Phys 506 lecture 1: Spin and Pauli matrices

This lecture should be primarily a review for you of properties of spin one-half. I do suspect that some of the identities derived here, especially the exponential disentangling identity, will be new for you.

Spin operators and states

Recall the spin operators and eigenstates. The states |;z|\uparrow ; z\rangle and |;z|\downarrow ; z\rangle are eigenstates of Ŝz\hat{S}_z, which satisfy

Ŝz|;z=2|q;zŜz|;z=2|;z} eigenstates. \begin{aligned} & \left.\begin{array}{l} \hat{S}_{z}|\uparrow ; z\rangle=\frac{\hbar}{2}|q ; z\rangle \quad \\ \hat{S}_{z}|\downarrow ; z\rangle=-\frac{\hbar}{2}|\downarrow ; z\rangle \end{array}\right\} \text { eigenstates. } \end{aligned}

Next, we discuss the raising and lowering operators. Define Ŝ+\hat{S}_{+}and Ŝ\hat{S}_{-}to connect these states Ŝ+;z)=0,Ŝ+|;z=|;zŜ|;z=|;z,Ŝ;z=0.\begin{aligned} & \hat{S}_+\mid \uparrow ; z)=0, \quad \hat{S}_+|\downarrow ; z\rangle=\hbar|\uparrow ; z\rangle \\ & \hat{S}_-|\uparrow ; z\rangle=\hbar|\downarrow ; z\rangle, \quad \hat{S}_-\mid \downarrow ; z\rangle=0. \end{aligned}

Compute their commutators by acting the operator each, in turn, onto the states. Note that the commutator is computed via the action on the states themselves, using the above rule. We cannot determine them any other way. We have

(Ŝ+ŜŜŜ+)|;z=2|;z=2Ŝz|;z(Ŝ+ŜŜŜ+)|;z=2|;z=2Ŝz|;z\begin{aligned} & \left(\hat{S}_+\hat{S}_- - \hat{S}_-\hat{S}_{+}\right)|\uparrow ; z\rangle=\hbar^{2}|\uparrow ; z\rangle=2 \hbar \hat{S}_{z}|\uparrow ; z\rangle \\ & \left(\hat{S}_+\hat{S}_--\hat{S}_-\hat{S}_{+}\right)|\downarrow ; z\rangle=-\hbar^{2}|\downarrow ; z\rangle=2 \hbar \hat{S}_{z}|\downarrow ; z\rangle \end{aligned}

So [Ŝ+,Ŝ]=2Ŝz\boxed{\quad\left[\hat{S}_{+}, \hat{S}_{-}\right]=2 \hbar \hat{S}_{z}}

Similarly: (ŜzŜ+Ŝ+Ŝz)|;z=0=Ŝ+|;z(ŜzŜ+Ŝ+Ŝz)|;z=2|;z=Ŝ+|;z\begin{aligned} & \left(\hat{S}_{z} \hat{S}_{+}-\hat{S}_{+} \hat{S}_{z}\right)|\uparrow ; z\rangle=0=\hat{S}_{+}|\uparrow ; z\rangle \\ & \left(\hat{S}_{z} \hat{S}_{+}-\hat{S}+\hat{S}_{z}\right)|\downarrow ; z\rangle=\hbar^{2}|\downarrow ; z\rangle=\hbar \hat{S}_{+}|\downarrow ; z\rangle \end{aligned}

So [Ŝz,Ŝ+]=Ŝ+\boxed{\left[\hat{S}_{z}, \hat{S}_{+}\right]=\hbar \hat{S}_{+}}. You can verify yourself that [Ŝz,Ŝ]=Ŝ\boxed{\left[\hat{S}_{z}, \hat{S}_{-}\right]=-\hbar \hat{S}_{-}}.
Please ensure that you are comfortable calculating the commutators in this way, by acting the operators onto the states in the order provided and determining what the final result is. Because we acted them on all the states in our Hilbert space, we can use this to determine the commutation rule for the operators themselves. This is how we obtained our summary equations.

These three commutation relations are the SU(2)S U(2) algebra.

Cartesian spin operators

Next, we move on to determine the Cartesian spin operators. If we define Ŝx=12(Ŝ++Ŝ)\hat{S}_{x}=\frac{1}{2}\left(\hat{S}_{+}+\hat{S}_{-}\right) and Ŝy=12i(Ŝ+Ŝ)\hat{S}_{y}=\frac{1}{2 i}\left(\hat{S}_{+}-\hat{S}_{-}\right), then Ŝ+=Ŝx+iŜy\hat{S}_{+}=\hat{S}_{x}+i \hat{S}_{y} and Ŝ=ŜxiŜy\hat{S}_{-}=\hat{S}_{x}-i \hat{S}_{y}. The algebra of the Cartesian spins follows:

[Ŝx,Ŝy]=14i[Ŝ++Ŝ,Ŝ+Ŝ]=14i[[Ŝ+,Ŝ+][Ŝ+,Ŝ]+[Ŝ,Ŝ+][Ŝx,Ŝ]}=14i[2Ŝz2Ŝz]=iŜz\begin{aligned} {\left[\hat{S}_{x}, \hat{S}_{y}\right] } & =\frac{1}{4 i}\left[\hat{S}_{+}+\hat{S}_{-}, \hat{S}_{+}-\hat{S}_{-}\right] \\ & =\frac{1}{4 i}\left[\left[\hat{S}_{+}, \hat{S}_{+}\right]-\left[\hat{S}_{+}, \hat{S}_{-}\right]+\left[\hat{S}_{-}, \hat{S}_{+}\right]-\left[\hat{S}_{x}, \hat{S}\right]\right\} \\ & =\frac{1}{4 i}\left[-2 \hbar \hat{S}_{z}-2 \hbar \hat{S}_{z}\right]=i \hbar \hat{S}_{z} \end{aligned} where we used the commutation relations we already knew to evaluate them.

You can verify (and should) that we have [Ŝi,Ŝj]=ikεijkŜk\left[\hat{S}_{i}, \hat{S}_{j}\right]=i \hbar \sum_{k} \varepsilon_{i j k} \hat{S}_{k} which is the standard form for angular momentum commutators. Here, εijk \varepsilon_{\text {ijk }} is the completely antisymmetric tensor, which satisfies ε123=ε231=ε312=1\varepsilon_{123}=\varepsilon_{231}=\varepsilon_{312}=1 and ε132=ε321=ε213=1\varepsilon_{132}=\varepsilon_{321}=\varepsilon_{213}=-1, and all others vanish. Note we freely use 1, 2, 3 or xx, yy, zz. It should be clear from the context what we mean.

From this relation, we can compute the commutation relation of the total spin squared with the Cartesian angular momentum operators. We have [Ŝ2,Ŝj]=0.\left[\hat{S}^{2}, \hat{S}_{j}\right]=0.

Proof: Ŝ2=iŜiŜi\quad \hat{S}^{2}=\sum_{i} \hat{S}_{i} \hat{S}_{i} so [Ŝ2,Ŝj]=i[ŜiŜi,Ŝj]=i(Ŝi[Ŝi,Ŝj]+[Ŝi,Ŝj]Ŝi)=iik(ŜiεijkŜk+εijkŜkŜi)=iikεijk(ŜiŜk+ŜkŜi).\begin{aligned} {\left[\hat{S}^{2}, \hat{S}_{j}\right] } & =\sum_{i}\left[\hat{S}_{i} \hat{S}_{i}, \hat{S}_{j}\right]=\sum_{i}\left(\hat{S}_{i}\left[\hat{S}_{i}, \hat{S}_{j}\right]+\left[\hat{S}_{i}, \hat{S}_{j}\right] \hat{S}_{i}\right) \\ & =i \hbar \sum_{i} \sum_{k}\left(\hat{S}_{i} \varepsilon_{i j k} \hat{S}_{k}+\varepsilon_{i j k} \hat{S}_{k} \hat{S}_{i}\right) \\ & =i \hbar \sum_{i k} \varepsilon_{i j k}\left(\hat{S}_{i} \hat{S}_{k}+\hat{S}_{k} \hat{S}_{i}\right). \end{aligned} The claim is that this is zero. To see this let iki \rightarrow k^{\prime} and kik \rightarrow i^{\prime} =iikεkji(ŜkŜi+ŜiŜk)=iik(εijk)(ŜiŜk+ŜkŜi)\begin{aligned} & =i \hbar \sum_{i^{\prime} k^{\prime}} \varepsilon_{k^{\prime} j i^{\prime}}\left(\hat{S}_{k^{\prime}} \hat{S}_{i^{\prime}}+\hat{S}_{i^{\prime}} \hat{S}_{k^{\prime}}\right) \\ & =i \hbar \sum_{i^{\prime} k^{\prime}}\left(-\varepsilon_{i^{\prime} j k^{\prime}}\right)\left(\hat{S}_{i^{\prime}} \hat{S}_{k^{\prime}}+\hat{S}_{k^{\prime}} \hat{S}_{i^{\prime}}\right) \end{aligned} Since εijk=εkji\varepsilon_{i j k}=-\varepsilon_{k j i}. Now drop the primes =iikεijk(ŜiŜk+ŜkŜi)=-i \hbar \sum_{i k} \varepsilon_{i j k}\left(\hat{S}_{i} \hat{S}_{k}+\hat{S}_{k} \hat{S}_{i}\right) Anything equal to its negative must vanish, so [Ŝ2,Ŝj]=0\left[\hat{S}^{2}, \hat{S}_{j}\right]=0.

Let’s compute Ŝ2|σ;z\hat{S}^{2}|\sigma ; z\rangle with σ=or\sigma=\uparrow \text{or} \downarrow. First, we express the square of the spin in the spherical basis, because we know what those operators do when acting on states: Ŝ2=Ŝx2+Ŝy2+Ŝz2=14(Ŝ++Ŝ)214(ŜzŜ)2+Ŝz2=12(Ŝ+Ŝ+ŜŜ+)+Ŝz2\begin{aligned} \hat{S}^{2}=\hat{S}_{x}^{2}+\hat{S}_{y}^{2}+\hat{S}_{z}^{2} & =\frac{1}{4}\left(\hat{S}_{+}+\hat{S}_{-}\right)^{2}-\frac{1}{4}\left(\hat{S}_{z}-\hat{S}_{-}\right)^{2}+\hat{S}_{z}^{2} \\ & =\frac{1}{2}\left(\hat{S}_+\hat{S}_{-}+\hat{S}_{-} \hat{S}_{+}\right)+\hat{S}_{z}^{2} \end{aligned} Then we evaluate directly on states. So, acting on the up spin gives Ŝ2|;z=12(Ŝ+Ŝ+ŜŜ+)|;z+Ŝz2|;z=(122+142)|;z=342|;z\begin{aligned} \hat{S}^{2}|\uparrow ; z \rangle =\frac{1}{2}\left(\hat{S}_{+} \hat{S}_{-}+\hat{S}_- \hat{S}_{+}\right)| \uparrow ; z\rangle+\hat{S}_{z}^{2}|\uparrow ; z\rangle \\ =\left(\frac{1}{2} \hbar^{2}+\frac{1}{4} \hbar^{2}\right)|\uparrow ; z\rangle=\frac{3}{4} \hbar^{2}|\uparrow ; z\rangle \end{aligned} And acting on the down spin gives Ŝ2(;z)=12(Ŝ+Ŝ+ŜŜ+)|;z+Ŝz2|;z=342(;z)\begin{aligned} \hat{S}^{2}(\downarrow ; z) & =\frac{1}{2}\left(\hat{S}_{+} \hat{S}_{-}+\hat{S}_{-} \hat{S}_{+}\right) \rvert\, \downarrow ; z\rangle+\hat{S}_{z}^{2}|\downarrow ; z\rangle \\ & =\frac{3}{4} \hbar^{2}(\downarrow ; z) \end{aligned}
as well. Hence, |σ;z|\sigma ; z\rangle is an eigenstate of both Ŝ2\hat{S}^{2} and Ŝz\hat{S}_{z} !

Matrix representation of spin

We next look at the matrix representations of spin. Assume we have an arbitrary superposition of states |ψ=α|;z+β|;z|\psi\rangle=\alpha|\uparrow ; z\rangle+\beta|\downarrow ; z\rangle We let the column vector (αβ)\binom{\alpha}{\beta} denote the state |ψ|\psi\rangle. The operators Ŝi\hat{S}_{i} are represented by two-dimensional matrices in this space. For example $$\begin{array}{rl} \hat{S}_{z} \mid \hat{\uparrow} ; z\rangle=\frac{\hbar}{2}|\uparrow ; z\rangle \\ \hat{S}_{z} \mid \downarrow ; z \rangle=\frac{\hbar}{2}|\downarrow ; z\rangle \end{array} \Leftrightarrow \frac{\hbar}{2}\left(\begin{array}{cc} 1 & 0 \\ 0 & -1 \end{array}\right)\binom{1}{0}=\frac{\hbar}{2}\binom{1}{0} \\ \text{ and } \frac{\hbar}{2}\left(\begin{array}{cc} 1 & 0 \\ 0 & -1 \end{array}\right)\binom{0}{1}=-\frac{\hbar}{2}\binom{0}{1}.$$ The matrix representing the operator is Mσσ=σ;z|Ŝi|σ;zM_{\sigma \sigma^\prime}=\left\langle\sigma ;z\right| \hat{S}_{i}\left|\sigma^{\prime} ; z\right\rangle which includes 4 numbers for the matrix from the different choices of σ\sigma and σ\sigma'.

Let’s compute the matrix for Ŝy\hat{S}_{y}

Sσσy=σ;z|Ŝy|σ;z=12iσ;z|Ŝ+Ŝ|σ;zσ=:(Ŝ+Ŝ);z=;zSy=0, Sy=i2σ=:(Ŝ+Ŝ);z=|;zSy=i2,Ŝy=0.\begin{aligned} & S_{\sigma \sigma^{\prime}}^{y}=\langle\sigma ; z| \hat{S}_{y}\left|\sigma^{\prime} ; z\right\rangle=\frac{1}{2 i}\langle\sigma ; z| \hat{S}_{+}-\hat{S}_{-}\left|\sigma^{\prime} ; z\right\rangle \\ & \sigma^{\prime}=\uparrow:\left(\hat{S}_{+}-\hat{S}_{-}\right) \mid \uparrow ; z\rangle=-\hbar \mid \downarrow ; z\rangle \Rightarrow S_{\uparrow \uparrow}^{y}=0,~ \text{ } S_{ \downarrow \uparrow}^{y}=\frac{i \hbar}{2} \\ & \sigma^{\prime}=\downarrow:\left(\hat{S}_{+}-\hat{S}_{-}\right) \mid \downarrow ; z\rangle=\hbar|\uparrow ; z\rangle \Rightarrow S_{\uparrow \downarrow}^{y}=-\frac{i \hbar}{2},~ \text{} \hat{S}_{\downarrow \downarrow}^{y}=0. \end{aligned} So Ŝy=2(0ii0)\hat{S}_{y}=\frac{\hbar}{2}\left(\begin{array}{cc}0 & -i \\ i & 0\end{array}\right). Similarly, we have Ŝx=2(0110)andŜz=2(1001)\hat{S}_{x}=\frac{\hbar}{2}\left(\begin{array}{ll}0 & 1 \\ 1 & 0\end{array}\right) \quad \text{and} \quad\hat{S}_{z}=\frac{\hbar}{2}\left(\begin{array}{cc}1 & 0 \\ 0 & -1\end{array}\right).
We call σi\sigma_{i} the Pauli matrices σx=(0110)σy=(0ii0)σz=(1001)\sigma_{x}=\left(\begin{array}{ll} 0 & 1 \\ 1 & 0 \end{array}\right) \quad \sigma_{y}=\left(\begin{array}{cc} 0 & -i \\ i & 0 \end{array}\right) \quad \sigma_{z}=\left(\begin{array}{cc} 1 & 0 \\ 0 & -1 \end{array}\right) Note that σi2=(1001)Sσσ2=24×3×(1001)\sigma_{i}^{2}=\left(\begin{array}{ll}1 & 0 \\ 0 & 1\end{array}\right) \Rightarrow S_{\sigma \sigma^\prime}^2=\frac{\hbar^{2}}{4} \times 3 \times\left(\begin{array}{cc}1 & 0 \\ 0 & 1\end{array}\right)
These matrices also anticommute. To see this go back to our spin commutators ŜxŜyŜyŜx=iŜz.\hat{S}_{x} \hat{S}_{y}-\hat{S}_{y} \hat{S}_{x}=i \hbar \hat{S}_{z}. Multiply on left by Ŝx\hat{S}_{x} and on right by Ŝx\hat{S}_{x} Ŝx2ŜyŜxŜyŜx=iŜxŜzŜxŜyŜxŜyŜx2=iŜzŜx.\begin{array}{r} \hat{S}_{x}^{2} \hat{S}_{y}-\hat{S}_{x} \hat{S}_{y} \hat{S}_{x}=i \hbar \hat{S}_{x} \hat{S}_{z} \\ \hat{S}_{x} \hat{S}_{y} \hat{S}_{x}-\hat{S}_{y} \hat{S}_{x}^{2}=i \hbar \hat{S}_{z} \hat{S}_{x}. \end{array} Now substitute in the Pauli matrices 38[σx2σyσxσyσx]=i34σxσz38[σxσyσxσyσx2]=ii34σzσx.\begin{aligned} & \frac{\hbar^{3}}{8}\left[\sigma_{x}^{2} \sigma_{y}-\sigma_{x} \sigma_{y} \sigma_{x}\right]=i \frac{\hbar^{3}}{4} \sigma_{x} \sigma_{z} \\ & \frac{\hbar^{3}}{8}\left[\sigma_{x} \sigma_{y} \sigma_{x}-\sigma_{y} \sigma_{x}^{2}\right]=i \frac{i \hbar^{3}}{4} \sigma_{z} \sigma_{x}. \end{aligned} Add (and recall σx2=1\sigma_{x}^{2}=\mathbb{1} ): 38[σyσy]=i34(σxσz+σzσx)0=σxσz+σzσx\begin{gathered} \frac{\hbar^{3}}{8}\left[\sigma_{y}-\sigma_{y}\right]=i \frac{\hbar^{3}}{4}\left(\sigma_{x} \sigma_{z}+\sigma_{z} \sigma_{x}\right) \\ 0=\sigma_{x} \sigma_{z}+\sigma_{z} \sigma_{x} \end{gathered} and it is obvious this holds for other permutations too.
Hence, we have derived that σiσj=12[σiσj+σjσi+σiσjσjσi]=12δij×21+12i2εijkσk or σiσj=δij1+iεijkσk\begin{aligned} \sigma_{i} \sigma_{j} & =\frac{1}{2}\left[\sigma_{i} \sigma_{j}+\sigma_{j} \sigma_{i}+\sigma_{i} \sigma_{j}-\sigma_{j} \sigma_{i}\right] \\ & =\frac{1}{2} \delta_{i j} \times 2 \mathbb{1}+\frac{1}{2} i 2 \varepsilon_{i j k} \sigma_{k} \\ \text{ or } \sigma_{i} \sigma_{j} & =\delta_{i j} \mathbb{1}+i \varepsilon_{i j k} \sigma_{k} \end{aligned} This last equation is an important relation to remember about the product of two Pauli spin matrices.

Another interesting identity is the product of all 3 σxσyσz=iε123σzσz=i1σxσyσz=i1.\begin{gathered} \sigma_{x} \sigma_{y} \sigma_{z}=i \varepsilon_{123} \sigma_{z} \sigma_{z}=i \mathbb{1} \\ \sigma_{x} \sigma_{y} \sigma_{z}=i \mathbb{1}. \end{gathered}
Any 2×22 \times 2 matrix can be expressed in terms of the identity and the 3 Pauli matrices. This is called completeness over the space of 2×22\times 2 matrices.
Check: α1+βσx+γσy+δσz=(α+δβiγβ+iγαδ)\quad \alpha \mathbb{1}+\beta \sigma_{x}+\gamma \sigma_{y}+\delta \sigma_{z}=\left(\begin{array}{cc}\alpha+\delta & \beta-i \gamma \\ \beta+i \gamma & \alpha-\delta\end{array}\right)
So to find (abca)\left(\begin{array}{ll}a & b \\ c & a\end{array}\right), we set a=α+δb=βiγa=\alpha+\delta \quad b=\beta-i\gamma c=β+iγd=αδc=\beta+i \gamma d=\alpha \delta
or α=a+d2β=b+c2γ=i2(bc)andδ=ad2\alpha=\frac{a+d}{2} \quad \beta=\frac{b+c}{2} \quad \gamma=\frac{i}{2}(b-c) \quad \text{and}\quad \delta=\frac{a-d}{2}.
We often write this as M=α1+vσv=(b+c2,i2(bc),ad2)M=\alpha \mathbb{1} +\vec{v} \cdot \vec{\sigma} \quad \vec{v}=\left(\frac{b+c}{2}, \frac{i}{2}(b-c), \frac{a-d}{2}\right).

Working with spin matrices

Let’s get some practice working with these objects (Aσ)(Bσ)=ijAiBjσiσj=ijAiBj(δij1+iεijkσk)=AB1+iεijkAiBjσk=AB1+i(A×B)σ\begin{aligned} (\vec{A} \cdot \vec{\sigma})(\vec{B} \cdot \vec{\sigma}) & =\sum_{i j} A_{i} B_{j} \sigma_{i} \sigma_{j} \\ & =\sum_{i j} A_{i} B_{j}\left(\delta_{i j} \mathbb{1}+i \varepsilon_{i j k} \sigma_{k}\right) \\ & =\vec{A} \cdot \vec{B} \mathbb{1}+i \varepsilon_{i j k} A_{i} B_{j} \sigma_{k} \\ & =\vec{A} \cdot \vec{B} \mathbb{1}+i(\vec{A} \times \vec{B}) \cdot \vec{\sigma} \end{aligned}
Note that if A\vec{A} is parallel to B\vec{B}, then (Aσ)(Bσ)=AB1(\vec{A} \cdot \vec{\sigma})(\vec{B} \cdot \vec{\sigma})=\vec{A} \cdot \vec{B} \mathbb{1}. This identity is a useful one to remember.
Next, we evaluate the matrix exponential: exp[ivσ]=n=0(i)nn!(vσ)n.\operatorname{exp}[i \vec{v} \cdot \vec{\sigma}]=\sum_{n=0}^{\infty} \frac{(i)^{n}}{n!}(\vec{v} \cdot \vec{\sigma})^{n}. First separate out into even and odd powers eivσ=n=0(1)n(2n)!(vσ)2n+in=0(1)n(2n+1)!(vσ)2n+1.e^{i \vec{v} \cdot \vec{\sigma}}=\sum_{n=0}^{\infty} \frac{(-1)^{n}}{(2 n)!}(\vec{v} \cdot \vec{\sigma})^{2 n}+i \sum_{n=0}^{\infty} \frac{(-1)^{n}}{(2 n+1)!}(\vec{v} \cdot \vec{\sigma})^{2 n+1}.
But recall (vσ)(vσ)=v21(\vec{v} \cdot \vec{\sigma})(\vec{v} \cdot \vec{\sigma})=v^{2} \mathbb{1} so eivσ=n=0(1)n(2n)!(v2)n1+ivσn=0(1)n(2n+1)!(v2)neivσ=cos|v|1+ivσ|v|sin|v|\begin{aligned} & e^{i \vec{v} \cdot \vec{\sigma}}=\sum_{n=0}^{\infty} \frac{(-1)^{n}}{(2 n)!}\left(v^{2}\right)^{n} \mathbb{1}+i \vec{v} \cdot \vec{\sigma} \sum_{n=0}^{\infty} \frac{(-1)^{n}}{(2 n+1)!}\left(v^{2}\right)^{n} \\ & \boxed{e^{i \vec{v} \cdot \vec{\sigma}}=\cos |v| \mathbb{1}+i \frac{\vec{v} \cdot \vec{\sigma}}{|v|} \sin |v|} \end{aligned}
This is called the generalized Euler identity.

Let’s compute the similarity transformation of a Pauli matrix (corresponding to a rotation) eivσσjeivσ.e^{i \vec{v} \cdot \vec{\sigma}} \sigma_{j} e^{-i \vec{v} \cdot \sigma}. In general, such terms can involve an infinite series, as we will show in a later lecture via the Hadamard lemma, but in this case, we can explicitly calculate it. Just use our generalized Euler identity for each exponential, followed by our product rule for pairs of Pauli matrices: eivσσjeivσ=(cos|v|1+ivσ|v|sin|v|)σj(cos|σ|1ivσ|v|sin(v))=cos2|v|σjicos|v|sin|v||v|(σjvσvσσj)+sin2|v||v|2vσσjvσ=cos2|v|σjicos|v|sin|v||v|i[σj,σi]vi+sin2|v||v|2ikvivkσiσjσk=cos2|v|σj+cos|v|sin|v|vik2εjikviσk+sin2|v||σ|2ikvivkσiσjσk\begin{aligned} & e^{i \bar{v} \cdot \vec{\sigma}} \sigma_{j} e^{-i \vec{v} \cdot \vec{\sigma}}=\left(\cos |v| \mathbb{1}+i \frac{\vec{v} \cdot \vec{\sigma}}{|v|} \sin |v|\right) \sigma_{j}\left(\cos |\sigma| \mathbb{1}-i \frac{\vec{v} \cdot \vec{\sigma}}{|v|} \sin (v)\right) \\ & =\cos ^{2}|v| \sigma_{j}-i \frac{\cos |v| \sin |v|}{|v|}\left(\sigma_{j} \vec{v} \cdot \vec{\sigma}-\vec{v} \cdot \vec{\sigma} \sigma_{j}\right) +\frac{\sin ^{2}|v|}{|v|^{2}} \vec{v} \cdot \vec{\sigma} \sigma_{j} \vec{v} \cdot \vec{\sigma} \\ & =\cos ^{2}|v| \sigma_{j}-i \frac{\cos |v| \sin |v|}{|v|} \sum_{i}\left[\sigma_{j}, \sigma_{i}\right] v_{i}+\frac{\sin ^{2}|v|}{|v|^2} \sum_{i k} v_{i} v_{k} \sigma_{i} \sigma_{j} \sigma_{k} \\ & =\cos ^{2}|v| \sigma_{j}+\frac{\cos |v| \sin |v|}{v} \sum_{i k} 2 \varepsilon_{j i k} v_{i} \sigma_{k}+\frac{\sin ^{2}|v|}{|\sigma|^{2}} \sum_{i k} v_{i} v_{k} \sigma_{i} \sigma_{j} \sigma_{k} \end{aligned}

But ikvivkσiσjσk=ikvivk[δij1+iεijlσl]σk\sum_{i k} v_{i} v_{k} \sigma_{i} \sigma_{j} \sigma_{k}=\sum_{i k} v_{i} v_{k}\left[\delta_{i j} \mathbb{1}+i \varepsilon_{i j l} \sigma_{l}\right] \sigma_{k}. So, we have that $$\begin{aligned} \sum_{i k} v_{i} v_{k} \sigma_{i} \sigma_{j} \sigma_{k}&=v_{j} \vec{v} \cdot \vec{\sigma}+i \sum_{i k l} v_{i} v_{k} \varepsilon_{i j l}\left(\cancelto{0}{\delta_{l k}}+i \varepsilon_{l k m} \sigma_m\right) \\ &=v_{j} \vec{v} \cdot \vec{\sigma}-\sum_{i k l m} v_{i} v_{k} \varepsilon_{i j l} \varepsilon_{l k m} \sigma_{m} \\ &=v_{j} \vec{v} \cdot \vec{\sigma}-\sum_{i k m} v_{i} v_{k}\left(\delta_{i k} \delta_{j m}-\delta_{i m} \delta_{j k}\right) \sigma_{m} \\ &=v_{j} \vec{v} \cdot \vec{\sigma}-v^{2} \sigma_{j}+\vec{v} \cdot \vec{\sigma} v_{j} \\ \text{So that}~~~~~&\\ e^{i \bar{v} \cdot \vec{\sigma}} \sigma_{j} e^{-i \vec{v} \cdot \vec{\sigma}}&=\cos ^{2}|v| \delta_{j}+2 \frac{\cos |v| \sin |v|}{|v|}(\vec{v} \times \vec{\sigma})_{j}+\frac{\sin ^{2}(v)}{|v|^{2}}\left(2 \vec{v} \cdot \vec{\sigma} v_{j}-v^{2} \sigma_{j}\right). \end{aligned}$$

Exponential disentangling

Our last topic is on exponential disentangling. This is an identity most have not seen before. It is derived by factorizing the exponential of a combination of Pauli matrices into a product of three special Pauli matrices. First note that σ+=σx+iδy=(0200)\sigma_{+}=\sigma_{x}+i \delta_{y}=\left(\begin{array}{ll}0 & 2 \\ 0 & 0\end{array}\right) so

$$\begin{aligned} \exp \left(\alpha \sigma_{+}\right) & =1+\alpha \sigma_{+}+\frac{1}{2} \alpha^{2}\left(\cancelto{0}{\sigma_{+}^2}\right)+\cdots \\ & =\left(\begin{array}{cc} 1 & 2 \alpha \\ 0 & 1 \end{array}\right) \\ \exp \left(\alpha \sigma_{-}\right) & =\left(\begin{array}{cc} 1 & 0 \\ 2 \alpha & 1 \end{array}\right) \end{aligned}$$
That is, the exponential of raising or lowering Paulis is given by a sum of just two terms. Now, let’s consider the following exponential exp(vσ)=(cos|σ|+ivz|σ|sin|v|(ivx|v|+vy|v|)sin|v|(ivx|v|vy|v|)sin|v|cos|v|ivz|v|sin|v|)\exp (\vec{v} \cdot \vec{\sigma})=\left(\begin{array}{ll} \cos |\sigma|+i \frac{v_{z}}{|\sigma|} \sin |v| & \left(i \frac{v_{x}}{|{v}|}+\frac{v_{y}}{|v|}\right) \sin |v| \\ (\frac{i v_{x}}{|v|} - \frac{v_{y}}{|v|}) \sin|v| & \cos |v|-i \frac{v_{z}}{|v|} \sin |v| \end{array}\right)
We want to re-write it as exp[ασ+]exp[βσz]exp[γσ]\exp [ \alpha \sigma_{+}] \exp \left[\beta \sigma_{z}\right] \exp \left[\gamma \sigma_{-}\right], so we have (12α01)(eβ00eβ)(102γ1)=(eβ2αeβ0eβ)(102γ1)=(eβ+4αγeβ2αeβ2γeβeβ).\begin{aligned} & \left(\begin{array}{cc} 1 & 2 \alpha \\ 0 & 1 \end{array}\right)\left(\begin{array}{cc} e^{\beta} & 0 \\ 0 & e^{-\beta} \end{array}\right)\left(\begin{array}{cc} 1 & 0 \\ 2 \gamma & 1 \end{array}\right) \\ = & \left(\begin{array}{cc} e^{\beta} & 2 \alpha e^{-\beta} \\ 0 & e^{-\beta} \end{array}\right)\left(\begin{array}{cc} 1 & 0 \\ 2 \gamma & 1 \end{array}\right)=\left(\begin{array}{cc} e^{\beta}+4 \alpha \gamma e^{-\beta} & 2 \alpha e^{-\beta} \\ 2 \gamma e^{-\beta} & e^{-\beta} \end{array}\right). \end{aligned}
Comparing to the exponential, we have cos|v|+ivz|v|sin|v|=eβ+4αγeβ(ivx|σ|+vy|v|)sin|v|=2αeβ(ivx|v|vy|v|)sin|v|=2γeβcos|v|ivz|v|sin|v|=eβ,\begin{aligned} & \cos |v|+i \frac{v_{z}}{|v|} \sin |v|=e^{\beta}+4 \alpha \gamma e^{-\beta} \\ & \left(i \frac{v_{x}}{|\sigma|}+\frac{v_{y}}{|v|}\right) \sin |v|=2 \alpha e^{-\beta} \\ & \left(i \frac{v_{x}}{|v|}-\frac{v_{y}}{|v|}\right) \sin |v|=2 \gamma e^{-\beta} \\ & \cos |v|-i \frac{v_{z}}{|v|} \sin |v|=e^{-\beta}, \end{aligned}

First note that eβ(1+4αγe2β)=1cos|v|ivz|v|sin|v|(1+sin2|v|(vz2vy2|v|2))e^{\beta}\left(1+4 \alpha \gamma e^{-2 \beta}\right)=\frac{1}{\cos|v|- i\frac{v_{z} }{|v|} \sin |v|}\left(1+\sin ^{2}|v| (-\frac{v_{z}^{2}-v_{y}^{2}}{| v|^{2}}\right))
=cos|v|+ivz|v|sin|v|cos2|v|+vz2|v|2sin2|v|(1+sin2|v|(vz2|v|21))=cos|v|+ivz|v|sin|v|\begin{aligned} & =\frac{\cos |v|+i \frac{v_{z}}{|v|} \sin |v|}{\cos ^{2}|v|+\frac{v_{z}^{2}}{|v|^{2} \sin ^{2} |v|}}\left(1+\sin ^{2}|v|\left(\frac{v_{z}^{2}}{|v|^{2}}-1\right)\right) \\ & =\cos |v|+i \frac{v_{z}}{|v|} \sin |v| \end{aligned}

So if we solve the bottom three equations, the top automatically holds! This then gives us
β=ln[cos|v|ivz|v|sin|v|]α=12(ivx|v|+vy|v|)sin|v|cos|v|ivz|v|sin|v|γ=12(ivx|v|vyv)sin|v|cos|v|ivz|v|sin|v|\begin{aligned} &\boxed{ \beta=-\ln \left[\cos |v|-i \frac{v_{z}}{|v|} \sin |v|\right] }\\ &\boxed{\alpha=\frac{1}{2}\left(\frac{i v_{x}}{|v|}+\frac{v_{y}}{|v|}\right) \frac{\sin |v|}{\cos |v|-\frac{i v_{z}}{|v|} \sin |v|} }\\ &\boxed{\gamma=\frac{1}{2}\left(\frac{i v_{x}}{|v|}-\frac{v_{y}}{\mid v^{\prime}}\right) \frac{\sin |v|}{\cos |v|-i \frac{v_{z}}{|v|} \sin |v|} } \end{aligned} This identity is called the exponential disentangling identity and it is an important one.

An important special case is when vx=vz=0\boxed{v_{x}=v_{z}=0}, then β=lnsecvyα=12tanvyγ=12tanvy\boxed{ \beta=\ln \sec v_{y} \quad \alpha=\frac{1}{2} \tan v_{y} \quad \gamma=-\frac{1}{2} \tan v_{y} } This is your first taste of exponential disentangling. We will see more of it soon.

What is amazing about exponential disentangling is that is holds for any angular momentum, not just spin one half. But, we will have more to say about that later.