Lecture 2: State space and vector spaces

Rajat Mittal

IIT Kanpur

We will start by looking at the first question? How do we represent data/information stored inside a
quantum computer/system. In other words, how do we mathematically capture the state of a quantum
computer at any instance of the computation.
Exercise 1. How do we capture the state for a classical computer?
Simply, the state of a computer can be represented by a sequence of 0’s and 1’s, where each such number
is called a bit.
For the quantum case, the answer is given by the 1st and 4th postulate of quantum mechanics. To
understand these postulates, we will need to go through few concepts in linear algebra first. These notes will
assume that the reader is familiar with the concept of vector space, basis and linear independence. Strang’s
book, Linear Algebra and its applications, is a good source to brush up these concepts.
This exposition will focus on vectors and matrices. Dirac’s notation is used widely in quantum computing
to represent these linear algebraic quantities, because it simplifies the understanding of quantum mechanical
concepts. We will switch between the standard vector notation and Dirac notation in these notes.
Exercise 2. Read about vector space, basis, and linear independence if you are not comfortable with these
words.

1 Vector spaces
One of the most fundamental concept in linear algebra is that of a vector space. You must have seen vector
spaces over real numbers. The vector space Rn (dimension n) consists of vectors with n coordinates such
that each coordinate is a real number.
We will mostly concern ourselves with the vector space Cn , the vector space of dimension n over the field
of complex numbers. The vectors in Cn are going to be our states of a quantum computer. This means that
the scalars used in these vector spaces are complex numbers, i.e., every coordinate will contain a complex
number. For example, C2 is a vector space of dimension 2. All elements in this vector space can be written
as a linear combination of two standard basis vectors,
   
1 0
e0 = & e1 = .
0 1

So, every vector v ∈ C2 can be written as α0 e0 + α1 e1 , where α0 , α1 are two complex numbers. In Dirac’s
notation these two standard basis vectors are written as e0 = |0⟩ and e1 = |1⟩.
A column vector is the most basic unit of a vector space. Using Dirac’s notation, a column vector will
be denoted by |ψ⟩. Suppose {|v1 ⟩, |v2 ⟩, · · · , |vn ⟩} is the basis of the vector space, then any vector |v⟩ can be
written as
|v⟩ = a1 |v1 ⟩ + · · · + an |vn ⟩,
where all ai ’s are complex numbers.
For a vector space with dimension n, the standard basis is denoted by |0⟩, |1⟩, · · · , |n − 1⟩. Here you can
think of |i⟩ as the vector with 1 at the (i + 1)-th position and 0 otherwise. For example, a 3-dimensional
space will have standard basis elements |0⟩, |1⟩ and |2⟩.
     
1 0 0
|0⟩ =  0  & |1⟩ =  1  & |2⟩ =  0 
0 0 1
 Remember that a bit in a classical computer is a number in {0, 1}. Analogously, a qubit in a quantum
computer is going to be a unit vector in C2 .
Exercise 3. What is a unit vector?
In other words, the state of a qubit can be written as α|0⟩ + β|1⟩, where α2 + β 2 = 1. You can think
of α, β as weights on classical states |0⟩, |1⟩ respectively. Notice that we did not say probability, more on it
later.
Exercise 4. Can you come up with a state of qubit which is not classical? Are your α, β real numbers?
Exercise 5. What is the difference between vector |0⟩ and vector 0?
There is a small difference between vector |0⟩ and vector 0 (the vector with all entries 0). First one is a
basis vector with the first entry 1 and rest 0.
The notation ⟨ψ| denotes the row vector whose entries are complex conjugate of the entries of the vector
|ψ⟩ (also known as the adjoint, |ψ⟩ = ⟨ψ|∗ ). If |ψ⟩ = (x1 x2 · · · xn )T and |ϕ⟩ = (y1 y2 · · · yn )T , the vector
space Cn is equipped with the natural inner product (like dot product),
n
X
⟨ψ|ϕ⟩ = ((x1 , x2 , · · · , xn ), (y1 , y2 , · · · , yn )) = x∗i yi .
i=1

Here x∗ denotes the complex conjugate of a complex number x. In usual vector notation, the inner product
between ψ and ϕ will be denoted by ψ ∗ ϕ.
For a vector space V ∈ Cn , its orthogonal complement V ⊥ is defined as,

V ⊥ := {w ∈ Cn : ⟨v|w⟩ = 0 ∀v ∈ V }.

In the beginning, you can convert expressions in Dirac notation to the usual vector notation. Slowly, it
might become easier to directly manipulate expressions in Dirac notation.
√ √
Exercise 6. Suppose a qubit is in state |ψ⟩, where ψ T = (1/ 2, −i/ 2). Is it a valid qubit state? Write it in
Dirac notation.

2 Operators
A quantum computer will not be very useful if it remains in the same state. It needs to move to desired
states by application of transformations (like gates in a classical computer). Since states are vectors, and all
transformations in quantum are linear, it is time to study linear operators on vectors.
Given two vector spaces, V and W over C, a linear operator M : V → W is defined as an operator
satisfying the following properties.
– M (x + y) = M (x) + M (y).
– M (αx) = αM (x), ∀α ∈ C.
These conditions imply that the zero of the vector space V is mapped to the zero of the vector space W .
Also,
M (α1 x1 + · · · + αk xk ) = α1 M (x1 ) + · · · + αk M (xk )
Where x1 , · · · , xk are elements of V and αi ’s are in C. Because of this linearity, it is enough to specify the
value of a linear operator on any basis of the vector space V . In other words, a linear operator is uniquely
defined by the values it takes on any particular basis of V .
Let us define the addition of two linear operators as (M + N )(u) = M (u) + N (u). Similarly, αM (scalar
multiplication) is defined to be the operator (αM )(u) = αM (u). The space of all linear operators from V to
W (denoted L(V, W )) is a vector space in itself. The space of linear operators from V to V will be denoted
by L(V ).

2
Exercise 7. Given the dimension of V and W , what is the dimension of the vector spaces L(V, W )?

Exercise 8. Can you think of a linear operator which takes |0⟩ to √1 (|0⟩ + |1⟩) and |1⟩ to √1 (|0⟩ − |1⟩)?
2 2

One of the issue is, how to represent a linear operator. You can give its action on all the elements in
vector space, but that will take infinite space.

2.1 Matrices as linear operators

Given two vector spaces V = Cn , W = Cm and a matrix M of dimension m × n, the operation v ∈ V →

M v ∈ W is a linear operation. Here, if |v⟩ = a1 |v1 ⟩ + · · · + an |vn ⟩, we represent it as a vector,
 
a1
 a2 
 
 .. 
 . 
an

So, a matrix acts as a linear operator on the corresponding vector space.

To ask the converse, can any linear operator be specified by a matrix?
Let f be a linear operator from a vector space V (dimension n) to a vector space W (dimension m).
Suppose {e1 , e2 , · · · , en } is a basis for the vector space V . Denote the images of elements of this basis under
f as {w1 = f (e1 ), w2 = f (e2 ) · · · , wn = f (en )}.

Exercise 9. What is the lower-bound/ upper-bound on the dimension of the vector space spanned by
{w1 , w2 , · · · , wn }?

Define Mf to be the matrix with columns w1 , w2 , · · · , wn . Notice that Mf is a matrix of dimension m × n.

It is a simple exercise to verify that the action of the matrix Mf on a vector v ∈ V is just Mf v. Here we
assume that v is expressed in the chosen basis {e1 , e2 , · · · , en }.

Exercise 10. Convince yourself that M v is a linear combination of columns of M .

Notice that the matrix Mf and the operator f act exactly the same on the basis elements of V . Since
both the operations are linear, they are exactly the same operation. This proves that any linear operation
can be specified by a matrix.

Exercise 11. Can you now think of a linear operator which takes |0⟩ to √1 (|0⟩ + |1⟩) and |1⟩ to √1 (|0⟩ − |1⟩)?
2 2

In Dirac’s notation, we denoted inner product between two vectors |ψ⟩, |ϕ⟩ by ⟨ψ|ϕ⟩. The expression
A = |ψ⟩⟨ϕ| is a matrix which takes |v⟩ to ⟨ϕ|v⟩|ψ⟩. The analog of this expression in the simple vector
notation would be, A = ψϕ∗ . If a linear operator M takes an orthonormal basis {v1 , v2 , · · · , vn } to vectors
{w1 , w2 , · · · , wn }, then the matrix representation of M is
n
X
M= |wi ⟩⟨vi |.
i=1

Exercise 12. What will be (|w⟩⟨v|)|v⟩. Prove that M defined in the equation above will have the action as
intended.

The equivalence of matrices and linear operators does not depend upon the chosen basis. We can pick
our favorite bases of V and W , and the linear operator can similarly be written in the new basis as a matrix
(The columns of this matrix are images of the basis elements of V ). In other words, given bases of V and W
and a linear operator f , it has a unique matrix representation.

3
 To compute the action of a linear operator, express v ∈ V in the preferred basis and multiply it with
the matrix representation. The output will be in the chosen basis of W . We will use the two terms, linear
operator and matrix, interchangeably in future (bases will be clear from the context). In Dirac’s notation,
action of M on a vector |v⟩ is just M |v⟩. Also, ⟨u|M v⟩ = ⟨M ∗ u|v⟩ is denoted by ⟨u|M |v⟩.
For a matrix A, AT denotes the transpose of the matrix and A∗ denotes the adjoint of the matrix (take
complex conjugate and then transpose).

Exercise 13. Why is matrix multiplication defined the way it is? Why can’t it be defined in the more natural
way of entry-wise multiplication?

Let us look at some simple matrices which will be used later.

– Zero matrix: The matrix with all the entries 0. It acts trivially on every element and takes them to the
0 vector.
– Identity matrix: The matrix with 1’s on the diagonal and 0 otherwise. It takes v ∈ V to v itself.
– All 1’s matrix (J): All the entries of this matrix are 1.

Exercise 14. What is the action of matrix J?

2.2 Extra reading: Kernel, image and rank

For a linear operator/matrix (from V to W ), the kernel is defined to be the set of vectors which map to 0.

ker(M ) = {x ∈ V : M x = 0}

Here 0 is the zero vector in space W .

Exercise 15. What is the kernel of the matrix J?

The image is the set of vectors which can be obtained through the action of the matrix on some element
of the vector space V .
img(M ) = {x ∈ W : ∃y ∈ V, x = M y}

Exercise 16. Show that img(M ) and ker(M ) are subspaces.

Exercise 17. What is the image of J?

Notice that ker(M ) is a subspace of V , but img(M ) is a subspace of W . The dimension of img(M ) is
known as the rank of M (rank(M )). The dimension of ker(M ) is known as the nullity of M (nullity(M )).
For a matrix M ∈ L(V, W ), by the famous rank-nullity theorem,

rank(M ) + nullity(M ) = dim(V ).

Here dim(V ) is the dimension of the vector space V .

Proof. Suppose u1 , · · · , uk is the basis for ker(M ). We can extend it to the basis of V : u1 , · · · , uk , vk+1 , · · · , vn .
We need to prove that the dimension of img(M ) is n − k. It can be proved by showing that the set
{M vk+1 , · · · , M vn } forms a basis of img(M ).
Exercise 18. Prove that any vector in the image of M can be expressed as linear combination of M vk+1 , · · · , M vn .
Also any linear combination of M vk+1 , · · · , M vn can’t be zero vector.

Given a vector v and a matrix P

M , it is easy to see that the vector M v is a linear combination of columns
of M . To be more precise, M v = i Mi vi where Mi is the ith column of M and vi is the ith co-ordinate of
v. This implies that any element in the image of M is a linear combination of its columns.

4
Exercise 19. Prove the rank of a matrix is equal to the dimension of the vector space spanned by its columns
(column-space).
The dimension of the column space is sometimes referred as the column-rank. We can similarly define
the row-rank, the dimension of the space spanned by the rows of the matrix. Luckily, row-rank turns out to
be equal to column-rank and we will call both of them as the rank of the matrix. This can be proved easily
using Gaussian elimination.

2.3 Operations on matrices

Lets look at some of the basic operations on these matrices.

– Trace: The trace of a square matrix A is the sum of all the diagonal elements.
X
tr(A) = A[i, i]
i

– Entry-wise multiplication: The entry-wise multiplication of two matrices is known as Hadamard product
and only makes sense when both of them have same number of rows and columns. The Hadamard product
of two matrices A, B is
(A ◦ B)[i, j] = A[i, j]B[i, j].
The related operation is when you add up the entries of this Hadamard product.
X
(A • B) = A[i, j]B[i, j]
i,j

Notice that A • B is a scalar and not a matrix.

Exercise 20. Given a matrix, express • operation in terms of multiplication and trace operation.
– Inverse: Inverse of a matrix M is the matrix M −1 , s.t., M M −1 = M −1 M = I. The inverse only exists if
the matrix has full rank (columns of M span the whole space).
Exercise 21. What is the inverse of matrix J (all 1’s matrix).
Exercise 22. Show that the inverse of a matrix exists iff it has full rank.

3 First postulate: state of a system

The postulates of quantum mechanics provide us the mathematical formalism over which the physical theory
is developed. For people studying quantum computing, it gives the basic laws according to which any quantum
system (or a quantum computer) works.
These postulates were agreed upon after a lot of trial and error. We won’t be concerned about the
physical motivation of these postulates. Most of the material for this lecture is taken from [1]. It is a very
good reference for more details.
As discussed before, in this lecture note, we are interested in postulates which allow us to represent
data/information in a quantum computer (the state of a quantum computer at an instance). The first
postulates specifies, what is meant mathematically by the state of a quantum system.

Postulate 1: A physically isolated system is associated with a Hilbert space, called the state space of the
system. The system, at a particular time, is completely described by a unit vector in this Hilbert space,
called the state of the system.

Intuitively, Hilbert space is a vector space with enough structure so that we can apply the techniques of
linear algebra and analysis on it.

5
Exercise 23. Read more about Hilbert spaces.

For this course, we will only be dealing with vector spaces over complex numbers with inner product
defined over them. In almost all these cases, the dimension is going to be finite (say n). In particular, we
will assume that our state space is Cn for some n (n is the dimension of this state space).
The simplest non-trivial state space would be C2 (dimension being 2), the state space of a qubit. Re-
member that a qubit is a generalization of bit, the way we store information in a classical computer. It
will be spanned by two standard basis vectors, |0⟩ and |1⟩. These two, |0⟩ and |1⟩, represent classical states
in a quantum computer. For an example of a non-classical state in this basis, take |+⟩ = √12 (|0⟩ + |1⟩) or
|−⟩ = √12 (|0⟩ + |1⟩). These states, |+⟩ and |−⟩, are going to be pretty useful in quantum computation.

Exercise 24. Find another basis of C2 .

Exercise 25. Write states |0⟩, |1⟩ in terms of |+⟩, |−⟩.

Any state in this system can be written as,

2 2
|ψ⟩ = α|0⟩ + β|1⟩, |α| + |β| = 1.

The coefficients, α and β, are called the amplitude. Specifically, α (β) is the amplitude of the state |ψ⟩
for |0⟩ (|1⟩) respectively. When α and β are non-zero, we say that |ψ⟩ is in superposition of states |0⟩ and
|1⟩.
The property of superposition seems to be one of the major reasons behind the power of quantum
computing (other is entanglement, described later). It allows us to compute aggregate properties of an input
much faster than the classical computer (we will see this in action very soon, Deutsch’s algorithm).

2
Note 1. Many people interpret this as, the state |ψ⟩ is in state |0⟩ with probability |α| and in state |1⟩ with
2
probability |β| . This is only a consequence of |ψ⟩ = α|0⟩ + β|1⟩ and not equivalent to it.

Exercise 26. Why is it not equivalent?

1
Exercise 27. Suppose 3 ((1 + 2i)|0⟩ + (ai)|1⟩) is a quantum state. What is the value of a?

We have talked about the case when C2 is our Hilbert space, and that case will be very useful. Though,
in general, if there are n different classical states, the quantum state would be a unit vector expressed in an
orthonormal basis {|0⟩, |1⟩, · · · , |n − 1⟩}.
Remember that the standard basis is one of the most convenient basis to represent a state, but definitely
not the only basis to represent a state. We can have any basis {|v0 ⟩, |v1 ⟩, · · · , |vn − 1⟩} and a state |ψ⟩ can
be written as,
X 2
|ψ⟩ = α0 |v0 ⟩ + α1 |v1 ⟩ + · · · + αn−1 |vn−1 ⟩, |αi | = 1.
i

We will say that the state |ψ⟩ is in superposition of basis states {|v0 ⟩, |v1 ⟩, · · · , |vn − 1⟩} (ideally using only
those states whose amplitude is non-zero).
You might already guess from the discussion that the operators on these quantum states will be matrices
(linear operators over the vector space). It will turn out that not all linear operators are allowed. Though,
that will be discussed in the next lecture (second postulate).
Before that, we have to look at one more postulate. Notice that the state of a classical computer is
described by multiple bits (not just a single bit). How can we describe state of multiple qubits? That takes
us to the concept of tensor product.

6
4 Tensor product
We have described the state of a system as a vector in a Hilbert space. What happens if we have multiple
systems. For a classical computer, the answer is pretty simple, you just describe the state of both systems
independently.
Interestingly, we can have a state of a composite quantum systems such that the individual state of the
constituent systems can’t be described. This property is known as entanglement and is the reason behind
many weird properties of quantum mechanics. To understand this phenomenon, we need to understand the
concept of tensor products.
Suppose there is a ball which can be colored blue or red. The state of a quantum ball is a unit vector in
two dimensions,
|v⟩ = α|r⟩ + β|b⟩.
Where |r⟩, |b⟩ represent the classical states, the ball being red or blue, the coefficients α, β follow the
usual law.
How about if there are two different balls. The classical states possible are |rr⟩, |rb⟩, |br⟩, |bb⟩, i.e., we take
the set multiplication of possible states of individual system.
What are the possible states if this system is quantum? By analogy with one quantum ball case, any
linear superposition of these classical states (normalized) should be a possible quantum state:
|v⟩ = α|rr⟩ + β|rb⟩ + γ|br⟩ + δ|bb⟩,
where |v⟩ is a unit vector.
This idea motivates the definition of tensor product. Given two vector spaces V, W equipped with an
inner product and spanned by the orthonormal basis v1 , v2 , · · · , vn and w1 , w2 , · · · , wm , the tensor product
V ⊗ W is the space spanned by the mn vectors (v1 ⊗ w1 ), · · · , (v1 ⊗ wn ), (v2 ⊗ w1 ), · · · , (vn ⊗ wm ).
Exercise 28. What is the dimension of space V ⊗ W ?
Formally, tensor product of two vector spaces is equipped with a bilinear map ⊗ : V × W → V ⊗ W
which satisfies the following conditions.
– Scalar multiplication: for α ∈ C, v ∈ V and w ∈ W ,
α(v ⊗ w) = (αv) ⊗ w = v ⊗ (αw).
– Linearity in the first component: for v1 , v2 ∈ V and w ∈ W ,
(v1 + v2 ) ⊗ w = v1 ⊗ w + v2 ⊗ w.
– Linearity in the second component: for v ∈ V and w1 , w2 ∈ W ,
v ⊗ (w1 + w2 ) = v ⊗ w1 + v ⊗ w2 .
For our purposes, we can define the tensor product of two vectors in a canonical way for the vector spaces
Cn and Cm . The tensor product of two vectors a = (a1 , · · · , an ) ∈ V and b = (b1 , · · · , bm ) is the vector
a ⊗ b ∈ V ⊗ W = Cmn ,  
a1 b1
 a1 b2 
 .. 
 
    . 
a1 b

 a1 bm 
 a2 b  
  .. 

a⊗b= . = .
 
 ..    . 

 .. 
an b  
 an b1 
 
 . 
 . .
an bm

7
Exercise 29. What is the tensor product a ⊗ b, where
   
1 0
0 1
a =  ,b =  
   .
0 0
0 0
In Dirac’s notation, we further simplify |v⟩ ⊗ |w⟩ to |vw⟩ when v and w are symbols. For example, state
|0⟩ ⊗ |1⟩ is same as |0⟩|1⟩ = |01⟩.
Exercise 30. What is the basis of state of two qubits? What about three qubits?
Exercise 31. Show that (v + w) ⊗ (a + b) = v ⊗ a + v ⊗ b + w ⊗ a + w ⊗ b. In other words, (|v⟩ + |w⟩)(|a⟩ + |b⟩) =
|va⟩ + |vb⟩ + |wa⟩ + |wb⟩.
We can define the inner product on the tensor product space in the natural way,

⟨a ⊗ b|c ⊗ d⟩ = ⟨a|c⟩⟨b|d⟩. (1)

In other words, we have defined the tensor product between two Hilbert spaces V and W . First, look at
them as vector spaces and define the vector space V ⊗ W . Then, we define the inner product on V ⊗ W by
the equation above (Eq. 1).
Carrying the analogy further, given two linear operators A ∈ L(V ) and B ∈ L(W ), their tensor product
A ⊗ B can be defined in the space L(V ⊗ W ). Precisely, its action is specified by,
(A ⊗ B)(a ⊗ b) = Aa ⊗ Bb.
We can extend this by linearity to define the action on the complete space V ⊗ W .
 
⊗2 1 1 1
Exercise 32. Write out the matrix representation of H = H ⊗ H where H = 2 √ is the Hadamard
1 −1
 
Matrix. What is H ⊗2 √12 (|00⟩ + |11⟩) ?

Exercise 33. Given the matrix representation of A, B; come up with the matrix representation of A ⊗ B.
You can easily verify the following properties,
– (A ⊗ B)(C ⊗ D) = AC ⊗ BD
– (A ⊗ B)∗ = A∗ ⊗ B ∗
A ⊗ B are linear operators in L(V ⊗ W ). Can there be other linear operators? P
The sum of two linear operators is a linear operator. So, any operator of the form i ci (Ai ⊗ Bi ) is also
a linear operator.
Are there any more linear operators? It turns out that these are the only linear operators in L(V ⊗ W ),
you can prove this by dimensionality argument.
The description of tensor product is given in a very simplified manner in terms of basis vectors, sufficient
for use in our course. Readers are encouraged to check out the formal definitions.

5 Fourth postulate: composite Systems

The fourth postulate deals with composite systems, systems with more than one part. We will cover second
and third postulates in the later sections. We will use tensor products for the sake of describing multiple
systems.

Postulate 4: Suppose the state space of Alice is HA and Bob is HB , then the state space of their combined
system is HA ⊗ HB . If Alice prepares her system in state |a⟩ and Bob prepares it in |b⟩, then the combined
state is |a⟩ ⊗ |b⟩, succinctly written as |ab⟩.

8
 √
3
Exercise 34. If Alice’s qubit is in state 21 |0⟩ + 2 |1⟩ and Bob’s qubit is in √1 |0⟩
2
− √i |1⟩,
2
what is the state
of the combined system?

Similarly, if operator A is applied on Alice’s system and operator B is applied on Bob’s system, then
operator A ⊗ B is applied to the combined system. This follows from the property,

(A ⊗ B)(|a⟩ ⊗ |b⟩) = A|a⟩ ⊗ B|b⟩.

Generally, it is quite clear which part of the system belongs to which party. In case of confusion, we will
use subscripts to resolve it. So if A is an operator on first system and B is an operator on second system,
the combined operator is A1 ⊗ B2 .
The most useful example will be of k qubits.

Exercise 35. What will be the dimension of the state space of k qubits?
k
The state space for k qubits is C2 . It is a vector space of dimension 2k . A natural way to represent
this vector space is through basis |0⟩, |1⟩, |2⟩, · · · , |2k − 1⟩. Though, a better way to represent the basis is by
binary strings of length k. This way, we keep the structure of k qubits and not a general vector space with
2k dimension.
To take an example, the state space of 2 qubits is spanned by |00⟩, |01⟩, |10⟩, |11⟩. The state space of k
qubits is spanned by |Bn ⟩, where Bn is the binary representation of number n and n ranges from 0 to 2k − 1.
These basis states are product states, e.g., state |001⟩ = |0⟩ ⊗ |0⟩ ⊗ |1⟩.
Notice the rise in the dimension of the state space for k qubits. If we described the state space of k bits,
it will have dimension k (over the field of two elements). For k qubits, it rises to dimension 2k over complex
numbers. There is an exponential growth, which makes it expensive to simulate quantum computer on a
classical computer.

Exercise 36. Are the two states √1 (|00⟩ + |11⟩) and |+⟩ ⊗ |+⟩ same?
2

The tensor product structure of the composite system gives rise to a very interesting property called
entanglement. As explained before, there are states in the composite system which cannot be decomposed
into the states of their constituent systems. Such states are called entangled states.
The most famous example of an entangled state is called the Bell state,
1
√ (|00⟩ + |11⟩).
2
Exercise 37. Show that the Bell state can’t be written as |ψ⟩ ⊗ |ϕ⟩.
Pl
It is clear that every state in the composite system H1 ⊗ H2 can be written as i=1 |ψi ⟩ ⊗ |ϕi ⟩ (Why?).

Exercise 38. Prove a bound of dim(H1 ) × dim(H2 ) on l for any state in the composite system.

Can you give a better bound? Read about Schmidt decomposition for a better bound. We have defined
when a state is entangled and when is it not. But how can we quantify entanglement? In other words, how
entangled is an state? These are very interesting questions and lot of research is currently being done to
answer them.

6 Assignment
Exercise 39. Prove that rank(AB) ≤ rank(A).

Exercise 40. Prove that rank(A) = rank(A∗ A) without using singular or spectral decomposition.

Hint: rank(A) ≥ rank(A∗ A) is easy. For the other direction, reduce A to its reduced row echelon form.

9
 Aij vi∗ wj .
P
Exercise 41. Show that ⟨v|A|w⟩ = ij

Exercise 42. Show that tr(A|v⟩⟨v|) = ⟨v|A|v⟩.

Exercise 43. Suppose M, N are two square matrices, show that M N = I ⇒ N M = I. Notice that it is not
true if matrix is not square, find a counterexample.

Exercise 44. Write the matrix representation of the operator which takes √1 (|0⟩+|1⟩) to |0⟩ and √1 (|0⟩−|1⟩)
2 2
to |1⟩.
|00⟩+|11⟩
Exercise 45. Read about Pauli matrices X,Y and Z. Let ψ = √
2
, calculate the value of ⟨ψ|X1 ⊗ Z2 |ψ⟩.
 √ 
1 3
Exercise 46. What is 2 |0⟩ + 2 |1⟩ ⊗ |−⟩.

References
1. M. A. Nielsen and I. L. Chuang. Quantum Computation and Quantum Information. Cambridge, 2010.

10