CMPポリッシングパッド：技術と比較

1. Why Polishing Pads Matter in CMP

The CMP polishing pad is the physical interface between the tool and the wafer. It is responsible for distributing the applied downforce across the wafer surface, transporting fresh slurry into the contact zone, and carrying spent slurry and reaction by-products away from the polishing interface. In doing so, the pad exerts a dominant influence over material removal rate, within-wafer uniformity, defect levels, and the cost structure of the CMP operation.

Despite its critical role, the polishing pad is often treated as a secondary variable in process development — receiving less attention than the slurry. This is a mistake. In practice, the pad and slurry form an inseparable system: the same slurry can produce dramatically different performance profiles when used with different pad types, groove patterns, or conditioning programs. Understanding pad technology in depth is essential for any engineer tasked with optimizing or troubleshooting a CMP process.

2. Pad Microstructure: What’s Inside a CMP Pad

The vast majority of commercial CMP pads are fabricated from ポリウレタン, a polymer selected for its chemical resistance, tuneable mechanical properties, and the controllability of its pore structure during manufacturing. Understanding the three-dimensional microstructure of a polyurethane CMP pad explains why its properties evolve during use and why conditioning is necessary to maintain process stability.

Closed-Cell Foam Pads (Hard Pads)

Hard CMP pads are produced by a cast polyurethane process in which a blowing agent or hollow microspheres (typically PTFE or glass) are incorporated into the polymer matrix during cure. The result is a closed-cell foam structure with pores of 20–80 µm diameter distributed throughout the pad thickness. These pores serve two functions: they reduce the effective elastic modulus of the pad (making it slightly more compliant than solid polyurethane) and they act as micro-reservoirs that hold slurry within the contact zone.

The surface of the pad as manufactured is relatively smooth — the pores are subsurface. Pad conditioning by a diamond disc opens the pore structure at the surface, creating the micro-textured asperity landscape that is essential for effective polishing. This is why pad conditioning is a prerequisite for stable polishing performance, and why pad break-in is required before a new pad reaches its steady-state MRR.

Open-Cell and Felt-Based Pads (Soft Pads)

Soft polishing pads use either a non-woven felt base impregnated with polyurethane, or an open-cell foam structure. These materials have significantly lower Shore hardness than hard pads and much higher porosity (often 50–70%). Their open pore structure provides excellent slurry absorption and retention, and their low modulus allows them to conform closely to the local wafer topography — the property that gives them their low-defect, low-scratch performance advantage.

3. Pad Types: Hard, Soft, Stacked & Fixed-Abrasive

4. Groove Pattern Engineering

The macro-scale groove pattern machined or embossed into the pad surface has a major influence on slurry transport, lateral pressure uniformity, and wafer-edge performance. Groove design is one of the most underappreciated variables in CMP pad selection, yet it can produce WIWNU differences of 1–3% across the wafer when changed while holding all other parameters constant.

Groove width (typically 0.25–0.75 mm), groove depth (0.3–0.8 mm), and groove pitch (1–3 mm) all interact with platen rotation speed and slurry flow rate to determine the effective slurry film thickness under the wafer. Optimization of these parameters is typically done empirically using blanket wafer MRR mapping experiments combined with computational fluid dynamics (CFD) modeling at the pad supplier’s application engineering center.

5. Key Physical Properties and Their Process Impact

Selecting a CMP pad requires understanding the relationship between measurable physical properties and on-tool process performance. The following properties are the most diagnostically useful for pad characterization and comparison:

6. Head-to-Head Comparison: Major Pad Families

7. Pad Break-In Protocols

A new CMP pad does not deliver its nominal performance from the first wafer. The pad surface as received from the manufacturer has a relatively smooth, closed microstructure that has not yet been opened by diamond conditioning. Before the pad reaches its steady-state MRR and uniformity, it must go through a structured break-in procedure that progressively opens the pad surface and stabilizes the conditioning response.

Break-in protocol duration and dummy wafer count vary by pad type: hard IC1000-type pads typically require 10–20 dummy wafers; softer pads may stabilize in fewer. Always follow the pad supplier’s recommended break-in procedure for the specific product, as formulation changes between pad generations can alter break-in behavior significantly. For a detailed treatment of the conditioning process and its relationship to pad performance, see our guide on CMPパッド・コンディショナーとコンディショニング・プロセス.

8. Pad Lifetime Modeling and Replacement Triggers

Pad lifetime is not a fixed number — it is a function of the conditioning program, the downforce and velocity used in polishing, the slurry chemistry aggressiveness, and the specific pad formulation. Effective lifetime management requires defining measurable replacement triggers and monitoring them continuously across the pad’s useful life.

Primary Pad Replacement Triggers

• MRR drift beyond specification (most common trigger): When the pad-normalized MRR on a reference blanket wafer falls below 90% of the initial qualified value, the pad is flagged for replacement. This criterion is objective, measurable, and directly linked to process performance.
• WIWNU degradation: As the pad surface texture becomes non-uniform through uneven conditioning or localized wear, within-wafer uniformity degrades — typically at the wafer edge first. An increase in 1σ WIWNU beyond the specification limit triggers pad replacement independent of MRR status.
• Pad thickness below minimum: Each conditioning cycle removes a small amount of pad material. When the remaining pad thickness drops below ~0.8 mm (or the supplier’s minimum for that product), structural integrity and sub-pad support behavior change significantly. Thickness is monitored using a non-contact eddy-current or ultrasonic gauge.
• Defect event (scratch spike): A sudden increase in scratch or particle defect count — detected by in-line inspection — may indicate pad surface delamination, groove edge cracking, or conditioner diamond shedding onto the pad surface. This triggers immediate pad inspection and typically replacement.
• Elapsed time / wafer count limit: Many fabs implement a preventive replacement schedule based on wafer count (e.g., every 1,000 wafer passes) as a risk management measure, regardless of whether performance metrics have been breached.

9. Pad Selection Guide by Application

10. Cost of Ownership Optimization

Polishing pad cost is the second-largest consumable line item in the CMP budget after slurry. Optimizing pad cost of ownership requires looking beyond unit price to the full system cost per wafer pass — which includes pad lifetime, conditioning consumable wear, throughput impact, and yield risk from pad-related defect events.

The Key Levers for Pad COO Reduction

• Maximize pad lifetime through conditioning optimization: Conditioning too aggressively wears the pad faster than necessary. Reducing conditioner downforce by 20–30% while maintaining MRR stability (confirmed by reference wafer measurement) can extend pad life significantly without process penalty.
• Match pad hardness to actual step-height requirement: Using a harder-than-necessary pad for a low-topography application increases both pad wear rate and conditioning frequency without benefit. Matching pad hardness to the actual planarization requirement is the most straightforward way to extend pad life.
• Implement in-situ MRR monitoring: Real-time friction current monitoring or optical reflectance endpoint detection allows precise identification of the MRR transition point, preventing both under-polish (yield loss) and over-polish (dishing + pad wear). Tight endpoint control reduces mean over-polish time and therefore mean pad wear per wafer.
• Use pad health metrics for predictive replacement: Rather than replacing pads on a fixed schedule, implementing reference wafer MRR measurement at defined intervals allows pads to remain in service until they actually approach their performance limits. This statistical approach can extend average pad lifetime by 15–25% over fixed-schedule replacement.

JEEZ application engineers can conduct a detailed cost-of-ownership analysis for your specific CMP tool fleet and production volume. Our pad products are designed with pad lifetime and conditioning compatibility as primary engineering objectives — not just initial MRR. Contact us to discuss a pad evaluation program.

11.よくあるご質問

How often should I condition a CMP pad?

Conditioning frequency depends on pad type, application, and the MRR stability requirement. For hard pads in oxide CMP, in-situ conditioning running concurrently with every wafer pass is standard. For softer pads in Cu buff applications, ex-situ conditioning between runs (or after every 5–10 wafers) is typically sufficient. The right frequency is determined empirically by measuring MRR stability as a function of wafers-between-conditioning cycles — the goal is the minimum conditioning frequency that maintains MRR within ±5% of target.

What causes CMP pad glazing, and how do I detect it?

Pad glazing occurs when the surface asperities are progressively smoothed by the polishing process, reducing the real contact area between pad and wafer. The primary indicator is a progressive MRR decline over a polishing run — typically 10–30% from beginning to end of pad life if not managed by conditioning. Glazing can also be detected visually (a shiny, reflective pad surface) or by surface profilometry showing reduced asperity height. Proper in-situ conditioning prevents glazing from accumulating to performance-limiting levels.

Can I use the same pad for both Cu Step 1 and Step 2?

In principle, a single pad type can cover both steps, but it is rarely optimal. Step 1 benefits from a harder, more aggressive pad to maximize Cu bulk removal rate, while Step 2 requires a softer, more selective pad to minimize dishing and erosion during barrier clearing. Using a hard pad for Step 2 risks excessive oxide erosion; using a soft pad for Step 1 reduces throughput. Most advanced logic fabs use a dedicated pad type for each step, typically on separate platens of a multi-platen CMP tool.

How does pad groove pattern affect slurry utilization efficiency?

Groove patterns strongly influence how efficiently slurry is transported across the pad-wafer interface. X-Y cross-hatch patterns typically provide the most uniform lateral and radial slurry coverage and are favored for copper CMP where uniform slurry chemistry across the wafer is critical. Concentric groove patterns rely on centrifugal force for radial distribution and can develop slurry depletion zones at specific radii under certain flow conditions. Groove pitch and depth must be matched to the slurry flow rate: too coarse a groove at low flow rates leads to slurry starvation in the contact zone.