Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
adjoints_and_hermitian_operators [2021/02/24 07:20] – created adminadjoints_and_hermitian_operators [2022/10/06 00:45] (current) – [Projection Operators] admin
Line 16: Line 16:
 Note, in mathematics, the Hermitian adjoint is usually denoted $^*$.  This makes sense as, in many ways it is the generalization of complex conjugation to operators.  However, in physics we normally reserve $^*$ for taking the operator formed by taking the complex conjugates of each of its elements in a given basis.  This is not the same thing as the Hermitian adjoint.  We will discuss how to represent operators in a basis in section 2.iv. Note, in mathematics, the Hermitian adjoint is usually denoted $^*$.  This makes sense as, in many ways it is the generalization of complex conjugation to operators.  However, in physics we normally reserve $^*$ for taking the operator formed by taking the complex conjugates of each of its elements in a given basis.  This is not the same thing as the Hermitian adjoint.  We will discuss how to represent operators in a basis in section 2.iv.
  
-You should now do the first in class activity to determine some properties of Hermitian adjoints.+===== Properties of Hermitian Adjoints ===== 
 + 
 +In an in class activity, you will show that 
 +\[\left ( \hat{A}^{\dagger}\right )^{\dagger} = \hat{A},\] 
 +and 
 +\[\left ( \hat{A}\hat{B} \right )^{\dagger} = \hat{B}^{\dagger}\hat{A}^{\dagger},\] 
 +so taking adjoints reverses the order of products. 
 + 
 +===== Taking Adjoints of General Expressions =====
  
 It is convenient to define $\left ( \ket{\psi} \right )^{\dagger} = \bra{\psi}$ and $\left ( \bra{\psi} \right )^{\dagger} = \ket{\psi}$.  Then, we can write things like It is convenient to define $\left ( \ket{\psi} \right )^{\dagger} = \bra{\psi}$ and $\left ( \bra{\psi} \right )^{\dagger} = \ket{\psi}$.  Then, we can write things like
Line 37: Line 45:
 \[\left ( \hat{A}\hat{B}\hat{C}\hat{D} \ket{\psi}\right )^{\dagger} = \bra{\psi}\hat{D}^{\dagger}\hat{C}^{\dagger}\hat{B}^{\dagger}\hat{A}^{\dagger}\] \[\left ( \hat{A}\hat{B}\hat{C}\hat{D} \ket{\psi}\right )^{\dagger} = \bra{\psi}\hat{D}^{\dagger}\hat{C}^{\dagger}\hat{B}^{\dagger}\hat{A}^{\dagger}\]
 \[\left ( \ketbra{\psi}{\phi} \right )^{\dagger} = \ketbra{\phi}{\psi}\] \[\left ( \ketbra{\psi}{\phi} \right )^{\dagger} = \ketbra{\phi}{\psi}\]
 +
 +===== Operators Acting Inside Bras and Kets =====
 +
 +It is convenient to define the action of operators inside bras and kets as
 +\begin{align*}
 +\ket{a\hat{A}\psi} & = a\hat{A}\ket{\psi}, & \bra{a\hat{A}\psi} = a^* \bra{\psi}\hat{A}^{\dagger}.
 +\end{align*}
 +In other words, you can think of the symbol $\bra{\text{expression}}$ as an instruction to take the Hermitian adjoint of $\text{expression}$.
 +
 +With this convention, we have
 +
 +\[\sand{\phi}{\hat{A}}{\psi} = \braket{\hat{A}^{\dagger}\phi}{\psi} = \braket{\phi}{\hat{A}\psi}.\]
 +
 +Personally, I tend to avoid writing operators inside of bras and kets as much as possible because it can get confusing to keep track of where the $^{\dagger}$'s have to go.  In rare cases, the notation is useful though.
 +
 +===== Hermitian and Skew-Hermitian Operators =====
 +
 +An operator $\hat{A}$ is //**Hermitian**// if
 +\[\hat{A}^{\dagger} = \hat{A},\]
 +or equivalently
 +\[\sand{\phi}{\hat{A}}{\psi} = \sand{\psi}{\hat{A}}{\phi}^*,\]
 +for all vectors $\ket{\psi}$, $\ket{\phi}$.
 +
 +It is //**skew-Hermitian**// or //**anti-Hermitian**// if
 +\[\hat{A}^{\dagger} = -\hat{A},\]
 +or equivalently
 +\[\sand{\phi}{\hat{A}}{\psi} = -\sand{\psi}{\hat{A}}{\phi}^*,\]
 +for all vectors $\ket{\psi}$, $\ket{\phi}$.
 +
 +Note that this implies that the expectation value of a Hermitian operator is always real, since
 +\[\sand{\psi}{\hat{A}}{\psi} = \sand{\psi}{\hat{A}}{\psi}^*.\]
 +It should come as no surprise that the operators representing physical quantities like position, momentum, and energy are all Hermitian, since their expectation values need to be real numbers.
 +
 +Similarly, the expectation value of an anti-Hermitian operator is always imaginary, since
 +\[\sand{\psi}{\hat{A}}{\psi} = -\sand{\psi}{\hat{A}}{\psi}^*.\]
 +
 +For any operator, $\hat{A}$, we can define its //**Hermitian part**// as
 +\[\hat{A}_h = \frac{\hat{A} + \hat{A}^{\dagger}}{2},\]
 +which is a Hermitian operator, and its //**anti-Hermitian part**// as
 +\[\hat{A}_a = \frac{\hat{A} - \hat{A}^{\dagger}}{2i},\]
 +which is also a Hermitian operator.
 +
 +Then, the operator $\hat{A}$ can be written as
 +\[\hat{A} = \hat{A}_h + i\hat{A}_a.\]
 +You should think of this decomposition as analogous to writing a complex scalar as
 +\[a = \text{Re}(a) + i\text{Im}(a).\]
 +
 +Note, the imaginary part of a complex number is a real number (not an imaginary number) and the anti-Hermitian part of an operator is Hermitian (not anti-Hermitian).  This terminology may be confusing, but at least it is consistent between complex numbers and operators. 
 +===== Projection Operators =====
 +
 +Let $V$ be a subspace of a Hilbert space $\mathcal{H}$.  The //**projection operator**// or //**projector**// $\hat{P}_V$ onto $V$ is defined via
 +\begin{align*}
 +\hat{P}_V \ket{\psi} & = \ket{\psi},\,\,\text{for all }\ket{\psi} \in V, \\
 +\hat{P}_V \ket{\psi} & = \boldsymbol{0},\,\,\text{for all }\ket{\psi} \in V^{\perp},
 +\end{align*}
 +where $V^{\perp}$ is the orthogonal complement of $V$ in $\mathcal{H}$.
 +
 +In order to completely define the operator, we need to say what it does to vectors that are neither in $V$ or $V^{\perp}$.  However, since $\mathcal{H} = V\oplus V^{\perp}$, we can write any vector $\ket{\psi}$ as $\ket{\psi} = a\ket{\phi} + b\ket{\phi^{\perp}}$, where $\ket{\phi}\in V$ and $\ket{\phi^{\perp}}\in V^{\perp}$.  Then, if we want $\hat{P}_V$ to be a //linear// operator, we must have
 +\[\hat{P}_V \ket{\psi} = \hat{P}_V \left ( a\ket{\phi} + b\ket{\phi^{\perp}} \right ) = a\hat{P}_V\ket{\phi} + b\hat{P}_V \ket{\phi^{\perp}} = a\ket{\phi}.\]
 +In other words, the projection operator $\hat{P}_V$ simply returns the component of the vector $\ket{\psi}$ that lies in the subspace $V$.
 +
 +A trivial example of a projection operator is the projector $\hat{P}_{\mathcal{H}}$ onto the entire Hilbert space $\mathcal{H}$.  Since the orthogonal complement of $\mathcal{H}$ only consists of the zero vector $\boldsymbol{0}$, the projector $\hat{P}_{\mathcal{H}}$ maps every vector $\ket{\psi}$ to itself.  In other words, it is just the identity operator $\hat{P}_{\mathcal{H}} = \hat{I}$.  
 +
 +Another trivial example of a projector is $\hat{0}$ which just multiplies a vector by the scalar $0$, i.e. 
 +\[\hat{0}\ket{\psi} = 0 \ket{\psi} = \boldsymbol{0},\]
 +for all vectors $\ket{\psi}\in\mathcal{H}$.  This is the projector onto the zero-dimensional subspace that just consists of the zero vector $\boldsymbol{0}$.  This subspace is the orthogonal complement of $\mathcal{H}$.
 +
 +We already know that, if $\ket{\phi_1}, \ket{\phi_2},\cdots$ is an orthonormal basis for $\mathcal{H}$ then
 +\[\hat{I} = \sum_j \proj{\phi_j}.\]
 +
 +This result can be extended to arbitrary projectors.  If $\ket{\phi_1},\ket{\phi_2},\cdots$ is an orthonormal basis for a //subspace// $V$ of $\mathcal{H}$ then
 +\[\hat{P}_V = \sum_j \proj{\phi_j}.\]
 +To see this, let note that any vector $\ket{\psi}$ can be written as $\ket{\psi} = \ket{\phi} + \ket{\phi^{\perp}}$ where $\ket{\phi} \in V$ and $\ket{\phi^{\perp}} \in V^{\perp}$.  Since all of the basis vectors $\ket{\phi_j}$ are in $V$, they must satisfy $\braket{\phi_j}{\phi^{\perp}} = 0$.  Also, since $V$ is itself a Hilbert space $\hat{P}_V = \sum_j \proj{\phi_j}$ is the identity operator on $V$, so $\hat{P}_V \ket{\phi} = \ket{\phi}$.
 +
 +All projection operators are Hermitian $\hat{P}_V^{\dagger} = \hat{P}_V$ and //**idempotent**//, which means $\hat{P}_V^2 = \hat{P}_V$.  In fact, the converse is also true: //all// operators that are Hermitian and idempotent are projectors, so it is common to //define// a projection operator to be an idempotent Hermitian operator.  We shall prove the converse in section 2.v, but for now we will show that projectors have these two properties.
 +
 +To show that $\hat{P}_V$ is Hermitian, let $\ket{\psi_1}$ and $\ket{\psi_2}$ be two vectors in $\mathcal{H}$, we can write them both as
 +\begin{align*}
 +\ket{\psi_1} & = \ket{\phi_1} + \ket{\phi_1^{\perp}}, & \ket{\psi_2} & = \ket{\phi_2} + \ket{\phi_2^{\perp}},
 +\end{align*}
 +where $\ket{\phi_1},\ket{\phi_2} \in V$ and $\ket{\phi_1^{\perp}},\ket{\phi_2^{\perp}} \in V^{\perp}$.  Then,
 +\begin{align*}
 +   \sand{\psi_1}{\hat{P}_V}{\psi_2} & = \left ( \bra{\phi_1} + \bra{\phi_1^{\perp}} \right ) \hat{P}_V \left (\ket{\phi_2} + \ket{\phi_2^{\perp}} \right ) \\
 +   & = \sand{\phi_1}{\hat{P}_V}{\phi_2} + \sand{\phi_1}{\hat{P}_V}{\phi_2^{\perp}} + \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2} + \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2^{\perp}}. 
 +\end{align*}
 +Now, the terms $\sand{\phi_1}{\hat{P}_V}{\phi_2^{\perp}}, \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2}$ and $\sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2^{\perp}}$ are all zero because $\ket{\phi_1^{\perp}}$ and $\ket{\phi_2^{\perp}}$ are in a subspace orthogonal to $V$, so $\hat{P}_V\ket{\phi_1^{\perp}} = \hat{P}_V \ket{\phi_2^{\perp}} = 0$.  In particular, since the complex conjugate of zero is zero, this means that
 +\begin{align*}
 +  \sand{\phi_1}{\hat{P}_V}{\phi_2^{\perp}} & = \sand{\phi_2^{\perp}}{\hat{P}_V}{\phi_1}^*, &
 +  \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2} & = \sand{\phi_2}{\hat{P}_V}{\phi_1^{\perp}}^*, &
 +   \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2^{\perp}} & = \sand{\phi_2^{\perp}}{\hat{P}_V}{\phi_1^{\perp}}^*.
 +\end{align*}
 +In addition
 +\begin{align*}
 +   \sand{\phi_1}{\hat{P}_V}{\phi_2} & = \braket{\phi_1}{\phi_2} \\
 +   & = \braket{\phi_2}{\phi_1}^* \\
 +   & = \sand{\phi_2}{\hat{P}_V}{\phi_1}^*.
 +\end{align*}
 +This is because $\ket{\phi_1},\ket{\phi_2}\in V$, so $\hat{P}_V \ket{\phi_1} = \ket{\phi_1}$ and $\hat{P}_V \ket{\phi_2} = \ket{\phi_2}$.
 +
 +All together then, we have
 +\begin{align*}
 +  \sand{\psi_1}{\hat{P}_V}{\psi_2} & =\sand{\phi_1}{\hat{P}_V}{\phi_2} + \sand{\phi_1}{\hat{P}_V}{\phi_2^{\perp}} + \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2} + \sand{\phi_1^{\perp}}{\hat{P}_V}{\phi_2^{\perp}} \\
 +  & = \sand{\phi_2}{\hat{P}_V}{\phi_1}^* + \sand{\phi_2^{\perp}}{\hat{P}_V}{\phi_1}^* + \sand{\phi_2}{\hat{P}_V}{\phi_1^{\perp}}^* + \sand{\phi_2^{\perp}}{\hat{P}_V}{\phi_1^{\perp}}^* \\
 +  & = \left [ \left (  \bra{\phi_2} + \bra{\phi_2^{\perp}}\right ) \hat{P}_V \left ( \ket{\phi_1} + \ket{\phi_1^{\perp}}\right ) \right ]^* \\
 +  & = \sand{\psi_2}{\hat{P}_V}{\psi_1}^*,
 +\end{align*}
 +so $\hat{P}_V^{\dagger} = \hat{P}_V$.
 +
 +To prove that projection operators are idempotent, write an arbitrary vector $\ket{\psi}$ as
 +\[\ket{\psi} = \ket{\phi} + \ket{\phi^{\perp}},\]
 +with $\ket{\phi} \in V$ and $\ket{\phi^{\perp}} \in V^{\perp}$.  Then we have
 +\[\hat{P}_V \ket{\psi} = \ket{\phi},\]
 +but also
 +\[\hat{P}_V^2 \ket{\psi} = \hat{P}_V \hat{P}_V \ket{\psi} = \hat{P_V} \ket{\phi} = \ket{\phi},\]
 +since $\ket{\phi} \in V$.  Since this is true for an arbitrary vector $\ket{\psi}$, we have $\hat{P}_V^2 = \hat{P}_V$.
 +
 +We conclude with some more basic properties of projectors.
 +  * If two projectors $\hat{P}_1$ and $\hat{P}_2$ commute, then $\hat{P}_1\hat{P}_2$ is also a projector, since
 +  \[\left ( \hat{P}_1 \hat{P}_2\right )^{\dagger} = \left ( \hat{P}_2 \hat{P}_1\right )^{\dagger} = \hat{P}_1^{\dagger} \hat{P}_2^{\dagger} = \hat{P}_1\hat{P}_2\]
 +  \[\left ( \hat{P}_1 \hat{P}_2\right )^2 = \hat{P}_1\hat{P}_2\hat{P}_1\hat{P}_2 = \hat{P}_1\hat{P}_1\hat{P}_2\hat{P}_2 = \hat{P}_1^2 \hat{P}_2^2 = \hat{P}_1 \hat{P}_2.\]
 +  In fact, $\hat{P}_1\hat{P}_2$ is the projector onto the intersection of the subspaces that $\hat{P}_1$ and $\hat{P}_2$ project onto.
 +  * Two projectors are //**orthogonal**// if $\hat{P}_1\hat{P}_2 = 0$ (this is a special case of two commuting projectors).  It means that the subspaces that they project onto are orthogonal.
 +  * A sum of two orthogonal projectors is a projector, since
 +  \[\left ( \hat{P}_1 + \hat{P}_2 \right )^{\dagger} = \hat{P}_1^{\dagger} + \hat{P}_2^{\dagger} = \hat{P}_1 + \hat{P}_2\]
 +  \[\left ( \hat{P}_1 + \hat{P}_2 \right )^2 = \hat{P}_1^2 + \hat{P}_2^2 + \hat{P}_1\hat{P}_2 + \hat{P}_2\hat{P}_1 = \hat{P}_1 + \hat{P}_2.\]
 +  The projection operator $\hat{P}_1 + \hat{P}_2$ projects onto the linear span of the subspaces that $\hat{P}_1$ and $\hat{P}_2$ project onto.
 +  * A sum of non-orthogonal projectors is not a projector.
 +
 +{{:question-mark.png?direct&50|}}
 +====== In Class Activities ======
 +
 +  - **Properties of the Hermitian Adjoint**
 +    - Prove that $\left ( \hat{A}^{\dagger} \right )^{\dagger} = \hat{A}$.
 +    - Prove that $\left ( \hat{A}\hat{B} \right )^{\dagger} = \hat{B}^{\dagger}\hat{A}^{\dagger}$.\\ 
 +    HINT: You will find it useful to do the Dirac notaty.  Insert an identity operator somewhere and decompose it as $\hat{I} = \sum_j \proj{e_j}$ for some orthonormal basis $\ket{e_j}$.