr. were not too far from tributed throughout a uniform and static distribution of the beginning of their research so Walter invited them electrically neutralizing positive charge (Kohn and Net- both to join him in La Jolla. Woll chose to accompany tel 1960). Emile Daniel and former post-doc Seymour his advisor; Glasser remained in Pittsburgh and finished Vosko used many-body perturbation theory to study the his degree with Assistant Professor J. Michael Radcliffe sharpness of the Fermi surface for a fully interacting elec- (Glasser 2013). Post-docs Hiroshi Hasegawa and Robert tron gas at zero temperature and Tony Houghton studied Howard had arrived the previous fall from Tokyo and Ox- the specific heat and spin susceptibility of a dilute al- ford, respectively, and worked together on a problem mo- loy (Daniel and Vosko 1960, Houghton 1961). [82] The tivated by Kohn’s experience with shallow donor states numerical calculations of Daniel and Vosko confirmed in semiconductors (Hasegawa 2004). Hasegawa accom- Quin Luttinger’s analytic demonstration that electron- panied Kohn to San Diego while Howard moved on to a electron interactions do not destroy the sharpness of the permanent position at the National Bureau of Standards Fermi surface of an electron gas. This result was reported in Washington, D.C. Post-docs Emile Daniel and An- by Luttinger at a (retrospectively famous) meeting at- thony Houghton were former PhD students of Jacques tended by Kohn and Vosko on “The Fermi Surface” held Friedel in Paris and Geoffrey Chester (in the group of at Cooperstown, New York on August 22-24, 1960. (Lut- Rudolf Peierls) in Birmingham, respectively. Both be- tinger 1960, Harrison & Webb 1960). gan research projects in metal alloy physics before mov- The La Jolla campus of the University of California ing to the west coast. Kohn himself taught a graduate opened for business in the fall of 1960. There were no course in advanced solid state physics and gave a talk undergraduates (until 1964), but sixteen physics grad- on “The Electron Theory of Solids” at a one-day “Solid uate students arrived and began taking classes. Walter State Symposium” in New York City sponsored by the Kohn taught a course from 4:30 PM - 6:00 PM on Thurs- American Institute of Physics for the benefit of science days and from 9:30 AM - 11:00 AM on Saturdays using writers from national magazines and newspapers. [77] the textbooks Thermodynamics by Herbert Callen (1960) and Elements of Statistical Mechanics by Dirk ter Haar Even before leaving Pittsburgh, Walter worked hard (1954). [83] The smallness of the new department and the to recruit faculty members to his new Physics Depart- setting just steps from the Pacific Ocean fostered consid- ment in San Diego. He had immediate luck with three erable informality in the early days. Arnold Sherwood, a Bell Laboratories scientists, the statistical mechanician member of the first cohort of graduate students, recalled Harry Suhl, the nuclear magnetic resonance experimenter that “Kohn had a charming European formality even George Feher, and the superconductivity experimenter when he was trying to be informal” (Sherwood 2013). Bernd Matthias.[78] In an October 26 1959 letter to According to faculty member George Feher (2002): Keith Brueckner, Kohn laments Quin Luttinger’s deci- sion to choose Columbia University over UCSD and sug- Occasionally a student would come to class in gests several solid state and/or low-temperature physi- a bathing suit or scantily dressed. We didn’t cists whom Brueckner should approach (WKP 1959b). mind it, except perhaps Walter Kohn, who On the theoretical side, he proposed J. Robert Schrief- was a bit more formal than the rest of us. fer, the junior author of a breakthrough paper on the But he couldn’t tell the student off lest he be theory of superconductivity published in 1957, Volker called a stuffy professor. He finally solved this Heine, a specialist in the electronic structure of metals problem by telling the students that he didn’t from the University of Cambridge, and Philippe Nozi`eres, mind their behavior but did not want them to an expert in many-body theory trained at Princeton by acquire bad habits because some stuffy pro- David Pines. On the experimental side, he suggested his fessor might take offense. 18

Kohn flew to New York City on February 1 1961 to at- Professor: Nothing too strange. That’s Wal- tend the annual joint meeting of the American Physical ter Kohn taking his students for a walk after Society and the American Association of Physics Teach- lunch. ers. He was there to accept the 1961 Oliver Buckley Sometime in the spring of 1961, Keith Brueckner an- Solid State Physics Prize “for having extended and elu- cidated the foundations of the electron theory of solids” nounced that he was stepping down as the chair of the Physics Department after one year of service. He had ac- (PT 1961). This prize, endowed in 1952 by Bell Tele- phone Laboratories and named in honor its former pres- cepted the position of vice-president and technical direc- tor of the Institute for Defense Analyses in Washington, ident and board chairman, is awarded each year by the APS to “recognize and encourage outstanding theoreti- D.C. (Brueckner 2013) Therefore, against his expressed desire and much to his chagrin, Walter Kohn found him- cal or experimental contributions to solid state physics”. Some measure of the esteem carried by this honor may be self in the position of chair for the 1961 fall semester. As a department head at a start-up university, Kohn en- judged from the fact that four of the eight persons who won the Buckley Prize before Kohn later won a Nobel joyed opportunities and faced challenges that do not ordi- narily arise for administrators at established universities. Prize (William Shockley, John Bardeen, Clifford Shull, and Nicolaas Bloembergen). He was permitted to hire a dozen new faculty members (which doubled the size of his department), but the of- A few months later, Kohn submitted for publication fices and laboratories he could offer new recruits occupied his forty-eighth scientific paper in sixteen years. The temporary space that would soon revert to the Scripps first paragraph of this paper gives a good indication of Institution of Oceanography. One interesting hire was his his style, perspective, and level of engagement at this friend Norman Rostoker, who had turned himself into a point in this career (Kohn 1961): plasma physicist at General Atomics since he and Walter There has been considerable interest in recent had collaborated at Carnegie Tech. The senior faculty months in the effects of the electron-electron members recruited by Brueckner and Kohn had excel- interaction on the cyclotron resonance fre- lent research records, but many came from industrial or quency and de Haas-van Alphen oscillations government laboratories with no teaching experience. It of a gas of electrons. As some of the theoreti- was Walter’s responsibility to ensure that competent in- cal treatments of these problems use very so- structors staffed the courses offered to the first few classes phisticated methods, and others are based on of graduate students. At the same time, he maintained incorrect qualitative reasoning, we wish here a research group of never less than five persons (gradu- to present some simple considerations which ate students, post-doctoral fellows, and visitors), served we think shed some light on what has been a as an Associate Editor of the Journal of Mathematical rather confusing situation. Physics, team-taught a course on “Advanced Solid State Physics”, presented a university-wide lecture on “New The paper Walter cited as bringing “very sophisticated Viewpoints in the Theory of Matter”, and submitted four methods” to the issue at hand was written by his good research papers. [84] friend and collaborator, Quin Luttinger (1961). A more Within his group, Edwin Woll, Jr., Kohn’s PhD stu- precise statement would be that Luttinger’s work em- dent from Carnegie Tech was making good progress with ployed very sophisticated mathematical methods. Kohn’s semi-quantitative calculations of Kohn’s phonon anomaly paper exploited much simpler mathematics but rather so- in the metals aluminum, sodium, and lead. Walter also phisticated physical reasoning to reach the same conclu- began working with Michael Greene and Max Luming, sions. two PhD candidates from the 1960 crop of UCSD grad- At UCSD, Kohn initiated a weekly Theoretical Solid uate students, and Chanchal Majumdar, a student from State Lunch where faculty members, post-docs, and long- the 1961 class. Greene was tasked to use scattering the- term visitors made presentations to anyone who brought ory to compute the resistivity of liquid alkali metals. Max a bag lunch and cared to listen. If a short-term visitor Luming involved himself in calculations of the orbital sus- addressed the group, Kohn would take the visitor out ceptibility of dilute metal alloys but switched to theoret- to lunch and invite his personal research group to come ical particle physics after the publication of Kohn and along. After lunch, everyone was invited to take a walk Luming (1963). [85] Majumdar began a project on the around the campus. Twenty-five years later, this behav- theory of positron annihilation in metals. Overall, Kohn ior was parodied in a skit performed at a celebration acquired a reputation among potential theory students to commemorate the anniversary of the founding of the as a supervisor with very high standards who could as- UCSD Physics Department. The relevant dialog involves sign a thesis problem that might take a very long time to a visitor to the campus and a professor in the department complete (Feibelman 2012). This did not deter Michael (LJPS 1985): Greene (PhD 1965), who joined Kohn’s group (despite an initial lack of interest in solid state physics) because he Visitor: And what are those figures I see in was impressed by the thoughtful questions Walter asked the misty distance? It looks like a giant duck at seminars. Greene wanted to learn to think like Kohn followed in a straight line by giant ducklings. (Greene 2013). 19

tween the conducting state and the insulating state had been an issue for Kohn since his Bell Labs-inspired work with Luttinger on electrical transport (Kohn and Lut- tinger 1957, Luttinger and Kohn 1958). Subsequent pa- pers on the behavior of a point charge in a dielectric, the nature of Wannier’s spatially localized states in solids, and the electromagnetic properties of insulators contin- ued this theme (Kohn 1958, Kohn 1959a, Ambegaokar and Kohn 1960). His new work took seriously a sug- gestion by Nevill Mott (1949) that the many-body wave function in an insulator should be fundamentally differ- ent from the many-body wave function in a conductor. Kohn exploited a ground-breaking paper that focused at- tention on the gauge principle for the electromagnetic Walter Kohn at age 39 (1962). vector potential in quantum mechanics (Aharonov and Courtesy of Walter Kohn and Bohm 1959) and used a characteristically elegant method the John Simon Guggenheim Foundation. to calculate the electrical conductivity of a ring threaded by a line of magnetic flux. The result was a proof that the At the beginning of the fall 1962 semester, Walter learned that his colleague Norman Kroll would take over spatial organization of the electrons in an insulator cor- responds to a many-body wave function that breaks up as Physics chair beginning in the fall 1963 semester. into disconnected regions that do not overlap with one Right away, he made an application to the John Simon another. The published paper, Kohn (1964), has been Guggenheim Memorial Foundation for funds to support called a “a mine of ideas and methods” by no less an ex- a recuperative leave for the fall 1963 semester (Hohen- pert than Walter’s old Harvard classmate and Bell Labo- berg, Kohn, and Sham 1990). His plan was to spend ratories colleague Philip Anderson (Anderson 2012). [86] that time at the physics department of the Ecole´ Normale Now 40 years old, Walter Kohn was a mature solid Sup´erieure in Paris (Kohn, 1962). This was an ideal place state physicist whose scientific talent and taste in prob- to get back to full-time research. It was also an ideal place lems had produced results that were highly valued by his to renew his personal and scientific ties with three spe- peers. Two of those peers, David Pines and Charles Kit- cialists in his own field of theoretical solid state physics: tel, highlighted Kohn’s work four and eight times, respec- Jacques Friedel, Pierre-Gilles de Gennes, and Philippe tively in their (now classic) 1963 graduate level textbooks Nozi`eres. It was a bonus that the physics department Elementary Excitations in Solids and Quantum Theory of was just a short walk from the Luxembourg Gardens, his Solids. By the end of that summer, Walter’s manuscript favorite place in the whole world (Kohn 2001a). on the “Theory of the Insulating State” was ready for Kohn’s fellowship application proposed that he would submission and he had only to review some professional study the interaction of electrons and phonons in met- correspondence before he could depart for Paris. In ret- als (Kohn, 1962). This was a hot topic. The collisions rospect, the most important letter on his desk came from between electrons and the particle-like phonons had long Lu-Jeu Sham, a graduating PhD student from John Zi- been recognized as important for a proper description of man’s group at the University of Cambridge whom Kohn the electrical conductivity of metals (Ziman, 1960). How- had earlier recruited to become a post-doctoral fellow. ever, only five years earlier, John Bardeen, Leon Cooper, Kohn had written to Sham to inform him about his Paris and J. Robert Schrieffer had proposed a many-electron sabbatical and to urge him to come to San Diego as orig- wave function for a superconductor based on a model inally planned. Kohn proposed that Sham work on liq- where the electron-phonon interaction mediates an effec- uid metals with graduate student Mike Greene until he tive attractive interaction between pairs of electrons with (Kohn) returned to campus. The return letter from Sham opposite spin (Bardeen et al. 1957). Moreover, barely agreed to this plan (WKP 1963a). a year earlier, neutron scattering experiments had con- firmed Kohn’s own prediction of anomalies in the phonon spectra of metals (Kohn, 1959c). Accordingly, Kohn pro- IV. ALLOYS IN PARIS posed to spend the fall of 1963 generalizing the theory of Kohn anomalies. For good measure, he also proposed to develop a theory of the effect of electron-phonon interac- Walter Kohn’s base of operations in Paris was the tions on the optical properties of metals. Ecole Normale Sup´erieure, one of the elite grandes ´ecoles The Guggenheim Foundation responded positively to of the higher education system in France. His host was Walter’s application in the spring of 1963. This news the 31-year old Philippe Nozi`eres, an expert in many- must have been a great stimulant because he quickly body theory who had just collaborated with Joaquin Lut- completed a calculation which achieved “a new and more tinger to derive Landau’s theory of the Fermi liquid us- comprehensive characterization of the insulating state of ing diagrammatic perturbation theory (Nozi`eres and Lut- matter” (Kohn 1964). The fundamental difference be- tinger, 1962, Luttinger and Nozi`eres 1962).[87] The set- 20 ting inside the Physics Department at 24 rue Lhomond in found it, because his desire to reconcile these points of the Latin Quarter was not typical. According to Nozi`eres view was the immediate trigger for the creation of den- (Cheetham 1992, Nozi`eres 2012b), sity functional theory. The fundamental problem was to calculate the eigenfunctions and energy eigenvalues for a I had inherited a very magnificent office with binary alloy where A-type atoms replace a fraction of the an old desk made from an oak tree. It was gi- atoms in a perfect B-type crystal. If the replaced B-type gantic, eight times the size of a normal desk, atoms are chosen randomly, the resulting structure is no and I put Walter on one side and me on longer periodic and the Bloch theorem which underlies the other side, facing each other. The office conventional band structure theory is no longer valid. [91] had a very high ceiling, maybe 15 feet high, By the end of the 1950’s, approximate ways to analyze and an upper level balcony had been installed this situation had been proposed by Lothar Nordheim for a secretary, which I did not have. In- (1931), Harry Jones (1934), and Jacques Friedel (1954). stead, Pierre Hohenberg settled there, watch- All of them acknowledge a debt to the eminent En- ing Walter and I from above. glish physical metallurgist William Hume-Rothery and his 1931 book, The Metallic State. Pierre Hohenberg was a newly-arrived post-doc in Nozi`ere’s group. He places himself on the main floor of The first half of Hume-Rothery (1931) reviews years of “Philippe’s own very large office . . . and I remember it experimental effort to systematize the electric, thermo- to have been a general meeting place and thoroughfare, electric, and thermionic properties of metals and alloys. a little like trying to think deep thoughts in the middle The second half reviews the classical and quantum me- of Times Square” (Hohenberg 2003, 2012). chanical theories that had been devised to explain some Kohn began his research activities, but he did not work of these properties. A typical result reported in The on the electron-phonon interaction as he had proposed to Metallic State was the observation that many disordered the Guggenheim Foundation. Some months earlier, he substitutional alloys AxB1−x exhibit an electrical resis- had changed his mind and decided to think more deeply tivity that varies with the A-type atom concentration about the electronic structure of disordered metal alloys. as x(1 − x). Nordheim (1931) explained this by replac- More precisely, he asked himself how one might best de- ing the real alloy, where dissimilar potentials VA(r) and scribe the behavior of the electrons in a bulk metal com- VB(r) act on the valence electrons near lattice sites oc- posed of different types of atoms where there is at least cupied by A-type atoms and B-type atoms, respectively, some randomness in the identities of the atoms that oc- by a fictitious virtual crystal where the valence electrons cupy the sites of the underlying periodic lattice. [88] Un- near every lattice site feel the same average potential, ¯ like most theoretical solid state physicists in the United V (r) = xVA(r) + (1 − x)VB (r). By construction, the States, Kohn had followed developments in alloy physics potential energy function for the virtual crystal is peri- for more than a decade because of the intense experi- odic and any band structure method becomes applicable mental interest in this subject by his faculty colleagues to find the eigenfunctions and energy eigenvalues (Muto in the Physics and Metallurgy departments at Carnegie 1938). Tech. On the other hand, his personal contribution to Jones (1934) was concerned with some empirical ‘rules’ the field consisted of only two published papers and both deduced by Hume-Rothery which related the crystal concerned dilute alloys like AxB1−x where the fraction x structure of certain alloys to their ‘electron concentra- of A-type atoms dissolved in a host metal made of B-type tion’, i.e., the ratio of the total number of valence elec- atoms was very small (Kohn and Vosko 1960, Kohn and trons to the total number of atoms in the entire crystal. Luming 1963). He now turned his attention to concen- Jones focused on the dilute limit and made two indepen- trated alloys where the populations of A-type atoms and dent assumptions. First, he used a nearly-free electron B-type atoms could be comparable. Philippe Nozi`eres description where the energy spectrum and wave func- was not particulary interested in alloys (he was working tions of the host B-type metal was presumed to differ on liquid helium at the time), but a short 30 km train only slightly from the energy spectrum and wave func- ride took Kohn to the suburban campus of the Univer- tions of a collection of completely free electrons. Second, sity of Paris in Orsay where his old friend Jacques Friedel he made a rigid-band approximation which supposed that maintained his research group. Friedel was an acknowl- the sole effect of the A-type atoms was to contribute their edged expert in the theory of metals and alloys. [89] Also valence electrons to the pre-existing ‘sea’ of valence elec- present in Orsay at the time were Andr´eGuinier, an ex- trons contributed by the B-type atoms. This implied that perimentalist renowned for his x-ray diffraction studies of the electronic structure of the alloy was identical to the alloys, and Pierre-Gilles de Gennes, a theorist working on electronic structure of the host metal except that a few a set of problems he would soon collect and discuss in his energy states were occupied (empty) in the alloy com- book, Superconductivity in Metals and Alloys (1966). [90] pared to the host if the valence of the A-type atoms was Kohn immersed himself in the literature of metals and larger (smaller) than the valence of the B-type atoms. [92] alloys and soon discovered that two seemingly contradic- Using quantum mechanical perturbation theory, Jones tory points of view dominated discussions of their elec- estimated the change in total electronic energy when the tronic structure. I pause here to sketch the field as he Fermi surface of the rigid-band alloy contacts the Bril- 21 louin zone boundary for different crystal structures. In Slater opened the conference by remarking that (Slater this way, he was able to rationalize the Hume-Rothery’s 1955) electron concentration rules in a semi-quantitative way. The virtual crystal and rigid-band approximations The metallurgist expects the physicist to be share a ‘delocalized’ view of the electrons in a metal al- able to apply wave mechanics to the problem loy. This means that each electron in the conduction of the cohesive energy of metals. By this one band occupies an eigenstate whose Bloch-like wave func- means the energy of the metal as a function tion has a non-zero amplitude on every atomic site of of the positions of the nuclei. . . . Unfortu- the alloy [see the left side of Eq. (5)]. This perspec- nately, the errors in our present calculations tive gained popularity among practitioners because its of the energies of the isolated atoms and of successes were detailed in the first two research mono- the atoms combined into a metallic crystal graphs devoted exclusively to metal physics: The Theory are both considerably larger than the energy of Metals (1936) by Alan Wilson and The Theory of the difference between the two, which is the co- Properties of Metals and Alloys (1936) by Nevill Mott hesive energy which we hope to find. . . . and Harry Jones. This theory is not yet in a position to make calculations of the accuracy which the met- allurgists need and which they have been led to believe that they have been getting. Met- allurgists have been understandably anxious to get real guidance from physicists regard- ing their problems. A few papers written by theoreticians have led them to think that this guidance could be actually given in a quan- titatively satisfactory form. For instance, a large literature has grown up as a result of pa- pers [published] in the middle 1930’s by Jones . . . and Pauling has discussed metallic cohe- sion and ferromagnetism using methods that seem simple and quantitative. I wish to state my very firm opinion that these theories, as Jacques Friedel pioneered a spatially local des- far as they pretend to be quantitative, are cription of the electronic structure of alloys. based on approximations which are not really Courtesy of the AIP Emilio Segr`eVisual Archive. justified. They may have qualitative truth in them, but they do not represent quantitative A rather different, ‘localized’ point of view was devel- conclusions firmly based on fundamental the- oped by Jacques Friedel (1954). He considered an alloy ory. where ∆ZV is the valence difference between the solute A-type atoms and the solvent B-type atoms and recog- It is perhaps unsurprising that Slater’s research group at nized that each A-type atom with its valence electrons re- the time was exploring the quantitative accuracy of his moved amounts to a point-like impurity with charge ∆ZV own ‘augmented plane wave’ method for band structure with respect to the host crystal. The screening of this im- calculations (Slater 1953). purity [discussed earlier in connection with the work of In Paris in the fall of 1963, Walter Kohn was in an ex- Langer and Vosko (1959)] by the entire sea of conduction cellent position to learn the latest developments in both electrons implies that electronic charge density accumu- the localized and de-localized approaches to alloy the- lates in the immediate vicinity of the A-type atoms. The ory. The preceding January, Jacques Friedel and Andr´e de-localized electrons of the host scatter from these local Guinier had completed editing Metallic Solid Solutions, charge accumulations and Friedel used perturbation the- a book which documented the proceedings of an inter- ory to show that the energy shift of each scattering state national symposium on the electronic and atomic struc- (with respect to the Fermi energy) was the same as the ture of alloys held in Orsay in July 1962. [94] In one in- energy shift predicted by the rigid-band model. He went vited paper, Stanley Raimes of Imperial College noted on to use this ‘local’ perspective to rationalize other ex- that Fermi surface measurements for the noble metals perimental trends summarized in Hume-Rothery’s book. invalidated the nearly-free electron assumption used by An even more localized, covalent-bond approach to met- Jones (1934), but did not quash the rigid-band approx- als was advocated by Linus Pauling (1949). imation itself (Raimes 1963). Similarly, an invited pa- The semi-quantitative nature of all existing theories of per by Frank Blatt of Michigan State University asserted alloy electronic structure was criticized by John Slater that a broad range of transport measurements showed at an October 1955 conference devoted to the theory of that “the rigid-band model is a good approximation to alloy phases. [93] Speaking to an audience of physicists, the electronic structure of dilute alloys” (Blatt 1963). On physical chemists, metallurgists, and crystallographers, the other hand, Blatt continued, “it is difficult to over- 22 state the importance” of Friedel’s local screening model recent effort, both experimental and theoret- for the interpretation of not only resistivity and thermo- ical, has been focused primarily on proper- electric data for alloys, but also for data obtained from ties of one-electron character. In the present measurements of impurity diffusion, positron annihila- work, we have returned to the study of the tion, and the Knight shift.” Friedel himself reported an cohesive energy of metals. extension of his previous work to the case of transition- metal atom impurities where screening occurs by the oc- The remainder of this paper, written by Walter’s friend cupation of atomic-like orbitals spread out in energy into Morrel Cohen from the University of Chicago, displays an ‘virtual bound state resonances’ (Friedel 1963). exact formula for the total energy of a uniform (constant The most general conclusion to be drawn from the pa- density) electron gas due to Nozi`eres and Pines and then pers collected in Metallic Solid Solutions was that some generalizes it to the case of an arbitrary “non-uniform observations were best understood assuming that the system.” Both formulae involve an integral over a pa- conduction electrons of an alloy are delocalized through rameter λ with the quantities in the integrand computed the volume of the crystal while other observations were assuming a charge on the electron of λe. The limits of the best understood assuming that the most relevant elec- integral extend from λ = 0 (the non-interacting electron trons are localized in the immediate vicinity of the solute gas) to λ = 1 (the fully interacting electron gas). atoms. At least two prominent metallurgists regarded In his 1990 reconstruction, Kohn wrote a similar inte- these points of view as complementary rather than con- gral to compute ∆E, the difference in total energy be- th tradictory. In the 4 edition of their Structure of Metals tween a real AxB1−x alloy and the same alloy in the vir- and Alloys, William Hume-Rothery and Geoffrey Vincent tual crystal approximation. If ZA and ZB are the atomic Raynor write (Hume-Rothery and Raynor 1962) numbers of the A-type and B-type atoms, every ion in the virtual crystal has nuclear charge Z¯ = xZA + (1 − x)ZB. The covalency interpretation and the Bril- A B Then, replacing ∆Z = Z −Z by λ∆Z, Kohn evaluated louin zone picture each express a part of the the integral of dE/dλ from λ =0 to λ = 1 to second or- truth. In the case of the diamond structure, der in ∆Z (keeping Z¯ fixed). The resulting expression for for example, the covalency theory . . . gives ∆E depended on two quantities only: the electron den- the more correct picture of the probable cloud sity distribution of the virtual crystal,n ¯(r), and the elec- density of valence electrons in the crystal. tron density distribution of the real alloy, n(r). At this This concept by itself ignores the fact that point, “the question occurred to Kohn whether a knowl- electrons are free to move in the crystal, and edge of n(r) alone determined–at least in principle–the this freedom is emphasized by the zone theo- total energy” (Hohenberg, Kohn, and Sham 1990). In ries, which in their turn ignore the variation other words, could the many qualitative successes of the of the electron-cloud density in space. Both Friedel point of view, which put great emphasis on the concepts are required to express the whole space-varying electronic charge density inside an alloy, be truth. elevated to show that the exact charge density uniquely The complementarity expressed by this paragraph was predicts the exact total energy?[95] well-known to many solid-state physicists, and partic- It is important to understand the revolutionary nature ularly so to Walter Kohn who had published a funda- of this question. Notwithstanding the work of Cohen mental paper on the relationship between the delocal- mentioned above, most solid state theorists in the early ized Bloch functions and the localized Wannier func- 1960’s thought about the total energy of a solid along tions (Kohn 1959a). Kohn had also kept up with band the lines laid out by Frederick Seitz in his widely ad- structure calculations and could well appreciate the point mired textbook, The Modern Theory of Solids (1940). John Slater had made about the absence of quantitative One starts with the nuclear charge eZk and the fixed po- calculations in alloy theory (Callaway and Kohn 1962). sition Rk of each of the M nuclei in the system. The Kohn has written about his 1963 survey of the alloy Coulomb potential energy of interaction between pairs of literature on several occasions. His 1990 account, which nuclei is a classical quantity which poses no problem to is closest in time to the actual events, recalls the “rough compute. The remaining energy terms all contribute to and ready” rigid-band model and then recounts a calcula- the Schr¨odinger equation for the N-electron wave func- tion of the total energy of a disordered alloy (Hohenberg, tion: the kinetic energy of the electrons, the Coulomb Kohn, and Sham 1990). This is interesting because the potential energy of interaction between every pair of elec- total energy of a real metal was not a focus of research trons, and the Coulomb potential energy of interaction at the time. Indeed, the only paper in Metallic Solid between every electron and every nucleus. Solutions concerned with total energy begins with the It will be convenient to define v(r) as the potential statement (Cohen 1963), energy of interaction between an electron at the point r and all the fixed nuclei: In the early history of the theory of metals, calculation of the cohesive energy was a cen- M Z tral concern. Apart from very considerable v(r)= −e2 k . (7) |r − Rk| development of the theory of the electron gas, Xk=1 23

This potential energy appears in the Schr¨odinger equa- Cartesian variables? Looking for support, Walter asked tion, as do the operators for the electron kinetic energy himself if he knew any examples where complete knowl- and the electron-electron potential energy, edge of n(r) implied complete knowledge of v(r) (Hohen- berg, Kohn, and Sham 1990). One such example was the ~2 N N N 2 elementary Schr¨odinger equation, ˆ 2 ˆ 1 e T = − ∇k and U = . (8) 2m 2 |rk − rm| Xk=1 kX=1 mX=1 ~2 − ∇2ψ + [v(r) − E] ψ(r)=0. (11) The Schr¨odinger equation determines the N-electron 2m wave function, ψ(r1, r2,..., rN ), and then the electron density distribution, One can always choose ψ(r) real for this equation. There- fore, the electron density is n(r) = ψ2(r) and the inver- ∗ n(r)= ψ (r, r2,..., rN ) ψ(r, r2,..., rN ) dr2 ··· drN . sion we seek is a matter of algebra: Z (9) ~2 ∇2 n(r) If we omit the classical ion-ion energy, which does not v(r)= E + . (12) involve the electrons, the total energy is 2m pn(r) p E = n(r) v(r) dr + hψ|Tˆ|ψi + hψ|Uˆ|ψi, (10) Unfortunately, Eq. (11) applies only to a one-particle Z system and thus does not shed any light on the many- body problem. A more relevant example known to Kohn where the last two terms are expressed as averages (ex- was the Thomas-Fermi method, a semi-classical but self- pectation values) with respect to the ground state wave consistent approximation to the quantum theory of a function. many-electron system with a non-uniform density n(r) The conventional perspective outlined in the previ- (March 1975, Zangwill 2013). Nevill Mott had used this r ous paragraph shows that the external potential, v( ), method to calculate the screening of a point charge by the is the only term in the Schr¨odinger equation which dis- conduction electrons in a metal. For the problem consid- r tinguishes one alloy from another. Therefore, v( ) deter- ered here, Thomas-Fermi theory method shows that n(r) mines the wave function from the Schr¨odinger equation, determines v(r) through the relation [96] which in turn determines the electron density n(r) from Eq. (9) and the total energy E from Eq. (10). From this 2/3 1 3h3 n(r′) point of view, the energy amounts to a functional of the r 2/3 r 2 r′ v( )= E − n ( ) − e d r r′ . (13) external potential. Of course, the accuracy of the com- 2m  8π  Z | − | puted energy depends on the quality of the choice made for the form of the N-electron wave function used to solve the Schr¨odinger equation. Encouraged by these examples, Kohn sought to prove r Kohn now contemplated a radical inversion of this pro- that the ground state electron density n( ) uniquely deter- r cedure. Was it possible that the total energy depended mines the external potential v( ). Such a simple result only on the electron density n(r)? That is, could the total (if true) could not depend on the details of the many- energy be a functional of the density alone? If so, knowl- electron wave function. Therefore, he looked for a very edge of n(r) was sufficient to determine (implicitly) the general idea to exploit and found it with the Rayleigh- external potential, the N-particle wave function, and all Ritz variational principle (Hohenberg, Kohn, and Sham ground state properties, including the Green functions of 1990). The reader will recall that this was the first tool many-body theory! This was a very deep question. Wal- Walter had placed in his theoretical toolbox while still ter realized he wasn’t doing alloy theory anymore (Kohn a student at the University of Toronto. The reductio ad 2012b). absurdum proof is “disarmingly simple” (Parr and Yang 1989). Assume that a Hamiltonian (energy) operator Hˆ =v ˆ+ V. THE WORK OF HOHENBERG AND KOHN Tˆ +Uˆ produces a ground state energy E0, a ground state wave function ψ(r1,..., rN ) and an electron density n(r). Kohn’s proposition that the total energy of an elec- Contrary to what we wish to prove, assume that another ′ ′ tron system was a functional of the density seemed pre- Hamiltonian Hˆ =v ˆ + Tˆ + Uˆ produces a ground state ′ ′ posterous on the face of it. How could knowledge of the energy E0, a ground state wave function ψ (r1,..., rN ), electron density, which is a function of three Cartesian but the same electron density n(r). The idea is to use variables, be sufficient to compute the total energy when ψ′ as a trial function in the variational principle for Hˆ in the last two terms in Eq. (10) depend explicitly on the Eq. (1). Ignoring the possibility of degeneracy, this gives many-electron wave function, which is a function of 3N the strict inequality [97] 24

′ ′ ′ ′ ′ ′ ′ ′ ′ ′ E0 < hψ |Hˆ |ψ i = hψ |Hˆ |ψ i + hψ |Hˆ − Hˆ |ψ i = E0 + n(r)[v(r) − v (r)] dr. (14) Z

Similarly, we may use ψ as the trial function in the varia- with the primed and unprimed variables exchanged: tional principle for H′. The result is the same as Eq. (14)

′ ′ ′ ′ E0 < hψ|Hˆ |ψi = hψ|Hˆ |ψi + hψ|Hˆ − Hˆ |ψi = E0 + n(r)[v (r) − v(r)] dr. (15) Z

′ ′ Adding Eq. (14) to Eq. (15) yields E0 + E0 < E0 + Hohenberg’s first task was a literature search to dis- E0, which is impossible. Hence, our assumption that cover if the theorem Kohn had proved was already two external potentials correspond to the same electron known. Apparently not. However, it had been known density is false. for about a decade that a computation of the total en- ergy of a system of nuclei and electrons did not really require complete knowledge of the entire N-particle wave function (L¨owdin 1959, McWeeny 1960, Coleman 1963). It was sufficient to specify a quantity derived from the wave function called the second-order density matrix. [98] In principle, there exists a variational principle for the energy where the second-order density matrix is varied rather than the wave function. Unfortunately, it was not known what properties the density matrix had to possess to ensure that it was derivable from an N-particle wave function. This became known as the ‘N-representability problem’. In a similar way, Kohn’s theorem reminded Hohen- berg of “formal work on stationary entropy and renor- malization which had just been completed by Paul Mar- Pierre Hohenberg in 1965, soon after his work tin and Cyrano de Dominicis, also working together in with Kohn on density functional theory. Paris” (Hohenberg, Kohn, and Sham 1990). These au- Courtesy of Pierre Hohenberg. thors had studied the grand partition function of an ar- bitrary many-body system and used the mathematical Kohn was exhilarated by his proof, but “it seemed technique of Legendre transformations to effect the de- such a remarkable result that I did not trust myself” sired ‘renormalizations’. The latter eliminated the func- (Kohn 1998). He looked around for help and found it tional dependence on any one-body potential (like the ex- in the person of Pierre Hohenberg, the recently arrived ternal potential above) in favor of functional dependence post-doctoral fellow who was also ensconced in Philippe on a one-particle distribution function. It also eliminated Nozi`eres’ enormous office. Hohenberg had just completed the functional dependence on any two-particle potential a post-doctoral year doing many-body theory with Alexei (like the Coulomb interaction between pairs of electrons) Abrikosov and Lev Gor’kov in Moscow, but was having in favor of functional dependence on a two particle dis- some trouble getting the attention of Nozi`eres, his new tribution function (de Dominicis 1963, de Dominicis and post-doctoral supervisor (Hohenberg 2012). Walter pro- Martin 1964). Fifty years later, Hohenberg and Kohn posed that they work together and Pierre agreed. Like differ slightly in their recollection of whether Legendre Kohn’s previous collaborators, Res Jost and Quin Lut- transformations played any role in their work together. tinger, Hohenberg was a rather formal theoretical physi- Kohn recalls that he and Hohenberg recognized that his cist. He had been trained at Harvard by Paul Martin and theorem could be interpreted as a Legendre transforma- his 1962 PhD thesis, “Excitations in a dilute condensed tion only after the fact and notes that the final published Bose gas”, used density matrix and Green function meth- paper makes no mention of it (Kohn 2013b). Hohenberg ods to provide microscopic justification for phenomeno- (2012) recalls that logical theories that had been offered by Lev Landau and others to explain the properties of superfluid helium. we were certainly thinking in the language 25

of Legendre transformations, but we did not Accordingly, Hohenberg and Kohn proceeded on their need that idea in the end. It was character- own to the natural next step: a reformulation of the istic of Walter’s style to introduce [in print] Rayleigh-Ritz variational principle in terms of the den- only those theoretical ideas needed to solve sity rather than the many-body wave function. the problem at hand. Kohn’s theorem implies that the many-body wave function is a functional of the ground state electron den- Martin had returned to Harvard, but de Dominicis sity n(r). The same is true of the exact kinetic en- worked at the Centre d’Etudes Nucl´eaires in nearby ergy T [n] and the exact electron-electron potential en- Saclay. According to Hohenberg, “it took a number of in- ergy U[n]. Therefore, if tense but informative discussions with de Dominicis and his colleague Roger Balian to convince them that the F [n]= T [n]+ U[n] (16) procedure worked with the density rather than the dis- tribution function (Hohenberg, Kohn, and Sham 1990). is the sum of these two, the total energy in Eq. (10) is Philippe Nozi`eres finally joined the conversation when he learned of the larger theoretical issue that now engaged E[n]= n(r) v(r) dr + F [n]. (17) Hohenberg and Kohn. As he later recalled (Nozi`eres Z 2012b),

The three of us discussed it a lot. But I was Now let nT (r) ≥ 0 be a trial density which produces the not fully convinced. In my opinion, putting correct total number of electrons, N = nT (r) dr. By all the emphasis on the density did not ac- Kohn’s theorem, this density determines itsR own external count properly for exchange and correlations. potential, Hamiltonian, and ground state wave function I did not share the enthusiasm of Walter and ψT . In that case, the usual wave function variational Pierre and I stayed aside. principle expressed by Eq. (1) tells us that

hψT |Hˆ |ψT i = nT (r) v(r) dr + F [nT ]= E[nT ] ≥ E[n]. (18) Z

The variational principle in Eq. (18) establishes that the fidence that all “except some pathological distributions” energy in Eq. (17) evaluated with the true ground state will have this property but the mere fact that they draw density is minimal with respect to all other density func- attention to this point demonstrates how carefully the tions for the same number of particles. authors looked for holes in their arguments. [99] The proofs given just above appear early in Hohen- berg and Kohn (1964), the first foundational paper of HK make a point to remark that the functional F [n] density functional theory. In connection with Eq. (18), in Eq. (16) is “universal” in the sense that it is valid Hohenberg and Kohn (HK) use a footnote to warn the for any number of particles and any external potential. reader of a possible ‘v-representability problem’ because Even more significant for later work, they define yet an- “we cannot prove whether an arbitrary positive definite other functional G[n] by extracting from F [n] the clas- distribution n(r) which satisfies n(r) dr = N can be sical Coulomb energy of a system with charge density realized for some external potential.”R They express con- en(r). Doing this renders Eq. (17) in the form

2 ′ e n(r) n(r ) ′ E[n]= n(r) v(r) dr + dr dr + G[n]. (19) Z 2 Z Z |r − r′|

The functional G[n] in Eq. (19) includes the exact kinetic so would imply that they had solved the entire many- energy of the electron system and all the potential energy electron problem. associated with the electron-electron interaction that is not already counted by the classical electrostatic energy. The approach Hohenberg and Kohn took to analyze Needless to say, HK could not write down G[n]. Doing G[n] reflects the way Kohn chose to frame the final pub- lished paper. There is no mention of the alloy problem or 26 even of any desire to re-formulate the electronic structure son 1962). These papers are among the “important ad- problem for solids. Instead, the title of the HK paper is vances” noted by HK in the passage quoted just above. simply “Inhomogeneous electron gas” and the first line [100] of the abstract announces that “this paper deals with the As trained solid-state physicists, HK knew that the ground state of an interacting electron gas in an exter- entire history of research on the quantum mechanical nal potential v(r).” The Introduction goes on to note many-electron problem could be interpreted as attempts that “during the last decade there has been consider- to identify and quantify the physical effects described able progress in understanding the properties of a homo- by G[n]. For example, many years of approximate quan- geneous, interacting electron gas.”A footnote refers the tum mechanical calculations for atoms and molecules had reader to David Pines’ book Elementary Excitation in established that the phenomenon of exchange—a conse- Solids (1963) for the details. HK then remind the reader quence of the Pauli exclusion principle—contributes sig- about the Thomas-Fermi method, nificantly to the potential energy part of G[n]. Exchange reduces the Coulomb potential energy of the system by r in which the electronic density n( ) plays a tending to keep electrons with parallel spin spatially sep- central role. . . This approach has been arated. The remaining potential energy part of G[n] useful, up to now, for simple though crude takes account of short-range correlation effects. Cor- descriptions of inhomogeneous systems like relation also reduces the Coulomb potential energy by atoms and impurities in metals. Lately, there tending to keep all pairs of electrons spatially separated. have been some important advances along The effect of correlation is largest for electrons with anti- this second line of approach. . . . The present parallel spins because these pairs are not kept apart at all paper represents a contribution in the same by the exchange interaction. I note for future reference area. that the venerable Hartree-Fock approximation takes ac- HK confirmed that the Thomas-Fermi model of elec- count of the kinetic energy and the exchange energy ex- tronic structure follows from Eq. (19) by using an ex- actly but (by definition) takes no account of the correla- pression for G[n] which accounts approximately for the tion energy (Seitz 1940, L¨owdin 1959, Slater 1963). kinetic energy of the electrons but takes no account of the HK devoted the remainder of their time together to non-classical electron-electron potential energy. Specifi- studying G[n] for two cases: (i) an interacting electron cally, the kinetic energy density at a point r of the real gas with a nearly constant density; and (ii) an interacting electron gas with a slowly-varying density. For the nearly system is set equal to the kinetic energy density of an infi- r r nite gas of non-interacting electrons with a uniform den- constant density case, HK wrote n( ) = n0 +˜n( ) with r n˜(r)/n0 ≪ 1 and pointed out that G[n] admits a formal sity n = n( ). The latter is computed using elementary r statistical mechanics and one finds (March 1975, Zang- expansion in powers ofn ˜( ): will 2013) ′ ′ ′ G[n]= G[n0]+ K(|r − r |)˜n(r)˜n(r ) dr dr + ··· (21) 3 2/3 Z 3 3h 5/3 GTF[n]= n (r) dr. (20) ′ 10m  8π  Z K(|r − r |) is a linear response function for the uniform and interacting electron gas which had been studied in- After inserting Eq. (20) into Eq. (19) to produce ETF[n], tensely by experts in many-body theory (Pines 1963). In the variational principle in Eq. (18) directs us to mini- particular, the derivative of its Fourier transform K˜ (q) mize ETF[n] with respect to density. This is done using was known to diverge at a certain value of q (Pines 1963). Lagrange’s method to ensure that the total particle num- Walter knew this divergence well: it was responsible for ber, N = n(r) dr, remains constant. The final result is the oscillatory algebraic form of the Friedel density dis- the Thomas-FermiR expression in Eq. (13). turbance formula in Eq. (6). It was also responsible for The Thomas-Fermi model never went out of fashion the “Kohn anomalies” in the phonon spectrum of met- as a quick and easy way to gain qualitative information als. HK point this out and remark in passing that “the about atoms, molecules, solids, and plasmas. By 1957, density oscillations in atoms which correspond to shell the British solid state physicist Norman March was able structure . . . are of the same general origin.” to publish a 100-page review article surveying the suc- The divergence in K˜ (q) disappears if one retains only a cesses and failures of the model and its generalizations finite number of terms in its power series expansion. The (March 1957). Earlier, I labeled 1957 as the ‘magic year’ corresponding expansion of K(|r − r′|) reduces Eq. (21) when many-body Green functions and diagrammatic per- to a gradient expansion, turbation theory transformed the study of many-electron systems. Therefore, it is not surprising that several 2 G[n]= G[n0]+a n˜(r) dr+b |∇n˜(r)| dr+··· , (22) physicists–including Kohn’s old post-doctoral colleague Z Z Sidney Borowitz–applied these methods with the aim to systematically generalize the Thomas-Fermi model to in- where a and b are constants related to K˜ (q). Therefore, clude the effects of electron correlation and spatial inho- as HK pointed out, any ‘quantum oscillations’ produced mogeneity (Baraff and Borowitz 1961, DuBois and Kivel- by the divergence in K˜ (q) cannot be captured by any 27 low-order gradient expansion of G[n] like Eq. (22). This previous sentence that “actual electronic systems” have explained the failures of the many generalizations of the neither nearly constant densities nor slowly-varying den- Thomas-Fermi approximation which involved adding gra- sities, i.e., the situations where the partial summation dient terms. was expected to be valid. Walter and Pierre took some time off in the first week Any reader of Hohenberg and Kohn (1964) cannot of December 1963 to attend an All-Union Conference help but be struck by its understated and rather for- on Solid State Theory in Moscow. Kohn joined John mal tone. The Introduction is succinct, the basic theo- Bardeen and George Vineyard (Brookhaven National rems are proved quickly, most of the paper is taken up Laboratory) as the only senior Americans in attendance. with the gradient expansions, and no applications are dis- Hohenberg had left Moscow in July and he relished the cussed or proposed. Earlier, I quoted Kohn to the effect opportunity to visit his old friends there. The topics dis- that he understood that the basic theorem which drives cussed at the meeting were superconductivity, the the- the paper was “remarkable”. If so, he and Hohenberg ory of metals, semiconductors and dielectrics, lasers, the made a conscious decision not to ballyhoo the truly rev- magnetic properties of rare earths, neutron scattering, olutionary idea at its core: that the ground state electron the M¨ossbauer effect, phonons, dislocation theory, and density, in principle, determines all the properties of an the theory of radiation damage in crystals (Agranovich electronic system. That being said, we learn from their 1964). 1990 reminiscence that the authors spent at least a little Upon their return to Paris, Hohenberg and Kohn fo- time talking about applications (Hohenberg, Kohn, and cused on G[n] for a system with a slowly-varying den- Sham 1990): sity function. This assumption precluded variations with The question arose as to what the method short (spatial) wavelengths, but allowed for the possibil- might be good for, and Kohn suggested that ity of substantial variations in the overall magnitude of one could try using it to improve current tech- the density. For this case, the appropriate gradient-type niques for calculating the band structure of expansion is solids. Hohenberg’s immediate reaction was 2 to say, “But band structure calculations are G[n]= g0(n(r)) dr + g2(n(r)) |∇n(r)| dr + ··· , Z Z horribly complicated, isn’t that the sort of (23) stuff better left to professionals?” To this, where g0 and g2 are functions (not functionals) of n(r). Kohn simply replied, “Young man, I am the By specializing Eq. (23) to the previously studied case of Kohn of Kohn and Rostoker (1954) !” a nearly uniform electron gas, HK were able to express A recommendation from Kohn helped Hohenberg win these functions in terms of the properties of the uniform a job at Bell Telephone laboratories in the fall of 1964. and interacting electron gas. For example, g0(n) is the Recently, he recalled that “I gave a talk on my Paris work sum of the kinetic energy density, the exchange energy during my first few months at Bell Labs. Phil Anderson, density, and the correlation energy density for an inter- Bill McMillan, and Phil Platzman were in the audience acting electron gas with uniform density n. With this and there was no enthusiasm. They correctly understood information, HK performed a partial (infinite) summa- that our results would not help them solve the difficult tion of the entire gradient expansion in Eq. (23). Their many-body problems they were struggling with” (Hohen- final expression had the great virtue of recovering the sin- ˜ berg 2012). In the event, Hohenberg turned to hydrody- gularity in K(q) needed to describe quantum oscillations. namics and phase transitions as topics for research and Kohn left Paris in January 1964 and visited physicists worked productively in those areas over a thirty-year ca- in London, Cambridge, and Oxford before returning to reer at Bell. He was elected to the US National Academy California. He wrote up a first draft of a manuscript and of Sciences in 1989 and moved to Yale University in 1995 sent it to Hohenberg for his review. Pierre made sev- to accept the position of Deputy Provost for Science and eral suggestions, all of which were incorporated into the Technology. Since 2004, he has served at New York Uni- final version (WKP 1964). The paper was sent off to versity as a professor of physics and Vice Provost for the Physical Review in the second week of June 1964 and Research. published in the November 9 issue. It is notable that the ‘Concluding Remarks’ section does not remind the reader of the two theorem’s proved at the beginning of VI. THE WORK OF KOHN AND SHAM the paper. Instead the authors remark that they “have developed a theory of the electronic ground state which is Kohn returned to La Jolla in early 1964 “to find a exact in two limiting cases.” The importance Hohenberg group of postdocs and visitors eagerly awaiting a first- and Kohn ascribed to the ability of their theory to cap- hand account of the work” (Hohenberg, Kohn, and Sham ture quantum oscillations may be judged from the fact 1990). According to one of those post-docs, Vittorio Celli that their paper ends with the statement that ‘the most (2013), promising formulation of the theory . . . appears to be that obtained by partial summation of the gradient ex- Walter gave a full departmental colloquium pansion”. They say this despite warning the reader in the rather than just a technical seminar after his 28

return from Paris. Keith Brueckner (who had returned from Washington and was Dean at the time) said that colloquia were usually re- served for “foreign stars” but that today we have “our own star” to give a talk. I remem- ber thinking that the theory with Hohenberg was cute but would not have many conse- quences. I certainly did not think it com- pared in significance with the many-body cal- culations for the electron gas that had been obtained by Brueckner and his collaborators.

Lu-Jeu Sham (circa early 1980’s) was a Kohn proposed to another post-doc, N. David Mermin, post-doc with Walter Kohn from 1963-1966. that he exploit his knowledge of statistical mechanics to Courtesy of Lu-Jeu Sham. generalize the Hohenberg-Kohn results to non-zero tem- Quantum density oscillations were much on Walter’s perature. Mermin realized almost immediately that “a mind when he related the details of his Paris experience strange variational principle for the free energy that I to Sham. He was acutely aware that no oscillations would had formulated for an utterly unrelated purpose” was emerge from Eq. (22), the most natural choice for G[n] perfectly suited for the job. As a result, “it took me less when the external potential was slowly-varying function than an hour to check that the HK proof went through” of position. Therefore, in the same letter where Walter almost without change (Mermin 2004). Kohn was skepti- informed Pierre Hohenberg that the HK manuscript had cal, but the simplicity of the proof could not be assailed. been submitted for publication, he related that “Lu Sham On the other hand, the significance of Mermin’s work was and I have started looking at situations like a heavy atom not very clear and he was disinclined to write it up. It where one has a localized and rapidly-varying potential” took six months, his imminent departure from La Jolla, (WKP 1964). In principle, their goal was to develop a and Kohn’s insistence that “it someday might be impor- general theory of quantum density oscillations for use tant” for the manuscript of Mermin (1965) to get written in situations where the electron density is strongly non- (Mermin 2013). uniform. In practice, they developed a method to find the Kohn’s colloquium featuring the Hohenberg-Kohn the- leading quantum corrections to the Thomas-Fermi elec- orems generated skepticism from several of the many- tron density for a collection of non-interacting electrons body theorists who were in La Jolla to work with Keith moving in a one-dimensional potential. The published Brueckner. Most notably, “Nobuyuki Fukuda, visit- paper, Kohn and Sham (1965a), reports the results of ing from the University of Tokyo, constructed counter- an elegant Green function calculation which related the example after counter-example purporting to demon- electron density (which did exhibit the desired quantum strate the non-uniqueness of the potential given a density oscillations) to the potential in a way which generalized distribution. The job of resolving these fell to Lu Jeu Eq. (13). Unfortunately, the extension of their method to Sham” (Hohenberg, Kohn, and Sham 1990). three-dimensional periodic potentials presented a daunt- ing numerical challenge which did not lend itself to prac- Lu Sham was the graduating PhD student to whom tical calculations for real solids. Accordingly, Kohn and Kohn had written requesting that he begin his post- Sham dropped this line of investigation and moved in a doctoral fellowship at UCSD while Kohn was still in different direction, albeit one still motivated by the basic Paris. Sham, a native of Hong Kong, arrived in San results obtained by Hohenberg and Kohn. [101] Diego by way of Imperial College and Cambridge Univer- By the late fall of 1964, Kohn was thinking about al- sity where he had earned, respectively, an undergraduate ternative ways to transform the theory he and Hohen- degree in Mathematics and a graduate degree in physics. berg had developed into a practical scheme for atomic, His 1963 PhD thesis, “The electron-phonon interaction”, molecular, and solid state calculations. Happily, he was was supervised by John Ziman, a distinguished theorist very well acquainted with an approximate approach to whose books, Electrons and Phonons (1960) and Princi- the many-electron problem that was notably superior ples of the Theory of Solids (1964), helped train a gen- to the Thomas-Fermi method, at least for the case of eration of solid-state physicists. In La Jolla, Sham used atoms. This was a theory proposed by Douglas Hartree in the months before Kohn returned to write a paper on 1923 which exploited the then just-published Schr¨odinger the phonon spectrum of sodium metal. This quantita- equation in a heuristic way to calculate the orbital wave tive calculation used a pseudo-potential for the sodium functions φk(r), the electron binding energies ǫk, and the ionic potential and a self-consistent modification of the charge density n(r) of an N-electron atom (Park 2009, Hartree-Fock method to take approximate account of cor- Zangwill 2013). Hartree’s theory transcended Thomas- relation effects which act to screen (reduce) the exchange Fermi theory primarily by its use of the exact quantum- interaction (Sham 1965). mechanical expression for the kinetic energy of indepen- 29 dent electrons. The Hartree equations which define the Kohn suggested to Sham that he try to derive the theory for an atom with nuclear charge Z = N are Hartree equations from the Hohenberg-Kohn formalism. Walter had good reason to believe this could be done 2 ~ 2 (Kohn 2001a, Kohn 2012b). On the one hand, his work − ∇ φ + [veff (r) − ǫ ] φ (r)=0, k =1,...N, 2m k k k with Hohenberg had established the central role of the (24) electron density n(r) for a complete description of any where electronic system. On the other hand, the Hartree equa- ′ tions could be read as a self-consistent scheme to de- 2 ′ n(r ) veff (r)= v(r)+ e dr (25) duce an approximate expression for n(r). Therefore, it |r − r′| Z should be possible to derive the Hartree equations as an and v(r) = −Ze2/r. The electron density n(r) in example of the HK variational principle. Specifically, the variational minimization of some approximate form of Eq. (25) is calculated assuming that the electrons occupy thet N lowest energy eigenfunctions of the Schr¨odinger the total energy functional E[n] should lead to Hartree’s equations. Sham set to work with enthusiasm. His equation in Eq. (24). Therefore, if k =1, 2,...,N labels these N lowest energy orbitals, state of mind at that particular moment was noted by Philip Taylor, a fellow former-graduate student of John N Ziman’s at Cambridge University. Taylor had taken a 2 n(r)= |φk(r)| . (26) job at the Case Institute of Technology in Cleveland, Xk=1 Ohio and had recently published a quantitative analy- sis of Kohn anomalies in the phonon spectrum of metals The effective potential energy function in Eq. (25) shows (Taylor 1963). It was during a visit to La Jolla to con- that every electron interacts with the charge of the ‘exter- sult with Kohn and visit with Sham that Sham remarked nal’ nucleus and with the charge of the entire atomic elec- to Taylor that he was “thrilled” by the research project tron cloud taken as a whole. [102] Hartree stressed that Kohn had given to him (Taylor 2013). these equations must be solved self-consistently. That is, an iterative numerical method is required to ensure that Kohn and Sham recognized that the Hartree method the φk(r) generated by Eq. (24) are the same as the φk(r) regards each electron as moving independently in an ef- used in Eq. (26) to construct the particle density n(r). fective potential veff (r) which does not recognize the in- Slater (1930) and Fock (1930) had provided a rigor- dividual identity of the other electrons. Consistent with ous derivation of the Hartree equations. They used the this, the kinetic energy implied by Eq. (24) is correct only variational principle in Eq. (1) and evaluated the to- for independent and non-interacting electrons. This was tal energy using a many-electron wave function of the the key to progress because the Hohenberg-Kohn anal- form ψ(r1, r2,..., rN ) = φ1(r1)φ2(r2) ··· φN (rN ). Min- ysis implied that the kinetic energy of a strictly non- imizing this energy with respect to different choices for interacting system of electrons is also a functional of the the φk functions generates the Hartree equations. Slater density. If we call this functional TS[n], ordinary quan- and Fock also evaluated the total energy using a more tum mechanics specifies that [103] sophisticated many-body wave function (called a Slater determinant) which combines the same N orbital func- tions φk in such a way that the Pauli exclusion principle N 2 ∗ ~ 2 is obeyed automatically. With this choice, minimizing TS[n]= φ (r) − ∇ φk(r) dr. (27) Z k  2m  the total energy with respect to choices for the φk func- Xk=1 tions generates what are called the Hartree-Fock equa- tions. Hartree-Fock theory is superior to Hartree theory because the kinetic energy and the exchange energy are treated exactly. Unfortunately, the Hartree-Fock equa- The path was now open for Sham to derive the Hartree tions are significantly harder to solve than the Hartree equations from a density functional point of view (Kohn equations. 2001a). He chose the approximate total energy functional

2 ′ e n(r) n(r ) ′ EH [n]= TS[n]+ v(r) n(r) dr + dr dr , (28) Z 2 Z Z |r − r′|

and minimized it with respect to a density assumed to sity variation of the first term in Eq. (28) into variations have the form in Eq. (26). The latter transforms the den- with respect to the φk functions. The density variation 30 of the remaining terms in Eq. (28) is straightforward and reference system is unambiguously Hartree-like, so it is the final result is exactly the Hartree equations (24) and correct to use the exact kinetic energy for non-interacting (25). electrons in Eq. (27) and represent the density function Kohn and Sham (KS) now knew how to move forward as in Eq. (26). On the other hand, the eigenfunctions with the general many-electron problem. Motivated by φk(r) and the eigenvalues ǫk in Eq. (24) have no direct Eq. (28), they defined a functional Exc[n] by the partition physical meaning for the true interacting electron system. G[n]= TS[n]+ Exc[n]. This puts the exact total energy Finally, Kohn and Sham proposed an approximation functional in the form for Exc[n] which has come to be known as the local den- sity approximation (LDA). Namely, E[n]= EH [n]+ Exc[n]. (29)

Exc[n]= n(r) ǫxc(n(r)) dr, (32) The great virtue of Eq. (29) is that it has exactly the same Z structure as Eq. (28), even if we revert to Eq. (7) for v(r). Therefore, the interacting electron density n(r) which where ǫxc(n) is the exchange and correlation energy per minimizes the original total energy in Eq. (10) is pre- electron of a fully interacting electron gas with uniform cisely equal to the non-interacting electron density n(r) density n. Hohenberg and Kohn had introduced a similar which minimizes Eq. (29). Carrying out the latter min- approximation for G[n] in their paper and it was reason- imization explicitly produces the Kohn-Sham equations, able for Kohn and Sham to “regard ǫxc(n) as known from which are identical to the Hartree equations Eqs. (24) theories of the homogeneous electron gas.”[104] When and (26) with Eq. (25) replaced by Eq. (32) is used for Exc[n], the exchange-correlation po- tential in Eq. (31) becomes r′ r r 2 r′ n( ) r veff ( )= v( )+ e d ′ + vxc( ) (30) d Z |r − r | v (r)= [nǫ (n)] . (33) xc dn xc where The foregoing results were reported in a short δE [n] manuscript, “Exchange and correlation effects in an in- v (r)= xc . (31) xc δn(r) homogeneous gas” which Kohn and Sham submitted to Physical Review Letters in May of 1965. Samuel The exchange-correlation potential energy, vxc(r), ob- Goudsmit, one of the editors of Physical Review Letters tained in Eq. (31) from the functional derivative of at that time, informed Sham by letter that the Kohn- Exc[n], is a function of r and not a functional of n(r). Sham manuscript “deserves publication as an Article in Therefore, according to KS, a numerical procedure no the Physical Review, but it is not of such urgency to more difficult that Hartree’s original method is suffi- warrant speedy publication in Physical Review Letters” cient to compute the ground state electron density and (Goudsmit 1965). The authors responded by withdraw- thus the ground state total energy of an arbitrary many- ing the manuscript and, three weeks later, submitted to electron system subject to an external potential. If the Physical Review a longer and more detailed paper Exc[n] was known exactly, one could calculate n(r) and with a new title, “Self-consistent equations including ex- E[n] exactly as well. change and correlation effects”. The published version, The “exchange-correlation” energy functional Exc[n] in Kohn and Sham (1965b) is the second foundational pa- Eq. (29) is similar to G[n] in Eq. (19) in the sense that it per of density functional theory. It is also one of the accounts for all the energy associated with the Coulomb most highly cited papers in the history of physics. [105] interaction between electrons not already counted by the Interestingly, it was only at the page-proof stage of the classical Coulomb self-energy. However, while G[n] had longer paper that the authors realized that their Hartree- also to account for the total kinetic energy of the real in- like equations with Eq. (31) represented a formally exact teracting electron system [called T [n] in Eq. (16)], Exc[n] statement of the complete many-body problem (Sham has only to account for the difference between the kinetic 2014). For that reason, Eq. (31) appears only in a “Note energy of an interacting electron system and the kinetic Added in Proof” while Eq. (33) appears in the main ex- energy of a non-interacting electron system with exactly position. the same density n(r). Of course, the exact and univer- Kohn and Sham knew it was straightforward to use sal functional Exc[n] is no better known that G[n] for the Eq. (33) and write down an explicit and analytic for- interacting electron problem. mula for vxc(r) and incorporate it seamlessly into exist- Unlike the Slater-Fock methodology sketched earlier, ing computer programs to calculate the electronic struc- the foregoing derivation of the Hartree-like Kohn-Sham ture of atoms, molecules, and solids. Briefly, the sep- equations does not introduce a many-electron wave func- aration ǫxc(n) = ǫx(n)+ ǫc(n) known for the interact- tion at any stage. Instead, Kohn and Sham replace the ing and uniform electron gas implies that the exchange- true interacting electron system with a non-interacting correlation potential Eq. (33) separates similarly into electron reference system which has exactly the same vxc(r) = vx(r)+ vc(r). The exact exchange energy den- ground state electron density. The wave function of the sity, ǫx(n), had been calculated years earlier by Dirac 31

(1930) for the purpose of improving the Thomas-Fermi Walter left for his annual visit to Bell Labs after approximation. Using Dirac’s formula, KS reported their the June 1965 submission of the longer Kohn-Sham result for the LDA exchange potential: [106] manuscript. He collaborated with Quin Luttinger as usual and their efforts produced a prediction for a new 1/3 2 3 mechanism for superconductivity based on a presumed v LDA(r)= −e n(r) . (34) x, π  oscillatory interaction between pairs of electrons (Kohn and Luttinger 1965). Meanwhile, back in La Jolla, Lu This expression was consequential at the time because, Sham began work on two density functional projects. r when Eq. (34) replaces vxc( ) in Eq. (30), the Kohn-Sham One of these, which became the final paper he and Kohn equations become almost identical to a set of equations would publish together, examined the one-body Green John Slater had proposed in 1951 as a local approxima- function of many-body theory (Sham and Kohn 1966). It tion to the non-local Hartree-Fock equations. I say ‘al- was important for them to study this quantity because its most’ identical because the ad hoc local exchange poten- properties determine the energy, lifetime, and spatial ex- tial proposed by Slater was tent of single-particle-like excitations out of the ground 1/3 state of a many-body system. At the same time, the 3 2 3 vx,Slater(r)= − e n(r) . (35) Hohenberg-Kohn theory implies that the Green function 2 π  is as a functional of the ground state electron density. KS argue for the correctness of their proposed exchange Sham, who did most of the calculations, demanded potential and it is notable that the abstract of Kohn and that the Green function satisfy the requirements of Sham (1965b) devotes a sentence to announcing the fac- particle-number conservation and charge neutrality and r tor of 3/2 difference between Eqs. (34) and (35). The deduced thereby that an electron at a point in an atom, authors’ motivation to do this was surely their awareness molecule, or solid responds to the electrostatic poten- that the so-called ‘Hartree-Fock-Slater’ method was in tial at that point and to exchange and correlation effects wide use by physicists performing band structure calcu- which depend on the electron density distribution in the r lations for real solids (Callaway 1958, Herman 1964). immediate vicinity of only. For a slowly-varying den- The Kohn-Sham equations [with and without the local sity, this conclusion justifies a local density approxima- tion for the Green function, which in turn provides an density approximation for vxc(r)] are the reason for the enduring importance of Kohn and Sham (1965b). The approximation for the energy spectrum and an indepen- paper itself differs in tone from Hohenberg and Kohn dent justification for using the LDA with the Kohn-Sham (1964) in the sense that the abstract notion of an “inho- equations to calculate the ground state electron density r mogeneous electron gas” disappears from the title and n( ). from most of the text. Instead, there is the practi- Sham’s second post-Kohn-Sham density functional cal promise of “self-consistent equations” appropriate to project was done in collaboration with Bok Yin Tong, a “real systems (atoms, molecules, solids, etc.) [where] the thirty-year old graduate student who had begun to work electronic density is nonuniform.” The Introduction is with Kohn. Their goal was to solve the Kohn-Sham equa- even more specific and makes the point that tions numerically for several atoms and ions. Luckily for them, Frank Herman and Sherwood Skillman had just most theoretical many-body studies have published a computer program which solved the Hartee- been concerned with elementary excitations Fock-Slater equations for atoms (Herman and Skillman and as a result there has been little progress 1963). Tong and Sham needed only to replace Eq. (35) by in the theory of cohesive energies, elastic con- Eq. (34) in the program and add some code for the cor- stants, etc. of real metals and alloys. The relation part of vxc(r) in the LDA. For this they used an methods proposed here offer the hope of new interpolation formula for the correlation energy derived progress in the latter area. from the information given in Pines (1963). The pub- That being said, KS did not themselves report any cal- lished paper, Tong and Sham (1966), focused on total en- culations of the cohesive energy or elastic constants (or ergies, total energy differences, and charge densities. The any other measurable quantity) for any ‘real’ electronic final results with correlation omitted were gratifying, giv- system. Indeed, they did not even bother to write down ing “slightly better results for energies and substantially an explicit form for the correlation part of vxc(r) in the better results for densities that Slater’s method.” The local density approximation. This was a straightforward correlation correction worsened the results, “presumably exercise for anyone familiar with the electron gas liter- because the electronic density in atoms has too rapid a ature. As for the LDA itself, KS remark that it should spatial variation.” “give a good representation of exchange and correlation Lu Sham published three articles unrelated to density effects . . for metals, alloys, and small-gap insulators.” functional theory before beginning an Assistant Profes- On the other hand, they warn the reader that the LDA sorship at the University of California at Irvine in the fall should have “no validity [at] the ‘surface’ of atoms and of 1966. Two years later, he accepted an offer to return the overlap regions of molecules. . . We do not expect to La Jolla as an Associate Professor. He was promoted an accurate description of chemical binding.” to Professor in 1975, served as a Dean from 1985-1989, 32 and is currently Emeritus Professor of Physics. He was Chemistry. elected to the US National Academy of Sciences in 1998. An important point mentioned earlier is that a com- At UCSD, Sham developed a broad research program puter program written to solve the local exchange poten- with a particular expertise in the theory of the electronic tial equations of the Hartree-Fock-Slater (HFS) method and optical properties of semiconductor heterostructures. (Slater 1951) was easily adapted to solve the Kohn-Sham However, the twenty papers he published on density func- equations in the local density approximation. There- tional theory over the years show that he never com- fore, after the 1965 publication of the Kohn-Sham paper, pletely abandoned the main subject of his post-doctoral systematic calculations for atoms began to reveal that work. Eq. (34) was superior to Eq. (35) as a local approximation Walter Kohn remained at San Diego until 1979, when to the exact, non-local exchange potential (Herman, Van he accepted the position as founding Director of the In- Dyke, and Oretenburger 1969). HFS and LDA calcula- stitute for Theoretical Physics (ITP), a research facility tions for molecules and solids were more difficult to evalu- established and supported by the US National Science ate because computational exigencies encouraged the use Foundation at the University of California at Santa Bar- of a ‘muffin-tin approximation” where the effective po- bara. He served as Director for five years and contin- tential in Eq. (30) was replaced by its spherical average ued as a Professor of Physics at UCSB until 1991 when inside a set of touching spheres centered at the atoms. A he gained Emeritus status. Kohn and his collaborators constant potential was used outside the spheres. published over 150 papers between 1965 and 2006. One Beginning in the 1970’s, a small group of scientists third of these explore some aspect of density functional committed themselves to carrying out HFS and LDA theory, particulary its application to solid surfaces. An calculations for small molecules without imposing the equal number of papers concern Kohn’s pre-DFT inter- muffin-tin (or any other) constraint on the self-consistent ests including disordered states of matter, superconduc- potentials (Baerends, Ellis, and Ros 1973, Gunnarsson, tivity, Bloch and Wannier functions, scattering theory, Harris, and Jones 1977, Becke 1982). Over time, HFS and the transition between the conducting and insulat- calculations disappeared in favor of LDA calculations ing states of matter. Kohn was elected to the National and the basic conclusion was that one obtained reason- Academy of Sciences (1969) and was a recipient of a Na- able binding energies and predictions for molecular struc- tional Medal of Science (1998) before winning a share of tures that agreed well with experiment (Jones 2012). On the 1998 Nobel Prize in Chemistry. the other hand, the LDA was not capable of the kind of ‘chemical accuracy’ (±2 kcal/mole) that the best ab initio methods of traditional quantum chemistry could VII. DISCUSSION AND CONCLUSION achieve. Then, in the 1980’s, efforts to go beyond the lo- cal density approximation led to the proposal and testing In March 2001, Kohn addressed a symposium on “The of various non-local approximations for the exchange and History of the Electronic Structure of Atoms, Molecules, correlation functional (Langreth and Mehl 1983, Perdew and Solids” at a meeting of the American Physical So- 1986, Becke 1988, Lee, Yang, and Parr 1988). These, so- ciety. With an audience well-schooled to interpret the called “generalized gradient approximations” (GGA) re- lower case Greek letter ǫ as an infinitesimally small quan- placed the electron gas exchange-correlation energy den- tity, Walter was understood immediately when he char- sity in Eq. (32) with much more complicated functions of acterized the initial reception of density functional the- both the local density n(r) and the local density gradi- ory as “+ ǫ by theoretical physicists and zero by the- ent ∇n(r). [108] Systematic Kohn-Sham calculations for oretical chemists” (Kohn 2001a). The small reaction atoms and molecules using GGA, particulary a hybrid by physicists reflected the fact that Kohn’s theory did approach introduced by Axel Becke (1993), quickly be- not directly address the “big” issues that occupied many gan to approach chemical accuracy. solid-state physicists at the time: superconductivity, the A turning point occurred in 1991 at the VIIth Inter- Kondo effect, superfluid helium, disorder-induced local- national Congress of Quantum Chemistry in Menton, ization, the metal-insulator transition, and quasi-one- France (Kohn 2001a). John Pople gave the final talk and dimensional conductors. The positive response was lim- summarized the achievements of ‘G2’ theory, his most ited mostly to band structure theorists who were well- comprehensive ab initio attempt to improve the Hartree- positioned to carry out the numerical work needed to Fock approximation using perturbation theory (Curtiss, solve the Kohn-Sham equations for real materials. [107] Raghavachari, Trucks, and Pople 1991). On the other The lack of response (or negative response) from chemists hand, earlier in the week, the Congress had given its came mostly from their near-universal belief that no the- triennial “outstanding young scientist” award to Axel ory of electronic structure based on the particle density Becke for his manifestly non-ab-initio work with DFT. alone could possibly be correct. In a future publication, Specifically, “for unique advances in numerical methods I will trace the evolving response of both the physics in density functional theory as applied to molecules, and and chemistry communities to DFT. For the present, it for important developments in the understanding of the suffices to sketch very briefly the path which led from exchange-correlation functional that enters density func- virtually no response by chemists to a Nobel Prize in tional theory” (IAQMS 2014). Pople made a point in his 33 talk to remark that he found Becke’s results “stimulating 1960, no fewer than 1000 papers in the scientific lit- and intriguing” (Pople 1991). erature concerned themselves with the ‘many-electron’ Barely a year later, Pople’s group published a system- problem.[109] No fewer than 1000 more papers focused atic comparison of the best quantum chemical calcula- on this topic between 1961 and 1965. The abstracts of tions with DFT calculations performed using a variety of these papers reveal five principal approaches to this issue: exchange-correlation potentials for 32 molecules. They the Thomas-Fermi method, electron gas models, many- concluded that the most sophisticated non-local function- electron wave function methods, density matrix methods, als “outperformed correlated ab initio methods, which and quantum field theory methods. Of these, only the ap- are computationally more expensive. Good agreement proximate Thomas-Fermi method singles out the ground with experiment was obtained with a small basis set” state electron density n(r) as the primary quantity for (Johnson, Gill, and Pople 1992). DFT was promptly study. Attempts to build “quantum corrections” into the incorporated into Pople’s widely-used GAUSSIAN com- Thomas-Fermi model were current in the early 1960’s, puter program and, with this endorsement, the popular- but none of these suggested that the theory could be ity of DFT calculations among chemists began to grow “exactified” to reveal the density as a truly fundamental exponentially (see Fig. 2). Accordingly, when the No- quantity. A paper inspired by Friedel’s alloy work which bel Chemistry Committee decided it was time to honor ignores correlation and derives a formal perturbation se- quantum chemistry with a Prize, it was not difficult for ries for the density in term of the external potential is them to split the award between John Pople and Walter perhaps closest in this regard (March and Murray 1961). Kohn. In the late 1950’s and early 1960’s, the ground state At the beginning of this paper, I suggested that den- electron density was not very interesting to many theoret- sity functional theory might be unknown today if Wal- ical solid state physicists. They focused instead on the ter Kohn had not created it in the mid-1960’s. Such a excited states of solids and the powerful new methods counter-factual claim can never be proved. However, it is of quantum field theory which made their study possi- interesting to examine the evidence that supports it. We ble (Hoddeson et al. 1992). Those who did appreciate do this with full awareness of a long tradition which ex- the general importance of the charge density—primarily amines great scientific discoveries from the personalistic the practitioners of band structure calculations—saw no and naturalistic points of view (Boring 1950). The former reason and had no motivation to elevate it to the lofty focuses on the specific attributes of an individual whose status of the many-electron wave function (Herman 1958, “exceptional insight may lead to an original discovery Pincherle 1960). The latter attitude was shared by the which has not been anticipated by others and which is rel- quantum chemistry community who devoted enormous atively independent of the times.” The latter posits that efforts to solving the Schr¨odinger equation for molecules “the zeitgeist (scientific climate of the time) determines with greater and greater accuracy (Barden and Schaefer the great discovery and that he who makes the discovery 2000, Gavroglu and Sim˜oes 2012). An interesting excep- is great merely because the times employed him.” Here, tion is the Canadian Richard Bader, perhaps the greatest I begin with the zeitgeist of electronic structure theory champion of n(r) among theoretical chemists. Just a year in Kohn’s lifetime and then turn to his particularism for before the publication of Hohenberg and Kohn (1964), he the case of DFT. wrote (Bader and Jones 1963):

The manner in which the electron density is disposed in a molecule has not received the attention its importance would seem to merit. Unlike the energy of a molecular system, which requires a knowledge of the second- order density matrix for its evaluation, many of the observable properties of a molecule are determined in whole or in part by the sim- ple three-dimensional electron density distri- bution.

Despite his fondness for the density, even Bader could not deny the primacy of the second-order density matrix (L¨owdin 1959). This was the rock-solid quantum chem- Fig. 2. Numbers of papers that mention ical view that Hohenberg had discovered in Paris when DFT as found by the Web of Science. he reviewed the literature of many-electron theory. To my knowledge, the only work in the pre-1964 The modern concept of the electronic structure of electronic structure literature where the electron den- atoms, molecules, and solids began when the old quan- sity plays a fundamental role is a one-page paper by tum theory of Bohr gave way to the new quantum the distinguished theoretical chemist, E. Bright Wilson, theory of Schr¨odinger (Jammer 1966). From 1925 to Jr. (Wilson 1962). Wilson asks the rhetorical question, 34

“Does there exist some procedure for calculating n(r) [for experiences at the University of Toronto were probably an N-electron system] which avoids altogether the use of typical of first-rate academic institutions at the time. 3N dimensional space?” He then uses a device mentioned What was not typical was the unusually high calibre earlier (see Section IV) and defines n(r, λ) to be the ex- of the individuals who mentored him and who (because act ground state electron density of a many-electron sys- of their own professional interests) repeatedly empha- tem where the charge of every electron is taken to be sized variational principles for both general proofs and λe rather than e. Using just a few lines of calculation, for detailed numerical calculations. It is true that all Wilson shows that the total energy E in Eq. (10) can be well-trained theoretical physicists learn about variational written as an integral from λ =0 to λ = 1 of the sum of principles and many use them occasionally in their pro- the classical Coulomb potentials produced by n(r, λ) at fessional work. However, very few physicists who learn the positions of all the nuclei. [110] about them as undergraduates go on to work with a From the foregoing, I conclude that no scientists before doctoral supervisor like Julian Schwinger who attacked Walter Kohn in 1963 were even vaguely thinking about almost every problem from a variational point of view, using the ground state charge density as a fundamental and then write a thesis where variational principles are quantity from which to build an exact theory of a many- used (again) both to prove a general result and to ob- electron system. The idea was not “in the air” and all tain numerical results for a specific quantum situation. eyes were riveted either on the many-electron wave func- Fundamental and numerical variational calculations ap- tion, the first- and second-order density matrices, or the pear over and over again in Kohn’s solid state work at Green functions of many-body field theory. The zeitgeist Carnegie Tech and Bell Labs through the 1950’s. It of electronic structure theory was simply not moving in is little wonder that he turned quickly to this power- the direction of the electron density function for some ful tool when he sought to prove the first Hohenberg- particulary well-prepared scientist to exploit and earn the Kohn theorem, which states that the many-body wave accolades of discovery. function and everything calculable from it are function- r I now turn to Kohn himself. Section IV detailed how als of the ground state electron density n( ). The second Walter chose to focus his fall 1963 sabbatical leave on a Hohenberg-Kohn theorem, which states that the total r problem that was not under active investigation by many energy takes its minimum value when n( ) is the exact of his theoretical colleagues. Namely, how might one cal- ground state density, is explicitly a variational result. culate the electronic structure of a three-dimensional dis- The Hohenberg-Kohn paper is austere, elegant, and ordered metal alloy, a system with no underlying spatial deep. Like several other papers in his oeuvre, it demon- periodicity? [111] His study of the metallurgy and metal strates a characteristic of his work that a Kohn-watcher physics literature inspired him to ask whether the elec- of fifty years tenure summarized in this way (Langer tron density n(r) was sufficient to completely characterize 2003): a many-electron system. It must be admitted that Kohn He always has loved mathematical elegance, was a product of his scientific milieu as much as any other but he reserves it for situations where it is electronic structure theorist working at the time. There- truly necessary. His emphasis [is always] on fore, the mere fact that this question came to his mind the most important physical questions and and he took it seriously must be regarded as a legitimate the ways in which they could be answered ‘eureka moment’ which few are privileged to experience. with insight and confidence. That being said, I wish to argue further that his partic- ular history, style of research, and scientific tastes made This observation is interesting and important, but it does him unusually well-suited to exploit this insight and cre- not distinguish Kohn from a number of other theoretical ate from it the edifice of density functional theory. physicists with a taste for proving theorems. However, Two aspects of Kohn’s pre-college years (surveyed in I believe it is unlikely that any of them would have had Section II) bear on the narrow question of his future life either the interest or the inclination to derive the Kohn- as a physicist. First, the cataclysm of the Anschluss put Sham equations and suggest the local density approxima- the budding classics scholar into contact with Emil Nohel tion for practical calculations. To make this case, I have and Victor Sabbath, two high school teachers whose pas- surveyed the publications of the most active theoretical sion for their subjects converted him to an enthusiastic solid state physicists working between 1950 and 1980. As student of mathematics and physics. Second, the camp a matter of personal taste, three broad activities engage schools Kohn attended while interned in Canada exposed them: formal calculations and proofs of theorems, anal- him to sophisticated one-on-one instruction from profes- yses of model Hamiltonians, and numerical calculations sional scientists. A pedagogical experience of this kind is for specific materials systems. If an individual was active barely imaginable at a conventional high school, then or in more than one of these, it was most often the first and now. In a normal setting, there is little chance that the second activities or the second and third activities. Kohn teen-aged Kohn would have encountered (much less de- is unusual among his peers simply because he followed up voured) books like Hardy’s A Course in Pure Mathemat- a paper which asks and answers a deep theoretical ques- ics and Slater’s Introduction to Chemical Physics. [112] tion with a paper which constructs a practical tool to Walter’s undergraduate and master’s level classroom perform calculations for specific systems. 35

Kohn’s own early research history demonstrates a will- tographs. Dozens of people communicated with me about ingness to compute actual numbers for direct compar- their experiences with Kohn and his work and I collec- ison with experiment. The Kohn-Sham equations are tively thank all of them here. Ms. Pia Otte provided the vehicle for this activity in the context of electronic excellent translations from German into English and I structure theory. In practice, Kohn turned over much of acknowledge Andre Bernard (John Simon Guggenheim the explicit numerical work on DFT to his students and Foundation), Carlo Siochi (University of Toronto), Gre- post-docs, but there is complete agreement among this gory Giannakis (McGill University), and Felicity Pors cohort that he never regarded this activity as a less im- (Niels Bohr Archive) for sending me copies of original portant part of his group’s research. This is the reason documents. The staffs of the Department of Special Col- that a senior quantum chemist could remark that Kohn lections of the Library of the University of California at is “the least arrogant of the deep physicists” in the sense Santa Barbara and of the Vancouver Holocaust Educa- that he does not “give a lower standing to those parts tion Centre provided essential help. Special thanks go to of physics that deal with the complexities of phenomena my Georgia Tech colleague, Glenn Smith, who read every governed by known laws” (Baerends 2003). The ‘com- word and offered many helpful suggestions. plexities’ mentioned here are discovered only by carrying out numerical computations for specific systems using the ‘known laws’ described by the Kohn-Sham formalism. IX. REFERENCES In summary, Walter Kohn earned one-half of the 1998 Nobel Prize in Chemistry by asking himself a simple (yet Abragam, A. 1989. Time Reversal. Oxford: Clarendon deep) scientific question about the electronic structure Press. of matter. He answered that question in an elegant and Abrikosov, A.A., L.P. Gorkov, and I.E. Dzyaloshinski. thought-provoking manner and then exploited his result 1963. Methods of Quantum Field Theory in Statistical to re-formulate the quantum many-electron problem in a Physics Englewood Cliffs, N.J.: Prentice-Hall. manner which made calculations for real systems compu- Adhikari, S.K. 1998. Variational Principles and the tationally cheap and surprisingly accurate. These things Numerical Solution of Scattering Problems. New York: may have been achieved by someone else if Kohn and Wiley. his post-doctoral associates had not done so in the years Agranovich, V. 1964. At the all-union conference on solid 1963-1965, but that person would probably look very state theory. Soviet Atomic Energy 17: 863-865. much like Walter Kohn himself. Aharonov, Y. and D. Bohm. 1959. Significance of electromagnetic potentials in the quantum theory. Physical Review 115: 485-491. Allin, E.J. 1981. Physics at the University of Toronto 1843-1980. Toronto: University Press. Allis, W.P. and M.A. Herlin. 1952. Thermodynamics and Statistical Mechanics. New York: McGraw Hill. Ambegaokar, V. 2013. May 12 2013 telephone interview with the author. Ambegaokar, V. and W. Kohn. 1959. Connection between true effective mass and optical absorption in insulators. Physical Review Letters 2: 385. Ambegaokar, V. and W. Kohn. 1960. Electromagnetic properties of insulators. Physical Review 117: 423-431. Anderson, P.W. 1958. Absence of diffusion in certain The physicist Walter Kohn learns a new trade after random lattices. Physical Review 109: 1492-1505. winning one-half the 1998 Nobel Prize in Chemistry. Anderson, P.W. 1977. Nobel Prize autobiography. Drawing by Peter Meller. Courtesy of Walter Kohn. http://www.nobelprize.org/nobel\ textunderscore prizes/physics/ laureates/1977/anderson-bio.html VIII. ACKNOWLEDGMENTS Anderson, P.W. 1978. Local moments and localized states. Reviews of Modern Physics 50: 191-201. First and foremost, I thank Walter Kohn, who gra- Anderson, P.W. 1987. John Hasbrouk Van Vleck. ciously granted me two interviews, answered my follow- Biographical Memoir Washington, D.C.: National up questions by email, and gave me access to several Academy of Sciences Press. private documents and photographs. Prof. Kohn’s ad- Anderson, P.W. 1999. Interview of Philip Anderson by ministrative assistant, Ms. Chris Seaton, deserves spe- Alexei Kojevnikov on May 29 1999. Niels Bohr Library cial mention for her unfailing help. I am also grate- and Archives, American Institute of Physics, College ful to Pierre Hohenberg and Lu Jeu Sham, who shared Park, MD, USA, http://www.aip.org/history/ohilist/ with me both their memories and their personal pho- 23362 2.html 36

Anderson, P.W. 2011. Physics at Bell Labs, 1949-1984. Physics 98: 5648-5652. pp. 73-80 in: More and Different. Hackensack, New Beeby, J.L. 1964. Electronic structure of alloys. Jersey: World Scientific. Physical Review 135: A130-A143. Anderson, P.W. 2012. August 7 2012 correspondence Bellemans A. and M. De Leener. 1961. Ground-state with the author. energy of an electron gas in a lattice of positive point Anderson, P.W., R.M. Friedberg, and W. Kohn. 1997. charges. Physical Review Letters 6: 603-604. Joaquin M. Luttinger. Physics Today 50 (12) 89-90. Beller, S. 1989. Vienna and the Jews: 1867-1938. APS 2013a. American Physical Society. Prizes. Cambridge: University Press. Available at http://www.aps.org/programs/honors/ Berkley, G.E. 1988. Vienna and its Jews. Cambridge, prizes/index.cfm MA: Abt Books. APS 2013b. American Physical Society. Membership Blatt, F.J. 1963. Transport properties in dilute alloys. growth, 1899-2013. Available at http://www.aps.org/ pp. II-1-11 in: J. Friedel and A. Guinier (eds.) membership/statistics/upload/APSMembership Metallic Solid Solutions. New York: Benjamin. Growth 2013.pdf Bloch, F. 1928. Uber¨ die Quantenmechanik der Aronszajn N. and A. Weinstein. 1943. On the unified Elektronen in Kristallgittern. Zeitschrift f¨ur Physik theory of eigenvalues of plates and membranes. 52: 555-600. American Journal of Mathematics 64: 623-645. Bohr, N. 1951. Evaluation of Walter Kohn. Niels Bohr Arrott, A. 2013. July 19 2013 telephone interview Archive. Copenhagen. with the author and subsequent correspondence. Boring, E.G. 1950. Great men and scientific progress. Auger, M.F. 2005. Prisoners on the Home Front. Proceedings of the American Philosophical Society Vancouver: UBC press. 94: 339-351. Avery, D.H. 1998. The Science of War: Canadian Borowitz, S. and W. Kohn. 1949. On the electro- Scientists and Allied Military Technology during the magnetic properties of nucleons. Physical Review Second World War. Toronto: University Press. 76:818-827. Bader, R.F.W. and G.A. Jones. 1963. Electron density Br´echet, Y. 2008. Presentation of Professor Jacques distributions in hydride molecules. The ammonia Friedel. ‘Leonardo da Vinci Award of the European molecule. Journal of Chemical Physics 38: 2791- Academy of Sciences’. http://www.eurasc.org/ 2802. davinci/davinci2010.asp Baerends, E.J. Walter Kohn, the chemist. pp. 20-22 in: Brockhouse, B.N. and A.T. Stewart. 1958. Normal M. Scheffler and P. Weinberger (eds.) Walter Kohn– modes of aluminum by neutron spectroscopy. Reviews Personal Stories and Anecdotes. Berlin: Springer. of Modern Physics 30: 236-249. Baerends, E.J., Ellis, D.E., and P. Ros. 1973. Self- Brook, A.G. and W.A.E. McBryde. 2007. Historical consistent molecular Hartree-Fock-Slater calculations Distillates: Chemistry at the University of Toronto I. the computational procedure. Chemical Physics since 1843. Toronto: Dundurn Group. 2: 41-51. Brown, W.L., R.C. Fletcher, and K.A. Wright. 1953. Baraff, G.A. and S. Borowitz. 1961. Green’s function Annealing of bombardment damage in germanium: method for quantum corrections to the Thomas-Fermi experimental. Physical Review 92: 591-596. model of the atom. Physical Review 121: 1704-1713. Brueckner, K. 2000. Highlights of many-body physics. Bardeen, J., L.N. Cooper, and J.R. Schrieffer. 1957. pp. 25-29 in: R.F. Bishop, K.A. Gernoth, N.R. Walet, Theory of superconductivity. Physical Review 108: and Y. Xian (eds.) Recent Progress in Many-Body 1175-1204. Theories. Singapore: World Scientific. Barden, C.J. and H.F. Schaeffer III. 2000. Quantum Brueckner, K. 2013. Keith Brueckner and the chemistry in the 21st century. Pure and Applied founding of UCSD. Chronicles: Newsletter of the Chemistry 72: 1405-1423. UCSD Emeriti Association 12 (4): 6-7. Bassani, F. and M. Tosi. 1988. Theoretical research Bruch, L.W. 2013. April 10 2013 correspondence with in the physics of solids. pp. 129-136: in G. Giuliani the author. (ed.) The Origins of Solid State Physics in Italy: Buckingham, A.D. 2006. Sir John Anthony Pople. 1945-1960. Bologna: Italian Physical Society. Biographical Memoirs of Members of the Royal Society Baym, G. 1969. Lectures on Quantum Mechanics. 52: 299-314. Menlo Park: Benjamin Cummings. Callaway, J. 1958. Energy bands in solids. pp: 99-212 Becke, A.D. 1982. Numerical Hartree-Fock-Slater in: F. Seitz and D. Turnbull (eds.) Solid State Physics, calculations on diatomic molecules. Journal of volume 7, New York: Academic. Chemical Physics 76: 6037-6045. Callaway, J. and W. Kohn. 1962. Electron wave func- Becke, A.D. 1988. Density functional exchange energy tions in metallic lithium. Physical Review 127: 1913. approximation with correct asymptotic behavior. Callen, H.B. 1960. Thermodynamics. New York: John Physical Review A 38: 3098-3100. Wiley. Becke, A.D. 1993. Density functional thermochemistry. Casey, N.T. 1950. A critical analysis of Physics S-1b. III. The role of exact exchange. Journal of Chemical Walter Kohn Papers. Box 23, Folder 4. UArch FAcP 37

34. Department of Special Collections. University Cyclotron resonance of electrons and holes in silicon of California, Santa Barbara. and germanium crystals. Physical Review 98: 368-384. Celli, V. 2013. February 8 2013 telephone interview with Duff, G.F.D. 1969. Arthur Francis Chesterfield Steven- the author. son. 1899-1968. Proceedings of the Royal Society of Cesarini, D. and T. Kushner. 1993. The Internment of Canada. Series IV. 8: 104-107. Aliens in Twentieth Century Britain. London: Frank Edwards, S.F. 1998. Reminiscences. pp. 203-204 in: Cass & Co. Ltd. E.A. David (ed.) Nevill Mott, Reminiscences and Cheetham, A. 1992. Video interview of Walter Kohn and Appreciations. London: Taylor and Francis. Philippe Nozi`eres for the Vega Science Trust. Ehrenreich, H. and L.M. Schwartz. 1976. The electronic Available at http://vega.org.uk/video/programme/134 structure of alloys. pp. 149-286 in: H. Ehrenreich, Chien, C.L. and C.R. Westgate. 1980. The Hall Effect F. Seitz, and D. Turnbull (eds.) Solids State Physics, and Its Applications. New York: Plenum. volume 31, New York: Academic. Cohen, M.H. 1963. Interatomic interactions in metals. Eisinger, J. and G. Feher. 1958. Hfs anomaly of Sb121 pp. XI-1-9 in: J. Friedel and A. Guinier (eds.) Metallic and Sb123 determined by the electron nuclear double Solid Solutions. New York: Benjamin. resonance technique. Physical Review 109: 1172-1183. Coleman, A.J. 1963. Structure of fermion density Eisinger, J. 2003. For Rappa on his 80th Birthday from matrices. Reviews of Modern Physics 35: 668-687. Terry. pp. 63-65 in: M. Scheffler and P. Weinberger Crawford, T.D, Wesolowski, S.S., Valeev, E.F., King, (eds.) Walter Kohn–Personal Stories and Anecdotes. R.A., Leininger, M.L., and H.F Schaefer III. 2001. The Berlin: Springer. past, present, and future of quantum chemistry. pp. Eisinger, J. 2011. Einstein on the Road. Amherst, New 219-246 in: E. Keinan and I. Schechter (eds.) Chemis- York: Prometheus Books. try for the 21st Century. Weinheim: Wiley-VCH. Eisinger, J. 2013. July 24 2013 telephone interview by Curio, C. 2004. Invisible children: the selection and the author and subsequent correspondence. integration strategies of relief organizations. Shofar Ehrlich, G. 2003. Reunion in history. pp. 60-62 in: M. 23: 41-56. Scheffler and P. Weinberger (eds.) Walter Kohn– Curtiss, L.A, Raghavachari, K., Trucks, G.W., and Personal Stories and Anecdotes. Berlin: Springer. J.A. Pople. 1991. Gaussian-2 theory for molecular Faddeev, L. 1965. Mathematical Aspects of the energies of first- and second-row compounds. Three Body Problem in Quantum Scattering Theory. Journal of Chemical Physics 94:7221-7230. Jerusalem: Israel Program for Scientific Translations. Daniel, E. and S.H. Vosko. 1960. Momentum distri- Fast, V.K. 2011. Children’s Exodus: a History of the bution of an interaction electron gas. Physical Kindertransport. London: I.B. Tauris. Review 120:2041-2044. Feher, G. 2002. The creation of the physics department. DeWitt, B.S. 1951. Theoretical Physics. Physics Today Chronicles: Newsletter of the UCSD Emeriti 4(12): 22-23. Association 2 (2): 6-8. DeWitt, C.M. 2013. Program of the 1951 Summer School Feibelman, P.J. 2012. June 28 2012 telephone interview of Theoretical Physics, Les Houches, France. Courtesy with the author. Peter Feibelman was a PhD student of Prof. C´ecile deWitt, University of Texas. of Keith Brueckner at UCSD. Diaz, J.B. 1978. Alexander Weinstein Selecta. London: Feldberg, W.S. 1960. Bruno Mendel. 1897-1959. Pitman. Biographical Memoirs of Fellows of the Royal Dirac, P.A.M. 1930. Note on exchange phenomena in the Society 6: 190-199. Thomas atom. Proceedings of the Cambridge Fernandez, B. 2013. Unravelling the Mystery of the Philosophical Society 26: 376-385. Atomic Nucleus New York: Springer. Dirac, P.A.M. 1935. The Principles of Quantum Feynman, R.P. 1949. Space-time approach to quantum Mechanics 2nd edition. Oxford: Clarendon Press. electrodynamics. Physical Review 76: 769-789. Domb, C. 1996. The Critical Point: a Historical Fletcher, R.C. 2013. October 30 2013 correspondence Introduction to the Modern Theory of Critical with the author. Phenomena. London: Taylor & Francis. Florides, P.S. 2008. John Lighton Synge. 1897-1995. de Dominicis, C. 1963. Variational statistical Biographical Memoirs of Fellows of the Royal Society mechanics in terms of ‘observables’ for normal and 54: 401-424. superfluid systems. Journal of Mathematical Fock, V. 1930. Approximate methods for the solution Physics 4: 255-265. of the quantum mechanical many-body problem (in de Dominicis, C. and P.C. Martin. Stationary entropy German). Zeitschrift f¨ur Physik 61: 126-148. principle and renormalization in normal and superfluid Ford, J. 2013. June 19 2013 correspondence between the systems. Journal of Mathematical Physics 5: 14-30. author and Jon Ford, Head Teacher, Imberhorne Sec DuBois, D.F. and M.G. Kivelson. 1962. Quasi-classical ondary School, East Grinstead, England. theory of electron correlation in atoms. Physical Frank, P. 1947. Einstein, his Life and Times. New York: Review 127: 1182-1192. Alfred A. Knopf. Dresselhaus, G., A.F. Kip, and C. Kittel. 1955. Friedel, J. 1954. Electronic structure of primary solid 38

solutions. Advances in Physics 3: 446-507. Physics 67: 3970-3979. Friedel, J. 1958. Metallic alloys. Supplemento del ter Haar, D. 1954. Elements of Statistical Mechanics Nuovo Cimento 7: 287-311. Rinehart: New York. Friedel, J. 1963. The concept of the virtual bound state. Hanta, K. 1999. From exile to excellence: an interview pp. XIX-1-22 in: J. Friedel and A. Guinier (eds.) with Nobel prize laureate Walter Kohn. Metallic Solid Solutions. New York: Benjamin. Austria Kultur 9 (1). Available online at Frisch, O.R. 1979. What Little I Remember. Cambridge: http://www.auslandsdienst.at/de/projekt/ University Press. pressearchiv-1999/austria-kultur-vol-9-no1- Galitskii, V.M. and A.B. Migdal. 1958. Application of januaryfebruary-1999-exile-excellence-interview quantum field theory methods to the many-body Hardy, G.H. 1938. A Course in Pure Mathematics. problem. Journal of Experimental and Theoretical Seventh edition. Cambridge: University Press. Physics 34: 139-150. Harrison, W.A. and M.B. Webb. 1960. The Fermi G´asp´ar, R. 1954. About an approximation to the Surface John Wiley: New York. Hartree-Fock potential through a universal potential Hasegawa, H. 2004. Walter Kohn in Japan. pp. 93-95 function (in German). Acta Physica Hungarica 3: 263- in: M. Scheffler and P. Weinberger (eds.) Walter 286. Kohn–Personal Stories and Anecdotes. Berlin: Gavroglu, K. and A. Sim˜oes. 2012. Neither Physics nor Springer. Chemistry Cambridge, MA: MIT Press. Herman, F. 1958. Theoretical investigation of the Gell-Mann, M. 1996. Reminiscences. Philosophical electronic energy band structure of solids. Reviews of Magazine 74: 431-434. Modern Physics 30: 102-121. de Gennes, P.G. 1966. Superconductivity in Metals and Herman, F. 1964. Recent progress in energy band theory. Alloys. Reading, MA: Benjamin. pp. 3-22 in: Hulin, M. (ed.) Physics of Semi- Geoffroy, P.R. 1946. Report on the Magnetometer Survey conductors. Paris: Dunod. of the property of Dante Red Lake Gold Mines Ltd. Herman, F. and S. Skillman. 1963. Atomic Structure Available as file 52N04N9958.pdf from the Ontario Calculations. Englewood Cliffs, New Jersey: Ministry of Northern Development and Mines. Prentice-Hall. Giannakis, G. 2013. August 6 2013 correspondence Herman, F., Van Dyke, J.P., and I.B. Ortenburger. between the author and G. Giannakis, a member of 1969. Physical Review Letters 22: 807-811. the reference staff of the McGill University Archives. Hinman, G. and D. Rose. 2010. Edward Chester Creutz. Gilman, P. and L. Gilman. 1980. Collar the Lot: How Biographical Memoir Washington, D.C.: National Britain Interned and Expelled its Wartime Refugees. Academy of Sciences Press. London: Quartet Books. Hoddeson, L., E. Braun, J. Teichmann, and S. Weart. Glasser, M.L. 2013. July 12 2013 telephone interview by 1992. Out of the Crystal Maze. New York: Oxford the author and subsequent correspondence with former University Press. Carnegie Tech graduate student Larry Glasser. Hoddeson, L., H. Schubert, S.J. Heims, and G. Baym. Gold, A.V. 1958. An experimental determination of the 1992. Collective Phenomena. pp. 489-616: in Fermi surface in lead. Philosophical Transactions of Hoddeson, L., E. Braun, J. Teichmann, and S. Weart the Royal Society A 251: 85-112. (eds.) Out of the Crystal Maze. New York: Goldman, J.E. 1957. The Science of Engineering Oxford University Press. Materials. New York: John Wiley. Hodges, C.H. and Stott, M.J. 1972. Theory of electro- Goldstone, J. 1957. Derivation of Brueckner many-body chemical effects in alloys. Philosophical Magazine theory. Proceedings of the Royal Society A 239: 26: 375-392. 267-279. Hohenberg, P.C. 2003. A personal tribute to Walter Goudsmit, S.A. and G.L.Trigg. 1964. October 1964 Kohn on his 80th birthday. pp. 99-102 in: M. Memorandum from the Editors of Physical Review Scheffler and P. Weinberger (eds.) Walter Kohn– Letters. Personal Stories and Anecdotes. Berlin: Springer. Goudsmit, S.A. 1965. May 28 letter to L.J. Sham. Hohenberg, P.C. 2012. July 19 2012 interview with the Courtesy of L.J. Sham. author. Greene, M.P. 2013. March 30 2013 telephone interview Hohenberg, P. and Kohn, W. 1964. Inhomogeneous with the author. electron gas. Physical Review 136: B864-B871. Griffin, A. 2007. Many-body physics in the 1960’s: a Hohenberg, P.C., W. Kohn, and L.J. Sham. 1990. The golden age. A talk given at the symposium, ‘Fifty beginnings and some thoughts on the future. Years of Condensed Matter Physics’. June 16 2007. Advances in Quantum Chemistry 21: 7-26. Available at http://www.lassp.cornell.edu/ Hollander, N. 2000. Interview with Walter Kohn. Nobel lassp data/LASSP/50YearsCMP-VA2007.pdf Voices Video History Project. Archives Center, Gunnarsson, O., Harris, J. and R.O. Jones. 1977. Smithsonian National Museum of American History. Density functional theory and molecular bonding. I. Horsley, A. 2013. June 25 2013 telephone interview by first-row diatomic molecules. Journal of Chemical the author with Allen Horsley (son of Caperton B. 39

Horsley, the founder of the Sutton-Horsley Co.) and Memoir of the University of California, 1949-1967. subsequent email correspondence with his sisters, Volume 1. Academic Triumphs. Berkeley: University Rose Shelton Horsley Cruz and Lucile Horsley of California Press. Blanchard. Kittel, C. and A.H. Mitchell. 1954. Theory of donor and Houghton, A. 1961. Specific heat and spin susceptibility acceptor states in silicon and germanium. Physical of dilute alloys. Journal of the Physics and Chemistry Review 96: 1488-1493. of Solids 20: 289-293. Kittel, C. 1963 Quantum Theory of Solids New York: Hubbard, J. 1957. The description of collective motions Wiley. in terms of many-body perturbation theory. Kjeldaas, T. and W. Kohn. 1956. Interaction of con- Proceedings of the Royal Society (London) duction electrons and nuclear magnetic moments in A 239: 539-560. metallic sodium. Physical Review 101: 66-67. Hubbard, J. 1958. The description of collective motions Kjeldaas, T. 1959. Theory of Ultrasonic Cyclotron Reso- in terms of many-body perturbation theory III. The nance in Metals at Low Temperature. Ph.D. Thesis. extension of the theory to the non-uniform gas. University of Pittsburgh. Proceedings of the Royal Society (London) A 244: Klein, F. and A. Sommerfeld. 1898. Uber¨ die Theorie 199-211. des Kreisels Leipzig: Teubner. Hulth´en, L. 1946. The variational principle for Koch, E. 1980. Deemed Suspect. Toronto: Methuen. continuous spectra. pp. 201-206: in Dixi`eme Kohn, W. 1945. The spherical gyrocompass. Quarterly Congre`es des Math´ematiciens Scandinaves of Applied Mathematics 3: 87-88. Copenhagen: Julius Gjellerups Forlag. Kohn, W. 1946. Contour integration in the theory of the Hume-Rothery, W. 1931. The Metallic State. Oxford: spherical pendulum and the heavy symmetrical top. Clarendon Press. Transactions of the American Mathematical Hume-Rothery, W. and G.V. Raynor. 1962. The Society 59: 107-131. Structure of Metals and Alloys. 4th edition. Kohn, W. 1947. Two applications of the variational London: Institute of Metals. method to quantum mechanics. Physical Review 71: IAQMS. 2014. Awards website of the International 635-637. Academy of Quantum Molecular Science. Kohn, W. 1948. Variational methods in nuclear collision www.iaqms.org/awards.php problems. Physical Review 74: 1763-1772. Infeld, L. 1978. Why I Left Canada. 2nd edition. Kohn, W. 1952a. Validity of Born expansions. Montreal: McGill-Queen’s University Press. Physical Review 87: 539-540. Jammer, M. 1966. The Conceptual Development Kohn, W. 1952b. Variational methods for periodic of Quantum Mechanics. New York: McGraw-Hill. lattices. Physical Review 87: 472-481. Johnson, B.G., Gill, P.M.W., and J.A. Pople. 1992. Kohn, W. 1954. Interaction of conduction electrons and Journal of Chemical Physics 97: 7846-7848. nuclear magnetic moments in metallic lithium. Jones, H. 1934. The theory of alloys in the γ-phase. Physical Review 96: 590-592. Proceedings of the Royal Society of London A 144: Kohn, W. 1957a. Shallow impurity states in Si and Ge. 225-234. Solid State Physics. F. Seitz and D. Turnbull (eds.) Jones, R.O. 2012. Density functional theory: a personal pp:257-320. view. pp. 1-28: in M. Avella and F. Mancini (eds.): Kohn, W. 1957b. Effective mass theory in solids from a Strongly Correlated Systems: Theoretical Methods. many-particle point of view. Physical Review 105: Berlin: Springer-Verlag. 509-516. Jones, T. 1988. Both Sides of the Wire. Fredericton, Kohn, W. 1958. Interaction of charged particles in a New Brunswick, Canada: New Ireland Press. dielectric. Physical Review 110: 857-864. Jones, W., March, N.H., and S. Sampanthar. 1961. Kohn, W. 1959a. Analytic properties of Bloch waves and The energy and the Dirac density matrix of a non- Wannier functions. Physical Review 115: 809-821. uniform electron gas. Physics Letters 1: 303-304. Kohn, W. 1959b. Theory of Bloch electrons in a magnetic Jost, R. and W. Kohn. 1952a. Construction of a field: the effective Hamiltonian. Physical Review potential from a phase shift. Physical Review 87: 115: 1460-1478. 977-992. Kohn, W. 1959c. Image of the Fermi surface in the Jost, R. and W. Kohn. 1952b. Equivalent potentials. vibration spectrum of a metal. Physical Review Letters Physical Review 88: 382-385. 2: 393-394. Jost, R. and W. Kohn. 1953. On the relation between Kohn, W. 1961. Cyclotron resonance and de Haas- phase shift energy levels and the potential. Det van Alphen oscillations of an interacting electron gas. Kongelige Danske Videnskarbernes Selskab Physical Review 123: 1242-1244. Matematisk-fysiske Meddelelser 27: 1-19. Kohn, W. 1962. John Simon Guggenheim Memorial Kaiser, D. 2005. Drawing Theories Apart. Chicago: Foundation Application. Used by permission of Walter University Press. Kohn and the Guggenheim Memorial Foundation. Kerr, C. 2001. The Gold and the Blue: a Personal Kohn, W. 1964. Theory of the insulating state. Physical 40

Review 133: A171-A181. Kohn, W. and J.M. Luttinger 1960. Ground state of a Kohn, W. 1996. Tribute to Julian Schwinger. pp. 61-64: many-fermion system. Physical Review 118: 41-45. in Y. Jack Ng (editor): Julian Schwinger: the Kohn, W. and S.J. Nettel. 1960. Giant fluctuations in a Physicist, the Teacher, and the Man. Singapore. degenerate fermi gas. Physical Review Letters 5: 8-9. World Scientific. Kohn, W. and N. Rostoker. 1954. Solution of the Kohn, W. 1998. Nobel Prize autobiography. Available Schr¨odinger equation in periodic lattices with an online at http://www.nobelprize.org/nobel prizes/ application to metallic lithium. Physical Review chemistry/laureates/1998/kohn-autobio.html. 94: 1111-1120. Kohn, W. 1999. Nobel lecture: electronic structure of Kohn, W., D. Ruelle, and A. Wightman. 1992. Res Jost. matter–wave functions and density functionals. Physics Today 45: 120-121. Reviews of Modern Physics 71: 1253-1266. Kohn, W. and D. Schechter. 1955. Theory of acceptor Kohn, W. 2000. Letter of August 23 2000 from levels in germanium. Physical Review 99: 1903-1904. Walter Kohn to the Principal of the Imberhorne Kohn, W. and L.J. Sham. 1965a. Quantum density Secondary School, East Grinstead, England. oscillations in an inhomogeneous electron gas. Kohn, W. 2001a. A personal account of the history of Physical Review 137: A1697-A1705. density functional theory. Session S3 (History of Kohn, W. and L.J. Sham. 1965b. Self-consistent electronic structure theory in atoms, molecules, and equations including exchange and correlation effects. solids), American Physical Society annual meeting, Physical Review 140: A1133-A1138. March 14 2001, Seattle, WA. Sound recording AV Kohn, W. and S.H. Vosko. 1960. Theory of nuclear 2001-374z used with the permission of the Niels Bohr resonance intensity in dilute alloys. Physical Library & Archives, American Institute of Physics Review 119: 912-918. One Physics Ellipse, College Park, MD 20740, USA. Korringa, J. 1947. On the calculation of the energy of a Kohn, W. 2001b. Graduate School Days. p. 6 in the fall Bloch wave in a metal. Physica 13:392400. 2002 Newsletter of the American Physical Society Korringa, J. 1958. Dispersion theory for electrons Forum on Graduate Student Affairs. One Physics in a random lattice with applications to the electronic Ellipse, College Park, Maryland. structure of alloys. Journal of the Physics and Kohn, W. 2003. A Fireside Chat with Nobel Laureate Chemistry of Solids 7: 252-258. Professor Walter Kohn. An August 15 2003 broadcast Korringa, J. 1994. Early history of multiple scattering by the University of California Television. Online at theory for ordered systems. Physics Reports. 238: http://www.youtube.com/watch?v=VDNNiKdtyhg. 341-360. Kohn, W. 2004. My honored teachers in Vienna. pp. Kryachko, E.S. and C.V. Lude˜na. 1990. Energy Density 43-50 in: F. Stadler (editor): Osterreichs¨ Umgang Functional Theory of Many-Electron Systems mit dem Nationalsozialismus. Wien: Springer. Dordrecht: Kluwer Academic Publishers. Kohn, W. 2012a. Congratulation. Journal of Super- Kubo, R. 1957. Statistical mechanical theory of irrever- conductivity and Novel Magnetism 25: 551. sible processes. I. Journal of the Physical Society Kohn, W. 2012b. December 18 2012 interview conducted of Japan. 12: 570-586. by the author. Lambek, J. 1980. Reminiscences of Fritz Rothberger. Kohn, W. 2013a. March 3 2013 interview conducted by Published in the September 2000 issue of the Nina Krieger for the Internment Project of the Van- newsletter CMS Notes de la SMC 32: (5) 29. couver Holocaust Education Centre (VHEC). The Landau, L.D. 1956. The theory of a Fermi liquid. author thanks archivist Elizabeth Shaffer of the Journal of Experimental and Theoretical Physics VHEC for making a videotape of this interview 30: 1058-1064. available to him. Langer, J.S. 2003. Reminiscences on the occasion Kohn, W. 2013b. October 8 2013 interview conducted of Walter Kohn’s 80th birthday. pp. 124-126 in: by the author and subsequent correspondence. M. Scheffler and P. Weinberger (eds.) Walter Kohn– Kohn, W. and N. Bloembergen. 1950. Remarks on the Personal Stories and Anecdotes. Berlin: Springer. nuclear resonance shift in metallic lithium. Physical Langer, J.S. and S.H. Vosko. 1959. The shielding of a Review 80: 913. fixed charge in a high-density electron gas. Journal Kohn, W. and M. Luming. 1963. Orbital susceptibility of Physics and Chemistry of Solids 12: 196-205. of dilute alloys. Journal of the Physics and Langreth, D.C. and M.J. Mehl. 1983. Beyond the Chemistry of Solids 24: 851-862. local density approximation in calculations of Kohn, W. and J.M. Luttinger. 1955a. Hyperfine splitting ground state electronic properties. Physical Review B of donor states in silicon. Physical Review 97: 883-888. 28: 1809-1834. Kohn, W. and J.M. Luttinger. 1955b. Theory of donor Lee, C., Yang, W., and R.G. Parr. 1988. Develop- states in silicon. Physical Review 98: 915-922. ment of the Colle-Salvetti correlation energy into a Kohn, W. and J.M. Luttinger. 1957. Quantum theory of density functional of the electron density. Physical electrical transport phenomena. Physical Review 108: Review A 37: 785-789. 590-611. LJPS 1985. Proceedings of the La Jolla Physics 41

Symposium. September 6-8, 1985. pp. 142-145. Millman, S. 1983. A History of Engineering and Science http://www.physics.ucsd.edu/dept/department in the Bell System: Physical Science (1935-1980). history.pdf Murray Hill, New Jersey, Bell Telephone L¨owdin, P.-O. 1959. Correlation problem in many- Laboratories. electron quantum mechanics. Advances in Chemical Morse, P.M. 1982. John Clarke Slater. Biographical Physics 2: 207-322. Memoir Washington, D.C.: National Academy of Luttinger, J.M. 1960. Fermi surface and some Sciences Press. simple equilibrium properties of a system of interaction Mott, N.F. 1936. Resistivity of dilute solid solutions. fermions. Physical Review 119: 1153-1163. Proceedings of the Cambridge Philosophical Society 32: Luttinger, J.M. 1961. Theory of the de Haas-van Alphen 281-290. effect for a system of interacting electrons. Physical Mott, N.F. 1937. The energy of the superlattice in β Review 121: 1251-1258. brass. Proceedings of the Physical Society 49: Luttinger, J.M. and W. Kohn. 1955. Motion of electrons 258-263. and holes in perturbed periodic fields. Physical Mott, N.F. 1949. The basis of the electron theory of Review 97: 869-883. metals, with special reference to the transition metals. Luttinger, J.M. and W. Kohn. 1958. Quantum theory of Proceedings of the Physical Society (London) 62: electrical phenomena. II. Physical Review 109: 416-422. 1892-1909. Mott, N.F. and H. Jones. 1936. The Theory of the Luttinger, J.M. and P. Nozi`eres. 1962. Derivation of the Properties of Metals and Alloys. Oxford: University Landau theory of Fermi liquids. II. Equilibrium Press. Properties and Transport Equation. Physical Musher, J.I. 1966. Comment on some theorems of Review 127: 1431-1440. quantum chemistry. American Journal of Physics March, N.H. 1957. The Thomas-Fermi approximation in 34: 267-268. quantum mechanics. Advances in Physics 6:1-101. Muto, T. 1938. On the electronic structure of alloys. March, N.H. 1975. Self-Consistent Fields in Atoms. Scientific Papers of the Institute of Physical and Oxford: Pergamon Press. Chemical Research (Tokyo) 34: 377-390. March, N.H. and A.M. Murray. 1961. Self-consistent NBA 1951a. Niels Bohr Archive. Guest Book of the perturbation treatment of impurities and imperfec- Institute for Theoretical Physics. Copenhagen. tions in metals. Proceedings of the Royal Society A Denmark 261: 119-133. NBA 1951b. Niels Bohr Archive. Program and Markoff, J. 2011. Jacob E. Goldman, founder of Xerox List of Participants for the Conference on Problems lab dies at 90. New York Times December 23 2011: of Quantum Physics. July 6-10 1951. Institute for B17. Theoretical Physics. Copenhagen. Denmark. Martin, P.C. and S.L. Glashow. 1995. Julian Schwinger: Nesbet, R.K. 2003. Variational Principles and Methods prodigy, problem solver, pioneering physicist. Physics in Theoretical Physics and Chemistry. Cambridge: Today 48 (10): 40-46. University Press. McWeeny, R. 1960. Some recent advances in density Neuhaus, H. 2003. A class with class. pp. 173-176 in: matrix theory. Reviews of Modern Physics 32: M. Scheffler and P. Weinberger (eds.) Walter 335-369. Kohn–Personal Stories and Anecdotes. Berlin: Mehra, J. and K.A. Milton. 2000. Climbing the Moun- Springer. tain. Oxford: University Press. Nobel Media. 2013. The Nobel Prize in Chemistry 1998. Mermin, N.D. 1965. Thermal properties of the inhomo- Online at http://www.nobelprize.org/nobel prizes/ geneous electron gas. Physical Review 137: chemistry/laureates/1998/ A1441-A14443. Nordheim, L.H. 1931. Zur electronentheorie der metalle Mermin, N.D. 2003. Memorable moments with Walter II. Annalen der Physik 9: 641-678. Kohn. pp. 155-159 in: M. Scheffler and P. Weinberger Nozi`eres, P. 1963. Le Probl`eme a N Corps Paris: Dunod. (eds.) Walter Kohn–Personal Stories and Nozi`eres, P. 2012a. Sixty years of condensed matter Anecdotes. Berlin: Springer. physics: an everlasting adventure. Annual Reviews Mermin, N.D. 2013. February 12 2013 correspondence of Condensed Matter Physics 3: 1-7. with the author. Nozi`eres, P. 2012b. July 6 2012 correspondence with the Meyenn, K. von. 1989. Physics in the Making in Pauli’s author. Z¨urich. pp. 93-130 in: A. Sarlemijn and M.J. Sparnaay Nozi`eres, P. and J.M. Luttinger. 1962. Derivation of the (editors): Physics in the Making. Amsterdam: Landau theory of Fermi liquids I. formal prelimi- North-Holland. naries. Physical Review 127: 1423-1431. Miedema, A.R., F.R. de Boer, and P.F. de Chatel. 1973. Old, B.S. 1961. The evolution of the Office of Naval Empirical description of the role of electronegativity in Research. Physics Today 14(8): 30-35. alloy formation. Journal of Physics F: Metal Physics Pais, A. 1982. Subtle is the Lord: the Science and Life 3: 1558-1576. of Albert Einstein. Oxford: Clarendon Press. 42

Pais, A. 1996. Res Jost. pp. 1-9 in : K. Hepp (editor) the author. Das M¨archen vom Elfenbeinernen Turm: Reden Rowland, T.J. 1960. Nuclear magnetic resonance in und Aufs¨atze. Berlin: Springer. copper alloys: electron distribution around solute Park, B.S. 2009. Between accuracy and manageability: atoms. Physical Review 119: 900-912. computational imperatives in quantum chemistry. Rozental, S. 1967. The forties and the fifties. Historical Studies in the Natural Sciences pp. 149-190 in: S. Rozental (editor). Niels Bohr 39: 32-62. Amsterdam: North Holland. Parr, R.G. and W. Yang. 1989. Density-Functional R¨urup R. 2008. Schicksale und Karrieren Gottingen: Theory of Atoms and Molecules Oxford: Clarendon Wallstein. Press. Rudnick, J. 2003. It started with image charges. pp. Pauling, L. 1949. A resonating-valence-bond theory of 208-210 in: M. Scheffler and P. Weinberger (eds.) metals and intermetallic compounds. Proceedings of Walter Kohn–Personal Stories and Anecdotes. the Royal Society of London A 196: 343-262. Berlin: Springer. Peierls, R. 1933. Zur theorie des diamagnetismus von Sapolsky, H.M. 1990. Science and the Navy Princeton: leitungselektronen. Zeitschrift f¨ur Physik. University Press. 80: 763-791. Schwinger, J. 1947. A variational principle for scattering Percus, J.K. 1963 The Many-Body Problem. New York: problems. Physical Review 72: 742. Interscience. Schweber, S.S. 1994. QED and the Men who Made it: Perdew, J.P. 1986. Density functional approximation for Dyson, Feynman, Schwinger, and Tomonaga. the correlation energy of the inhomogeneous electron Princeton: University Press. gas. Physical Review B 33: 8822-8824. SDUT. 2010. Lois M. Kohn. Obituary. San Diego Union Perutz, M. 1985. Enemy Alien. The New Yorker. Tribune, January 22 2010. August 12 1985: 35-54. Sears E. 1990. The Life and Work of William S. Pincherle, L. 1960. Band structure calculations in Heckscher. Zeitschrift f¨ur Kunstgeschichte solids. Reports on Progress in Physics 23: 355-394. 53: 107-133. Pines, D. 1961. The Many-Body Problem. Reading, MA: Seitz, F. 1940. The Modern Theory of Solids. New York: Benjamin. McGraw-Hill. Pines, D. 1963. Elementary Excitations in Solids. New Seitz, F. 1994. On the Frontier: My Life in Science. New York: Benjamin. York: American Institute of Physics Press. Pippard, A.B. 1957. An experimental determination of Seitz, F. and D. Turnbull. 1955. Solid State Physics. the Fermi surface in copper. Philosophical Volume 1. New York: Academic. Transactions of the Royal Society A. 250: 325-357. Senft, G. 2003. Economic development and economic Pople, J.A. 1965. Two-dimensional chart of quantum policies in the St¨andestaat era. pp. 32-55 in: G. chemistry. Journal of Chemical Physics 43: S229-S230. Bischof, A. Pelinka and A. Lassner (eds.): The Pople, J.A. 1991. The computation of molecular energies. Dollfuss-Schuschnigg Era in Austria: A Reassess- Video recording of an invited talk to the VIIth ment. New Brunswick, NJ: Transaction Publishers. International Congress of Quantum Chemistry. Sham, L.J. 1965. A calculation of the phonon frequencies Courtesy of Prof. Axel Becke. in sodium. Proceedings of the Royal Society Powell, C.F. and G.P.S. Occhialini. 1947. Nuclear of London A 283: 33-49. Physics in Photographs. Oxford: Clarendon Press. Sham, L.J. 2014. January 10 2014 correspondence with PT 1961. APS-AAPT Annual Joint Meeting in New the author. York. Physics Today 14(3): 52-56. Sham, L.J. and W. Kohn. 1966. One-particle properties Pugh, E.M. and N. Rostoker. 1953. Hall effect in of an inhomogeneous interacting electron gas. Physical ferromagnetic materials. Reviews of Modern Physics Review 145: 561-567. 25: 151-157. Sherwood, A.I. 2013. April 30 2013 telephone interview Raimes, S. 1963. The rigid-band model. pp. X-1-9 in: with the author. Arnold Sherwood was a PhD student J. Friedel and A. Guinier (eds.) Metallic Solid of Maria Goeppert Mayer at UCSD. Solutions. New York: Benjamin. Shockley, W. 1950. Electrons and Holes in Semi- Ren, M.L. 2013. October 20 2013 telephone interview conductors. Princeton: Van Nostrand Co. with Max Luming Ren. Silverman, R.A. 1951. The Fermi Energy of Metallic Robinson, G. de B. 1979. The Mathematics Department Lithium. PhD thesis. Harvard University. in the University of Toronto 1827-1978 Toronto: Silverman, R. and W. Kohn. 1950. On the cohesive Department of Mathematics, University of Toronto. energy of metallic lithium. Physical Review 80: Rostoker, N. 2003. I am happy that the R stands for 912-913. Rostoker. pp. 206-207 in: M. Scheffler and P. Siochi, C. 2013. March 20 2013 correspondence between Weinberger (eds.) Walter Kohn–Personal the author and Carlos Siochi, University of Toronto Stories and Anecdotes. Berlin: Springer. Alumni Relations Officer for the Faculty of Arts and Rostoker, N. 2013. July 23 2013 telephone interview with Science. 43

Slater, J.C. 1939. Introduction to Chemical Physics. New of Technology. Department of Special Collections, York: McGraw Hill. Davidson Library, University of California, Santa Slater, J.C. 1951. A simplification of the Hartree-Fock Barbara. method. Physical Review 81: 385-390. WKP 1953b. Walter Kohn Papers. Final program of the Slater, J.C. 1953. An augmented plane wave method for July 1953 Gordon Research Conference on the Physics the periodic potential problem. Physical Review 92: and Chemistry of Metals. Department of Special 603-608. Collections, Davidson Library, University of Slater, J.C. 1956. Band theory of bonding in metals. California, Santa Barbara. pp. 1-12 in: Theory of Alloy Phases. Cleveland, OH, WKP 1955. Walter Kohn Papers. April 20 1955 American Society for Metals. letter from Walter Kohn to Harry Jones. Slater, J.C. 1963. The electronic structure of atoms– Department of Special Collections, Davidson Library, the Hartree-Fock method and correlation. Reviews of University of California, Santa Barbara. Modern Physics 35: 484-487. WKP 1957. Walter Kohn Papers. October 7 1957 Slichter, C.P. 2010. Frederick Seitz. Biographical letter from Frederick Reif to Walter Kohn. Memoir Washington, D.C.: National Academy of Department of Special Collections, Davidson Sciences Press. Library, University of California, Santa Barbara. Sonnert, G. and G. Holton. 2006. What Happened WKP 1958. Walter Kohn Papers. January 29 1958 to the Children Who Fled Nazi Persecution. New letter from Carnegie Tech President J.C. Warner York: Palgrave Macmillan. to Walter Kohn. Department of Special Stevenson, A.F. and M.F. Crawford. 1938. A lower limi Collections, Davidson Library, University of for the theoertical energy of the normal state of helium. California, Santa Barbara. Physical Review 54: 375-379. WKP 1959a. Walter Kohn Papers. October 21 1959 Swaminathan, S. 2000. Obituary of Fritz Rothberger. letter from Conyers Herring to Walter Kohn. CMS Notes de la SMC 32: (5) 29. Department of Special Collections, Davidson Taylor, P.L. 1963. Theory of Kohn anomalies in the Library, University of California, Santa Barbara. phonon spectra of metals. Physical Review 131: WKP 1959b. Walter Kohn Papers. October 26 1959 1995-1999. letter from Walter Kohn to Keith Brueckner. Taylor, P.L. 2013. March 10, 2013 interview with Department of Special Collections, Davidson the author. Library, University of California, Santa Barbara. Titchmarsh, E.C. 1949. Godfrey Harold Hardy. Obituary WKP 1959c. Walter Kohn Papers. October 9 1959 Notices of Fellows of the Royal Society 6: 446-461. letter from Keith Brueckner to Ben Mottelson. Tong, B.Y. and L.J. Sham. 1966. Application of a self- Department of Special Collections, Davidson consistent scheme including exchange and correlation Library, University of California, Santa Barbara. effects in atoms. Physical Review 144: 1-4. WKP 1963a. Walter Kohn Papers. June 10 1963 UCSDA 1960. University of California San Diego letter from Lu Jeu Sham to Walter Kohn. Department Archives. University Communications News Releases. of Special Collections, Davidson Library, University of RSS 6020. September 15 1960. Mandeville Special California, Santa Barbara. Collections Library. WKP 1963b. Walter Kohn Papers. August 6 1963 Varley, J.H.O. 1954. The calculation of heats of forma- letter from Alan R. Liss (Vice President, Academic tion of binary alloys. Philosophical Magazine 45: Press, Inc.) to Walter Kohn. Department of Special 887-916. Collections, Davidson Library, University of Wannier, G.H. 1937. The structure of electronic California, Santa Barbara. excitation levels in insulating crystals. Physical WKP 1964. Walter Kohn Papers. June 15 1964 letter Review 52: 191-197. from Walter Kohn to Pierre Hohenberg. Department Weinstein, A. 1941. Les vibrations et le calcul des of Special Collections, Davidson Library, variations. Portugaliae Mathematica 2: 36-55. University of California, Santa Barbara. Weinstein, A. 1942. The spherical pendulum and Woll, Jr., E.J. and W. Kohn. 1962. Images of the Fermi complex integration. American Mathematical surface in phonon spectra of metals. Physical Review Monthly. 49: 521-523. 126: 1693-1697. Wilson, A.H. 1936 The Theory of Metals Cambridge: Young, A. 2013. May 2 2013 correspondence between the University Press. author and Alice Carroll Young. Wilson, E.B. 1962. Four-dimensional electron density Zabloudil, J., R. Hammerling, P. Weinberger, and L. function. Journal of Chemical Physics 36: 2232-2233. Szunyogh. 2005. Electron Scattering in Solid Matter. WKP 1953a. Walter Kohn Papers. January 12 1953 Berlin: Springer. letter to Walter Kohn from R.A. Deller, Bell Telephone Zangwill, A. 2013. Hartree and Thomas: the forefathers Laboratory; February 13 letter to Walter Kohn from of density functional theory. Archive for History of H. Tate, McGill University; February 24 letter to Exact Sciences 67: 331-348. Walter Kohn from Edward Creutz, Carnegie Institute Ziman, J. 1960. Electrons and Phonons. Oxford: 44

Clarendon Press. Cambridge: University Press. Ziman, J. 1964. Principles of the Theory of Solids.

[1] John Anthony Pople (1925-2004) earned his BA (1946) for himself, working on a farm in Yorkshire and wash- and PhD (1951) degrees in Mathematics from Cam- ing dishes in a hotel in Brighton, until the internment bridge University. His PhD thesis, “Lone pair orbitals” roundup of 1940 (Eisinger 2013). was supervised by John Lennard-Jones, head of the De- [10] The Sobieski was the last of four ships assigned the task partment of Theoretical Chemistry. A native English- to transport internees to Canada. The Ettrick and the man, Pople spent more than a decade teaching and Duchess of York crossed the Atlantic safely. The Aran- conducting research at Cambridge before moving to the dora Star was torpedoed by a German submarine on United States in 1964 to take a position as Professor July 2 and sunk with a loss of more than 800 lives. of Chemical Physics at the Carnegie Institute of Tech- Public outcry over this incident contributed to the ces- nology in Pittsburgh, PA. During his thirty-year career sation of the internments and the first releases of de- at Carnegie Tech, Pople made a gradual transition for tainees from the Isle of Man (Gilman & Gilman 1980). the development of semi-empirical methods in molecu- [11] Camp B was built on the site of an abandoned, lar orbital theory to the development of computer codes depression-era, unemployment relief facility. to solve the Schr¨odinger equation for small molecules at [12] Kurt Martin Guggenheimer (1902-1975) studied chem- the Hartree-Fock level and beyond. He moved to North- istry at the University of Munich and physics at the Uni- western University in 1993 and earned one-half of the versity of Berlin. He earned his PhD in 1933 for work 1998 Nobel Prize in Chemistry for his achievements in on the ultraviolet absorption spectra of zinc, potassium, computational quantum chemistry (Buckingham 2006). and cesium under the direction of Fritz Haber. Guggen- [2] Walther Kohn began to use the name Walter Kohn in heimer pursued post-doctoral work in Paris under the 1940. direction of Paul Langevin and published his specula- [3] Beller (1989) paints a cultural portrait of the Jewish tions about nuclei while there. He returned to Munich in community in early twentieth century Vienna. 1935 but was arrested following Kristallnacht and spent [4] This would be Kohn’s sixth (and last) year of secondary several months at the Dachau detention center. He em- school in Vienna. Quite unusually for Austria at that igrated to England and was working at King’s College time, boys and girls were not separated for instruction (Cambridge) when he was interned as a enemy alien. in Walther’s small class at the Chajes school (Neuhaus After the war, Guggenheimer worked as a Lecturer at 2003). the University of Bristol and the University of Glasgow [5] Emil Eliezer Nohel grew up on a small farm in Bohemia. before retiring from academic life in 1967 (R¨urup 2008, He studied mathematics at the Karl-Ferdinand (Ger- Fernandez 2013). man) University in Prague and served as an assistant [13] Fritz Rothberger (1902-2000) was a native of Vienna to Albert Einstein when Einstein was a professor there who graduated from the Akademisches Gymnasium and from 1911 to 1914. Nohel’s descriptions of the difficult earned BS and PhD degrees in mathematics from the conditions endured by Jews in Bohemia awakened Ein- University of Vienna. He came under the influence of stein’s concern for the plight of his co-religionists. For Waclaw Sierpinksi in Warsaw and began his lifelong most of his career, Nohel taught mathematics at the work in combinatorial set theory. Rothberger emigrated Handelsakademie Wien, a business-oriented secondary to England just before the start of World War II and school in Vienna. The Anschluss precipitated his dis- he was working as a scholar at Trinity College (Cam- missal and he found work as the physics teacher (and bridge) when he was interned. After the war, he served then the principal) of the Zwi Perez Chajes Gymnasium. as a professor on the mathematics faculties of several Nohel was arrested on December 12 1942 and spent two Canadian universities: Acadia, Fredericton, Laval, and years at the Theresienstadt labor camp before he was Windsor (Swaminathan 2000). transferred to the Auschwitz concentration camp and [14] The subjects tested were English literature, English murdered by the Nazis (Frank, 1947, Pais 1982). composition, general history, elementary algebra, ele- [6] The unemployment rate in Austria stood at 20 percent mentary geometry, physics, Latin authors, Latin com- at the time of the Anschluss (Senft 2003). position, German grammar, German composition, inter- [7] The term ‘fifth column’ refers to the secret supporters mediate algebra, intermediate geometry, and trigonom- of an enemy who live openly within the territory being etry. By his own account, Kohn performed better in defended. Latin than in German and his worst subject was Cana- [8] Vienna native Otto Frisch was working in the physics dian history (Kohn 2013a). laboratory of Prof. Mark Oliphant at the University [15] Godfrey Harold Hardy (1877-1947) was a leading British of Birmingham when the roundup of aliens began. He mathematician of his time. He was a child prodigy who avoided internment only because his employers made trained at Trinity College (Cambridge) and served as a the case that he was engaged in important war work professor at both Cambridge and Oxford over the course (Frisch 1979). of his career. He authored or co-authored eleven books [9] Eisinger’s family smuggled him onto a Kindertransport and over 300 research papers, mostly in the fields of train out of Vienna after failing to find an official spon- analysis and number theory. He was particularly well- sor or a foster family for him in England. He fended known for his collaborations with Srinivasa Ramanujan 45

and John Edensor Littlewood, and for his disdain for in his small private laboratory. Mendel read the political any kind of “applied” work (Titchmarsh 1949). situation correctly and he took his wife and three chil- [16] John Clarke Slater (1900-1976) was the chair of the dren to Holland when Hitler came to power in 1933. In physics department at the Massachusetts Institute of 1937, he emigrated to Canada and became (at first) an Technology (MIT) from 1930-1951. He wrote an ex- unpaid faculty member at the Banting Institute for Can- perimental PhD thesis at Harvard, but then traveled cer Research of the University of Toronto. He returned to Europe where he made several important theoreti- to Europe in 1950 to accept a chair in Pharmacology at cal contributions to the early development of quantum the University of Amsterdam (Feldberg 1960). mechanics. Dissatisfaction with ‘formal theory’ led him [23] Eisinger followed Kohn to the University of Toronto the to develop a large research group at MIT devoted to following year (Eisinger 2013). solving the Schr¨odinger equation numerically to calcu- [24] Leopold Infeld (1898-1968) earned the first PhD in the- late the physical properties of atoms, molecules, and oretical physics awarded by a Polish university from solids. He authored or co-authored thirteen books and the Jagellonian University in his native city of Kr´akow. over 150 research papers. An important paper he pub- He taught physics at Jewish high schools for nearly a lished in 1951 turned out to bear directly on Kohn’s decade before finding a senior assistantship in theoreti- density functional theory (Morse 1982). cal physics at Lwow University. Infeld knew the impor- [17] A Course in Pure Mathematics is an elegant and rig- tance of contacts with foreign physicists and successfully orous introduction to mathematical analysis for serious gained two-year visiting positions with Max Born in first-year college students. The subject matter includes Cambridge and Albert Einstein in Princeton. With the the notions of limit and convergence applied to series, latter, he co-authored The Evolution of Physics, a pop- sequences, functions, derivatives, and integrals. All the ular account of the history of ideas in physics. In 1938, main theorems of the calculus of a real variable are dis- Infeld accepted a position at the University of Toronto cussed, as is the general theory of logarithmic, exponen- where he worked on a variety of theoretical problems in tial and sinusoidal functions. general relativity and cosmology. He returned to Poland [18] The undergraduate course Slater taught at MIT using in 1950 to found an Institute of Theoretical Physics at Introduction to Chemical Physics proved to be too dif- the University of Warsaw (Infeld 1980). ficult for most of its intended audience. In later years, [25] Samuel Beatty (1881-1970) earned the first PhD in a separate “modern physics” course devoted to atoms, mathematics awarded by a Canadian university at the molecules, and the structure of matter became a prereq- University of Toronto. He remained at Toronto and from uisite for a senior-level “thermal physics” course which 1911-1959 rose from Lecturer in Mathematics to Profes- retained Slater’s idea to combine thermodynamics and sor and chair of Mathematics, to Dean of the College of statistical mechanics in a single presentation. A text- Arts & Science, and finally to Chancellor of the entire book written specifically for the latter by two of Slater’s University. He published 30 research papers, mostly in MIT colleagues expanded his treatment of thermody- the field of algebraic functions, and devoted the bulk of namics and contracted his treatment of statistical me- his energy to teaching and to building the Mathematics chanics (Allis and Herlin 1952). department (Robinson 1979). [19] The future Nobel laureate Max Perutz (1914-2002) [26] Frank B. Kenrick (1874-1951) trained as a physical spent only six months as an internee in Canada. His chemist with Wilhelm Ostwald in Leipzig. He served memoir Enemy Alien recounts how he was arrested by as chair of the Chemistry Department at the University the British just four months after earning his PhD under of Toronto from 1937-1944. Like his mentor and prede- the direction of Sir Lawrence Bragg. He was deported to cessor as department chair, William Lash Miller (1866- the Cove Fields internment camp in Quebec City (Camp 1949), Kenrick favored an approach to chemistry that L) where he organized a camp school with a faculty that denied the reality of individual atoms and molecules. included the physicist Klaus Fuchs and the future astro- The war-related work carried out in Kenrick’s depart- physicists Hermann Bondi and Thomas Gold. Fuchs was ment included experiments to develop the new explosive recruited to the atomic bomb project after his release RDX as a replacement for TNT and the development of and gained notoriety in 1950 when it was discovered detectors for poison gas (Brook and McBryde 2007, Av- that he had betrayed the secrets of the Los Alamos lab- ery 1998). oratory to the Soviet Union (Perutz 1985). [27] Dean Beatty apparently “regularized” Kohn’s ‘special [20] The Canadian government announced in May 1941 that student’ status at some point because his official tran- any internee under the age of 21 cleared by Scotland script inaccurately states that he completed an intro- Yard would be released and given the opportunity to ductory chemistry course and a chemistry laboratory continue his education in Canada if he could find a spon- course (Siochi 2013). sor willing to pay a fee of two thousand dollars (Jones [28] Kohn and Josef Eisinger had volunteered for (and been 1988). rebuffed by) the Canadian Air Force immediately after [21] Eisinger was released in early January. Kohn was re- their release from internment (Eisinger 2013). leased a few weeks later after a short stay at Camp [29] Kohn worked the summers of 1942-1943 for the Sutton- N, an internment camp outside the town of Sherbrooke, Horsley Company, a Toronto x-ray equipment manu- Quebec, about 130 km east of Montreal (Eisinger 2011). facturer which began producing signalling lamps and [22] Bruno Mendel (1897-1959) was the son of a research- instrument panels for fighter and bomber aircraft af- active medical doctor who trained in Berlin and became ter the war began. His specific task was the design and a research-active physician himself. His medical practice testing of compensation circuits to ensure that cock- slowly became less important as he increased the time pit instruments gave accurate readings when operated he spent researching the metabolism of the cancer cell at unusually high and low temperatures (Kohn 1998, 46

Horsley 2013, Kohn 2013b). Kohn worked the summers of chemical physics, quantum electronics, solid state of 1945-1946 for the mineral surveying and exploration physics, and magnetism. He published 169 scientific ar- geophysics company Koulomzine, Geoffroy, Brossard & ticles and two books and served on the faculty at Har- Company of Val D’Or, Quebec. His job was to conduct vard for forty-six years. In 1977, he was awarded a one- magnetic field surveys in suspected gold-bearing regions third share of the Nobel Prize in physics for his work in of northern Ontario. A typical survey consisted of a grid magnetism (Anderson 1987). of about 1000 magnetometer measurements with a grid- [37] Kohn had learned from his fellow graduate students spacing of 100 meters (Geoffroy 1946, Kohn 1998, Kohn about Wolfgang Pauli’s famously negative view that 2013b). solid-state physics was insufficiently fundamental and [30] See footnote 2. too approximate to attract the attention of a serious [31] John Lighton Synge (1897-1995) is often regarded as young theoretical physicist (von Meyenn 1989). the greatest mathematician of Irish descent since Sir [38] It is necessary to replace Ef by the complex number William Rowan Hamilton. Synge studied mathematics Ef + iǫ in Eq. (3) and let ǫ → 0 at the end to ensure at Trinity College Dublin and accepted a position as As- that ψ behaves like an outgoing spherical wave (Baym sistant Professor at the University of Toronto in 1920. 1969). There he began a lifelong interest in Einstein’s theory [39] This difficult and subtle problem was solved in the early of relativity and in geometrical methods to analyze dy- 1960’s by the Russian mathematical physicist Ludvig namical systems. The peripatetic Synge subsequently Faddeev (Faddeev 1965). held positions in Dublin, Toronto (again at the time [40] At some point before he wrote up his thesis, Kohn Kohn was there), Ohio State, Carnegie Tech, and finally learned that the Swedish physicist Lamek Hulth´en had the Dublin Institute for Advanced Studies. He published independently derived a variational principle for the 11 books and over 200 hundred papers (Florides 2008). scattering phase shift very similar to his own. Hulth´en [32] Alexander Weinstein (1897-1979) was a PhD student (1946) and Kohn (1948) begin with the same variational of Hermann Weyl, who considered him to be his most functional but propose slightly different variational pro- talented student. He worked with Tullio Levi-Civit`ain cedures. The Kohn-Hulth´en variational principle later Rome and Jacques Hadamard in Paris before the Ger- found wide application in atomic, molecular, and nu- man occupation of France in 1940 drove him from Eu- clear scattering problems (Adhikari 1998, Nesbet 2003). rope permanently. Weinstein was a member of the Ap- [41] This did not prevent Schwinger from later characterizing plied Mathematics faculty of the University of Toronto Kohn as “my most illustrious student” to one of Kohn’s from 1941-1946, worked for some time at the US Naval former graduate students (Rudnick 2003). Ordnance Laboratory, and spent 18 productive years at [42] Josef Eisinger remained a close friend. He did his gradu- the Institute for Fluid Dynamics and Applied Mathe- ate work at MIT, just two miles down the Charles river matics at the University of Maryland (Diaz 1978). from Harvard, and earned his PhD in physics in 1951 [33] Arthur Francis Chesterfield Stevenson (1899-1968) ac- for an experimental determination of the magnetic mo- cepted a position in the Mathematics department of the ment of K40. He spent thirty years at Bell Laborato- University of Toronto immediately after earning his BA ries where he made a transition from solid state physics from Trinity College, Cambridge in 1922. He returned to biophysics. From 1985 until his retirement in 1998, to Cambridge in 1928 where he worked under the su- he taught and conducted research at the Mount Sinai pervision of Ralph Fowler on a problem in theoretical School of Medicine in New York City (Eisinger 2013). spectroscopy which eventually led to his PhD. He re- [43] An evaluation of Kohn written by Ms. Norine T. Casey turned to Toronto where he published original research provides insight into Kohn’s teaching style and philos- and lectured on atomic physics, quantum mechanics, ophy of physics. Ms. Casey was a 1949 Wellesley grad- electromagnetic theory, scattering theory, and the dif- uate pursuing an MA in teaching at Harvard. As part ferential equations of mathematical physics. He spent of her curriculum, she attended Kohn’s summer 1950 the last dozen years of his academic career on the faculty class (devoted to introductory optics, electricity, mag- of Wayne State University in Detroit, Michigan (Duff, netism, atomic physics, and nuclear physics) and wrote 1969). a four-page evaluation of her experience. Kohn received [34] It is necessary here that trial function satisfies a copy of her report (Casey 1950), which states that “Dr. ∗ ∗ R dτψ ψ = 1 and that the integral R dτψ Hψ con- Kohn’s lectures were clear and concise. Demonstrations verges. accompanied every lecture and were given with great [35] Julian Seymour Schwinger (1918-1994) was one of the enthusiasm. . . . It was obvious from the beginning that greatest theoretical physicists of the 20th century. By [Dr. Kohn’s] interest was not his own mastery of the the age of 21, he had earned his PhD under the mathematics, but in our understanding of the physics. . (nominal) supervision of Isador Rabi and published . . It was not infrequent that he read from source mate- ten research papers in quantum mechanics and nuclear rial giving direct quotes [such] as Newton’s relating his physics. He spent the war years working on waveguides discovery of the diffraction of light.” for radar applications before turning his attention to [44] This technique had recently been introduced to study quantum electrodynamics. This work earned him a one- cosmic rays using plates produced by another company third share of the 1965 Nobel Prize in physics. He su- and Polaroid wanted to enter the business (Powell and pervised 73 PhD students over a long academic career Occhialini 1947). at Harvard (1945-1972) and UCLA (1972-1994) (Martin [45] Frederick Seitz (1911-2008) was one of the creators and & Glashow 1995, Mehra & Milton 2000). intellectual leaders of the American solid state physics [36] John Hasbrouk Van Vleck (1899-1980) was an eminent community. He was Eugene Wigner’s first PhD student theoretical physicist who contributed widely to the fields at Princeton and built influential research groups at four 47

different universities between 1935 and 1965. Later, he [53] For fifteen years after World War II ended, most solid served as president of Rockefeller University and presi- state physics research in the United States was funded dent of the National Academy of Sciences. Seitz worked by the Office of Naval Research. Most nuclear physics on a wide range of materials problems during World research was supported by the United States Atomic War II and was the author or co-author of more than Energy Commission (Old 1961, Sapolsky 1990). 100 scientific papers (Slichter 2010). [54] The actual computing was performed by a ‘compu- [46] Silverman’s PhD thesis states that “The author wishes tress’ named Alice Watson who operated a Friden Model to express his indebtedness to Professor Walter Kohn STW-1 Electro-Mechanical Calculator. Although her for suggesting the problem and for invaluable guidance equipment changed to an IBM 650 digital computer in during a large portion of the work (Silverman 1951). 1956, she continued to do computing tasks for Kohn the A generation of English-speaking solid-state physicists entire time he worked at Carnegie Tech (Rostoker 2003, have Silverman to thank for his translation into En- Young 2013). glish of Methods of Quantum Field Theory in Statisti- [55] Korringa (1994) relates that “computers were rare in the cal Physics by A.A. Abrikosov, L.P. Gorkov, and I.E. Netherlands in 1946 and a cost estimate [for a numerical Dzyaloshinski (1963). application] exceeded the annual research budget of our [47] Edward Chester Creutz (1913-2009) earned his PhD theory group.” in experimental nuclear physics from the University of [56] Kohn left the development and application of the Wisconsin in 1938. He moved to Princeton as an Instruc- Korringa-Kohn-Rostoker (KKR) method to others. It tor and used their cyclotron for nuclear physics projects eventually became a standard method of band structure until the Manhattan Project redirected his efforts to the calculation (Zabloudil et al. 2005). synthesis of plutonium and the triggering of the pluto- [57] Hoddeson et al. (1992) is a history of solid state physics nium bomb. After the war, Creutz joined the faculty up to about 1960. The autobiography of Frederick Seitz at Carnegie Tech to direct the construction of a proton (1994) provides a broad view from the perspective of a synchrocyclotron and to build an experimental nuclear major player in the development of solid state physics physics group. After nine years (seven as chair), Creutz as the discipline matured through the 1950’s and 1960’s. moved to San Diego, California to help found the Gen- [58] Jacob E. Goldman (1921-2011) was born in Brook- eral Atomics division of General Dynamics Corporation. lyn, New York and studied physics at Yeshiva Univer- He spend 15 years there before concluding his career as sity and the University of Pennsylvania. His expertise Associate Director of the National Science Foundation in magnetism led him to the Westinghouse Research (Hinman and Rose 2010). Laboratory in 1945 before he joined the faculty of the [48] One former Carnegie Tech graduate student remem- Carnegie Institute of Technology in 1951. He moved to bers Kohn’s solid state physics lectures as well-prepared, the nascent Scientific Research Laboratory of the Ford clearly delivered, and mathematically precise. He was Motor Company in 1955 and eventually became head surprised when the final exam avoided mathematical is- of all Ford’s corporate research and development. He sues and focused entirely on qualitative aspects of the joined the Xerox corporation in 1969 and one year later subject (Arrott 2013). founded their Palo Alto Research Center (PARC). The [49] Aage Bohr and Christian Møller were already fixtures first modern personal computer and the first graphical at the Institute. Kohn’s Harvard classmate Ben Mottel- user interface were invented at PARC a few years later son arrived in the fall of 1950 and never left. A partial (Markoff 2011). list of visitors who overlapped with Kohn for at least [59] Kohn never worked personally in the field of conven- some time includes Hendrik Casimir, Freeman Dyson, tional thermal phase transitions. Nevertheless, he often Ugo Fano, Nicolas Kemmer, Louis Michel, Wladyslaw supported a post-doctoral fellow trained in statistical Swiatecki, Jean Valatin, Nicolaas van Kampen, and mechanics to work on this class of problems. See Domb Arthur Wightman (NBA 1951a). (1996) for a history of this subject. [50] A partial list of attendees includes Hans Bethe, Homi [60] Joaquin Mazdak Luttinger (1923-1997) earned his BS Bhabha, L´eon Brillouin, Richard Dalitz, Paul Dirac, in physics at the Massachusetts Institute of Technology Maria Mayer, Dirk ter Haar, Werner Heisenberg, Walter and remained there to complete a PhD thesis (1947) Heitler, L´eon van Hove, Lamek Hulth´en, Egil Hylleraas, in statistical physics under the supervision of Lazslo Hendrik Kramers, Ralph Kronig, Jens Lindhard, Lise Tisza. He worked on quantum electrodynamics as the Meitner, Wolfgang Pauli, Rudolf Peierls, L´eon Rosen- first American post-doctoral fellow of Wolfgang Pauli feld, John Slater, Ernst Stueckelberg, Victor Weisskopf, but reverted to problems in many-body theory, solid Gregor Wentzel, John Wheeler, and Gian-Carlo Wick state physics and statistical mechanics for the rest of (NBA 1951b). his career as a professor, primarily at Columbia Univer- [51] Kohn was an emergency replacement for Mario Verde, a sity (1960-1993) (Anderson et al. (1997). nuclear physicist who had fallen ill (DeWitt 1951, 2013). [61] Sixty years later, the Bell Laboratories experimenter [52] Res Jost (1918-1990) wrote his PhD thesis under the su- Robert Fletcher recalled that “Walter was a very kind pervision of Gregor Wentzel and spent three years as the and thoughtful person to work with. I never had the im- principal assistant to Wolfgang Pauli. He was a senior pression he looked down on us experimenters as some fellow at the Institute for Advanced Study in Prince- theorists are inclined to do” (Fletcher 2013). Another ton for six years (1949-1955) before accepting a profes- Bell Laboratories experimenter who published papers sorship at the Eidgen¨ossische Technische Hochschule in related to the Kohn-Luttinger theory was Walter’s old Z¨urich. Jost’s research focused on mathematical physics friend Josef Eisinger (Eisinger and Feher 1958). and quantum field theory, particularly axiomatic ver- [62] Another graduate student, James Montague, never sions of the latter. (Kohn et al. 1992, Pais 1996). quite finished a thesis devoted to deep (strongly bound) 48

impurity levels in semiconductors (Glasser 2013). ground state many-body wave function) of various quan- [63] Kjeldaas pursued his Ph.D. part-time at the University tum mechanical operators. The Green function for the of Pittsburgh. His 1959 thesis thanks Kohn and West- Schr¨odinger partial differential equation, Eq. (2), is re- inghouse solid state theorist Theodore Holstein for act- lated to some of the Green functions used in many-body ing as co-supervisors (Kjeldaas 1959). Westinghouse was theory when all the forces between the particles are a lively place for solid-state physics in the mid-1950’s turned off. under the leadership of its Director of Science, Clarence [71] Samuel F. Edwards, another Schwinger PhD student Zener, himself a solid-state theorist. The theorists he who switched from nuclear physics to solid state physics, recruited to complement Holstein included Edward has remarked that ‘the Green function formalism is very Neufville Adams, Petros Argyres, William Mullins, and good to write down solutions in abstract exact form, Yako Yafet. The experimenters hired by Zener included which gives unassailable answers when used in compar- Raymond Bowers, Robert Keyes, Colman Goldberg, atively simple situations’ (Edwards 1998). and John Rayne. Walter Kohn had a consulting ar- [72] Kohn became an Associate Editor of the Journal of rangement with Westinghouse. Holstein, Adams, and Mathematical Physics in 1961. Yafet occasionally taught classes at Carnegie Tech (Am- [73] A phonon is a quantized lattice vibration in a crys- begaokar 2013, Arrott 2013). tal. The Fermi surface is the constant energy surface [64] By “screening” or “shielding”, we mean that the elec- in the Bloch k-space for the most energetic electrons in trons nearest to the positive charge are attracted to it a metal. and thereby partially neutralize the Coulomb force ex- [74] In this way, Langer became the Peierls PhD student erted by the positive charge on distant electrons and Kohn would have been if he had not gone to Harvard to ions. work with Julian Schwinger (Kohn 2013). [65] Visitors to the theory group that summer included [75] Langer and Vosko used a formulation of many-body per- Elihu Abrahams, Kerson Huang, David Pines, J. Robert turbation theory due to John Hubbard (1957). Schrieffer, and Philippe Nozi`eres (Anderson 1978). [76] The regents of the University of California committed [66] The Hall effect refers to a voltage that appears across a unprecedented financial resources so UCSD could re- current-carrying sample when a magnetic field is applied cruit senior scientists like Kohn to its nascent faculty. in a direction perpendicular to the direction of current In its first few years, 50 percent of new faculty hires at flow (Chien and Westgate 1980). UCSD were made at the full professor level or above, as [67] Kohn and Luttinger did not ultimately address the opposed to 15 percent for the University of California problem of the effect of a large concentration of impu- system as a whole (Kerr 2001). rities on the electrical conductivity of a semiconductor. [77] The Symposium was organized by Conyers Herring from This was done by Phil Anderson (1958). Simultaneous Bell Telephone Laboratories. Besides Kohn, the lectur- with Kohn and Luttinger’s work on quantum transport, ers were John Fisher from the General Electric Research the Japanese physicist Ryogo Kubo proposed a theo- Laboratory, Jack Goldman from (by then) the Ford Mo- retical approach to the same problem which ultimately tor Company Research Laboratory, and Frank Herman became standard (Kubo 1957). from RCA Laboratories (WKP 1959a). [68] There is no definitive history of this revolution. Differ- [78] Suhl and Feher arrived in San Diego in 1960. Matthias ent points of view can be found in Pines (1961), Hodde- waited a year because “during the first year there will son et al. (1992), Gell-Mann (1996), Brueckner (2000), be too many administrative chores” (Feher 2002). and Kaiser (2005). Percus (1963) is the proceedings of [79] Offers were also proffered to (and declined by) Kohn’s a January 28-29 1957 meeting convened at the Stevens old Harvard friend Ben Mottelson, a nuclear physicist at Institute of Technology “for the purpose of bringing to- Bohr’s Institute for Theoretical Physics, and the French gether workers in the numerous rapidly moving fields of magnetic resonance expert, Anatole Abragam (WKP many-particle physics.” The contributions to this vol- 1959c, Abragam 1989). ume (particularly the roundtable discussions) paint a [80] An APS Fellow is judged by his peers to have made vivid picture of the first months of the revolution. The “exceptional contributions to the physics enterprise”. Bardeen-Cooper-Schrieffer theory of superconductivity In 1960, the total number of Fellows was 1653 out of a appeared later the same year, but Bardeen et al. (1957) total Society membership of 16,157 (APS 2013b). makes no explicit use of field theoretic methods. [81] A visiting French scientist, Jacques des Cloizeaux, ar- [69] Keith Allen Brueckner (1924 - ) studied mathemat- rived in the fall of 1960 and worked on a statistical me- ics at the University of Minnesota before earning his chanics problem. Ph.D in physics (1950) from the University of Califor- [82] A “sharp” Fermi surface has the property that a quan- nia (Berkeley) under the supervision of Robert Serber. tum state labeled by the quantum number wave vector As a professor at Indiana University and the University k is occupied by an electron if that wave vector lies in- of Pennsylvania in the 1950’s, Brueckner made many side the volume of the k-space enclosed by the Fermi significant contributions to the theories of nuclear mat- surface and unoccupied if k lies outside that volume. ter and the electron gas. In 1959, he became the first [83] The unusual hours were chosen for “the convenience of member of the physics department at the newly created students employed in industry” (UCSDA 1960). University of California, San Diego. Brueckner divided [84] Kohn taught semiconductor physics and the transport his professional activities between academia, industry, and optical properties of metals for five weeks in the and the government until his retirement from UCSD in fall of 1962 (Bruch 2013). His April 1962 campus-wide 1991. lecture was part of a series delivered by senior UCSD [70] The Green functions used in many-body theory are professors from various departments. the ground state expectation values (averages over the [85] Luming (later Luming Ren) switched his PhD supervi- 49

sor from Kohn to Assistant Professor David Wong be- published in the October 1962 issue of Journal de cause he felt that particle physics was “more fundamen- Physique et le Radium. Reading them may have been tal” than solid state physics. He came to rue his decision the trigger for Kohn to change the subject of his Paris when he was unable to find permanent employment as research from the electron-phonon interaction to alloy a particle physicist and made his career as a systems physics. engineer for the Hughes Aircraft Company (Ren 2013). [95] In later recollections, Kohn emphasizes the concept of [86] Philip Warren Anderson (1923 - ) is generally regarded charge transfer in alloys. His Nobel Prize autobiogra- as one of the preeminent theoretical physicists of the phy reports that he read some “metallurgical literature second half of the twentieth century. He attended Har- in which the concept of an effective charge e∗ of an vard University as an undergraduate, worked at the U.S. atom in an alloy was prominent, which characterized Naval Research Laboratory during World War II and re- in a rough way the transfer of charge between atomic turned to Harvard to earn his PhD in 1949 under the su- cells” (Kohn, 1998). His Nobel Prize lecture notes sim- pervision of John van Vleck. Anderson spent a long and ilarly that “there is a transfer of charge between . . . productive career at Bell Telephone Laboratories before unit cells on account of their chemical differences. The moving to Princeton University in 1984. He shared the electrostatic interaction energies of these charges is an 1977 Nobel Prize in Physics with Nevill Mott and John important part of the total energy. Thus in consider- Van Vleck for ‘fundamental theoretical investigations of ing the energetics of this system there was a natural the electronic structure of magnetic and disordered sys- emphasis [in the literature] on the electron density dis- tems’ (Anderson 1977). tribution n(r)” (Kohn 1999). Despite these remarks, I [87] Philippe Nozi`eres (1932- ) graduated from the Ecole have been unable to find any significant discussion of Normale Sup´erieure (ENS) in 1955 and earned his PhD charge transfer or “effective charge” in the pre-1963 lit- two years later from the University of Paris, albeit under erature of metallurgy or metal physics. Indeed, if they the supervision of David Pines at Princeton University. mention this type of charge transfer at all, review ar- He spent a decade at the ENS before moving to the In- ticles of the period consistently refer to the same two stitut Laue-Langevin (ILL) in Grenoble. He also lectures papers—one by Nevill Mott (1937) which concerns or- at the Coll`ege de France in his capacity (since 1983) as dered alloys and one by John Henry Oliver Varley (1954) Professor of Statistical Physics (Nozi`eres 2012a). which demonstrates that the electrostatic energy asso- [88] In July, Kohn had written to an editor at Academic ciated with charge transfer is negligible in disordered Press confirming his interest to contribute to a book alloys. On the other hand, less than ten years after the about “Impurities in Metals” and indicating that he events narrated here, charge transfer became an impor- would be “working in this field” during his stay in Paris tant variable in two proposed theories of binary alloy (WKP 1963b). formation (Hodges and Stott 1972, Miedema, de Boer, [89] Jacques Friedel (1921 - ) is a fourth-generation French and de Chatel 1973). I have found no literature of the scientist who was educated at the Ecole´ Polytechnique time that emphasizes the spatially-varying electron den- (1944-46) and the Ecole´ Nationale Sup´erieure des Mines sity distribution n(r) in alloys except in the most qual- (1946-48) before earning his PhD in 1952 under the su- itative terms (see the passage by Hume-Rothery and pervision of Nevill Mott at the University of Bristol. Raynor quoted earlier in this section). Friedel began his academic career at the Sorbonne, but [96] The coefficient of the kinetic energy term in Eq. (13) moved in 1959 to the Orsay campus of the University includes the electron mass m and Planck’s constant h. of Paris, now the University of Paris-Sud. Friedel’s life- [97] Later analysis showed that removing the assumption of long interests in metallurgy and the physics of metals no degeneracy does not invalidate the final result (Parr resulted in over 200 theoretical publications, most of and Yang 1989). them characterized by the use of simple models and ele- [98] Another necessary quantity, the first-order density ma- mentary mathematics (Br´echet 2008). Kohn and Friedel trix, is derivable from the second-order density matrix. met at a July 1953 Gordon Conference in Laconia, New [99] The term ‘v-representability problem’ was coined in Hampshire devoted to the Chemistry and Physics of 1975. See Parr and Yang (1989) and Kryachko and Metals (Kohn 2012a). Lude˜na (1990) for extensive discussion of this issue. [90] Pierre-Gilles de Gennes (1932-2007) changed fields after [100] It is curious that Hohenberg and Kohn do not refer to the publication of his superconductivity book and won papers by Hubbard (1958), Bellemans and de Leener the 1991 Nobel Prize in Physics for his work on the (1961), and Jones, March, and Sampanthar (1962), all statistical physics of liquid crystals and polymers. of whom studied the energy of an electron gas in the [91] Alloys of this kind are called disordered. The Bloch the- presence of a lattice of positive charges using methods orem remains valid for ordered alloys where the A-type superior to the Thomas-Fermi approximation. atoms form a spatially periodic structure of their own. [101] The Web of Science database (accessed January 2014) [92] This follows from the eigenstate occupation rules of lists 230 citations to the quantum density oscillations quantum mechanics which dictate that states are pop- paper, Kohn and Sham (1965a). A review of these cita- ulated by electrons in order of increasing energy begin- tions shows that the vast majority of the citing papers ning with the lowest. (> 90%) seem unaware of its actual content. They in- [93] This is the same John Slater whose book Introduction correctly cite it along with Hohenberg and Kohn (1964) to Chemical Physics was purchased and read by Wal- and Kohn and Sham (1965b) as one of the foundational ter Kohn while he awaited release from internment in papers of density functional theory. wartime Canada. See Section II. [102] In the actual equations written by Hartree (1923), each [94] The unedited proceedings of this conference were also electron feels the classical electrostatic potential pro- duced by every electron except itself. Therefore, in con- 50

trast to Eq. (25), each electron in Hartree’s theory feels [109] Data collected from Google Scholar in January 2014. a slightly different electrostatic potential. The true number of papers is higher because not all [103] The correctness of Eq. (27) requires that the functions authors used the term ‘many-electron’ in their writing. φi and φj be orthonormal, which means that the integral [110] According to Musher (1966), Wilson’s method goes back ∗ R φi (r) φj (r) dr is one when i = j and zero when i =6 j. to Pauli. [104] The local density approximation for G[n] retains only [111] Early work on this problem published by Korringa the first term on the right hand side of Eq. (23). (1958) and Beeby (1964) blossomed into a full-scale the- [105] The Web of Science database (accessed January 2014) ory of the electronic structure of disordered alloys in the lists 21,372 citations to Kohn and Sham (1965b). Only late 1960’s and early 1970’s (Ehrenreich and Schwartz four physics-related papers have more citations and all 1976). four of them are density functional papers which owe [112] It is impossible to know how the frightful loss of his their existence to the Kohn-Sham paper. home and parents to Nazi terror motivated Kohn to [106] Kohn and Sham were unaware that the Hungarian succeed in later life. A statistical study of Viennese chil- physicist Rezs¨oG´asp´ar had derived Eq. (34) ten years dren who had similar experiences during World War earlier by similarly computing the variational derivative II and then emigrated to America shows that those of Dirac’s exchange energy with a local density approx- who entered the sciences achieved success (by conven- imation (G´asp´ar 1954). tional measures) more than twice as often as native- [107] Despite the warning in Sham and Kohn (1966) that the born American scientists of the same generation. The ǫk parameters should not be interpreted as one-electron study quotes several participants who said “they felt a energies, “the temptation to use [them] as band struc- great responsibility to make the most of their lives be- tures in solids proved irresistible” (Hohenberg, Kohn cause their survival was such a rare and unlikely event.” and Sham 1990). (Sonnert and Holton 2006) [108] The generalized gradient approximations go far beyond the simple gradient expansions analyzed by HK and KS.