Nano-Optical Polarization Control

← Back to Knowledge Share

From Target Polarization Phase to Structure

Step 1: Defining the Goal

Imagine a single pixel on your metasurface. You have decided on an arbitrary orthogonal polarization basis—let's say, two specific elliptical states, \(|\lambda^+\rangle\) and \(|\lambda^-\rangle\).

Your goal as the designer is to apply a specific phase delay \(\Phi^+\) to the first state, and a completely independent phase delay \(\Phi^-\) to the second state.

Step 2: The Complex Conjugate Constraint (The Tricky Part)

You want your metasurface pixel (which has a Jones matrix \(J\)) to do this:

\[J |\lambda^+\rangle \rightarrow \text{add phase } \Phi^+\]

However, a dielectric nanopillar is a standard, reciprocal waveplate. A fundamental property of a waveplate is that when it alters a polarization state, it flips its "handedness" (helicity). For example, if you send in Left-Circular light, a half-waveplate spits out Right-Circular light.

Mathematically, flipping the handedness is represented by the complex conjugate (\(*\)). Therefore, the true physical equation for what the pillar does to your target state is:

\[J |\lambda^+\rangle = e^{i\Phi^+} |\lambda^{+*}\rangle\]

\[J |\lambda^-\rangle = e^{i\Phi^-} |\lambda^{-*}\rangle\]

Step 3: Solving for the "Target Matrix" (\(J_{req}\))

Now we have a system of equations. We know the inputs (\(|\lambda\rangle\)), and we know the desired outputs. We can combine these two equations into a single matrix block:

\[J \begin{bmatrix} |\lambda^+\rangle & |\lambda^-\rangle \end{bmatrix} = \begin{bmatrix} e^{i\Phi^+}|\lambda^{+*}\rangle & e^{i\Phi^-}|\lambda^{-*}\rangle \end{bmatrix}\]

To isolate \(J\), we simply multiply both sides by the inverse of the input matrix:

\[J_{req} = \begin{bmatrix} e^{i\Phi^+}|\lambda^{+*}\rangle & e^{i\Phi^-}|\lambda^{-*}\rangle \end{bmatrix} \begin{bmatrix} |\lambda^+\rangle & |\lambda^-\rangle \end{bmatrix}^{-1}\]

If you compute this, you end up with a \(2 \times 2\) symmetric matrix of complex numbers. This is your \(J_{req}\). It represents the exact optical transformation that *must* happen at that specific pixel on your metasurface to achieve your independent phase goals.

Step 4: Bridging Math to the Physical Pillar (Diagonalization)

Here is the punchline. You have a mathematical matrix \(J_{req}\), but you need to send a GDSII layout to a nanofabrication facility. How do you turn a matrix into a pillar?

We know from basic optics that the Jones matrix of a simple rectangular pillar, rotated by an angle \(\theta\), is always written as:

\[J_{pillar} = R(-\theta) \begin{pmatrix} e^{i\phi_x} & 0 \\ 0 & e^{i\phi_y} \end{pmatrix} R(\theta)\]

*(Where *\(\phi_x\)* and *\(\phi_y\)* are the propagation phases determined by the pillar's width and length, and *\(R(\theta)\)* is the rotation matrix).*

Linear algebra states that any symmetric unitary matrix (like our \(J_{req}\)) can be diagonalized into this exact format.

So, you take the \(J_{req}\) you calculated in Step 3 and you run an eigendecomposition on it. The math directly spits out two things:

  1. The Eigenvalues: These are \(e^{i\phi_x}\) and \(e^{i\phi_y}\). They tell you the exact propagation phases you need. You look at your simulation lookup table to find which physical widths (\(L_x\) and \(L_y\)) produce those phases.
  2. The Eigenvectors: These form your rotation matrix \(R(\theta)\). They tell you the exact angle \(\theta\) to physically rotate that pillar on the substrate.

An Example

Suppose we want to design a metasurface pixel that completely decouples Left-Circular Polarization (LCP) and Right-Circular Polarization (RCP).

  • Goal 1: If RCP hits the pixel, it passes through with no phase delay (\(\Phi^+ = 0\)).
  • Goal 2: If LCP hits the pixel, it passes through with a \(\pi\) phase delay (\(\Phi^- = \pi\)).

Step 1: Define the Input Vectors

We start with the standard Jones vectors for our chosen orthogonal basis:

  • RCP input: [|R\rangle = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 \\ -i \end{bmatrix}]
  • LCP input: [|L\rangle = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 \\ i \end{bmatrix}]

Step 2: Set Up the "Target" Equations

Remember the physical constraint: the pillar will flip the handedness (complex conjugate) of the output. We multiply our inputs by the desired phase delays (\(e^{i0} = 1\) and \(e^{i\pi} = -1\)) and flip their helicity.

  • What must happen to RCP: [J_{req} |R\rangle = (1) \cdot |L\rangle = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 \\ i \end{bmatrix}]
  • What must happen to LCP: [J_{req} |L\rangle = (-1) \cdot |R\rangle = \frac{-1}{\sqrt{2}}\begin{bmatrix} 1 \\ -i \end{bmatrix} = \frac{1}{\sqrt{2}}\begin{bmatrix} -1 \\ i \end{bmatrix}]

Step 3: Calculate the Required Matrix (\(J_{req}\))

We combine those two rules into a single matrix equation:

\[J_{req} \times \frac{1}{\sqrt{2}}\begin{bmatrix} 1 & 1 \\ -i & i \end{bmatrix} = \frac{1}{\sqrt{2}}\begin{bmatrix} 1 & -1 \\ i & i \end{bmatrix}\]

To solve for \(J_{req}\), we multiply both sides by the inverse of the input matrix. Skipping the intermediate fractional arithmetic, the result beautifully simplifies to:

\[J_{req} = \begin{bmatrix} 0 & i \\ i & 0 \end{bmatrix}\]

This is the exact optical transformation needed at this specific coordinate on the metasurface.

Step 4: Extract the Physical Geometry (Diagonalization)

Now we must translate [\begin{bmatrix} 0 & i \\ i & 0 \end{bmatrix}] into a physical nanopillar. We find the eigenvalues and eigenvectors of this matrix.

1. The Eigenvalues = The Pillar Dimensions (\(L_x, L_y\))

Solving the characteristic equation gives two eigenvalues: \(\lambda_1 = i\) and \(\lambda_2 = -i\).

We convert these to pure phase:

  • \(i = e^{i\pi/2} \rightarrow\) Required propagation phase \(\phi_x = \frac{\pi}{2}\)
  • \(-i = e^{-i\pi/2} \rightarrow\) Required propagation phase \(\phi_y = -\frac{\pi}{2}\)

*(Note: The relative phase difference is exactly *\(\pi\)*, meaning the math has automatically deduced that a half-waveplate is required for this specific circular polarization task).*

2. The Eigenvectors = The Pillar Rotation Angle (\(\theta\))

Calculating the eigenvectors for those specific eigenvalues yields:

  • [\vec{v}_1 = \begin{bmatrix} 1 \\ 1 \end{bmatrix}] and [\vec{v}_2 = \begin{bmatrix} -1 \\ 1 \end{bmatrix}]

When normalized, these vectors form a standard rotation matrix for \(\theta = 45^\circ\).