Chemical mechanical planarization (CMP) combines surface chemistry and mechanical contact to remove material and reduce wafer topography. Removal depends on the material, pad, slurry, motion, and local pattern density; high features are often removed preferentially, but complete global flatness is not automatic. This article explains the pad–slurry–wafer interaction and the linked STI and dielectric CMP stations.
Where CMP Acts in the Flow
The free 40nm BSI CMOS Image Sensor flow names two CMP stations. Its public cross-sections show integration context, not measured flatness:
- Isolation polish: STI CMP follows trench filling; the figure does not measure the final height relative to the field.
- Dielectric polish: PMD 3 CMP is a later pre-metal dielectric example; the figure does not identify its pad or slurry.
Each step is publicly visible with its flow schematic from the 40nm flow overview; the flow is free.
Three named CMP-related stations from the free 40nm flow; the diagrams do not establish slurry, endpoint, or achieved flatness.
Process checkpoint
Understand STI CMP in context
Inspect the named STI CMP station without inferring its consumables.
Process context for “Chemical Mechanical Planarization: Principles, Mechanisms, and Advanced Node Integration”: 40nm BSI CMOS Image Sensor · STI · Step 42
Physics & Mechanism
CMP combines chemical surface reactions with mechanical contact among a wafer, pad, and slurry. Preston-type pressure–velocity scaling is a useful approximation in some regimes , but the effective removal rate also depends on local pattern, consumables, and material response. Elevated features may contact the pad more strongly; that does not guarantee a globally flat result.
A review of CMP scratches describes how contact and consumables can create defects . A separate copper CMP experiment compared electrochemical corrosion with total removal and found that, under its tested conditions, most copper was removed by sliding abrasive particles . Chemistry still changes the surface and its mechanical response. That result cannot be transferred without measurement to an oxide, STI, PMD, or different copper slurry.
Endpoint control may use time, in-situ signals, or post-polish measurements, depending on the station. The linked STI CMP and PMD 3 CMP cross-sections establish where polishing occurs, but neither gives the slurry, stop material, endpoint method, or measured removal profile.
Process Principles
- Specify the material pair: removal and selectivity depend on film and slurry, not merely the CMP label.
- Measure local and global outcomes: thickness, dishing, erosion, and scratches are distinct outputs .
- Bound removal models: pressure–velocity scaling and the copper experiment apply under stated conditions.
- Confirm the endpoint: a named CMP step does not reveal how the process stops.
Challenges & Failure Modes
Possible problems include scratches, residue, dishing, erosion, and unintended loss of a neighboring film . Their incidence depends on the particular station and must be inspected. The copper mechanism study does not quantify these defects at the linked 40nm steps.
From Principle to Production Flow
A CMP integration generally includes pad preparation, material-specific polishing, endpoint control, post-polish cleaning, and verification, with details chosen for the actual stack. The public 40nm flow overview shows two named CMP stations but does not document those controls. For the slurry half of the system, see CMP slurry and abrasives; for the module map its stations belong to, see planarization and the CMP integration overview; for the uniformity discipline it serves, see uniformity; and for the post-polish clean that closes the module, see wet cleaning.
Technology Node Evolution
Material and geometry changes can require different slurries, pads, endpoints, and defect controls. The linked two CMP stations are examples in one flow, not a measured history of polishing performance across nodes.
Related Processes
Related topics include CMP slurry and abrasives, planarization, uniformity, and post-polish cleaning. Their relationships depend on the material stack and CMP station.
Future Outlook
Better endpoint sensing and material-specific models may improve control. Their value must be checked against measured thickness, defects, and device outcomes for the chosen stack.
References
Scratch formation and its mechanism in chemical mechanical planarization (CMP)
T. Kwon, M. Ramachandran, Jin-Goo Park
Material Removal Mechanism during Copper Chemical Mechanical Planarization Based on Nano-Scale Material Behavior
Seungchoun Choi, F. Doyle, D. Dornfeld
Silicon VLSI Technology - Full
James D. Plummer, Michael D. Deal, Peter B. Griffin
Silicon VLSI Technology · ISBN 978-0130850379