Ferkans — Interactive Telecom Tutor

The Born Approximation — Cornerstone of RF Imaging

The Born approximation is the single most important linearization in this book. By replacing the unknown total field inside the scatterer with the known incident field, it transforms the nonlinear scattering problem into the linear forward model $\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w}$ that drives all subsequent chapters. Every reconstruction algorithm in Parts III--V assumes (explicitly or implicitly) that the Born approximation holds or that its violations can be managed.

,

Definition:
The Born Approximation

The first Born approximation replaces the total field $u$ inside the scatterer with the incident field $u^{\text{inc}}$ :

$u(\mathbf{r}') \approx u^{\text{inc}}(\mathbf{r}'), \quad \mathbf{r}' \in D.$

Substituting into the data equation yields the linearized scattering model:

$u^{\text{sca}}(\mathbf{r}_{j}) \approx \kappa_{0}^{2} \int_D G(\mathbf{r}_{j}, \mathbf{r}')\, \chi(\mathbf{r}')\,u^{\text{inc}}(\mathbf{r}')\,d\mathbf{r}'.$

This is linear in the contrast $\chi(\mathbf{r})$ — the integral is a linear operator applied to $\chi$ .

, ,

Born approximation

A first-order approximation to the scattering problem that replaces the total field inside the scatterer with the incident field, yielding a linear relationship between the contrast function and the scattered field.

Theorem: The Born Forward Model: $\mathbf{y} = \mathbf{A}\boldsymbol{\gamma} + \mathbf{w}$

Under the Born approximation with $M$ transmitter positions (incident fields $u_i^{\text{inc}}$ ) and $N$ receiver positions $\mathbf{r}_{j}$ , operating at $K$ frequencies, the measured scattered fields satisfy

$y(\mathbf{s}_{i}, \mathbf{r}_{j}; f_k) = \kappa_{0}^{2} \int_D G(\mathbf{r}_{j}, \mathbf{r}')\, u_i^{\text{inc}}(\mathbf{r}')\,\chi(\mathbf{r}')\,d\mathbf{r}' + w_{i,j,k}.$

Discretizing the domain $D$ into $Q$ voxels at positions $\mathbf{p}_{q}$ and defining the reflectivity vector $\ntn{refl_vec} \in \mathbb{C}^Q$ , this becomes

$\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w},$

where $\ntn{img} \in \mathbb{C}^{MNK}$ stacks all measurements and $[\mathbf{A}]_{(i,j,k),q}$ encodes the Green's function propagation and incident field illumination at each voxel.

Proof

Discretization of the volume integral

Partition $D$ into $Q$ cells of volume $\Delta V$ centered at $\mathbf{p}_{q}$ . Approximating the integral by a Riemann sum:

$u^{\text{sca}}(\mathbf{r}_{j}) \approx \kappa_{0}^{2} \sum_{q=1}^Q G(\mathbf{r}_{j}, \mathbf{p}_{q})\,u_i^{\text{inc}}(\mathbf{p}_{q})\, \chi(\mathbf{p}_{q})\,\Delta V.$

Defining $[\ntn{refl_vec}]_q = \chi(\mathbf{p}_{q})\Delta V$ and stacking the $MNK$ measurements into $\ntn{img}$ gives $\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w}$ . $\blacksquare$

,

Caire's Wavenumber Decomposition

Following Caire's unified framework, for a transmitter at $\mathbf{s}_{i}$ , receiver at $\mathbf{r}_{j}$ , and target centered at $\mathbf{p}_{0}$ , the far-field Born scattered signal has the form:

$x(\mathbf{s}_{i}, \mathbf{r}_{j}; f_k) \propto \int_{\tilde{\Omega}} c(\tilde{\mathbf{p}})\, e^{-j(\boldsymbol{\kappa}_s + \boldsymbol{\kappa}_r)^\mathsf{T}\tilde{\mathbf{p}}}\,d\tilde{\mathbf{p}},$

where $\boldsymbol{\kappa}_s = \kappa(\mathbf{p}_{0} - \mathbf{s}_{i})/d(\mathbf{s}_{i}, \mathbf{p}_{0})$ and $\boldsymbol{\kappa}_r = \kappa(\mathbf{p}_{0} - \mathbf{r}_{j})/d(\mathbf{p}_{0}, \mathbf{r}_{j})$ are the transmit and receive wavenumber vectors.

Each measurement samples the spatial Fourier transform of the reflectivity at the combined wavenumber $\ntn{comb_wavenum} = \boldsymbol{\kappa}_s + \boldsymbol{\kappa}_r$ . This is the Fourier diffraction theorem — the foundation for wavenumber-domain analysis in Chapter 6.

🎓CommIT Contribution(2023)

Unified Illumination and Sensing Model for RF Imaging

G. Caire, A. Rezaei, W. Jiang — CommIT Group, TU Berlin (research note)

Caire's research note provides the definitive derivation of the Born forward model in the RF imaging context. Starting from the scalar wave equation and applying the Born approximation under the far-field assumption, the note derives the discrete sensing model $\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w}$ with explicit expressions for the sensing matrix entries involving transmit/receive wavenumber vectors $\boldsymbol{\kappa}_s$ , $\boldsymbol{\kappa}_r$ , antenna gains $G^{\text{tx}}_i$ , $G^{\text{rx}}_j$ , and propagation delays. The key insight is that both the diffraction-tomography view (wavenumber-domain sampling) and the radar view (matched filtering) are two faces of the same Born-linearized forward operator.

RF imagingBorn approximationsensing matrixISAC

Definition:
Validity Conditions for the Born Approximation

The Born approximation is accurate when the scattered field is much weaker than the incident field inside the object:

$\left|\kappa_{0}^{2} \int_D G(\mathbf{r}, \mathbf{r}')\, \chi(\mathbf{r}')\,u^{\text{inc}}(\mathbf{r}')\,d\mathbf{r}'\right| \ll |u^{\text{inc}}(\mathbf{r})|, \quad \forall \mathbf{r} \in D.$

For a uniform sphere of radius $a$ and contrast $\chi_0$ , this reduces to the quantitative criterion:

$\kappa_{0} a \,|\chi_0| \ll 1.$

Low contrast ( $|\chi_0| \ll 1$ ): valid for large objects.
Small objects ( $\kappa_{0} a \ll 1$ ): valid even for high contrast.
Fails when both contrast and electrical size are large.

,

Example: Born Approximation Validity Check

Consider a 2D circular scatterer at $f = 1$ GHz ( $\lambda = 30$ cm), radius $a = 5$ cm, contrast $\chi_0 = 0.5$ .

Evaluate the Born validity criterion and determine whether the approximation is reliable.

Solution

Compute the Born parameter

$\kappa_{0} a |\chi_0| = \frac{2\pi}{0.3} \cdot 0.05 \cdot 0.5 = 0.52.$ $

This is on the boundary of validity — the Born approximation introduces noticeable error (typically 10--20% in amplitude).

Reduce contrast

Reducing to $\chi_0 = 0.1$ : $\kappa_{0} a |\chi_0| = 0.10$ , well within the valid regime ( $\ll 1$ ).

Born Approximation vs Exact Scattering

Compares the Born-approximated scattered field with the exact (Mie series) solution for a circular cylinder.

Left panel: Scattered field magnitude vs angle for both Born and exact solutions.

Right panel: Relative error $\|u^{\text{sca}}_{\text{Born}} - u^{\text{sca}}_{\text{exact}}\| / \|u^{\text{sca}}_{\text{exact}}\|$ as a color-coded indicator.

The dashed line marks the validity boundary $\kappa_{0} a |\chi_0| = 1$ .

Parameters

Object contrast

\chi_0

0.3

Frequency (GHz)1

Object radius (cm)5

Theorem: The Fourier Diffraction Theorem

Under the Born approximation with plane-wave illumination $\mathbf{k}_p$ and far-field measurement in direction $\mathbf{k}_q$ (with $|\mathbf{k}_p| = |\mathbf{k}_q| = \kappa_{0}$ ), the scattered field is proportional to the Fourier transform of the contrast:

$y_{pq} \propto \hat{\chi}(\mathbf{k}_p - \mathbf{k}_q).$

The set of accessible spatial frequencies $\{\mathbf{K} = \mathbf{k}_p - \mathbf{k}_q\}$ lies within the Ewald sphere of radius $2\kappa_{0}$ , imposing a fundamental resolution limit of $\lambda/2$ on any far-field imaging system.

Proof

Fourier transform of Born integral

For plane-wave illumination $u^{\text{inc}}(\mathbf{r}) = e^{j\mathbf{k}_p \cdot \mathbf{r}}$ and far-field observation, the Born scattered field becomes:

$u^{\text{sca}}_{\text{Born}} \propto \int_D \chi(\mathbf{r}')\, e^{j(\mathbf{k}_p - \mathbf{k}_q) \cdot \mathbf{r}'}\,d\mathbf{r}' = \hat{\chi}(\mathbf{k}_p - \mathbf{k}_q).$

Ewald sphere construction

Since $|\mathbf{k}_p| = |\mathbf{k}_q| = \kappa_{0}$ , the difference $\mathbf{K} = \mathbf{k}_p - \mathbf{k}_q$ satisfies $|\mathbf{K}| \leq 2\kappa_{0}$ . As we vary all illumination and observation angles, the accessible frequencies fill a ball of radius $2\kappa_{0}$ in Fourier space. By the Nyquist criterion, this limits the spatial resolution to $\Delta x \geq \pi/\kappa_{0} = \lambda/2$ . $\blacksquare$

,

Common Mistake: When Born Fails — Multiple Scattering Effects

Mistake:

Applying the Born approximation to high-contrast or electrically large objects, ignoring multiple scattering. Common failure modes:

Shadowing: A strong scatterer blocks illumination behind it.
Phase wrapping: For $\kappa_{0} a |\chi_0| > \pi$ , the total field phase wraps inside the object.
Resonances: Standing waves build up inside the object at certain frequencies.

Correction:

Check $\kappa_{0} a |\chi_0| \ll 1$ before applying Born. For larger values, use the Rytov approximation (Section 5.5) or iterative methods (Section 5.6).

Common Mistake: Born Validity Is Not Just About Weak Contrast

Mistake:

Assuming that $|\chi_0| \ll 1$ is sufficient for Born validity. A large object with $\chi_0 = 0.05$ but $\kappa_{0} a = 50$ gives $\kappa_{0} a |\chi_0| = 2.5 \gg 1$ — Born fails even though the contrast is small.

Correction:

The Born criterion involves the product of contrast and electrical size: $\kappa_{0} a |\chi_0| \ll 1$ . For large objects with small contrast, use Rytov instead.

Why This Matters: From Scattering Physics to Linear Sensing

The Born approximation is the bridge between electromagnetic physics and signal processing. The linear model $\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w}$ has the same structure as a MIMO communication model $\mathbf{y} = \mathbf{H}\mathbf{x} + \mathbf{w}$ , where the "channel matrix" $\mathbf{A}$ is determined by physics (Green's functions, antenna patterns, geometry) rather than propagation statistics. This allows all the estimation and detection tools from [?fsi] and linear algebra from Chapter 1 to be applied directly to imaging.

See full treatment in Chapter 6

⚠️Engineering Note

Born Approximation in Practical RF Imaging Systems

For typical indoor RF imaging at $f = 5$ GHz ( $\lambda = 6$ cm):

Drywall: $\chi \approx 1.5$ , thickness $L \approx 1$ cm. $\kappa_{0} L |\chi| \approx 1.6$ — Born is marginal.
Human body: $\chi \approx 40$ (at microwave). Born fails badly for through-body imaging.
Furniture (wood): $\chi \approx 0.5$ , size $\sim 50$ cm. $\kappa_{0} a |\chi| \approx 26$ — Born fails.

In practice, Born-based imaging works well for through-wall radar where the wall is thin and relatively low-contrast, and for sparse scenes where individual scatterers are small. For dense scenes, iterative methods or data-driven approaches are needed.

Practical Constraints

•
Born validity requires $\kappa_{0} a |\chi_0| \ll 1$
•
Indoor materials often violate this at microwave frequencies
•
Model mismatch appears as artifacts in reconstructed images

Quick Check

A dielectric sphere has radius $a = 2$ cm and contrast $\chi_0 = 0.2$ at $f = 10$ GHz ( $\lambda = 3$ cm). Is the Born approximation valid?

Yes — the contrast is small ( $0.2 \ll 1$ ).

Yes — $\kappa_{0} a |\chi_0| = (2\pi/0.03)(0.02)(0.2) \approx 0.84 < 1$ .

No — the frequency is too high.

No — the object is too large.

Correction:

Yes —

\kappa_{0} a |\chi_0| = (2\pi/0.03)(0.02)(0.2) \approx 0.84 < 1

.

The product is less than 1, so Born is marginally valid with modest error.

Key Takeaway

The Born approximation replaces $u^{\text{tot}}$ with $u^{\text{inc}}$ inside $D$ , yielding the linear forward model $\ntn{img} = \mathbf{A}\ntn{refl_vec} + \mathbf{w}$ . The sensing matrix $\mathbf{A}$ encodes propagation (Green's function) and illumination physics. Validity condition: $\kappa_{0} a |\chi_0| \ll 1$ . The Fourier diffraction theorem shows that Born measurements are Fourier samples of $\chi$ , with resolution limited to $\lambda/2$ (Ewald sphere). This linear model is the foundation for all reconstruction methods in subsequent chapters.

The Born Approximation

The Born Approximation — Cornerstone of RF Imaging

Definition: The Born Approximation

Born approximation

Theorem: The Born Forward Model: y=Aγ+w\mathbf{y} = \mathbf{A}\boldsymbol{\gamma} + \mathbf{w}y=Aγ+w

Discretization of the volume integral

Caire's Wavenumber Decomposition

Unified Illumination and Sensing Model for RF Imaging

Definition: Validity Conditions for the Born Approximation

Example: Born Approximation Validity Check

Compute the Born parameter

Reduce contrast

Born Approximation vs Exact Scattering

Parameters

Theorem: The Fourier Diffraction Theorem

Fourier transform of Born integral

Ewald sphere construction

Common Mistake: When Born Fails — Multiple Scattering Effects

Common Mistake: Born Validity Is Not Just About Weak Contrast

Why This Matters: From Scattering Physics to Linear Sensing

Born Approximation in Practical RF Imaging Systems

Quick Check

Key Takeaway

Definition:
The Born Approximation

Theorem: The Born Forward Model: $\mathbf{y} = \mathbf{A}\boldsymbol{\gamma} + \mathbf{w}$

Definition:
Validity Conditions for the Born Approximation