Advances in Linear Algebra & Matrix Theory, 2015, 5, 150-155 Published Online December 2015 in SciRes. http://www.scirp.org/journal/alamt http://dx.doi.org/10.4236/alamt.2015.54015

Trace of Positive Integer Power of Real 2 × 2 Matrices

Jagdish Pahade1*, Manoj Jha2 Department of Mathematics, Maulana Azad National Institute of Technology, Bhopal, India

Received 21 September 2015; accepted 4 December 2015; published 7 December 2015

Copyright © 2015 by authors and Scientific Research Publishing Inc. This work is licensed under the Creative Commons Attribution International License (CC BY). http://creativecommons.org/licenses/by/4.0/

Abstract The purpose of this paper is to discuss the theorems for the trace of any positive integer power of 2 × 2 real matrix. We obtain a new formula to compute trace of any positive integer power of 2 × 2 real matrix A, in the terms of Trace of A (TrA) and Determinant of A (DetA), which are based on definition of trace of matrix and multiplication of the matrixn times, where n is positive integer and this formula gives some corollary for TrAn when TrA or DetA are zero.

Keywords Trace, Determinant, Matrix Multiplication

1. Introduction Traces of powers of matrices arise in several fields of mathematics, more specifically, Network Analysis, Num- bertheory, Dynamical systems, Matrix theory, and Differential equations [1]. When analyzing a complex net- work, an important problem is to compute the total number of triangles of a connected simple graph. This num- ber is equal to Tr(A3)/6, where A is the adjacency matrix of the graph [2]. Traces of powers of integer matrices are connected with the Euler congruence [3], an important phenomenon in mathematics, stating that

rr−1 Tr( AApp) ≡ Tr( )(mod pr ) ,

for all integer matrices A, all primes p, and all r Z. The invariants of dynamical systems are described in terms of the traces of powers of integer matrices, for example in studying the Lefschetz numbers [3]. There are many applications in matrix theory and numerical linear∊ algebra. For example, in order to obtain approximations of the

*Corresponding author.

How to cite this paper: Pahade, J. and Jha, M. (2015) Trace of Positive Integer Power of Real 2 × 2 Matrices. Advances in Linear Algebra & Matrix Theory, 5, 150-155. http://dx.doi.org/10.4236/alamt.2015.54015

J. Pahade, M. Jha smallest and the largest eigenvalues of a symmetric matrix A, a procedure based on estimates of the trace of An and A−n, n Z, is proposed in [4]. Trace of a nn× matrix A = a is defined to be the sum of the elements on the main diagonal of A, i.e. ∊ ij = + ++ = n TrA aa11 22  ann ∑i=1 aii The computation of the trace of matrix powers has received much attention. In [5], an algorithm for compu- ting Tr( Ak ) ,k ∈ Z is proposed, when A is a lower Hessenberg matrix with a unit codiagonal. In [6], a sym- bolic calculation of the trace of powers of tridiagonal matrices is presented. Let A be a symmetric positive defi- q nite matrix, and let {λk } denote its eigenvalues. For q R, A is also symmetric positive definite, and it holds [7]. qq=∊ λ TrA ∑k k (1.1) This formula is restricted to the matrix A. Also we have other formulae [8] to compute the trace of matrix power such that nn= λ TrA ∑k k (1.2) But for many cases, this formula is time consuming. For example

11− 5 37± i Consider a matrix A =  and let we are to find TrA . Eigenvalues of A are , then by (1.2), 22 2 55 5 37+−ii  37  TrA =  + . 22   Computation of this value is time consuming. Therefore, other formulae to compute trace of matrix power are needed. Now we give new theorems and corollaries to compute trace of matrix power. Our estimation for the trace of An is based on the multiplication of matrix.

2. Main Result Theorem 1. For even positive integer n and 2 × 2 real matrix A, n 2 r n (−1) r nr−2 TrA=∑ nnr −+( 1)  nr −+( 2) [ up to r terms]⋅( DetAA) ( Tr ) r=0 r! ab Proof. Consider a matrix A =  where abcd,,, are real. cd Then TrA =ad + (2.1) and DetA =ad − bc (2.2) Now

abab a2 ++ bc b( a d ) 2 = = A 2 cd cd c( a++ d) bc d

Then

TrA2 = a22++ 2 bc d =+a222 ad +− d 22 ad + bc 2 (2.3) =(a + d) −−2( ad bc) 2 =(TrAA) − 2Det

151 J. Pahade, M. Jha

Now again a2 ++ bc b( a d ) ab 32= = A AA 2  c( a++ d) bc d  cd

a3+ abc + bc( a + d) a22 b ++ b c bd( a + d ) 3 = A 22 3 (2.4) ac( a++ d) bc + cd bc( a ++ d) bcd + d Then TrA333= a++ d3 bc( a + d ) =++a33 d333 ad( a +− d) ad( a ++ d) bc( a + d ) 3 =+−+(a d) 3( a d)( ad − bc) 3 = (TrA) − 3( Det AA)( Tr )

3 TrA3 =( Tr A) − 3( Det AA)( Tr ) (2.5) Now replace A by A2 in (2.3), we have

422 2 TrAA= Tr − 2DetA 2 =−=22 (TrA) – 2DetA 2( Det A)  Det AB( Det A)( Det B)

4 22 TrAA4 = ( Tr ) – 4DetAA(Tr A) + 2( Det ) (2.6) Again replace A by A2 in (2.5), we have

6 23 22 TrA= Tr A − 3Det AA( Tr )

2 3 22 =−−(TrAA) 2Det 3Det( AAA) ( Tr) − 2Det  =−62 2−− 3 (TrA) 6( Tr A) Det AA( Tr) 2Det A 8( Det A) 22 3 −+3( DetAA) ( Tr) 6( Det A)

6 4 22 3 TrA6 =−+−( Tr A) 6Det AA( Tr) 9( Det A) ( Tr A) 2( Det A) (2.7) Now again replace A by A2 in (2.6), we have

8 24 22 22 2 TrA=−+ Tr A 4Det AA  Tr 2 Det A  42 =−−2 22 −+4 (TrAA) 2Det 4Det( AAAA) ( Tr) 2Det 2Det( )

=−8 242+++ 442 (TrAAAAA) 8Det( )( Tr) ( Tr) 4Det( ) 16Det( AAA) 24Tr( ) ( Det ) −24 − 2 ++2 4 4Det( AA) ( Tr) 4Tr( AA) ( Det) 4Det( A) 2Det( A)

8 6 24 32 4 TrA8 =−+−( Tr A) 8( Det AA)( Tr) 20( Det A) ( Tr A) 16( Det A) ( Tr A) + 2( Det A) (2.8) Now we observe from (2.3), (2.6), (2.7) and (2.8) that 01 (−−11) 0 220−× ( ) 1 221−× TrAA2 = (DetAA) (Tr ) + 2(Det ) ( Tr A) 0! 1!

01 2 (−−11) 0 420−× ( ) 1 421−× ( − 1) 2 422−× TrAA4 = (DetAA) (Tr ) + 4(Det ) ( Tr A) +−4( 4 3)(DetA) ( TrA) 0! 1! 2!

152 J. Pahade, M. Jha

01 2 (−−11) 0 620−× ( ) 1 621−× ( − 1) 2 622−× TrAA6 = (DetA) (Tr ) +6(DetAA) ( Tr A) +−6( 6 3)(Det ) ( Tr A) 0! 1! 2!

3 (−1) 3 623−× +66( −− 4(65)) (DetA) ( TrA) 3!

01 2 (−−11) 0 820−× ( ) 1 821−× ( − 1) 2 822−× TrAA8 = (DetAA) (Tr ) +8(Det ) ( Tr A) +−8( 8 3)(Det A) ( Tr A) 0! 1! 2! 34 (−−1) 3 823−× ( 1) 3 824−× +88485( −−)( )(DetA) ( TrAA) +86586( −−−)( )(8 7)(DetA) ( Tr ) 3! 4! Continuing this process up to n terms we get

01 2 (−−11) 020nn−× ( ) 12−×1( − 1) 222 n−× TrAAn =(DetAA) (Tr ) +n(Det ) (Tr A) +−nn( 3)(Det A) ( Tr A) 0! 1! 2! r (−1) r nr−×2 ++nnr −+( 1)  nr −+( 2)  ( up to r terms)⋅(DetA) ( TrA) (2.9) r! n 2 (−1) n2 nn−× 22 ++nnn( 2+ 1)  n −( n 2 + 2)  ( up ton 2 terms)⋅(DetA) (TrA) n 2!

Finally from above, we get

n 2 r n (−1) r nr−2 TrA=∑ nnr −+( 1)  nr −+( 2) [ up to r terms]⋅( DetAA) ( Tr ) (2.10) r=0 r! Hence the proof is completed. Theorem 2. For odd positive integer n and 2 × 2 real matrix A,

(n−12) r n (−1) r nr−2 TrA=∑ nnr−+( 1)  nr −+( 2) [ up to r terms]⋅( DetAA) ( Tr ) r=0 r! Proof. Consider a matrix A as in theorem 1, we have from (1.4) and (1.6). a3+ abc + bc( a + d) a22 b ++ b c bd( a + d) a2 + bc b( a + d ) 5= 32 = A AA  22 3 2 ac( a++ d) bc + cd bc( a ++ d) bcd + d c( a + d) bc + d 

5 2 3 22 TrA = (a+ bc)  a + abc + bc( a ++ d)  c( a + d) a b ++ b c bd( a + d )  22 2 3 +b( a + d)  ac( a ++ d) bc + cd +( bc + d)  bc( a ++ d) bcd + d 3 =++a5 d 5252 bc( a 3 + d 3) + b 22 c( a ++ d) bc( a + d)( a2 + d 2) + bc( a + d ) 3 = a5++ d 5 5 bc( a + d) −10 abcd( a ++ d) 5 b22 c( a+ d ) 3 =++a5 d 5 5 adad( +) − 5 adad22( +−) 5( adad+ )3++5ad 22( a d) 3 +5bc( a + d) −10 abcd( a ++ d) 5 b22 c( a + d ) 5 32 =+−(a d) 55( ad − bc)( a ++ d) ad( ad − bc) ( a + d )

5 32 TrA5 =−+(TrA) 5Det AA( Tr ) 5( Det A) ( TrA) (2.11) Now we observe from (2.5) and (2.11) that

01 (−−11) 0 320−× ( ) 1 321−× TrAAA3 = ( Det) ( Tr ) + 3( DetAA) ( Tr ) 0! 1!

153 J. Pahade, M. Jha

01 2 (−−11) 0 520−× ( ) 1 521−× ( − 1) 2 522−× TrAAA5 =( Det) ( Tr ) +5( DetAA) ( Tr ) +−5( 5 3)( DetAA) ( Tr ) 0! 1! 2! Now we continuing this as in Theorem 1, we get TrAn same as Theorem 1. But here r varies up to (n −12) . Hence the theorem follows. n Corollary 1: For any positive integer n and 2 × 2 real singular matrix A, TrAAn = ( Tr ) . Proof: For singular matrix A, DetA = 0. Hence proof follows from Theorem 1 and Theorem 2. Corollary 2: For 2 × 2 real matrix A with TrA = 0. n 2 1) TrAn =2( −D etA) when n is even and; 2) TrAn = 0 when n is odd. Proof. Proof follows from theorem 1 and theorem 2. Corollary 3: For 2 × 2 real matrix A with TrA = 0 and DetA = 0. TrAn = 0 where n is any positive integer. Proof. Proof follows from Corollary 2.

11− 5 Example 1. Consider a matrix A =  and let we are to find TrA . 22 Here Dte A = 2 +=24 and TrA =+= 1 2 3. then by Theorem 2, we have

5 3 2 5 32 TrA5 =−( Tr A) 5Det AA( Tr) + 5( Det A) ( Tr A) =−+=−( 3) 5( 4)( 3) 5( 4) ( 3) 57

35− 10 Example 2. Consider a matrix A =  and let we are to find TrA . 12−  Here Dte A = −+6 51= − and TrA =−= 3 2 1. then by Theorem 1, we have

10 8 2 6 3 4 TrA10 =−+( Tr A) 10Det AA( Tr) 35( Det A) ( Tr A) − 50( Det A) ( Tr A) 42 5 +−25Det( AA) ( Tr) 2Det( A) =+1 10 + 35 + 50 + 25 += 2 123.

20 199 2015 Example 3. Consider a matrix A =  and let we are to find TrA . −−2 20 Here TrA = 0, DetA = −2 and n = 2015, which is odd, hence by corollary 2, we get TrA2015 = 0.

15 21 100 Example 4. Consider a matrix A =  and let we are to find TrA . −−10 14 Here A is a singular matrix with Trace 1, and then by Corollary 1, we have

100 100 TrAA100 =( Tr ) =(1) = 1.

Conclusion and Future Work After to discuss Theorems 1 and 2, Corollaries 1, 2 and 3, we are able to find trace of any integer power of a 2 × 2 real matrix. In future, we can be developed similar results for 3 × 3 real matrices.

Acknowledgements We would like to hardly thankful with great attitude to Director, Maulana Azad National Institute of Technology, Bhopal for financial support and we also thankful to HOD, Department of Mathematics of this institute for giv- ing me opportunity to expose my research in scientific world.

References [1] Brezinski, C., Fika, P. and Mitrouli, M. (2012) Estimations of the Trace of Powers of Positive Self-Adjoint Operators by Extrapolation of the Moments. Electronic Transactions on Numerical Analysis, 39, 144-155.

154 J. Pahade, M. Jha

[2] Avron, H. (2010) Counting Triangles in Large Graphs Using Randomized Matrix Trace Estimation. Proceedings of Kdd-Ldmta’10, 2010. [3] Zarelua, A.V. (2008) On Congruences for the Traces of Powers of Some Matrices. Proceedings of the Steklov Institute of Mathematics, 263, 78-98. [4] Pan, V. (1990) Estimating the Extremal Eigenvalues of a Symmetric Matrix. Computers & Mathematics with Applica- tions, 20, 17-22. [5] Datta, B.N. and Datta, K. (1976) An algorithm for Computing Powers of a Hessenberg Matrix and Its Applications. Linear Algebra and its Applications, 14, 273-284. [6] Chu, M.T. (1985) Symbolic Calculation of the Trace of the Power of a Tridiagonal Matrix. Computing, 35, 257-268. [7] Higham, N. (2008) Functions of Matrices: Theory and Computation. SIAM, Philadelphia. [8] Michiel, H. (2001) Trace of a Square Matrix. Encyclopedia of Mathematics, Springer. https://en.wikipedia.org/wiki/Trace_

155