← Exercise library← All exercises for this theme
Theme 2 — Hilbert Spaces and Dirac Notation

Hermitian inner product in C3

Keywords: Hermitian inner product · norm · orthogonality · Cauchy--Schwarz inequality · antilinearity

Exercise 1 : Hermitian inner product in C3

We work in C3\C^3, equipped with the Hermitian inner product ⟨u|v⟩=∑i=13ui∗vi\braket{u}{v}=\sum_{i=1}^3u_i^*v_i, and consider the vectors

∣u⟩=(1i1),∣v⟩=(i11−i).\ket u=\begin{pmatrix}1\\ i\\ 1\end{pmatrix},\qquad \ket v=\begin{pmatrix}i\\ 1\\ 1-i\end{pmatrix}.
Question 1
Compute ∥u∥\norm{u}, ∥v∥\norm{v}, ⟨u|v⟩\braket{u}{v} and ⟨v|u⟩\braket{v}{u}. What relation connects the last two numbers?

Solution
The square of the norm is the sum of the squared moduli of the components:

∥u∥2=∣1∣2+∣i∣2+∣1∣2=3,∥v∥2=∣i∣2+∣1∣2+∣1−i∣2=1+1+2=4,\norm{u}^2=|1|^2+|i|^2+|1|^2=3,\qquad \norm{v}^2=|i|^2+|1|^2+|1-i|^2=1+1+2=4 ,

and hence ∥u∥=3\norm u=\sqrt3 and ∥v∥=2\norm v=2. To compute the inner product, the components of the first vector must be conjugated, so ⟨u∣=(1, −i, 1)\bra u=(1,\,-i,\,1):

⟨u|v⟩=1⋅i+(−i)⋅1+1⋅(1−i)=i−i+1−i=1−i.\braket uv=1\cdot i+(-i)\cdot1+1\cdot(1-i)=i-i+1-i=1-i .

Similarly, with ⟨v∣=(−i, 1, 1+i)\bra v=(-i,\,1,\,1+i),

⟨v|u⟩=(−i)⋅1+1⋅i+(1+i)⋅1=−i+i+1+i=1+i.\braket vu=(-i)\cdot1+1\cdot i+(1+i)\cdot1=-i+i+1+i=1+i .

We find that ⟨v|u⟩=⟨u|v⟩∗\braket vu=\braket uv^*: this is the Hermitian symmetry of the inner product.

Question 2
Verify the Cauchy—Schwarz inequality ∣⟨u|v⟩∣≤∥u∥ ∥v∥|\braket uv|\leq\norm u\,\norm v.

Solution
We have ∣⟨u|v⟩∣=∣1−i∣=2≃1,41|\braket uv|=|1-i|=\sqrt2\simeq1{,}41, and ∥u∥ ∥v∥=23≃3,46\norm u\,\norm v=2\sqrt3\simeq3{,}46. The inequality is satisfied, and it is strict: the two vectors are not collinear. Recall that equality holds only if one of the vectors is a scalar multiple of the other.

Question 3
Compute ⟨λu|v⟩\braket{\lambda u}{v} and ⟨u|λv⟩\braket{u}{\lambda v} for λ=i\lambda=i. Comment on the result.

Solution
The vector λ∣u⟩=i∣u⟩\lambda\ket u=i\ket u has components (i, −1, i)(i,\,-1,\,i), and its bra is (−i, −1, −i)(-i,\,-1,\,-i). Thus

⟨iu|v⟩=(−i)⋅i+(−1)⋅1+(−i)(1−i)=1−1−i−1=−1−i=−i ⟨u|v⟩.\braket{iu}{v}=(-i)\cdot i+(-1)\cdot1+(-i)(1-i)=1-1-i-1=-1-i=-i\,\braket uv .

By contrast, ⟨u|iv⟩=i⟨u|v⟩=i(1−i)=1+i\braket{u}{iv}=i\braket uv=i(1-i)=1+i. The inner product is linear in the second vector but antilinear in the first: ⟨λu|v⟩=λ∗⟨u|v⟩\braket{\lambda u}{v}=\lambda^*\braket uv. This is the convention used by physicists.

Question 4
Determine a non-zero vector ∣w⟩\ket w orthogonal to both ∣u⟩\ket u and ∣v⟩\ket v. Verify the result.

Solution
We seek ∣w⟩=(w1,w2,w3)\ket w=(w_1,w_2,w_3) such that ⟨u|w⟩=0\braket uw=0 and ⟨v|w⟩=0\braket vw=0, that is, using the bras computed above,

w1−iw2+w3=0,−iw1+w2+(1+i)w3=0.w_1-iw_2+w_3=0,\qquad -iw_1+w_2+(1+i)w_3=0 .

The first equation gives w1=iw2−w3w_1=iw_2-w_3. Substituting this into the second gives

−i(iw2−w3)+w2+(1+i)w3=w2+iw3+w2+(1+i)w3=2w2+(1+2i)w3=0,-i(iw_2-w_3)+w_2+(1+i)w_3=w_2+iw_3+w_2+(1+i)w_3=2w_2+(1+2i)w_3=0,

and hence w2=−1+2i2w3w_2=-\frac{1+2i}{2}w_3. Choosing w3=2w_3=2 gives w2=−1−2iw_2=-1-2i and w1=i(−1−2i)−2=−iw_1=i(-1-2i)-2=-i:

∣w⟩=(−i−1−2i2).\ket w=\begin{pmatrix}-i\\ -1-2i\\ 2\end{pmatrix}.

Let us verify this: ⟨u|w⟩=−i+(−i)(−1−2i)+2=−i+i−2+2=0\braket uw=-i+(-i)(-1-2i)+2=-i+i-2+2=0, and ⟨v|w⟩=(−i)(−i)+(−1−2i)+(1+i)⋅2=−1−1−2i+2+2i=0\braket vw=(-i)(-i)+(-1-2i)+(1+i)\cdot2=-1-1-2i+2+2i=0. The vector ∣w⟩\ket w is determined up to a complex factor; its norm is 1+5+4=10\sqrt{1+5+4}=\sqrt{10}.

Question 5
Compute the quantity ∑iui2\sum_iu_i^2, obtained by omitting complex conjugation. Why can it not be used to define a norm? Give an example of a non-zero vector in C2\C^2 for which it vanishes.

Solution
We find ∑iui2=1+i2+1=1\sum_iu_i^2=1+i^2+1=1, whereas ∥u∥2=3\norm u^2=3. Without complex conjugation, the resulting quantity is not always real or positive: for (1,i)∈C2(1,i)\in\C^2, it is 1+i2=01+i^2=0, even though the vector is non-zero. It therefore cannot define a norm; the components of the first vector are conjugated precisely to ensure that ⟨u|u⟩=∑i∣ui∣2≥0\braket uu=\sum_i|u_i|^2\geq0.