Quantum Mechanics Term 1
Contents
- 1 Elements of Linear Algebra
- 2 Principles of Quantum Mechanics
- 3 Time Evolution of Quantum Mechanical states
- 4 Generalisation to infinite dimensional Hilbert spaces
- 5 Simple Harmonic Oscillator
- 6 Symmetries in Quantum Mechanics
1 Elements of Linear Algebra
1.1 First Principle of Quantum Mechanics
A quantum mechanical state is represented by a unit vector inside a Hilbert space .
If and , then .
1.2 Notation
-
•
For vectors in , we denote them using kets i.e. .
-
•
For the null vector, we use , we can also say that
-
•
For two vectors , we write the inner product as
The properties of the inner product reformulated in this notation is as follows:
-
1.
-
2.
-
3.
iff
-
4.
If , then ,
1.3 Expansion in Orthonormal basis and Column notation
Suppose that and given a basis . We write
for any and .
If the basis is orthornormal i.e. , then we can find the constants using the inner product
Hence, for the product of two vectors ,
Relating this to our matrix notation, it is intuitive for us to write as a row vector i.e.
so that works exactly like matrix multiplication.
1.4 Dual Space
The dual space of is the set of linear functionals s.t.
With this definition, it is clear that is a linear space itself since if and for any
The dimension of is since in order to uniquely determine an element we need to know how the basis gets mapped to under . Hence, .
Given a basis of , we can always choose a dual basis of s.t. .
So for .
1.5 Adjoint Vector
Given an inner product, we can define the adjoint of a vector . The defining property is
We call the inner product a braket as it is a product between the adjoint (a ’bra’) and the vector (a ket).
If , we find using the above definition.
It follows that the dual basis of an orthonormal basis is .
As a consequence, we have that if
-
1.
-
2.
1.6 Linear Operators
A linear operator is a linear map .
Similar to the discussion in dual spaces, the linear operator is completely fixed by its action on a basis . Let . Then .
Since is a basis, there has to be s.t.
We can also represent linear operators in matrix form. For any ,
In order to find these , we again use the inner product with the basis vectors.
Example:
Consider the operator which is a counterclockwise rotation by around on s.t. is the x-axis, is the y-axis and is the z-axis.
First we determine the action on these basis vectors.
whereby is any vector in with basis .
For the matrix form,
is the inverse of iff .
Also, for any , we can define a linear operator . Then for any ,
1.6.1 Projection Operators
The defining property of a projection operator is .
Hence an obvious example is the identity operator.
Consider .
We can decompose any s.t. whereby and .
Therefore we define ,
We can generalise this to an arbitrary .
If , we define s.t.
Example:
If the subspace is spanned by the orthonomal basis , . Then,
Proof:
Consider the complement i.e. whereby is spanned by with the following relations,
i.e. form an orthonormal basis for and for any ,
We now need to check that the property for projection operators hold under this definition.
There is an important special case that we will use often in the future which is the case that , then
We can now use the projection operators to find probabilities of observing a specific state.
Example:
Consider a quantum mechanical coin whereby . For ,
Alternatively,
1.6.2 Adjoint Operator
The adjoint of satisfies the following definition
for .
If , then
Also, under matrix representation, is the complex conjugate of the transpose matrix of since,
1.6.3 Hermitian Operators and the Eigenvalue Problem
An operator is Hermitian if .
The eigenvalue problem for is the set of all and s.t. whereby is the eigenvalue and .
The eigenvectors of a given value form a linear subspace , the eigenspace corresponding to the eigenvalue .
If , then is non-degenerate.
If , then is called the degeneracy of the eigenvalue
Properties of eigenvalues of Hermitian Operators:
-
1.
The eigenvalues of Hermitian are all real numbers
-
2.
The different eigenspaces of are orthogonal, i.e. for and , then
-
3.
The eigenvectors of can be used to construct an orthonormal basis of . If are all the different eigenspaces of , then
This is known as the completeness of eigenvectors of Hermitian Operators
Proof:
1) If , we want .
Consider and take the complex conjugate.
2) By definition, and . Do the same as the proof above but by considering .
In order to solve the eigenvalue problem, we need to solve .
For to be non-trivial, we need to be non-invertible i.e. having determinant 0.
Hence the roots of the characteristic polynomial are the eigenvalues.
Since has degree , we can say
whereby we have different eigenvalues with corresponding multiplicity and .
Also .
1.7 Commutator
We can define the product between two linear operators through composition i.e. .
However the order matters and the commutator measures the difference,
Identities:
-
1.
-
2.
-
3.
1.8 Simultaneous Diagonalisation of two Hermitian Operators
Theorem:
If and are two commuting Hermitian Operators i.e. , then there exists a common eigenbasis or a common eigenspace decomposition.
Suppose that i.e. . Then,
This means that and that . Hence,
whereby is the number of distinct eigenvalues of appearing in the eigenspace of .
In essence, this means that after we have decomposed using the eigenvalues of , if we can find a commuting operator we can further decompose in the eigenspaces of .
This helps in reducing the dimension of each individual eigenspace.
Example:
Consider
They are obviously Hermitian since they only have real entries and are symmetric. One can check that they commute by doing matrix multiplication. Also, we are given that has 2 eigenvalues with the corresponding eigenspaces:
Although there is not much point in decomposing the first eigenspace since it is non-degenarate already, we can see that .
So now, the main issue is we want to solve the eigenvalue problem of in the eigenspace in order to decompose it. I.e. we want to show,
In order to solve the eigenvalue problem, we need the new matrix form of constrained in this space.
Solving for its eigenvalues we have that they are with corresponding eigenvectors .
Note that we chose the constants s.t. the eigenvector is a unit vector.
Hence, we have decomposed ,
1.9 Spectral Decomposition
If is Hermitian and are its eigenvalues, then
is the projection operator on the eigenspace .
Proof:
For any ,
If is a function admits a taylor expansion. Then
Proof: We know that since different eigenspaces are orthogonal, we must have
(1) |
From equation (1) we more generally have that
(2) |
for any two natural numbers and .
where we used equation (2) in expanding .
Properties:
1) are also eigenspaces of
2) The corresponding eigenvalues are
Proof:
2 Principles of Quantum Mechanics
-
1.
Quantum, states form a Hilbert space with physical states having norm 1, i.e. .
-
2.
Physical observables are represented by Hermitian Operators .
-
3.
A measurement of given a state can only return one of its eigenvalues . The probability is .
-
4.
Right after a measurement which yielded the eigenvalue , the state collapses to
is the normalising constant
In order for to make sense, we need .
We can also derive the formula for expectation.
More generally,
It is also important to note that measurements cannot completely fix a quantum mechanical state.
Suppose that in the case of the QM coin, .
The overall phase is not important but a relative phase remains unfixed.
3 Time Evolution of Quantum Mechanical states
Quantum Mechanical systems have a special Hermitian operator , the Hamiltonian operator which obeys the Schrodinger equation.
(3) |
However, we assume that we are dealing with isolated systems i.e. is independent of time. Also, we need to find s.t. it satisfies Schrodinger’s equation and . We try the following,
Note that in general .
We can represent in terms of the eigenbasis of .
Method 1:
We use spectral decomposition of which has eigenvalues and are the respective projection operations.
If is an orthonormal basis of , then
Method 2:
The initial state can be written as a linear combination of the eigenbasis,
Recall that,
We also observe that if the initial state is an eigenvector of with eigenvalue , then .
3.1 Time evolution of expectation values
Ehrenfest’s Theorem:
(4) |
Proof:
Observations:
We can write .
This means that the norm of a state remains invariant under time evolution.
If commutes with , then is independent of time and is a conserved quantity. Hence, operators that commute with the Hamiltonian has its expectation being a conserved quantity and can be simultanueously diagonalised with .
Example: (Spin in an external magnetic field)
An electron has a two-dimensional ’internal’ spin Hilbert space. It is described by 3-pauli matrices . They satisfy
We study the time evolution of spin according to
We choose a magnetic field s.t. , hence .
This means that does not change with time i.e. conserved.
So we have two couple ODEs which we can solve simultaneously.
The constants are fixed by initial conditions and we have showed that the expectations transform in time by a rotation of .
3.2 Unitary operators
Any operator s.t. is called unitary. Unitary operators preserve the inner product.
Suppose that .
These operators generalise rotation in Euclidean space.
We saw before that the norm of the state does not change under time evolution.
We can show that time evolution is unitary in general by writing and demanding that it satisfies Schrodinger’s equation.
We now consider the derivative of .
This means that does not change in time.
Since .
4 Generalisation to infinite dimensional Hilbert spaces
4.1 Position Operator
So far we are given an observable and knowing the possible outcomes gives us the spectrum (eigenvalues), there is a complete basis which is also an eigenbasis of . Hence, for a particle on the real line, we have a Hermitian position operator with its discrete spectrum and corresponding eigenbasis .
We promote to eigenvalues and . We should also have . The s are now a complete basis.
whereby we denote is the wavefunction which will be useful in expressing the probability of finding the eigenvalue in a measurement of .
By extension, the probability to find the particle between and is
To discuss the continuum limit we segment the line of length by taking intervals of length i.e. .
This is problematic, we show this by considering an arbitrary vector
Hence in the limit, the norm blows up and is not normalisable. So, we modify the inner product to
More generally, for the inner product
This modifications mean that there is change for the product between basis vectors . When we take , we are replacing the discrete by a continuous and .
The are still eigenvalues for a Hermitian . Also, for .
We now need to check for the case that by using completeness,
The probability to find the particle in the interval wiil then be
Note that we derived the expression for the expectation by considering as the probability density for position.
Alternatively,
4.2 Momentum Operator
In classical mechanics, momentum is the generator of translations.
Suppose that translates states by If , . Then
We observe that forms a group since and .
For small translations,
is defined as the generator of the translation. So we should look for a constant s.t. and it should be hermitian while having dimensions of momentum i.e. .
For hermiticity, we check the adjoint by considering its matrix representation.
Hence, is anti-Hermitian so for we should have so that
Since has units , we need to have units . Hence we choose .
(5) |
Last year, we had which is the same expression as the equation above.
Also, it makes sense to write but no sense to write but rather .
Now we need to inspect the commutator between the position and momentum operators.
For finite translations , we break into small pieces of and use the group property.
generates translations in space whereas before we have shown that generates translations in time. Hence similarly, is unitary since .
4.2.1 Momentum Eigenbasis
We want to find the eigenbasis s.t. it satisfies
In order to find we express it in terms of .
So we need to find
We fix the constants by demanding that the eigenbasis has unit norm.
4.3 The position and momentum representations
We have the and eigenbasis. So we can write any vector in these bases,
whereby is the analogue of s.t. is the probability density of momentum i.e.
(6) |
We can also examine how the operators act under these eigenbases.
4.4 Canonical Quantisation of Classical Systems
The independent and classical variables become in Quantum Mechanics.
The classical observables become Hermitian operators which can be constructed by substituting .
For 2 dimensions , and should be degenerate i.e.
In higher dimensions, we have cartesian coordinates with .
There are also the corresponding momentum operators satisfying
Examples:
The Hamiltonian of a particle in a 3D potential . The classical Hamiltonian is
Promoting to operators in the potential term is unambiguous since as operators they all commute with each other. E.g. if , then . However, in the following example, the choice matters.
The problem arises in the second term since as operators they do not commute. Do we choose or ? Actually we choose to preserve Hermiticity.
4.5 Schrodinger’s Wave Mechanics
Schrodinger’s wave equation of a 3D particle in under Dirac’s formalism is
(7) |
In Schrodinger’s wave mechanics, we want to describe the wave function . So we multiply the equation by .
(8) |
For the second term, we used the fact that and taking the adjoint. Also, we used the lemma that if is real then .
5 Simple Harmonic Oscillator
The Hamiltonian for a 1D particle is
(9) |
We are interested in the energy spectrum of the SHO, so we want to solve the eigenvalue problem . However, before we do that, we can deduce that all the eigenvalues of are non-negative.
We want to show that is a positive definite operator i.e. .
First we consider a general Hermitian operator .
Choose and
To solve the eigenvalue problem we need the creation and annihilation operators.
We can then use these to redefine .
For the last equation we used the fact that . Using these equations,
(10) |
So we just need to solve since which means .
We can also derive the commutator between with .
Using these, we can observe an interesting fact.
This shows that is an eigenvector of with eigenvalue . This implies that as there is no degeneracy in 1D.
As we can always apply to lower the eigenvalue, this means we can make the eigenvalues of negative. However, this contradicts with our previous finding that all the eigenvalues of are non-negative. Hence there has to be a ground state s.t. .
Opposite of , raises the eigenvalues and all the eigenstates of can be found using . These form the eigenbasis of .
We fix the constants by demanding unit norm as usual.
We used the facts that and which is easily proven by induction.
Now we just need to find the eigenvalues of from these eigenvectors.
5.1 The 2-D Isotropic Simple Harmonic Oscillator
The last equation suggests that we can diagonalise simultaneously w.r.t. , so we label the eigenbasis of with two integers and define the corrresponding creation-annihilation operators.
Our new ground state now satisfies
Our general eigenvector will have the following form and corresponding eigenvalue,
This shows that the 2D oscillator is degenerate.
5.2 The Fermionic Simple Harmonic Oscillator
The SHO we considered before are Bosonic which means that we can add as many ’quanta’ as we want by acting on the ground state. This is different for the fermionic case which obeys the following principle:
Pauli’s Exclusion Principle:
No two fermionic quanta can occupy the same state
The Hamiltonian for this system is
There is one thing we can note about the operators ,
Analagously,
Let and consider .
This means that the eigenvalues of can only be 0 or 1, since
So this implies that the eigenvalues of can only be or . This is exactly Pauli’s exclusion principle as it allows for different states to have different quantas.
However, in principle we could have degeneracy i.e. we have more than one states with with . Yet here we assumed non degeneracy to simplify the system.
So we have that is an eigenvector of with eigenvalue 0.
So it only remains to check and to complete the system.
6 Symmetries in Quantum Mechanics
Suppose is a one-parameter family of Unitary operators and which acts on our states . This process is called active transformation.
An example was in the case of translations whereby
Close to the identity , we can expand using Taylor’s approximation
whereby is the generator. If we demand that is unitary, then
This means that the generator has to be Hermitian.
We want to ask for any observable , how does its expectation transform
This is called passive transformation whereby we transform the operator.
The basic operators are which transform like
Hence for any observable, we transform them as follows
Example:
Take .
In fact in general this is true for any series:
How does the position and momentum operator transform under a translation. Before we address this question, we need the following two facts:
The first can be proven through induction whereas the prove of the second can be proven as follows:
And this is what we expect, by translating we are adding to the position of the particle.
Again, translation does not affect the momentum of the particle.
Definition:
The operator is invariant under the transformation if .
Under this assumption, if is the generator of the transformation , then .
Proof:
For ,
Noether’s Theorem in Quantum Mechanics
If is invariant under a family and is the generator, then defines a conserved quantity.
Proof:
Since , this means that have a common eigenbasis and they can be diagonalised simultaneously.
A basis of eigenstates will be labelled by the eigenvalues of and
For a third operator, we need . Then we can have our vectors labelled as .
6.1 Rotations in 2-D
Classically for an anticlockwise rotation,
We translate this to Quantum Mechanics by considering the action of rotation on the position eigenstates .
We are mostly interested in the generator of the transformation . Under small angle appoximations,
So we need an expression for in terms of .
This corresponds to classical angular momentum which we have shown to be the generator of rotations.
As a sanity check, we can see if commutes with rotationally symmetric Hamiltonians.
This gets rather complicated so we evaluate the commutator by parts.
Since this a possible measurement, we are also interested in its eigenvalues.
This is a PDE whch we can solve by going to polar coordinates.
We have one constraint on which is .
This means that the eigenvalues of the angular momentum are quantised.