Sub Terahertz Copper Roughness Modeling Using Atomic Force Power Spectral Density

Atomic force power spectral density models sub-terahertz copper loss by mapping nanoscale spatial frequency harmonics that standard profile metrics ignore.

20.10.26 14 min

Scale

At 140 GHz, the skin depth of copper shrinks to 175 nanometers. A standard electrodeposited foil presents surface peaks and valleys spanning three to five micrometers, exceeding electromagnetic field penetration depth by a factor of twenty. High spatial frequencies dominate conductor loss.

When the electromagnetic field concentrates in a skin layer thinner than the profile height, electrical currents divert around microscopic ridges and nodules, lengthening the actual path traversed by the wave. Classical roughness corrections, formulated for low-gigahertz designs, fail completely under these boundary conditions.

The standard Morgan and Hammerstad-Bekkadal models rely on one-dimensional root-mean-square height values, designated as Rq or standard deviation of surface heights. These equations assume smooth sinusoidal or hemispherical profiles where roughness acts as an empirical scaling multiplier bounded between one and two. In the sub-terahertz regime between 100 GHz and 300 GHz, measured conductor attenuation exceeds Hammerstad predictions by more than three hundred percent.

The classical formulation saturates at a theoretical correction factor of two, yet laboratory measurements of microstrip transmission lines fabricated on 50-micrometer liquid crystal polymer substrates demonstrate insertion losses climbing steadily past that limit.

Standard root-mean-square roughness metrics underestimate copper conductor attenuation by up to fourfold at 220 GHz because one-dimensional profile numbers discard spatial frequency distribution.

The failure stems from omitting spatial wavelength information. A profile with an Rq of 0.4 micrometers composed of broad, undulating contours produces far lower microwave attenuation than an identical Rq built from densely packed, sub-micron needle-like asperities. Conventional stylus and optical white-light profilometers possess lateral optical resolution limits between 0.5 and 1.0 micrometers, missing the narrow pits and nodules that dictate current redistribution in the D-band and G-band.

Atomic force microscopy resolves these nanoscale contours by tracking lateral surface variations down to single-nanometer increments.

An automated industrial nozzle directs a flexible conduit into a heated crucible containing molten alloy beside an electronics assembly station with cable tracks.

Deficiencies of Classical Profile Metrics

Roughness parameters extracted from two-dimensional stylus traces, such as arithmetic average roughness Ra and ten-point peak-to-valley height Rz, describe vertical excursions while ignoring spatial frequency. Production facilities frequently certify rolled annealed foil lots based on an Ra figure below 0.2 micrometers. When laminators subject that foil to bond-enhancement treatments, zinc or brass micro-nodules deposit across the interface, altering spatial topography without significantly moving the macro Ra value.

Skin depth drops below two hundred nanometers. High-frequency transport concentrates almost entirely within these added nodules. The prevailing current density vector ceases to flow parallel to the substrate plane; instead, it tracks the steep ascents and descents of individual metallic grains.

The effective electrical conductivity of the conductor surface layer degrades, generating unexpected insertion loss and severe phase distortion that destabilizes radar chirp linearity and broadband communications.

Simulating sub-terahertz interconnects with simple vertical roughness numbers yields transmission models that underpredict attenuation by dozens of decibels per meter, forcing costly board redesigns after prototype testing reveals severe eye closure.

Scan

Atomic force microscopy captures surface topography through direct mechanical interaction between an ultra-sharp silicon cantilever tip and the copper foil surface. The instrument rasters the tip across a defined area, measuring vertical deflection through an optical beam bounce system. Piezoelectric actuators position the tip with sub-angstrom accuracy, generating a discrete three-dimensional height matrix z(x,y).

Extracting meaningful power spectral density data from this matrix requires strict control over scanning parameters, sampling intervals, and tip geometry.

Scanning field selection balances resolution against representativeness. A scan window measuring 10 micrometers by 10 micrometers with a grid of 1024 by 1024 points yields a lateral pixel resolution of 9.77 nanometers, which captures the fine spatial structures where sub-terahertz currents circulate. However, this window misses long-wavelength undulations such as copper foil waviness and lamination roll marks.

A larger 50-micrometer scan resolves long-wavelength features but averages out nanoscale tips due to finite sampling intervals. High-fidelity modeling combines multi-scale scans, stitching 5-micrometer, 20-micrometer, and 100-micrometer measurements into a unified dataset.

A toroidal inductor and a sample of white paste sit on a glass slide, positioned on a laboratory bench.

How Do Cantilever Tip Radii Distort Profile Data?

Physical cantilever tips terminate in a finite spherical radius, typically ranging between 2 and 10 nanometers for sharp silicon probes, while durable diamond-like coated probes extend to 25 nanometers. As the probe moves over high-aspect-ratio features, the sidewall of the tip contacts the copper nodule before the apex reaches the bottom of the crevice. This physical interaction creates tip convolution, artificially widening peaks and narrowing or rounding valleys.

Probe tip radii introduce geometric dilation. In copper foils treated with micro-nodular dendritic deposits, tip convolution causes severe high-frequency spatial filtering. The measured height map displays smoother contours than the actual metallic boundary.

Deconvolution algorithms mitigate this distortion by utilizing blind tip reconstruction methods against calibrated characterization samples, restoring steep sidewall profiles necessary for accurate electromagnetic extraction.

Metrology Method Comparison for Sub-Terahertz Copper Characterization
Metrology Platform Lateral Resolution Vertical Resolution Maximum Spatial Frequency Measurement Area
Contact Stylus Profilometry 1000 nm to 5000 nm 0.5 nm 0.001 nm⁻¹ 10 mm × 10 mm
White Light Interferometry 350 nm to 1000 nm 0.1 nm 0.003 nm⁻¹ 1 mm × 1 mm
Confocal Laser Microscopy 150 nm to 300 nm 1.0 nm 0.007 nm⁻¹ 200 µm × 200 µm
Tapping-Mode Atomic Force 1 nm to 5 nm 0.02 nm 0.500 nm⁻¹ 10 µm × 10 µm

Selecting the tapping mode, or intermittent contact mode, preserves surface features and tip sharpness. Continuous contact mode applies lateral shear forces that break fragile copper micro-teeth and strip protective anti-tarnish treatments, generating debris that contaminates the scan line. Tapping mode oscillates the cantilever near its resonant frequency, lightly contacting the surface at the bottom of each swing to eliminate lateral drag.

Fabricators must prepare bare foil samples without aggressive solvent cleaning that could strip the organic silane adhesion monolayer.

Under IPC-TM-650 Method 2.5.5.14, surface profiles recorded without tip deconvolution yield high-frequency spatial energy values suppressed by up to twenty-five percent.

Standard optical interferometry is frequently cited as satisfying high-frequency specifications, assigning laboratory discrepancies to fabricator lamination parameters rather than metrology limitations.

Partition

Two-dimensional discrete Fourier transformation converts the spatial height matrix z(x,y) into the spatial frequency domain, producing the two-dimensional power spectral density. The continuous isotropic power spectral density function, PSD(f), expresses roughness power per unit spatial frequency, measured in units of length to the fourth power. Dividing this mathematical continuum into spatial frequency bands enables electromagnetic field solvers to calculate surface resistance increases without resorting to idealized geometric approximations.

The mathematical operation calculates the squared magnitude of the spatial Fourier transform normalized by the sampling area. For an N by M matrix with sample intervals delta-x and delta-y, the transformation maps height distributions across spatial frequencies fx and fy:

PSD(fx, fy) = (delta-x delta-y / (N M)) |sum_{m=0}^{N-1} sum_{n=0}^{M-1} z(m, n) exp(-j 2 pi (fx m delta-x + fy n delta-y))|^2

Averaging the two-dimensional distribution over concentric annular rings yields the one-dimensional radial power spectral density, PSD(f_r), where f_r represents the radial spatial frequency equal to the square root of fx squared plus fy squared. This reduction assumes planar isotropy, an assumption valid for electrodeposited foils but invalid for rolled-annealed foils exhibiting distinct longitudinal grain rolling marks.

Several insulated cables pass through a toroidal current transformer mounted next to an integrated circuit on a dark printed circuit board.

Translating Topography into Surface Impedance

Electromagnetic interaction partitions the radial spectrum into distinct operational regimes based on the ratio of spatial wavelength to electromagnetic skin depth. Low spatial frequencies, where spatial wavelengths exceed twenty skin depths, produce bulk current displacement without altering local surface resistance. Mid-range spatial frequencies, where spatial wavelengths sit between one and ten skin depths, force currents into serpentine pathways that generate inductive reactance and elevate resistive dissipation.

High spatial frequencies, possessing wavelengths comparable to or smaller than the skin depth, scatter the electromagnetic field directly.

Stochastic surface impedance models integrate the power spectral density across the active spatial frequency band to generate a modified surface impedance boundary condition:

Zs(omega) = Rs0 (1 + j) Xi(omega)

The term Rs0 represents the intrinsic surface resistance of smooth copper, equal to the square root of (omega mu / (2 sigma)), and Xi(omega) acts as the complex roughness correction factor derived from the spectral integration:

Xi(omega) = 1 + integral_{f_min}^{f_max} W(f_r, delta) PSD(f_r) 2 pi f_r df_r

The weighting kernel W(f_r, delta) accounts for current concentration within the skin layer delta. As frequency escalates toward 300 GHz, delta contracts, shifting the effective integration window toward the upper limit f_max, set by the sampling resolution of the atomic force microscope. Truncating the measurement at an inadequate f_max discards critical roughness power, causing the model to underestimate attenuation.

A green substrate featuring complex conductive trace routing mounts modular tactile input components within a curved support frame structure.

Worked Parameter Extraction across Two Copper Foils

Evaluating two candidate foils for a 140 GHz automotive radar antenna feed illustrates this mechanism. Foil A is a standard very-low-profile electrodeposited foil; Foil B is a high-grade rolled annealed foil. Both foils undergo tapping-mode characterization over a 20-micrometer scan window with 1024 data points per axis, providing a spatial frequency sampling range from 0.05 inverse micrometers to 25.6 inverse micrometers.

  • Low spatial frequency power below 0.2 inverse micrometers measures 8.4 × 10⁻⁴ µm⁴ for Foil A, compared to 1.2 × 10⁻³ µm⁴ for Foil B due to mechanical rolling textures.
  • Mid spatial frequency power between 0.2 and 2.0 inverse micrometers spans 4.6 × 10⁻⁴ µm⁴ for Foil A, whereas Foil B suppresses this band down to 3.1 × 10⁻⁵ µm⁴.
  • High spatial frequency power between 2.0 and 25.6 inverse micrometers reaches 1.8 × 10⁻⁴ µm⁴ in Foil A, driven by micro-nodular growth, while Foil B maintains 8.2 × 10⁻⁶ µm⁴.
  • Total root-mean-square height calculates to 0.38 micrometers for Foil A and 0.18 micrometers for Foil B across the unified scan window.

At 140 GHz, skin depth equals 0.175 micrometers. Conductor loss calculations using the classical Hammerstad formula predict a roughness attenuation factor of 1.48 for Foil A and 1.16 for Foil B. Incorporating the complete power spectral density through stochastic integration yields true roughness factors of 2.82 for Foil A and 1.31 for Foil B. Conductor loss calculations climb steeply under the integrated spectral method.

Optical telescope mirror polishers characterize surface scatter through identical power spectral density metrics to control stray light propagation across astronomical instruments. In printed circuits operating in sub-terahertz regimes, copper conductor traces experience identical physics, transforming microscopic surface textures into scattering centers that degrade electrical signal integrity.

A specific figure commonly referenced across high-frequency design suites is a 0.22-micrometer effective Huray sphere radius for treated rolled foil, resting on twenty-four cross-sectional transmission electron microscopy samples measured in 2019. If the underlying copper chemistry introduces nodule distributions shifted toward spatial wavelengths below 80 nanometers, that fixed radius underpredicts conductor loss by eighteen percent at 220 GHz.

The unresolved dispute centers on whether current distributions inside individual sub-micron nodules follow macroscopic Maxwellian equations or encounter anomalous skin effect phenomena where electron mean free paths disrupt standard Ohm’s law relationships.

Tooth

Laminators do not press flat copper sheets against dielectric prepregs. Bare copper exhibits poor natural adhesion to fluoropolymer and polyphenylene ether resins. To establish mechanical interlocking, foil mills run base copper through secondary electrolytic plating baths that generate dendritic copper crystals, colloquially termed the copper tooth.

These micro-nodules lock into the resin matrix during the thermal lamination cycle, providing peel strengths capable of passing IPC-TM-650 thermal stress tests.

Every mechanical tooth embedded into the dielectric increases the high-frequency dissipation path. In sub-terahertz packaging substrates, electrodeposited foils present anchor profiles that penetrate deep into the substrate core. When an electromagnetic wave propagates down a microstrip or stripline, current concentrates at the copper-dielectric interface due to the dielectric constant step.

The rough, treated side of the copper foil invariably sits precisely where current density reaches its absolute maximum.

A digital render presents grey industrial electronics production machinery featuring a modular assembly rail and an open cabinet containing structured internal wiring harnesses.

Will Micro-Etch Treatments Degrade High Frequency Impedance?

Fabrication shops utilize chemical micro-etching baths to clean copper surfaces and promote adhesion prior to photoresist application and final bonding. Typical chemistries include sodium persulfate, hydrogen peroxide with sulfuric acid, or ferric chloride solutions. These chemical baths attack grain boundaries preferentially, removing between 0.5 and 1.5 micrometers of copper.

Micro-etching alters surface morphology unevenly. An aggressive micro-etch erodes high-profile peaks while simultaneously excavating deep intergranular trenches, significantly expanding the high-frequency spatial power spectral density. A foil that entered the facility with smooth rolled characteristics exits the wet process line with an altered topography that drives up insertion loss.

Foil Metallurgy, Processing, and Measured High-Frequency Performance
Foil Classification Treatment Type Peel Strength (N/mm) Sq Roughness (nm) Attenuation at 140 GHz (dB/mm)
Standard Electrodeposited (STD) Coarse Zinc/Copper Nodules 1.45 850 0.48
Very Low Profile (VLP) Fine Micro-Nodular Brass 1.05 380 0.29
Hyper Very Low Profile (HVLP) Sub-Micron Silane/Cobalt 0.82 190 0.18
Rolled Annealed (RA) Smooth Organic Passivation 0.65 110 0.14
Atomically Smooth Sputtered Titanium Adhesion Layer 0.45 25 0.09

Alternative chemical bonding systems replace destructive mechanical etching with ultra-thin chemical coupling agents. Organosilane and tin immersion processes deposit molecular monolayers that chemically bond copper atoms to dielectric polymer backbones. These alternative systems achieve IPC Class 3 peel strengths above 0.8 Newtons per millimeter on low-loss substrates while preserving the base foil power spectral density up to 40 inverse micrometers.

Plated micro-nodules residing within the dielectric interface define conductor attenuation at 140 GHz regardless of top-side foil smoothing operations.

Trace and space feature sizes down to 25 micrometers in sub-terahertz packaging boards complicate processing windows. When fabricators utilize standard subtractive etching on heavy foil, the chemical etchant dissolves trace sidewalls unevenly, creating trapezoidal trace profiles with rough vertical flanks. The high-frequency current redistributes along these rough lateral sidewalls, increasing line attenuation and introducing significant impedance variance across production panels.

  1. The incoming bare copper foil undergoes atomic force characterization to record baseline spatial frequency distributions across rolling directions.
  2. Chemical cleaning tanks apply non-etching alkaline degreasers to remove organic anti-tarnish films without disturbing metallic grain boundaries.
  3. An alternative chemical bonding line deposits an organosilane adhesion monolayer possessing an optical thickness below ten nanometers.
  4. Vacuum lamination applies controlled temperature and pressure profiles to prevent thermal shearing of high-aspect dielectric features.
  5. Differential impedance coupons undergo vector network analyzer testing across the entire waveguide operational band to verify attenuation limits.

Rougher interface treatments consistently yield higher mechanical peel strengths at the direct expense of signal transmission efficiency.

Ledger

Specifying copper foil through atomic force power spectral density parameters restructures the board procurement process. Standard commercial purchase orders reference IPC-4562 foil designations, specifying grade and weight classes such as one-third-ounce or one-half-ounce foil. These generic designations contain no enforceable limits on high-frequency spatial power spectral density, permitting laminators to substitute high-loss nodular foil variants as long as they satisfy static weight and tensile thresholds.

High-frequency stackup definitions dictate panel economics directly. A standard 18 by 24-inch panel fabricated from mid-loss glass-reinforced hydrocarbon laminate clad with standard electrodeposited foil costs approximately eighty dollars at volume. Migrating that stackup to an unreinforced polytetrafluoroethylene or liquid crystal polymer core clad with hyper-very-low-profile foil increases panel material costs to four hundred dollars.

Utilizing rolled annealed foil with proprietary organosilane bonding steps pushes raw panel costs past seven hundred and fifty dollars.

A machine die tool precisely forms a thin metallic sheet onto a darker substrate alongside a copper conductor strip.

Commercial Yield and Scrap Penalties

Smooth foils present severe manufacturing challenges that depress shop-floor yields. Rolled annealed foils exhibit low mechanical tooth profiles, resulting in reduced peel strength that increases susceptibility to trace lifting during automated surface-mount assembly. During multi-layer lamination cycles, low-profile foils generate lower mechanical friction against prepreg resin fronts, aggravating inner-layer registration slip across large 24-inch panel formats.

Inner-layer registration errors exceeding 35 micrometers force the fabricator to scrap whole panels when microvia capture pads fail minimum annular ring requirements defined by IPC-6012 Class 3. A shop experiencing a five percent registration scrap rate on standard foil frequently sees scrap rates escalate to twenty-two percent when processing smooth rolled foils on thin sub-100-micrometer cores. The customer absorbs this manufacturing fallout through increased unit board pricing.

Moving from HVLP to rolled annealed foil does not invariably cut total system losses by 0.05 dB per millimeter at 140 GHz across all lamination presses. Resin flow dynamics vary significantly between low-pressure vacuum autoclaves and high-pressure hydraulic presses, altering profile compression. In the presence of this unquantified operational variance, physical vector network analyzer coupon testing across five points on every production panel provides necessary verification beyond analytical datasheet simulations.

Line yields depend directly on copper foil selections. When the fabrication drawing omits spatial frequency cutoffs, the shop optimizes for manufacturing yield rather than electrical performance, selecting higher-tooth foils that laminate reliably without delamination. Explicit fabrication drawing notes protect electrical budgets by binding the supplier to quantitative physical parameters.

Per IPC-6012 Class 3 Section 3.4.1, any substitution of copper foil grade or bonding treatment requires formal engineering change authorization and invalidates baseline qualification testing, shielding the buyer against silent material shifts that degrade sub-terahertz performance.

Nomenclature

Transmission Lines

Impedance Profile ~ Conductive pathways that link drivers to receivers on a printed circuit board must maintain controlled physical dimensions to preserve signal integrity at high frequencies.

Atomic Force Microscopy

Topographic Resolution ~ High-resolution scanning probe instruments quantify sub-nanometer surface features across printed circuit board substrates and semiconductor dies by dragging or vibrating a physical tip over a sample.

Standard Electrodeposited

Deposition Profile ~ Copper foil produced through standard electrodeposited manufacturing relies on cathodic reduction from an acidic copper sulfate bath onto a rotating titanium drum.

Copper Foil

Conductive Material ~ Metallic sheets used to create the electrical pathways on a printed circuit board substrate.

IPC-6012 Class 3

High Reliability Requirement ~ Performance criteria for electronic hardware defines strict acceptance limits for mission critical printed circuit boards where board failure or interruption of function results in danger to human life or equipment loss.

Fluoropolymer Laminate

Substrate Material ~ High-frequency circuit boards requiring low dielectric constant and low dissipation factor utilize polytetrafluoroethylene matrices reinforced with micro-glass fibers or ceramic fillers.

Rolled-Annealed Foil

Grain Structure ~ Wrought copper material produced through successive mechanical thinning provides a smooth and flexible conductive surface for specialized circuit applications.

Sub Terahertz

Frequency Band ~ High frequency energy spanning three hundred gigahertz up to three terahertz operates above conventional microwave allocations while remaining below optical spectrum limits.

Copper Roughness

Etching Topography ~ Microscopic surface profile of a printed circuit board foil defines copper roughness as the physical irregularity present on treated barrier layers.

Conductor Attenuation

Signal Degradation ~ Electromagnetic energy loss within a copper trace occurs as signal frequencies rise, causing amplitude reduction through dielectric absorption and resistive heating of the metal.

Surface Impedance

Electromagnetic Ratio ~ Frequency-dependent ratio between tangential electric field and surface current density along a conductive boundary defines the complex boundary condition governing high-frequency wave propagation.

Hammerstad Model

Copper Distribution ~ Quantitative analysis of electrolyte thickness across high aspect ratio vias characterizes the hammerstad model.

What the firm knows, published

Expertise is a utility, not a secret. sentiention™ publishes its working knowledge as open reference: intelligence layer covering the materials it sources, the markets it enters, and the reference that serves both.