N A S A TN D-4857

Q! 1










by Howard C. Volkin
Lewis Resedrch Center Cleuehnd, Ohio

00 . Ohio 1 / .Volkip .TECH LIBRARY KAFB.. NM I111 111 1Ili1l1l ll1l 11II 1 lt l b ITERATED COMMUTATORS AND FUNCTIONS O F OPERATORS By Howard C.' / L NATIONAL AERONAUTICS AND SPACE ADMpdWF!%TlON-For s a l e by the Clearinghouse for Federal Scientific and T e c h n i c a l Information Springfield. Virginia 22151 .-Lewis R e s e a r c h Center Cleveland.CFSTI price $3.

By working only with the G ( j ) and their commutators. All iterated commutators of degree ( j + 1) in A and B a r e com­ bined in a suitable sum to form a new operator G(j). ii . the calculations a r e simplified and the r e s u l t s a r e given in an ex­ tremely compact notation.. .ABSTRACT A new method i s developed by which certain functions of the noncommuting operators A and B can be expressed in a systematic and efficient way in t e r m s of successively higher commutators of the operators.. exponential functions. The functions a r e converted into equivalent f o r m s involving the iterated commutators. and analytic functions of the operator (A + B). The procedure i s designed primarily for poly­ nomials. Certain commutation properties of the o p e r a t o r s A and B a r e represented by polynomials in A and B called iterated commutators. The method is related to the ordered-operator calculus of Feynman.

Then it is understood that the form of As does not depend on s. The procedure is designed primarily for polynomials. This method is related to the ordered-operator calculus of Feynman (ref. AsBs. The purpose of this report is to present a new method by which explicit f o r m s of the desired type can be developed in a systematic and efficient way for certain functions of interest. In Feyman's operator calculus. exponential functions. which does not otherwise depend on the ordering parameter s . 1). Compact expressions result because all iterated commutators of the same degree always appear i n the same linear combination. INTRODUCTION Functions of noncommuting operators have been studied extensively over the y e a r s . and s o forth. equals AB when s > s ' . has been given recently by Wilcox (ref. and analytic functions of the operator (A + B). A survey of results.ITERATED COMMUTATORS AND FUNCTIONS OF OPERATORS by Howard C. One approach in working with such a function is to express it as an expansion in successively higher commutators of the operators involved. and is undefined when s = s ' . An operator A. may be written As. but its o r d e r of operation does and is defined by the convention: The operator with smallest index operates first. A method is developed f o r system­ atically expressing functions of A and B i n t e r m s of the iterated commutators. The method is related to the ordered-operator calculus of Feynman. 2). a n ordering parameter indicates the sequence in which non­ commuting quantities appearing an any product a r e to be taken. In the more general case of an operator that is a function of the time. the convention is that t i m e is the . usually within the context of problems arising in mathematics and quantum physics. Thus. along with numerous references. Volkin Lewis Research Center SUMMARY Certain commutation properties of two operators A and B a r e represented by poly­ nomials in A and B called iterated commutators. Applications to the important case of exponential functions a r e given. equals BA when s' > s. with the ordering index attached as a subscript.

The method presented h e r e deals with this second aspect. exp(A1 + Bo) = exp(A) exp(B) (1) An example of a full functional is the well-known Dyson formula for the NeumannLiouville expansion of the unitary time development transformation. were ordinary commuting functions. ttDisentangling" is the name given by Feynman to the process of finding an operator expression for t h e functional in the usual positional notation. B(s). . the calculations are simplified and the results can be given in an extremely compact notation. complete details a r e given only where deemed necessary and a suitable method is indicated in all other cases. exponential functions. A s the derivations required a r e generally quite straight­ forward. the algebra and calculus of ordinary numbers may be applied. no general procedure for disentangling func­ tionals is available. Certain commutation properties of two operators A and B are represented by polynomials in A and B called iterated commutators. With the ordering parameter convention. any functional of the operator functions has a unique interpretation. In manipulating such a func­ . the iterated commutators are defined and their relevant prop­ N e r t i e s a r e given. tional. because the o r d e r of operations in any situ­ A(s). relations a r e developed that convert a suitable function of two operators into the desired form. Simple examples of disentangling a r e the following: . In the next section.. Moreover. The Feynman formalism permits all the r e s u l t s of ordinary analysis to be used in handling operator functionals. . ation is automatically specified by the parameter values. t h e r e does not appear to b e any direct way of exploiting any simple commutation properties that the operators may happen to possess. B(s).By working only with the G (j) and their commutators. where the o r d e r of operation is represented by the position in which the operators are written. We can use the notation A(s) to include either case. Applications to the important case of exponential operators a r e given in the section EXPONENTIAL FUNCTIONS.ordering parameter. The transformation is an exponential functional of a time-dependent Hamiltonian operator function. as though A(s). The procedure given converts polynomials. However. All iterated commutators of degree (j + 1)i n A and B a r e combined in a suitable sum to form a new operator G di 1. 2 . and analytic functions of (A + B) into equivalent f o r m s in t e r m s of the iterated commutators. which may b e omitted when the Feynman notation is adopted.. In the section EXPANSION I ITERATED COMMUTATORS. The ordering is usually exhibited explicitly by the chronological operator.

[[A. .B]. . . . B]] . . (B) I = [A. [A. . [A. . n commutations n commutations ITERATED COMMUTATORS The simple iterated commutator is { (A)". . [A. .] . B]. . B] . . .SYMBOLS operator operator with ordering index s operator which may depend on ordering parameter s operator operator in eq. . B] ] . . 1. . [. (29) operator in Zassenhaus formula operator in expansion of C analytic function of x linear combination of iterated commutators operator in BCH formula operator in expansion of H A +B F[X(A + B)]-lFIA(A1 + Bo)] operator in expansion of P F [ W + B)] - F [ m 1 + Bo)] ordering parameter [A. [A.

The f o r m of equation (3) is a polynomial whose t e r m s each have n factors of A and one of B . The case n = 0 is specified by It is seen that Clearly.which defines a linear operation on B obtained by n successive commutations with the operator A . B = ] ( : ) @)}An-' r=l r=1 A familiar example in which these iterated commutators occur is the L i e expansion. The following four theorems are estab­ lished by induction: n [An. one of which is raised to a power. (3) and (4)) arise when two operators. the iterated commutators (eqs. which is the power series exp(hA) B exp(-XA) = 4 . Similarly. are transposed.

convert any polynomial in A and B into a sum of t e r m s . followed by the remaining (n-1) iterations of the other operator. An R-type commutator. that is.There is no unique way to prescribe a commutator {(A)". they describe the commutation of a power of one operator with a power of another operator. then equation (3) becomes the nth derivative of B. f o r example. (9) and (10)) consist of a s u m of products each having m factors of A and n factors of B. (3)) with the m-fold iteration of the operator A (the operator appearing in the left-hand position) is formed first. each of which is a product of iterated commutators and an ordered product m n A B . (B)" ] that reduces to the f o r m s in equations (3) and (4) f o r the special c a s e s n = 1 and m = 1. respectively. and (8) follow directly the reduction formulas n m n m Equations (7a) and (12). dnB dtn 5 . f o r which the prescription is reversed. B A is the differentiation operator with respect to a parameter on which the oper­ a t o r B depends. Various generalizations suggest themselves. the simple commutator (eq. (7). From equations (6). is easily seen f r o m equa­ tion (5) to satisfy the relation The iterated commutators (eqs. (9)). B. The extended definition of iterated commu­ ator is chosen to be In the "L-type" of extended commutator (eq.

After using. (B)n}Lt= (-l)m+n-l{ (At)m. When each t e r m is written in the usual positional con­ vention. the second t e r m on the right-hand side of equation (18) becomes r=l t = l 6 p=O . both notations have been used on the right. ( 6 ) . and (9). (Bt)n}L (15) EXPANSION IN ITERATED COMMUTATORS First some useful disentangling expressions involving the quantities (A1 be developed. (7a). These "ordered powers" of (A + B) are simply + Bo)n will ( A +B ~ ~= ) ~ r=O C (:) n A ~ . ­ I . in suc­ cession. equations (7b). powers of A appear always to the left of powers of B. Bt ] and equation (10) gives the following f o r the Hermitian adjoint: { (A)m. Employing in turn the identities [A. they a r e obtained as ordinary binomial expansions in which A1 and Bo are treated as though they commuted. BIT = -[A t . Multiplying both sides of equation (17) from the left by (A + B) gives n (A2 + B2)(A1 + Bo)n = (A1 + r=l (r!l)p+l-r. f o r convenience.and equation (9) describes the commutation properties of the derivatives with the powers of B.~ B ~ that is.B]Br-l where.

If only a finite number of iterated com­ with values of j l a r g e r than a certain value mutators a r e nonvanishing. B is w i l l vanish. consider a polynomial in x of degree I . The subscript j is bracketed to signify that it has no ordering significance.t separate the summation into independent sums over 1 and k.( L -r+ r L-r m . an ordinary number. It is found. With the help of the identity [pm.r . that On the right-hand side of equation (19). all the iterated commutators (eq.The substitution? j = p + t . (20)) f o r which j = m + n . then all G ( j the case where A is proportional to p2 . A s an illustration. f o r example. and [p. . (9)) whose indices m and n add up to the s a m e value appear in the single operator (eq.1 and 1 = n + 1 .r 1)c x p it is found that 7 .x2] =E Z ( L (y) r=l m 1). The result allows equation (18) to be written in the form where i-1 k=O The operator G is a homogeneous polynomial of degree ( j + 1) in the operators A (j and B. k = t .x] = c .1 and in none of the others. .

1 = n-2 (r. Thefinal t e r m SN is (n . each left commutator (eq. when n is even. . with ordered powers of (A + B) appearing on the right i n place of ordinary powers: 8 . With the help of the recursion relation (eq.2m + r. .where qr(x) is a polynomial in x of degree 2 . The iterated commutator 2 j-m+l 2m vanishes if m > 2 . On the other hand.3) . when j 2 22 + 1. and is the sum t e r m f o r r equal to (1/2) (n f l). when n is odd.~ (1") + G r=l r=2 i+j=r (r. the following disentangling formula can be verified directly by induction: (A1 + Bo)n .~ .l)(n . { (p ) .2) ('f')'A +B)n-2-rG(i)G(j) +x n-3 r = 3 h+i+j=r (rr3) ( j )( r+2 h+i+l i )(A B)n-3-rG(h)G(i)G(j) The number N of summation t e r m s on the right equals the largest integer that is l e s s than or equal to (1/2)n.(A + B)" n. (x ) }L vanishes if j 2 2m + 1. Hence. (19)). (5)(3)[G(11] (1/2)n . when m 5 2.l)(~ + B ) " . A dual f o r m of equation (21) holds. (9)) appearing in G(j) vanishes and GU) = 0.

2)r+i+1) (Al+Bo)n-3-rG . ~F o r a simple function of operators.(A1 + BOP- (A + B)" + >: n-3 r = 3 h+i+j =r (r:3)r. . A and B are Hermitian. + (-l)N-lSN (22) The form of the dual expression (eq. If then G L ) = (-l$GG). beginning with a plus sign on the first t e r m . is obtained from the form (20) by substituting AT. (22)) is similar to equation (21) except that (1) the o r d e r of the subscripts on the G factors is reversed and (2) the sign on successive (j 1 sum t e r m s alternates. Let the analytic function be F(x) =>: fnxn n=O n! (23) In the formula F[IX(A + B)] = FIIX(A1 + Bo)] + R (24) the t e r m R h a s the series expansion 9 .) . Bo)? z F(A&.l ) J . the result is clearly F(A1. BI). Equa­ tion (15) shows that G?. The Hermitian adjoint of a functional that is written in the Feynman notation is gotten by replacing every operator A(s) by A ? (s) and exactly reversing the ordering convention on the p a r a m e t e r s of the adjoint operators. and inserting a n overall factor of ( . The application of equations (21) and (22) to analytic functions of (A + B) is straight­ forward. ( I(1) (h) . B 6) respectively. G . G +. B t f o r A. The o r d e r of the operator f a c t o r s in each t e r m of equation (21) can b e reversed by taking the Hermitian adjoint of each side.

A dual form of equation (26) can also be developed. and the G t ) * An alternative formulation is the disentangling formula i F[A(A~+ B ~ ) = F[A(A + B)]P ] where (26) Equating the coefficients of like powers of h in equation (26). with the help of equa­ The first few t e r m s a r e tion (21). Other expansion methods f o r analytic functions of (A + B) and references to earlier work in this a r e a a r e given by Kumar (ref. where P-' is expressed in t e r m s of ordered powers of (A + B).directly in t e r m s of the constant fn. This leads to a compact statement of the theorem exp(hA)exp(hB) = exph(A + B)exp(C) 10 (29) I . 3). namely. EXPONENTIAL FUNCTIONS The operators P (4 in equation (27) take a particularly simple f o r m when F is the exponential function. yields the operators P (n). i n which c a s e f n = 1 f o r all values of n. ordered Equation (22) gives the operator R(n) powers of (A + B).

where I equation (29) is written i n the form f the coefficient of (An/n! ) i n the power s e r i e s expansion of the left-hand side is given by equation (21).. (32)). The corresponding coefficient on the right is 1 m = l r=m a+b+. $(r:m)k?).r -m E E (a) (b) - * . * +i+j=r i+j+2 (h+i+j+3)(i+l ) h+l x ( A + B )n . The results through ninth o r d e r in h a r e 11 . The EU) a r e found by equating coefficients of the s a m e power of h on each side of equation (31). there appear m factors of E (j1 and m binomial coefficients. E(h)E(i)E(j) (32) In the mth product t e r m of the previous sum (eq. 2 >: .

However.. the operators C ci) of like powers of A . are obtained by equating coefficients With the help of equation (22). C contains the t e r m ( . 4) expb(A + Bg = exp(XA)exp(AB)exp . Nevertheless. the expressions given f o r the operators and C are not altogether unique. they display concisely the iteratedE(j) (j ) commutator nature of the t e r m s in the expansions of equations (30) and (33). the degree of the iterated commutators in G (j) * The formula of Zassenhaus (ref. (j) This procedure may also be applied to the Baker-Campbell-Hausdorff (BCH) formula 1 exp(XA)exp(XB) = exp(H). (3 4) 12 .G(j-2) .x t ) e.1 / 2 ) j ( j .Note that each of the t e r m s occurring in E is essentially predetermined to (j ) within a numerical coefficient.( 3 p$ . F o r example. No attempt has been made to obtain general expressions. f o r certain types of t e r m s the general form of the numerical coefficients can be inferred from the specific values already given. the r e s u l t s out to the ninth o r d e r in X are Because of the Jacobi and other identities. then the numbers s o obtained add up to (j + l ) .l)[G(l). (33) corresponds to the complete factorization of exp(-C) in the inverse form of equation (29). . Again. Consider any single t e r m in the sum of t e r m s that makes up the complete expression for a given E I the subscript on every G f (j) (i) appearing in the t e r m has its numerical value increased by unity. $ .

where H = AHo + ($)H(l) + ($)H(2) +.. . than either the E U ) o r the C (j). Moreover. (3 5) Ho=A+B In the case of the BCH formula. 4 and 5) that have been developed specifically f o r this problem. the H(j) can be computed within the framework of the procedure given herein without introducing additional elements (such as operator derivatives o r auxiliary operator func­ tions) that the m o r e specialized methods employ. Equating the coefficient of (hn/n!) in the expansions of [exp(hH) .exp(hHo)l to the expression given by equation (21) yields H(l) G(1) = When expressed in t e r m s of the G(j). the method offers no great advantage over previous methods (refs. the H(j) contain considerably more t e r m s The additional t e r m s are the ones containing Ho. 13 . however. However. the relations and 'No such simple relations exist for higher G(j).

1967. 84. Kumar. Math. 2. written: For example. as follows directly f r o m the fact that the right-hand-side expansions of equa­ tions (34) and (36) are the s a m e up to the nth power of A. J. : Exponential Operators and P a r a m e t e r Differentiation in Quantum Physics. no. : An Operator Calculus Having Applications in Quantum Elec­ trodynamics. Kailash: Expansion of a Function of Noncommuting Operators. 1951. Ohio. Wilcox. pp. vol. no.. Phys. vol. 14 . Dec. pp. Feynman. 1. Richard P. R. J. no. the operators in the first exponential on the right are those of the BCH formula. - REFERENCES 1. pp. 4 . all of which can be written as with equation (29) by n = 1 and equation (33) by the limit n 03. P h y s . M. 108-128. . vol. 1968. 1923-1927. The operators on the right-hand side of equation (35) can be evaluated by the procedure given herein. (34)) and equation (29). Math. August 21. 962-982. 8. National Aeronautics and Space Administration. the following can be There exist any number of intermediate formulas between the BCH f o r m (eq. Phys. Cleveland. 129-02-07-07-22.can b e used to reduce the expressions f o r the H (j). Oct. Apr. 3. 12. Rev. 1965. 1. 6.. In particular. Lewis Research Center.

G. H. 5 Weiss. P u r e Appl. Comm. 1968 . 1962. pp. 4. in Crystal Physics. July-Aug.. Operator. 771-777. A. A. Wilhelm: On the Exponential Solution of Differential Equations f o r a Linear . no. 1954. vol. Math. Nov. no. : The Baker-Hausdorff Formula and a Problem . Phys.4 Magnus. vol. Math. 3.19 E-4469 15 . 649-673. NASA-Langley. 4. J.. and Maradudin. 7 . . pp.

TECHNICAL NOTES: Information less broad in scope but nevertheless of importance as a contribution to existing knowledge. Publications include Tech Briefs. Technology Utilization R~~~~~~nd N a ~ ~ and Technology Surveys. and special bibliographies. ~ ~ . complete. monographs. or other reasons. sourcebooks. CONTRACTOR REPORTS: Scientific and technical information generated under a NASA contract or grant and considered an important contribution to existing knowledge. t o the expansion of hzimzan Lnowl­ edge of phenomena in the at?iiosphere and space.” -NATIONAL AERONAUTICS S P A C E ACT OF 1958 AND NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS TECHNICAL REPORTS: Scientific and technical information considered important. handbooks. TECHNOLOGY UTILIZATION PUBLICATIONS: Information on technology used by NASA that may be of particular interest in commercial and other non-aerospace applications. TECHNICAL TRANSLATIONS: Information published in a foreign language considered to merit NASA distribution in English. . C. SPECIAL PUBLICATIONS: Information derived from or of value to NASA activities. . T h e Administration shull provide for the widest practicable and appropriate dissemination of inforuintion concerning its activities and the r e s ? h thereof. D. OFFICIAL BUSINESS FIRST CLASS MAIL POSTAGE A N D FEES PAID NATIONAL AERONAUTICS A N D SPACE ADMINISTRATION POSTMASTER: If Undeliverable (Section 158 Postal Manual) D o Not Return “The aeronautical and space activities of the h i t e d States shall Je condncted so as to contribute . 20546 . 20546 D. and a lasting contribution to existing knowledge. Details on the availability of these publications may be obtained from: SCIENTIFIC AND TECHNICAL INFORMATION DIVISION NATIONAL AERONAUTICS AND SPACE ADMINISTRATION Washington. security classifica­ tion.NATIONAL AERONAUTICS SPACE ADMINISTRATION AND WASHINGTON.C. TECHNICAL MEMORANDUMS: Information receiving limited distribution because of preliminary data. data compilations. Publications include conference proceedings.

Sign up to vote on this title
UsefulNot useful