Hard vs. Soft CMP Polishing Pads: The Definitive Selection Guide
A rigorous, application-mapped guide to choosing between hard and soft CMP polishing pads — covering the fundamental physics of the trade-off, process-step-by-step recommendations, stacked pad strategy, and a practical decision framework for fab engineers.
Verified
- The Core Trade-off: Planarity vs. Uniformity
- Hard Pads: Properties and Strengths
- Soft Pads: Properties and Strengths
- Head-to-Head Comparison Table
- Application-by-Application Selection Map
- The Stacked Pad Strategy
- Decision Framework: Which Pad for Your Step?
- Qualifying a New Pad Hardness
- Jizhi’s Hard and Soft Pad Range
- FAQ
Every CMP process engineer eventually faces the same fundamental decision: hard pad or soft pad? The question seems deceptively simple — but it sits at the intersection of contact mechanics, surface chemistry, and wafer-scale uniformity physics. Getting it wrong can mean poor planarization (hard pad selected where soft was needed) or excessive defects and uniformity excursions (soft pad selected where hard was required).
This guide provides the definitive answer — not as a simple rule, but as a structured framework that maps pad hardness to process requirements, film type, node, and defect budget. If you want to first understand how pad hardness affects material removal at a mechanistic level, see: How CMP Polishing Pads Work. For a broader overview of pad material classes beyond just hardness, see: CMP Pad Materials: Polyurethane vs Other Options.
1. The Core Trade-off: Planarization Efficiency vs. Within-Wafer Uniformity
The hard-vs-soft pad decision is governed by a single, inescapable physical trade-off: a harder pad is better at removing topographic features, but a softer pad is better at distributing removal uniformly across the wafer. These two objectives are in direct tension, and no pad can simultaneously maximize both. Understanding why — at the mechanical level — is the foundation of informed pad selection.
Why Hard Pads Plannarize Better
A hard pad (Shore D 55–65) has a high Young’s modulus, typically 200–500 MPa. When pressed against a wafer surface with raised topographic features (hills created by underlying device structures), the rigid pad surface bridges over the valleys and concentrates contact force on the hilltops. This selective loading means that material removal preferentially occurs at the high points, progressively reducing step height and approaching global planarization. The phenomenon is analogous to a stiff ruler pressed across a bumpy surface — only the bumps touch the ruler.
Why Soft Pads Are More Uniform
A soft pad (Shore D 28–45) has a low Young’s modulus, typically 10–60 MPa. Under the same applied down-force, it deforms and conforms to the wafer surface topography rather than bridging over it. Contact force is distributed more evenly across both hills and valleys. The result is more uniform material removal — but at the cost of lower step-height reduction. Additionally, at the wafer scale, a soft pad conforms to the bow and warp of a 300 mm production wafer (commonly 20–80 µm peak-to-valley), reducing the edge-to-center pressure differential that causes non-uniform removal profiles with hard pads.
2. Hard Pads: Properties, Strengths, and Limitations
- High planarization efficiency — step-height reduction >80% in one pass typical
- Stable, predictable removal rate over extended conditioning campaigns
- Better selectivity between high and low features — ideal for damascene steps
- Stiffer surface resists asperity flattening — maintains MRR over longer pad life
- Higher Preston coefficient Kp — higher throughput per pressure unit
- Better compatibility with high-selectivity ceria slurries for STI oxide CMP
- Higher scratch density — stiff asperities transmit more force to abrasive particles
- Poor edge-center uniformity on bowed or warped wafers
- Risk of low-k dielectric delamination under high shear forces
- Higher sensitivity to pressure non-uniformity from retaining ring geometry
- Requires more aggressive conditioning to prevent glazing — higher conditioner wear
- Not suitable for ultra-thin films where over-polishing risk is high
Typical Hard Pad Specifications (Production Grade)
| パラメータ | Specification Range | Test Method |
|---|---|---|
| Shore D hardness | 55–65 (±2 within lot) | ASTM D2240, 5-point wafer map |
| Compressibility | 0.5–2.5% | % thickness change at 25 kPa, 60 s |
| Elastic recovery | >70% | % recovery 60 s after load removal |
| Mean pore diameter | 20–45 µm | Optical cross-section, image analysis |
| Pore size CV (%) | <18% | Standard deviation / mean × 100 |
| パッド厚 | 2.0–2.5 mm (±0.05 mm) | 5-point contact gauge |
| Groove depth | 0.5–0.8 mm | Profilometer cross-section |
| Groove width | 0.3–0.6 mm | Profilometer cross-section |
3. Soft Pads: Properties, Strengths, and Limitations
- Superior within-wafer uniformity — conforms to wafer bow and warp
- Low shear force on fragile films — safe for low-k dielectrics (k < 2.5)
- Lower scratch and micro-scratch density — critical for Cu and barrier CMP
- Excellent for final surface finishing steps requiring Ra < 0.5 nm
- Better results on 300 mm wafers with high bow/warp from stress films
- Forgiving to minor slurry flow variations and recipe perturbations
- Low planarization efficiency — poor step-height reduction on rough incoming surfaces
- Faster MRR decay as pad glazes — requires more frequent conditioning
- Lower Tg — more susceptible to thermal softening at elevated process temperatures
- Higher compressibility variation wafer-to-wafer in early pad life
- Not suitable for STI, PMD, or any step requiring >50% step-height reduction
- Sensitive to conditioning parameters — over-conditioning dramatically increases MRR
4. Head-to-Head Comparison: Every Key Metric
| メートル | Hard Pad (Shore D 55–65) | Soft Pad (Shore D 28–45) | Winner |
|---|---|---|---|
| Planarization efficiency | High — bridges topography, removes selectively from high points | Low — conforms to topography, removes uniformly | Hard |
| Within-wafer uniformity (WIWNU) | Moderate — sensitive to wafer bow and retaining ring geometry | High — conforms to wafer-scale shape variations | Soft |
| Scratch defect density | Higher — stiff asperities transmit higher local stress | Lower — compliant asperities reduce peak contact stress | Soft |
| Low-k film safety | Risk of delamination at standard pressures | Safe at standard pressures (<3 psi) | Soft |
| Material removal rate | Higher MRR at same P × V — better throughput | Lower MRR — longer polishing times needed | Hard |
| Pad lifetime (wafers/pad) | 500–2,000 wafers — stiffer surface resists wear | 300–1,000 wafers — softer surface glazes faster | Hard |
| Slurry utilization | Moderate — closed-cell pores provide good retention | High — open-cell structure absorbs and releases slurry efficiently | Soft |
| Conditioning sensitivity | Lower — MRR change per unit conditioner force is smaller | Higher — small conditioning changes cause significant MRR shifts | Hard |
| Thermal stability | Higher Tg (90–120°C) — better for high-pressure processes | Lower Tg (55–80°C) — softens faster under thermal load | Hard |
| Cost (unit price) | Baseline (1.0×) | Slightly lower (0.8–1.1×) — depends on formulation | Similar |
5. Application-by-Application Selection Map
The following application map provides pad hardness recommendations for the most common CMP steps in advanced semiconductor manufacturing. Each recommendation is grounded in the process physics described above and reflects April 2026 fab best practice. For a broader view of how CMP pads are used across the full IC process flow, see: Semiconductor CMP Polishing Pads.
6. The Stacked Pad Strategy: Getting the Best of Both
The most practically important development in CMP pad engineering over the past decade is the widespread adoption of stacked pad configurations — combining a hard polishing top pad with a compliant foam subpad — to simultaneously achieve planarization efficiency and within-wafer uniformity. This strategy directly addresses the hard-vs-soft trade-off by decoupling the two functions into separate layers.
How the Stack Works
In a stacked pad configuration, the hard polyurethane top pad (Shore D 55–65) provides the polishing surface. Its high Young’s modulus ensures that contact with the wafer surface is dominated by the asperity-level mechanics that deliver planarization efficiency. Beneath the top pad, a soft foam subpad (typically Shore A 30–55, 0.5–1.5 mm thick) is laminated directly to the platen. The subpad’s role is purely mechanical: its bulk compliance absorbs wafer-scale bow and warp, redistributing the contact force from the carrier head more evenly across the wafer surface. The subpad does not contact the slurry or the wafer directly.
Stack Configuration Naming Conventions
| Stack Type | Top Pad | Subpad | Application Sweet Spot |
|---|---|---|---|
| Hard / Hard | Shore D 60–65 | Shore D 45–55 (stiffer foam) | Maximum planarization, mature node oxide — wafer bow not a concern |
| Hard / Soft (standard stack) | Shore D 55–62 | Shore A 35–50 (soft foam) | 300 mm advanced node oxide and W CMP — industry standard configuration |
| Medium / Soft | Shore D 45–55 | Shore A 25–40 (very soft foam) | Cu bulk step — balance of MRR and uniformity, moderate low-k protection |
| Soft / Very Soft | Shore D 28–42 | Shore A 20–30 (ultra-soft foam) | Ultra-thin low-k finishing, Cu barrier buff — maximum defect protection |
7. Decision Framework: Choosing the Right Pad Hardness
The following decision tree provides a systematic path from process step description to pad hardness recommendation. Work through the questions in order — the first applicable branching point gives the recommendation.
8. Qualifying a New Pad Hardness in Production
Switching pad hardness — even within the same product family — is a significant process change that requires structured qualification. A change in Shore D hardness of even 5 points can shift the Preston coefficient by 8–15%, requiring recipe pressure adjustments. Here is the standard qualification protocol:
Establish Baseline Metrics on Qualified Pad
Run a minimum of 3 qualification lots (25 wafers each) on the currently qualified pad at the locked production recipe. Record mean removal rate, WIWNU (1σ), post-CMP scratch density (from KLA/Hitachi inspection), and electrical test results (if applicable). These become the acceptance criteria for the new hardness.
Run Initial Characterization Splits
Polish a minimum of 50 monitor wafers on the new-hardness pad at the existing production recipe parameters. Do not adjust the recipe yet. Compare MRR (target: within ±15% of baseline), WIWNU (target: within ±0.5% 1σ), and scratch density (target: within ±20%). Expect MRR to shift — the Preston coefficient changes with hardness. Recipe adjustment will be needed.
Adjust Recipe to Match Baseline MRR
If switching to a harder pad (MRR increased), reduce down-force pressure proportionally. If switching to a softer pad (MRR decreased), increase pressure. Use the Preston equation (MRR = Kp × P × V) as a first-order guide — the Kp of the new pad relative to the baseline can be estimated from the initial characterization data. Re-run 3 lots at the adjusted recipe to confirm MRR within ±8% of baseline.
Validate Conditioning Protocol
Pad conditioning parameters (conditioner down-force, sweep speed, in-situ vs. ex-situ ratio) optimized for the original hardness may need adjustment. Harder pads require more aggressive conditioning to prevent glazing; softer pads are more sensitive to over-conditioning. Optimize the conditioning protocol independently before committing to production qualification lots.
Run Full Qualification Lot and Engineering Sign-Off
Run 3 full-size production lots (25 wafers each) at the optimized recipe. All metrics must meet acceptance criteria. Obtain engineering and process owner sign-off. Update the process specification (process spec) with the new pad hardness, recipe parameters, and conditioning protocol before any production release.
For detailed guidance on pad conditioning protocols — which differ between hard and soft pad types — see: CMP Pad Conditioning and Lifespan Management. For the relationship between pad hardness and material removal rate at a quantitative level, see: CMP Material Removal Rate and Pad Parameters.
9. Jizhi’s Hard and Soft Pad Product Range
Jizhi Electronic Technology manufactures both hard and soft CMP polishing pads using proprietary polyurethane formulations developed through in-house R&D. Our product range is engineered to provide qualified alternatives to IC1000-type hard pads and Politex / Suba-type soft pads, with full process characterization data provided for each product.
| Product Series | パッドタイプ | Shore D | Primary Application | Availability |
|---|---|---|---|---|
| JZ-H60 Series | 硬質ポリウレタン | 58–62 | Oxide ILD, STI, PMD, W plug — IC1000 equivalent | In stock |
| JZ-H65 Series | Hard polyurethane (high hardness) | 63–67 | High-topography oxide CMP, aggressive step-height applications | In stock |
| JZ-S38 Series | Soft polyurethane subpad | 35–42 | Cu bulk removal (BEOL Step 1), stacked pad subpad | In stock |
| JZ-S28 Series | Very soft polyurethane | 26–32 | Cu / barrier buff (BEOL Step 2), ultra-thin low-k finishing | In stock |
| JZ-SiC Series | Specialty hard (SiC-optimized) | 60–68 | SiC and GaN substrate CMP, 3rd-generation semiconductors | In stock / custom |
| JZ-Custom OEM | Customer-specified | Any range | Custom hardness, groove, and formulation per customer spec | 3–6 week lead time |
