TOPICS
Search

Symmetric Polynomial


A symmetric polynomial on n variables x_1, ..., x_n (also called a totally symmetric polynomial) is a polynomial that is unchanged by any permutation of its variables. In other words, the symmetric polynomials satisfy

 f(y_1,y_2,...,y_n)=f(x_1,x_2,...,x_n),
(1)

where y_i=x_(pi(i)) and pi is an arbitrary permutation of the indices 1, 2, ..., n.

For fixed n and coefficients in a field, the symmetric polynomials form an infinite-dimensional algebra generated freely by the n elementary symmetric polynomials. These generators are algebraically independent; they are not a vector-space basis. They occur, with alternating signs, as the coefficients of the monic polynomial whose roots are x_1, ..., x_n.

There are four common homogeneous bases for the symmetric polynomials over Q, indexed by appropriate partitions lambda (Dumitriu et al. 2007). Letting l be the length of lambda, the elementary symmetric polynomials e_lambda, complete homogeneous symmetric polynomials h_lambda, and power-sum functions p_lambda are defined for l=1 by

e_(lambda_1)=sum_(j_1<j_2<...<j_(lambda_1))x_(j_1)...x_(j_(lambda_1))
(2)
h_(lambda_1)=sum_(m_1+...+m_n=lambda_1)product_(j=1)^(n)x_j^(m_j)
(3)
p_(lambda_1)=sum_(j=1)^(n)x_j^(lambda_1),
(4)

where m_1, ..., m_n are nonnegative integers, and for l>1 by

 s_lambda=product_(i=1)^ls_(lambda_i),
(5)

where s is one of e, h or p. In addition, the monomial symmetric polynomials m_lambda are defined as

 m_lambda=sum_(sigma in S_lambda)x_(sigma(1))^(lambda_1)x_(sigma(2))^(lambda_2)...x_(sigma(m))^(lambda_m),
(6)

where S_lambda is the set of permutations giving distinct terms in the sum and lambda is padded with zeros to length m=n.

As several different abbreviations and conventions are in common use, care must be taken when determining which symmetric polynomial is in use.

The elementary symmetric polynomials Pi_k(x_1,...,x_n) (sometimes denoted sigma_k or e_lambda) on n variables {x_1,...,x_n} are defined by

Pi_1(x_1,...,x_n)=sum_(1<=i<=n)x_i
(7)
Pi_2(x_1,...,x_n)=sum_(1<=i<j<=n)x_ix_j
(8)
Pi_3(x_1,...,x_n)=sum_(1<=i<j<k<=n)x_ix_jx_k
(9)
Pi_4(x_1,...,x_n)=sum_(1<=i<j<k<l<=n)x_ix_jx_kx_l
(10)
|
(11)
Pi_n(x_1,...,x_n)=product_(1<=i<=n)x_i.
(12)

The kth elementary symmetric polynomial is implemented in the Wolfram Language as SymmetricPolynomial[k, {x1, ..., xn}]. SymmetricReduction[f, {x1, ..., xn}] gives a pair of polynomials {p,q} in x_1, ..., x_n where f=p+q, p is a symmetric polynomial, and q is the remainder.

Alternatively, Pi_j(x_1,...,x_n) can be defined as the coefficient of x^(n-j) in the generating function

 product_(1<=i<=n)(x+x_i).
(13)

For example, on four variables x_1, ..., x_4, the elementary symmetric polynomials are

Pi_1(x_1,x_2,x_3,x_4)=x_1+x_2+x_3+x_4
(14)
Pi_2(x_1,x_2,x_3,x_4)=x_1x_2+x_1x_3+x_1x_4+x_2x_3+x_2x_4+x_3x_4
(15)
Pi_3(x_1,x_2,x_3,x_4)=x_1x_2x_3+x_1x_2x_4+x_1x_3x_4+x_2x_3x_4
(16)
Pi_4(x_1,x_2,x_3,x_4)=x_1x_2x_3x_4.
(17)

The power sum S_p(x_1,...,x_n) is defined by

 S_p(x_1,...,x_n)=sum_(k=1)^nx_k^p.
(18)

The relationship between S_p and Pi_1, ..., Pi_p is given by the so-called Newton-Girard formulas. The related function s_p(Pi_1,...,Pi_n) with arguments given by the elementary symmetric polynomials (not x_n) is defined by

s_p(Pi_1,...,Pi_n)=(-1)^(p-1)S_p(x_1,...,x_n)
(19)
=(-1)^(p-1)sum_(k=1)^(n)x_k^p.
(20)

It turns out that s_p(Pi_1,...,Pi_n) is given by the coefficients of the generating function

ln(1+Pi_1t+Pi_2t^2+Pi_3t^3+...)=sum_(k=1)^(infty)(s_k)/kt^k
(21)
=Pi_1t+1/2(-Pi_1^2+2Pi_2)t^2+1/3(Pi_1^3-3Pi_1Pi_2+3Pi_3)t^3+...,
(22)

so the first few values are

s_1=Pi_1
(23)
s_2=-Pi_1^2+2Pi_2
(24)
s_3=Pi_1^3-3Pi_1Pi_2+3Pi_3
(25)
s_4=-Pi_1^4+4Pi_1^2Pi_2-2Pi_2^2-4Pi_1Pi_3+4Pi_4.
(26)

In general, s_p can be computed from the determinant

 s_p=(-1)^(p-1)|Pi_1 1 0 0 ... 0; 2Pi_2 Pi_1 1 0 ... 0; 3Pi_3 Pi_2 Pi_1 1 ... 0; 4Pi_4 Pi_3 Pi_2 Pi_1 ... 0; | | | | ... 1; pPi_p Pi_(p-1) Pi_(p-2) Pi_(p-3) ... Pi_1|
(27)

(Littlewood 1958, Cadogan 1971). In particular,

S_1(x_1,...,x_n)=sum_(k=1)^(n)x_k=Pi_1
(28)
S_2(x_1,...,x_n)=Pi_1^2-2Pi_2
(29)
S_3(x_1,...,x_n)=Pi_1^3-3Pi_1Pi_2+3Pi_3
(30)
S_4(x_1,...,x_n)=Pi_1^4-4Pi_1^2Pi_2+2Pi_2^2+4Pi_1Pi_3-4Pi_4
(31)

(Schroeppel 1972), as can be verified by plugging in and multiplying through.


See also

Chromatic Quasisymmetric Function, Complete Homogeneous Symmetric Polynomial, Fundamental Theorem of Symmetric Functions, Jack Polynomial, McLeod Conjecture, Monomial Symmetric Polynomial, Newton-Girard Formulas, Orthogonal Polynomials, Power Sum, Shareshian-Wachs Conjecture, Symmetric Function, Vieta's Formulas, Zonal Polynomial

Portions of this entry contributed by David Terr

Explore with Wolfram|Alpha

References

Borwein, P. and Erdélyi, T. Polynomials and Polynomial Inequalities. New York: Springer-Verlag, p. 5, 1995.Cadogan, C. C. "The Möbius Function and Connected Graphs." J. Combin. Th. B 11, 193-200, 1971.Dumitriu, I.; Edelman, A.; and Shuman, G. "MOPS: Multivariate Orthogonal Polynomials (Symbolically)." J. Symb. Comput. 42, 587-620, 2007. https://doi.org/10.1016/j.jsc.2007.01.005.Littlewood, J. E. A University Algebra, 2nd ed. London, England: Heinemann, 1958.Schroeppel, R. Item 6 in Beeler, M.; Gosper, R. W.; and Schroeppel, R. HAKMEM. Cambridge, MA: MIT Artificial Intelligence Laboratory, Memo AIM-239, p. 4, Feb. 1972. https://www.inwap.com/pdp10/hbaker/hakmem/geometry.html#item6.Séroul, R. "Newton-Girard Formulas." §10.12 in Programming for Mathematicians. Berlin, Germany: Springer-Verlag, pp. 278-279, 2000.

Referenced on Wolfram|Alpha

Symmetric Polynomial

Cite this as:

Weisstein, Eric W., with contributions by David Terr. "Symmetric Polynomial." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/SymmetricPolynomial.html

Subject classifications