This paper focuses on showing that much of the theoretical part of linear algebra works fairly well without determinants and provides proofs for most of the major structure, theorems of linear algebra without resorting to determinants.

1. SUMMER PROJECT REPORT
RAVINDER SINGH, IMS13114
NAME : Ravinder Singh, ID: IMS13114.
GUIDE : Dr. Sachindranath Jayaraman, IISER-TVM.
I reported as a Summer Project Student to Dr. Sachindranath Jayaraman, IISER-
TVM, on May 1,2015. I started reading the book Introduction to Linear Algebra, by
Prof.Gilbert Strang and I also watched all the video lectures of the online course Linear
Algebra 18.06, available at ocw.mit.edu. Then I started reading the paper Down with
determinants!, by Sheldon Axler, published in the American Mathematical Monthly [1].
This paper focuses on showing that much of the theoretical part of linear algebra works
fairly well without determinants and provides proofs for most of the major structure
theorems of linear algebra without resorting to determinants. A brief summary of the
important results I learnt from the above paper as well as from other references are
described below.
Definition 1.1. Given an m ∗ n matrix A having columns A1, A2, ....An and rows
a1, a2, ....am, let there be another matrix B of size n∗p and have columns B1, B2, ....Bp
and rows b1, b2, ....bn and a matrix C = A∗B which has columns C1, C2, ....Cp and rows
c1, c2, ....cm.Now the matrix C can be obtained in the following ways.
(a) Ci = A ∗ Bi ∀ i ∈ {1, 2, ....p}
(b) cj = aj ∗ B ∀ j ∈ {1, 2, ....m}
(c) Cij = ai ∗ Bj ∀ i ∈ {1, 2, ....m} and ∀ j ∈ {1, 2, ....p}
(d) C =
i=1,2,....n
(Ai ∗ bi)
Lemma 1.2. Steinitz Exchange Lemma : If {v1, v2, ....vm} is a set of m linearly
independent vectors in a vector space V , and {w1, w2, ....wn} span V then m n and,
possibly after reordering the wi, the set {v1, v2, ....vm, wm+1, wm+2, ....wn} spans V .
Then using the Steinitz Exchange Lemma , I proved the following theorem.
Definition 1.3. Given an m ∗ n matrix A, its Column rank (rc) is defined as the num-
ber of independent columns of matrix A and similarly its Row rank (rr) is defined as
1

2. 2 RAVINDER SINGH, IMS13114
the number of independent rows of matrix A.
Theorem 1.4. Given an m ∗ n matrix A.Then rc = rr.
Theorem 1.5. Given an m ∗ n matrix A having rank = r.Then A ∗ x = b has solution
as follows:
(a) If r = m = n , then there exists 1 unique solution.
(b) If r = n < m , then there exists 0 or 1 solutions.
(c) If r = m < n , then there exists ∞ solutions.
(d) If r < m and r < n , then there exists 0 or ∞ solutions.
After this, I learned 4 different types of matrix decompositions. Given an m ∗ n
matrix A having rank r, then followings are the descriptions of its different types of
decompositions.
(a) A = L ∗ U, where L is a lower triangular matrix and U is an upper triangular
matrix.
(b) A = B ∗C, where B is a full column rank matrix and C is the full row rank matrix.
(c) A = Q ∗ R, where Q is an orthogonal matrix and R is an upper triangular matrix.
(d) A = S ∗ Λ ∗ S−1, where S is an eigenvector matrix and Λ is the diagonal matrix
with eigenvalues as its diagonal entries.
Theorem 1.6. Fundamental Theorem of Linear Algebra : Given an m ∗ n matrix
A, the fundamental theorem of linear algebra is a collection of results relating various
properties of the four fundamental matrix subspaces of A. In particular:
(a) dim(R(A)) = dim(R(AT )) and dim(R(A)) + dim(N(A)) = n where here, R(A)
denotes the range or column space of A, AT denotes its transpose, and N(A) denotes
its null space.
(b) The null space N(A) is orthogonal to the row space R(AT ).
After this, I proved the Euler s Formula for graphs, which is stated below:
Given a graph with n number of nodes, e number of edges, l number of loops, then
n − e + l = 1.
I also learned that for a matrix A to have a set of orthogonal eigenvectors, it should
have a property that A ∗ AT = AT ∗ A.

3. SUMMER PROJECT REPORT 3
Now, I started reading the paper Down with determinants!, by Sheldon Axler, pub-
lished in the American Mathematical Monthly [1].
Definition 1.7. A complex number λ is called an eigenvalue of a linear operator T if
T − λ ∗ I is not injective.
Theorem 1.8. Every linear operator on a finite-dimensional complex vector space has
an eigenvalue.
Proposition 1.9. Non-zero eigenvectors corresponding to distinct eigenvalues of T are
linearly independent.
Definition 1.10. A vector v ∈ V is called a generalisedeigenvector of linear operator
T if (T − λ ∗ I)k = 0, for some eigenvalue λ of T and some positive integer k.
Lemma 1.11. The set of generalised eigenvectors of T corresponding to an eigenvalue
λ equals ker(T − λ ∗ I)n, where n is the dimension of the V .
Proposition 1.12. The generalised eigenvectors of T span V .
Proposition 1.13. Non-zero generalised eigenvectors corresponding to distinct eigen-
values of T are linearly independent.
Theorem 1.14. Let λ1, ....., λm be the distinct eigenvalues of T, with U1,.....,Um de-
noting the corresponding sets of generalized eigenvectors. Then
(a) V =U1 ⊕ · · · ⊕ Um;
(b) T maps each Uj into itself;
(c) each (T − λjI)|Uj is nilpotent;
(d) each T|Uj has only one eigenvalue, namely λj.
Definition 1.15. The Minimal Polynomial : There is a smallest positive integer k
such that I, T, T2, ...., Tk are not linearly independent. Thus there exist unique complex
numbers a0, a1, ....ak−1 such that a0 ∗I +a1 ∗T +a2 ∗T2 +....ak−1 ∗Tk−1 +Tk = 0. The
polynomial a0+a1∗z+a2∗z2+....ak−1∗zk−1+zk is called the minimal polynomial of T.

4. 4 RAVINDER SINGH, IMS13114
Theorem 1.16. Let λ1, λ2, ....λm be the distinct eigenvalues of T, let Uj denote the set
of generalised eigenvectors corresponding to λj, and let αj be the smallest positive inte-
ger such that (T −λj ∗I)αj ∗v = 0 for every v ∈ Uj. Let p(z) = (z−λ1)α1 ....(z−λm)αm .
Then
(a) p is the minimal polynomial of T;
(b) p has degree at most dimV ;
(c) if q is a polynomial such that q(T) = 0, then q is a polynomial multiple of p.
Definition 1.17. Let λ1, λ2, ....λm be the distinct eigenvalues of T, with correspond-
ing multiplicities β1, β2, ....βm..The polynomial (z − λ1)β1 ....(z − λm)βm is called the
characteristic polynomial of T.
Theorem 1.18. Let q denote the characteristic polynomial of T. Then q(T) = 0.
Lemma 1.19. Suppose T is nilpotent. Then there is a basis of V with respect to which
the matrix of T contains only 0’s on and below the main diagonal.
Theorem 1.20. Let λ1, λ2, ....λm be the distinct eigenvalues of T. Then there is a
basis of V with respect to which the matrix of T has the form
M =




















λ1 ∗ 0 0 0 0 0 0 0 0
0 . ∗ 0 0 0 0 0 0 0
0 0 . ∗ 0 0 0 0 0 0
0 0 0 λ1 ∗ 0 0 0 0 0
0 0 0 0 . ∗ 0 0 0 0
0 0 0 0 0 . ∗ 0 0 0
0 0 0 0 0 0 λm ∗ 0 0
0 0 0 0 0 0 0 . ∗ 0
0 0 0 0 0 0 0 0 . ∗
0 0 0 0 0 0 0 0 0 λm




















Lemma 1.21. If T is normal, then kerT = kerT∗.
Proposition 1.22. Every generalised eigenvector of a normal operator is an eigenvec-
tor of the operator.

5. SUMMER PROJECT REPORT 5
Proposition 1.23. Eigenvectors of a normal operator corresponding to distinct eigen-
values are orthogonal.
Theorem 1.24. There is an orthonormal basis of V consisting of eigenvectors of T if
and only if T is normal.
Proposition 1.25. Every eigenvalue of a self-adjoint operator is real.
Theorem 1.26. Every linear operator on an old-dimensional real vector space has a
real eigenvalue.
Theorem 1.27. Suppose U is a real inner product space and S is a linear operator on
U. Then there is an orthonormal basis of U consisting of eigenvectors of S if and only
if S is self-adjoint.
Theorem 1.28. An operator is invertible if and only if its determinant is non-zero.
Proposition 1.29. The characteristic polynomial of T equals det(z ∗ I − T).
The classical definition of determinant goes as follows: The determinant of T is
defined as det(T) =
π∈Sn
(signπ)tπ(1),1 . . . tπ(n),n, where Sn denotes the permutation
group on n symbols. This is less intuitive at first. However, a nice and simple geomet-
ric meaning of the determinant is through the change of variable formula, the statement
of which goes as follows.
Lemma 1.30. Let S be a linear operator on a real inner product space U. Then there
exists a linear isometry A on U such that S = A ∗
√
S∗S.
Theorem 1.31. Let S be a linear operator on Rn. Then vol S(E) = det(S) ∗ volE for
E ⊂ Rn.
Theorem 1.32. The Schur decomposition reads as follows: if A is a n ∗ n square ma-
trix with complex entries, then A can be expressed as A = Q ∗ U ∗ Q−1, where Q is a
unitary matrix (so that its inverse Q−1 is also the conjugate transpose Q∗ of Q), and

6. 6 RAVINDER SINGH, IMS13114
U is an upper triangular matrix, which is called a Schurform of A.
Since U is similar to A, it has the same multiset of eigenvalues, and since it is trian-
gular, those eigenvalues are the diagonal entries of U.
Theorem 1.33. Matrix polar decomposition : The polar decomposition of a square
complex matrix A is a matrix decomposition of the form A = U ∗ P, where U is a
unitary matrix and P is a positive-semidefinite Hermitian matrix.
Intuitively, the polar decomposition separates A into a component that stretches the
space along a set of orthogonal axes, represented by P, and a rotation (with possible
reflection) represented by U. The decomposition of the complex conjugate of A is given
by A = U * P. This decomposition always exists; and so long as A is invertible, it is
unique, with P positive-definite.
Theorem 1.34. Singular V alue Decomposition : Suppose M is a m∗n matrix whose
entries come from the field K, which is either the field of real numbers or the field of
complex numbers. Then there exists a factorization of the form M = U ∗ Σ ∗ V ∗, where
U is an m m unitary matrix over K (orthogonal matrix if K = R), is a m∗n diagonal
matrix with non-negative real numbers on the diagonal, and the n ∗ n unitary matrix
V ∗ denotes the conjugate transpose of the n ∗ n unitary matrix V . Such a factorization
is called a singular value decomposition of M.
The diagonal entries σi of Σ are known as the singular values of M. A common
convention is to list the singular values in descending order. In this case, the diagonal
matrix Σ is uniquely determined by M (though the matrices U and V are not).
Theorem 1.35. If E is a subset of Rn and S is a linear transformation on Rn, then
the volume of S(E) must equal the volume of E multiplied by the absolute value of the
product of the eigenvalues of S, counting multiplicity.

7. SUMMER PROJECT REPORT 7
References
1. S. Axler, Down with determinants!, The American Mathematical Monthly, Vol. 102(2)(1995), 139-
154.
2. S. Axler, Linear Algebra Done Right, 3rd edition, Springer verlag, 2010.
3. Gilbert Strang, Introduction to Linear Algebra, 4th edition, Wellesley Cambridge Press, 2009.
4. Gilbert Strang, Linear Algebra 18.06, ocw.mit.edu(Video Lectures), fall 1999.
Indian Institute of Science Education and Research,Thiruvananthapuram-695016,Kerala,India
E-mail address: ravi-singh5851113@iisertvm.ac.in