Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Vasiliki Tsanaktsidou (s232489@dtu.

dk)
DTU Compute – Mathematical Modeling and Computation
01715 – Functional Analysis
Problem 3
This exercise is based on the definition of the partially ordered set ([K, Def. 4.1-1]). We
will also use some definitions from [K, page 210].
Define the relation ≺ on ℂ by
𝑧 ≺ 𝑤 ⟺ 𝑅𝑒(𝑧) ≤ 𝑅𝑒(𝑤) 𝑎𝑛𝑑 𝐼𝑚(𝑧) ≤ 𝐼𝑚(𝑤) (1)
(i) Show that ≺ defines a partial ordering on ℂ
(ii) Give an example of two complex numbers 𝑧, 𝑤 ∈ ℂ that are comparable and
two complex numbers 𝑧, 𝑤 ∈ ℂ that are incomparable.
Now consider the subset 𝑊of ℂ given by
𝑊 ≔ {𝑥 + 𝑖𝑦 ∈ ℂ |𝑥, 𝑦 ∈ ℝ, |𝑥| < 1, |𝑦| < 1}
(iii) Does 𝑊 have an upper bound in ℂ with respect to the ordering ≺?
(iv) Does 𝑊 have a maximal element? Note that by definition, a maximal
element is an element in 𝑊 itself.
Solution
(i)
To show that the relation ≺ is a partial ordering on ℂ, we have to verify three properties.
• Reflexivity: For any complex number 𝑧 ∈ ℂ, we need to show that 𝑧 ≺ 𝑧
maintains. We know that 𝑅𝑒(𝑧) ≤ 𝑅𝑒(𝑧) 𝑎𝑛𝑑 𝐼𝑚(𝑧) ≤ 𝐼𝑚(𝑧), so 𝑧 ≺ 𝑧.
Therefore, the relation ≺ is reflexive.
• Antisymmetry: For any complex numbers 𝑧, 𝑤 ∈ ℂ, if 𝑧 ≺ 𝑤 and 𝑤 ≺ 𝑧, then
we have that 𝑧 = 𝑤. If 𝑅𝑒(𝑧) ≤ 𝑅𝑒(𝑤) 𝑎𝑛𝑑 𝐼𝑚(𝑧) ≤ 𝐼𝑚(𝑤) and 𝑅𝑒(𝑤) ≤
𝑅𝑒(𝑧) 𝑎𝑛𝑑 𝐼𝑚(𝑤) ≤ 𝐼𝑚(𝑧), then we have that:
𝑅𝑒(𝑧) = 𝑅𝑒(𝑤) 𝑎𝑛𝑑 𝐼𝑚(𝑧) ≤ 𝐼𝑚(𝑤)
So, 𝑧 = 𝑤. Therefore, the relation ≺ is antisymmetric.
• Transitivity: For any complex numbers 𝑥, 𝑦, 𝑧 ∈ ℂ, if 𝑥 ≺ 𝑦 and 𝑦 ≺ 𝑧, then
we have that 𝑥 ≺ 𝑧. If 𝑅𝑒(𝑥) ≤ 𝑅𝑒(𝑦), 𝑅𝑒(𝑦) ≤ 𝑅𝑒(𝑧), 𝐼𝑚(𝑥) ≤ 𝐼𝑚(𝑦) and
𝐼𝑚(𝑦) ≤ 𝐼𝑚(𝑧), then we have 𝑅𝑒(𝑧) ≤ 𝑅𝑒(𝑧) and 𝐼𝑚(𝑥) ≤ 𝐼𝑚(𝑧).
This means that 𝑥 ≺ 𝑧.
So, ≺ is a partial ordering on ℂ.
(ii)
• Comparable complex numbers: 𝑧 = 1 + 2𝑖 and 𝑤 = 2 + 5𝑖. We can observe
that 𝑅𝑒(𝑧) = 1 ≤ 2 = 𝑅𝑒(𝑤) and 𝐼𝑚(𝑧) = 2 ≤ 5 = 𝐼𝑚(𝑤).
Therefore, 𝑧 ≺ 𝑤.
• Incomparable complex numbers: 𝑧 = 2 + 4𝑖 and 𝑤 = 4𝑖. We can observe that
𝑅𝑒(𝑧) = 2 ≥ 0 = 𝑅𝑒(𝑤) and 𝐼𝑚(𝑧) = 4 = 𝐼𝑚(𝑤).
Therefore, 𝑧 ⊀ 𝑤 and 𝑤 ⊀ 𝑧.

(iii)
Let’s now consider an element 𝑤 ∈ 𝑊 and a complex number 𝑣 such that 𝑤 ≺ 𝑣
(respecting the ordering ≺).
We can take 𝑣 = 1 + 𝑖 and 𝑤 = 𝑥 + 𝑖𝑦 ∈ 𝑊. So, we have that:
|𝑥| < 1 ≤ 𝑅𝑒(𝑣) = 1, |𝑦| < 1 ≤ 𝐼𝑚(𝑣) = 1
Also, 𝑅𝑒(𝑤) ≤ 𝑅𝑒(𝑣), 𝐼𝑚(𝑤) ≤ 𝐼𝑚(𝑣). In other words, 𝑤 ≺ 𝑣. So, 𝑣 = 1 + 𝑖 is an
upper bound of 𝑊.
(iv)
A maximal element in a partially ordered set is an element such that there is no other
element greater than it. We can always find an element 𝑤 ′ = (𝑥 + 𝜀) + 𝑖(𝑦 + 𝜀) ∈ 𝑊
with 𝜀 > 0 (very small). In this case 𝑅𝑒(𝑤 ′ ) > 𝑅𝑒(𝑤), 𝐼𝑚(𝑤 ′ ) > 𝐼𝑚(𝑤) and 𝑤 ≠ 𝑤′.
So, we have that 𝑤 ≺ 𝑤′. Thus, 𝑊 does not have a maximal element.

Problem 31
Let ℋ denote a Hilbert space, with orthonormal basis {𝑒𝑘 }∞
𝑘=1 . Let 𝑁 ∈ ℕ and consider
𝑁
the subspace 𝑀 ≔ 𝑠𝑝𝑎𝑛{𝑒𝑘 }𝑘=1
(i) Show that for any 𝑓 ∈ ℋ and any coefficients 𝑐1 , … , 𝑐𝑁 ∈ ℂ,
𝑁 2 𝑁 ∞

‖𝑓 − ∑ 𝑐𝑘 𝑒𝑘 ‖ = ∑|𝑐𝑘 − 〈𝑓, 𝑒𝑘 〉|2 + ∑ |〈𝑓, 𝑒𝑘 〉|2 .


𝑘=1 𝑘=1 𝑘=𝑁+1

(ii) Fix any 𝑓 ∈ ℋ. What is the unique element 𝑔 ∈ 𝑀 that minimizes the
distance to 𝑓?
Solution
(i)
Given any 𝑓 ∈ ℋ and any coefficients 𝑐1 , … , 𝑐𝑁 ∈ ℂ, consider 𝜑 = ∑𝑁
𝑘=1 𝑐𝑘 𝑒𝑘 .

Then, we have the following:


𝑁 ∞ 𝑁

𝑓 − 𝜑 = 𝑓 − ∑ 𝑐𝑘 𝑒𝑘 = ∑〈𝑓, 𝑒𝑘 〉𝑒𝑘 − ∑ 𝑐𝑘 𝑒𝑘
𝑘=1 𝑘=1 𝑘=1

because the basis vectors are orthonormal and have ‖∙‖1, so they are linearly
independent, so any element in a Hilbert space can be expressed as a linear combination
of these basis vectors.
Now, taking and expanding the right side, we get that:
𝑁 ∞

𝑓 − 𝜑 = ∑(〈𝑓, 𝑒𝑘 〉 − 𝑐𝑘 )𝑒𝑘 + ∑ 〈𝑓, 𝑒𝑘 〉𝑒𝑘


𝑘=1 𝑘=𝑁+1

So, now we can take the required norm:


𝑁 ∞ 2

‖𝑓 − 𝜑‖2 = ‖∑(〈𝑓, 𝑒𝑘 〉 − 𝑐𝑘 )𝑒𝑘 + ∑ 〈𝑓, 𝑒𝑘 〉𝑒𝑘 ‖


𝑘=1 𝑘=𝑁+1

But we have orthogonality, so we get that:


𝑁 ∞ 2

‖∑(〈𝑓, 𝑒𝑘 〉 − 𝑐𝑘 )𝑒𝑘 + ∑ 〈𝑓, 𝑒𝑘 〉𝑒𝑘 ‖ =


𝑘=1 𝑘=𝑁+1
𝑁 ∞

∑|〈𝑓, 𝑒𝑘 〉 − 𝑐𝑘 |2 + ∑ |〈𝑓, 𝑒𝑘 〉|2


𝑘=1 𝑘=𝑁+1

Therefore, we got the given equation:


𝑁 2 𝑁 ∞

‖𝑓 − ∑ 𝑐𝑘 𝑒𝑘 ‖ = ∑|𝑐𝑘 − 〈𝑓, 𝑒𝑘 〉|2 + ∑ |〈𝑓, 𝑒𝑘 〉|2 ∎


𝑘=1 𝑘=1 𝑘=𝑁+1

(ii)
Due to the fact that Hilbert spaces are complete spaces, there is a unique element
𝑔 ∈ 𝑀 ⊂ ℋ such that ‖𝑓 − 𝑔‖ becomes minimum.
Moreover, this closest element must be orthogonal to the subspace 𝑀, which means that
𝑓 − 𝑔 is orthogonal to every element of 𝑀. In other words, it is orthogonal to the entire
subspace 𝑀 ≔ 𝑠𝑝𝑎𝑛{𝑒𝑘 }∞ 𝑘=1 .

We can assume that there is a 𝑔′ ∈ 𝑀 such that ‖𝑓 − 𝑔′‖ < ‖𝑓 − 𝑔‖. So, 𝑔′ is closer
to 𝑓 than 𝑔. Since 𝑔′ ∈ 𝑀, it must be a linear combination of {𝑒𝑘 }. But 𝑓 − 𝑔 is
orthogonal to {𝑒𝑘 }, so it must be orthogonal to 𝑔′ . Therefore, 〈𝑓 − 𝑔, 𝑔′〉 = 0, but since
𝑔 ′ = ∑∞
𝑘=1 𝑐𝑘 𝑒𝑘 , we take that 〈𝑓 − 𝑔, 𝑔′〉 > 0. So, here we have a contradiction. So, 𝑔′
does not exist.
The outcome of all the above is that given any 𝑓 ∈ ℋ the unique element 𝑔 ∈ 𝑀 is
𝑃𝑀 (𝑓), the orthogonal projection of 𝑓 onto 𝑀. Since 𝑀 ≔ 𝑠𝑝𝑎𝑛{𝑒𝑘 }𝑁
𝑘=1 , we have that:
𝑁

𝑃𝑀 (𝑓) = ∑〈𝑓, 𝑒𝑘 〉𝑒𝑘 ∎


𝑘=1

You might also like