Mechanical cross-sectioning is still the workhorse of package-level failure analysis. This guide covers how we plan, mount, grind, lap, and polish semiconductor and microelectronic samples so the defect survives the preparation and the cross-section tells the truth on the first attempt.
FIB and other ion-beam methods get most of the attention, but most package-level failure analysis is still done on a polishing wheel. Solder joint fatigue, wire bond liftoff, die attach voiding, mold compound cracking, plated through-hole barrel cracks, and MLCC delamination are all features at the micrometer scale or larger. They need millimeters of section width, not tens of micrometers, and they are answered fastest and most economically by a well-executed mechanical cross-section.
The phrase that matters is well-executed. A mechanical section that smears copper across a crack, pulls out an intermetallic, or overshoots the target row of solder balls does more than waste a sample. It can produce a convincing image of the wrong answer. This article, Part 2 of the Watson FA Lab Cross-Sectioning Series, walks through the procedure we use for mechanical polishing and lapping of semiconductor and microelectronic samples, from planning the cut to the final polish, and the decisions that separate a reliable section from an artifact.
FA, reliability, and quality engineers who order or perform cross-sections and want to understand what drives section quality, turnaround, and interpretation. Part 1 of this series compares mechanical polishing with FIB milling, ion milling, and chemical etching; this article goes deep on the mechanical method.
When Mechanical Cross-Sectioning Is the Right Choice
The first decision in any cross-section is not how to polish but whether polishing is the right tool. Mechanical sectioning is the right primary technique when the feature of interest is about 1 µm or larger, when the section must span a large area (a full row of BGA balls, a complete wire bond from ball to stitch, or an entire via chain), and when the defect location is known to within tens to hundreds of micrometers rather than to a few nanometers.
- Best For
- Package, board, and interconnect features ≥ 1 µm
- Section Width
- Millimeters to several centimeters
- Depth Control
- ±5–25 µm with fiducials, X-ray, and inspection intervals
- Typical Bench Time
- 30 minutes to 3 hours per sample, plus cure time
It is the wrong primary technique for a single gate, contact, or via at an advanced node, for any target that requires nanometer depth placement, or for TEM preparation. In those cases mechanical work may still be the first step, removing bulk material to bring the target within FIB range, but the final face belongs to the FIB. Part 3 of this series covers that handoff in detail.
Before the First Cut: Planning the Section
Cross-sectioning is destructive and one-directional. Once a plane has been ground through, the material in front of it is gone. Every non-destructive question should therefore be answered before the sample is mounted. In our lab, a typical package cross-section is preceded by:
- External optical inspection and photo documentation of all surfaces, markings, and any visible damage, so the as-received condition is on record.
- X-ray inspection to map wire bonds, solder voids, die attach coverage, and internal structure, and to choose the plane that passes through the suspect feature.
- Scanning acoustic microscopy to locate delaminations and cracks at internal interfaces that X-ray cannot see.
- Electrical verification and, where needed, fault isolation, so the section is aimed at the failing net rather than at a general area.
With that information, define three things in writing before cutting: the target plane (for example, through the centers of balls A1 to A12), the approach direction (which edge you grind from, chosen so the most important feature is reached last and with the finest abrasive), and the stop criterion (what you expect to see on the face when you have arrived). A section plan agreed with the customer up front prevents the most expensive outcome in cross-sectioning: a beautiful polish through the wrong plane.
Rough Sectioning and Sample Sizing
Boards and large assemblies are first reduced to a coupon that fits the mounting cup, using a low-speed diamond wafering saw with coolant. Leave a generous margin, typically 1 to 3 mm, between the saw cut and the target plane. Saw damage, chipping, and heat extend well below the visible kerf, and every micrometer of that damage must be ground away before the target is reached. Cutting too close to save grinding time is a common way to crack a die or a ceramic capacitor before the real work begins.
Mounting and Encapsulation
The mount does three jobs: it gives you a flat, hand-sized body to hold square to the wheel, it supports fragile structures so they do not chip or crack, and it fills internal gaps so edges and voids are preserved rather than rounded or smeared. Most section quality problems that show up at the final polish were actually created at mounting.
Cold Mounting vs. Hot Compression Mounting
For semiconductor and electronic samples, cold-mount epoxy is the default. Hot compression mounting cures quickly and gives excellent edge retention, but it applies elevated temperature and significant pressure. That combination can crack thin die and ceramic parts, and it can alter or even melt low-temperature solders: tin-bismuth alloys melt near 138 °C, within the range used by many hot-mount cycles. Hot mounting also anneals and changes the microstructure that the analysis may be trying to characterize.
Within cold-mount systems, choose a low-shrinkage, low-exotherm epoxy. Fast-curing acrylics are convenient but shrink away from the sample and generate more heat, which opens a gap at the sample–resin interface. That gap fills with abrasive and debris, rounds the edges, and can look like a delamination in the final image.
Vacuum Impregnation
Packages and assemblies are full of voids: under-fill gaps, the space beneath a component, delaminations, cracks, and plated through-holes. If these are left empty, the abrasive drives debris into them and the walls collapse or smear during grinding. Vacuum impregnation pulls air out of these cavities and lets low-viscosity epoxy flow in before it cures, so cracks and delaminations are supported and preserved exactly as they existed in the failed part. For crack and delamination analysis, vacuum impregnation is not optional.
Adding a fluorescent dye to the impregnation epoxy makes pre-existing cracks and delaminations that were open to the epoxy glow under UV or fluorescence illumination. A crack filled with dyed epoxy existed before mounting. A crack with no epoxy in it was most likely created during preparation.
Edge Retention and Orientation
Features at the edge of the sample, such as the top surface of a die or the outer layer of a board, round off faster than material in the middle of the mount. Placing sacrificial pieces of similar hardness beside the critical edge, or using an epoxy with hard filler, keeps the polishing plane flat out to the edge. Orient the sample in the cup so the target plane is parallel to the mount face, using sample clips and a reference mark on the mount. Every degree of tilt at mounting turns into a gradient in depth across the section, which matters when you are trying to stop on the center of a row of 250 µm solder balls.
Grinding: Coarse Removal to the Approach Zone
Grinding removes the bulk of the material between the saw cut and the target. The principle that governs every step is simple: each abrasive step must remove the scratches and the subsurface damage left by the step before it. Subsurface damage extends below the visible scratch, so as a working rule remove at least two to three times the particle size of the previous abrasive before moving on. Skipping a grit to save time almost always costs more time later, because the deep scratches resurface at the fine polish.
Grind with water lubrication, moderate load, and moderate wheel speed. Excess load is the most common error: it generates heat, embeds abrasive in soft metals, cracks brittle layers, and tilts the sample. Rotate the sample 90° between grits so that the new scratch pattern runs perpendicular to the old one. When the old scratch direction has completely disappeared, the step is finished.
Coarse grinding should stop in the approach zone, typically 50 to 100 µm short of the target plane for package features and closer for fine features. From there, only fine grinding and lapping steps are used, so that the last material removed above the target is removed gently. The table below summarizes a typical sequence; actual steps depend on the materials in the section and on the size of the target.
Stage | Abrasive | Purpose | Move On When |
|---|---|---|---|
Coarse grind | SiC paper, P240–P400 | Remove saw damage; establish a flat plane | Plane is flat and within ~100 µm of the target |
Fine grind | SiC paper, P800–P1200 | Controlled approach; reduce damage depth | Within ~25–50 µm of the target; previous scratch direction gone |
Final grind | SiC paper, P2500–P4000 | Final approach to the target plane | Target just reached or within a few µm, confirmed optically |
Lapping / diamond polish | Diamond film or suspension, 9 → 3 → 1 µm | Remove grinding damage with minimal relief | Uniform fine scratch pattern at 200–500× optical |
Final polish | Colloidal silica or alumina, ~0.05 µm | Scratch-free face for SEM/EDX and optical imaging | No visible scratches at high magnification; no smear on metals |
Table 1. Typical mechanical cross-section sequence. Grit designations are FEPA P-grades. Distances to target are typical for package-level features and tighten for smaller targets.
Depth Control: Stopping on the Feature
Depth control is the central skill in mechanical cross-sectioning. A 250 µm solder ball sectioned 60 µm off center shows a narrower joint, a misleading intermetallic thickness, and possibly no crack at all, even though the crack is there in the center plane. A wire bond sectioned slightly off its axis can look lifted when it is not. The tools for getting the depth right are straightforward; the discipline is in using them every time.
- Measure, then grind. Measure the distance from the mount face to the target before each stage, using X-ray images of the mounted sample or a micrometer against a known reference, and calculate how much remains to be removed.
- Use the structure as a ruler. Ball diameters, bond wire diameters, via drill sizes, and die thickness are known dimensions. The width of a sectioned sphere or cylinder tells you exactly how far you are from its center.
- Inspect at short intervals near the target. Far from the target, inspect every few minutes. In the final 25 µm, inspect after every few passes. Grinding is fast and irreversible; inspection is cheap.
- Correct tilt early. If features at one end of the row are wider than at the other, the plane is tilted. Correct it with uneven pressure during fine grinding, while there is still material to work with.
- Leave room for polishing. Diamond and final polishing steps still remove a few micrometers. Stop grinding just short of the center plane, and let polishing take you the rest of the way.
If a critical feature is passed, the information in front of the plane is gone. When the sample is irreplaceable, as with a field return, a litigation exhibit, or a single flight part, slow down near the target and inspect more often. Going back for a second sample costs days. For a one-of-a-kind sample, there is no second chance.
Lapping and Diamond Polishing
Once the plane is at the target, lapping and diamond polishing remove the remaining grinding damage. Diamond lapping films on a rigid platen keep the face flat and minimize relief between hard and soft materials, which matters in a section that contains silicon, copper, solder, ceramic, and epoxy side by side. Diamond suspensions on low-nap or napless cloths are used for the 3 µm and 1 µm steps. Higher-nap cloths polish faster but round edges and exaggerate relief, so we reserve them for the final step on samples where edge retention is less critical.
Cleanliness between steps is as important as the steps themselves. A single 9 µm diamond particle carried onto the 1 µm cloth will leave a deep scratch across the final face. Rinse the sample and your hands thoroughly between every step, keep each abrasive size on its own dedicated cloth and platen, and cover cloths when they are not in use. Ultrasonic cleaning is effective, but use it with care on samples containing cracks or delaminations, because cavitation can enlarge them or dislodge loose material.
Final Polishing
The final polish produces the face that will be imaged. Colloidal silica, typically around 0.05 µm, combines mild chemical action with very fine abrasion. It is particularly effective at removing the smeared layer on copper, aluminum, and solder that diamond leaves behind, and it reveals grain structure and intermetallic boundaries that are otherwise hidden. Alumina suspensions are an alternative where the chemistry of colloidal silica is not wanted.
Keep final polishing short and controlled. Long final polishes create relief: soft phases such as solder recede below hard phases such as silicon and ceramic, and edges round. In SEM that relief shows up as shadowing and false topography at interfaces, which can be mistaken for gaps or delamination. Rinse immediately and thoroughly after colloidal silica. If it is allowed to dry, it forms a film that is difficult to remove and appears as a surface contaminant in EDX.
Recognizing and Preventing Preparation Artifacts
Every mechanical section contains some preparation artifacts. The question for the analyst is whether each feature in the image is real or created by preparation. Experienced reviewers ask that question about every void, crack, and gap in a section before it goes into a report. The table below lists the artifacts we see most often in semiconductor and electronic assemblies, with their causes and the checks that separate them from real defects.
Artifact | What It Can Be Mistaken For | Usual Cause | Prevention and Check |
|---|---|---|---|
Metal smearing | A crack that is not there (crack hidden under smeared metal), or a bridge or short between conductors | Ductile Cu, Al, Au, or solder deformed by diamond or excess load | Light loads; colloidal silica final polish; local FIB cleanup of the critical area if needed |
Pullout | Voids in solder, intermetallic, ceramic, or tungsten fill | Brittle grains or particles torn out during coarse steps | Vacuum impregnation; lower loads; real voids have smooth walls, pullouts have angular, fractured walls |
Preparation cracks | Die cracks, MLCC flex cracks, solder fatigue cracks | Sawing too close, hot mounting, excess grinding load | Cold mount with dyed epoxy; real cracks contain epoxy, preparation cracks do not |
Edge rounding and gaps | Delamination at die surface, mold compound, or board edge | Resin shrinkage, high-nap cloths, long final polish | Low-shrinkage epoxy; edge support pieces; correlate with acoustic microscopy before sectioning |
Relief | Interface gaps or steps in SEM images | Hard and soft phases polished at different rates | Rigid platens and films; short final polish; tilt the SEM stage to check whether the gap is real |
Embedded abrasive | Foreign particles or contamination; Si, C, or Al peaks in EDX of soft metals | SiC or diamond pressed into solder, gold, or aluminum | Light loads on soft materials; know which elements your abrasives contribute before interpreting EDX |
Table 2. Common mechanical preparation artifacts and how to tell them from real defects.
Material- and Structure-Specific Considerations
Solder Joints and Intermetallics
Solder fatigue cracks typically initiate at or near the intermetallic layer at the component or board side of the joint. Section through the center of the ball, because both crack length and intermetallic thickness are measured relative to the full joint width. Solder smears readily, so finish with colloidal silica and light loads, and expect to make a final pass after a brief inspection if the crack tip is not clearly resolved. For lead-free and low-temperature alloys, avoid any heat during mounting that could change the microstructure being evaluated.
Wire Bonds
Ball bonds are usually sectioned through the center of the ball, parallel to the wire, to show the intermetallic coverage at the bond pad, cratering beneath the pad, and the neck. Gold and copper wire are soft and smear easily, and a slight angle between the section plane and the wire axis can make a well-bonded ball look partially lifted. Align the wire axis to the section plane during mounting and confirm alignment as the ball first appears.
Die, Passivation, and Back-End-of-Line Structures
Silicon and dielectric stacks are brittle; low-k dielectrics in particular crack and delaminate under mechanical load. Mechanical sectioning is well suited to verifying passivation cracks, die cracks, and thick top-metal features, but individual contacts, vias, and gate structures at advanced nodes belong to the FIB. A common and efficient approach is to section mechanically to within a few micrometers of the target, then finish the critical site by FIB.
Printed Circuit Boards and Ceramic Capacitors
Plated through-holes and microvias are typically sectioned to the hole center to evaluate plating thickness, barrel cracks, corner cracks, and interconnect separation, following the microsection practices of IPC-TM-650 Method 2.1.1. Multilayer ceramic capacitors are sectioned to reveal flex cracks, which usually run diagonally from a termination into the active region, and internal delamination between electrode layers. Both are brittle, so vacuum impregnation and gentle coarse grinding are essential to keep preparation cracks out of the result.
Medical and High-Reliability Assemblies
Implantable and space-grade assemblies often combine dissimilar materials, such as hermetic glass-to-metal seals, thick-film ceramics, polymer coatings, and precious-metal interconnects, in a single section. These samples are frequently irreplaceable and the analysis may be subject to regulatory review. The section plan, the chain of custody, and every preparation step should be documented so that the result can be defended, and depth control should be at its most conservative.
From Polished Face to Answer: Imaging and Handoff
A polished section is imaged first with optical microscopy, which shows color differences between metals and intermetallics and gives a quick overview at low magnification. SEM imaging follows for higher resolution and for EDX microanalysis. Mounted samples are non-conductive, so they are either imaged at low voltage and low vacuum or given a thin conductive carbon coat and grounded with conductive tape. Carbon is preferred over metal coatings when EDX will be used, because a metal coat adds peaks that can obscure the elements of interest. Imaging and microanalysis of cross-sections are the subject of Part 4 of this series.
When the defect is smaller than mechanical preparation can resolve, or when smearing persists at a critical interface, the mechanically prepared sample becomes the starting point for a FIB cross-section or a TEM lamella. The mechanical section has already removed the bulk material, and the FIB only needs to cut the last few micrometers at a precisely chosen site.
Non-Destructive Inspection
External optical, X-ray, and acoustic microscopy locate the suspect feature and define the target plane.
Section Plan
Target plane, approach direction, and stop criterion are agreed with the customer before any material is removed.
Rough Cut and Mount
Low-speed saw cut with a 1–3 mm margin, then vacuum-impregnated cold-mount epoxy for support and edge retention.
Grind and Approach
Progressive SiC grinding with frequent measurement, stopping just short of the target plane.
Lap and Polish
Diamond lapping and polishing, then a short colloidal silica final polish for a scratch-free, low-relief face.
Image, Analyze, or Hand Off
Optical and SEM/EDX analysis of the section, or FIB finishing where the target is below about 1 µm.
How Watson Analytical Applies These Techniques
In our San Diego laboratory, mechanical cross-sectioning is planned from the non-destructive data. Each section is preceded by X-ray inspection and, where interfaces are in question, scanning acoustic microscopy, so that the section plane is chosen from evidence rather than guesswork. Polished sections go directly to SEM/EDX analysis in the same lab, and when a target is below the reach of mechanical preparation we finish the site with FIB cross-sectioning without shipping the sample to another facility.
We regularly section solder joints, wire bonds, die attach, flip-chip bumps, plated through-holes, and passive components for semiconductor, defense, space, and medical device customers, and for patent evidence of use investigations where a documented, defensible section is the deliverable. Work is performed under our ISO 9001:2015 quality system, and ITAR-controlled samples are handled under our ITAR registration. Typical turnaround for cross-sections is 2 to 3 days.
Need a cross-section that answers the question on the first attempt? Send us your failure description and we will propose a section plan.
Talk to Watson's FA Engineering Team →References
- Vander Voort, G.F. Metallography: Principles and Practice. ASM International, 1999.
- Shea, P. (Ed.). Microelectronics Failure Analysis Desk Reference, 7th ed. ASM International, 2019.
- ASTM E3, Standard Guide for Preparation of Metallographic Specimens. ASTM International.
- IPC-TM-650 Test Methods Manual, Method 2.1.1, Microsectioning. IPC.
- Goldstein, J.I. et al. Scanning Electron Microscopy and X-Ray Microanalysis, 4th ed. Springer, 2018.
Watson FA Lab Cross-Sectioning Series
- 01Overview of Cross-Sectioning Techniques
- 02Mechanical Polishing and Lapping — Procedures & Best Practices (this article)
- 03FIB Milling and Dual-Beam SEM/FIB — Procedures & TEM Lift-Out (coming soon)
- 04SEM/EDX Analysis of Cross-Sections — Imaging, Microanalysis & Interpretation (coming soon)
- 05Chemical & Wet Etching — Etchant Selection, Rate Control & Safety (coming soon)


