A 300 mm fab saw recurring photoresist defects. NH₃ in the litho bay had climbed to 15 ppb — 15× the 1 ppb process limit. How do you recover?

One De-Identified Industry Case

Biểu đồ 1: Khắc phục AMC tại fab 300 mm

Đỉnh NH₃ khu quang khắc → ổn định sau 3 tuần lắp lọc hóa H₃PO₄

Trước15×
NH₃ quang khắc
15 ppb
Yield cản quang: —
Sau−98 %
NH₃ quang khắc
0.3 ppb
Yield cản quang: +2.5 %

Ví dụ điển hình ẩn danh. Số liệu thực tế có thể khác.

Discovery

Recurring photoresist T-top defects in the litho bay. An IMS probe deployed on-site read NH₃ at 15 ppb — far above the 1 ppb process allowance.

Diagnosis

Traced to elevated NH₃ in MAU make-up air + operator-area amine buildup. Both fronts contributing simultaneously.

Remediation

  • ▸MAU intake → H₃PO₄-impregnated activated carbon chemical filter targeting bases
  • ▸FFU return → molecular-sieve sorbent module as a last line of defense

Result

  • ▸NH₃ dropped to below 0.3 ppb (~98 % reduction)
  • ▸Photoresist defect rate stabilized; yield up 2.5 % (at ~USD 20,000 per wafer at this node, annual recovery easily in the eight-figure USD range)
Key lesson: detecting AMC is fast (IMS reads in 3 minutes). The hard parts are finding all the sources and designing the correct filtration — and it's rarely a single source.

SEMI F21 AMC Classes — Quick Refresher

The semi industry sorts AMC into four classes under SEMI F21:

  • ▸Acids (MA) — HCl, HF, H₂SO₄, NOₓ, SOₓ: metal-line corrosion, copper tarnishing
  • ▸Bases (MB) — NH₃, Me₃N, NMP: DUV resist T-top failure
  • ▸Condensables (MC) — BHT, NMP, DOP: wafer hazing, optics contamination
  • ▸Dopants (MD) — AsH₃, B₂H₆, BF₃: shift semiconductor dopant concentration

Filter media must match the target:

  • ▸Acids → activated carbon impregnated with KOH / Na₂CO₃
  • ▸Bases → activated carbon impregnated with H₃PO₄
  • ▸Condensables / organics → un-impregnated high-surface-area activated carbon
  • ▸Dopants → specialty sorption resin or composite media

Using the wrong medium is worse than no filter — capture depends on the chemistry being right.

Core Strategy: Source Reduction × Path Interception

Biểu đồ 2: Hai mặt trận quản lý AMC — Giảm nguồn × Chặn đường

Thiếu một là rò. Chiến lược đầy đủ phải cả hai mặt

Giảm nguồn

Quản nguồn: giảm phát thải

  • ▸Vật liệu phát khí thấp
  • ▸Hút cục bộ tại máy
  • ▸Sàng lọc FOUP/găng/bao bì
  • ▸Kỷ luật thao tác
Chặn đường

Lọc đường đi: chặn trước khi đến wafer

  • ▸Lọc hóa MAU/OAU
  • ▸Sàng phân tử tại FFU
  • ▸Lọc mini-environment
  • ▸Giám sát IMS/GC-MS + chu kỳ thay

Lọc phải khớp chất gây ô nhiễm.

Single-front strategies always leak. Complete AMC control must do both at once:

Source reduction

  • ▸Low-outgassing specs for construction materials, paint, adhesives
  • ▸Local exhaust on process tools (prevent re-entrainment)
  • ▸Screening of FOUPs, gloves, and packaging
  • ▸Operator workflow (no ammonia-based cleaning agents)

Path interception

  • ▸MAU / OAU: upstream chemical filter (large-area, low-concentration continuous duty)
  • ▸FFU return: molecular sieve or activated carbon in the recirc loop
  • ▸Mini-environment / point-of-use: localized protection at critical tools
  • ▸Continuous IMS / GC-MS monitoring paired with replacement scheduling

Three Things to Evaluate Together When Sizing a Chemical Filter

1. Capacity vs concentration

Chemical filter life isn't "X months" — it's "mass of target species captured." At 1 ppb a filter may run for a year; at 100 ppb it might saturate in 2 months.

Approach: measure inlet concentration, then use the supplier's capacity-vs-concentration curve to estimate life. Do not just use the datasheet month count.

2. Pressure-drop budget

Chemical filters typically add 100–200 Pa of initial ΔP. Placing one in the MAU affects the full-fab airflow balance. Size the fan to include the chemical filter; retrofitting later often drops airflow by 15–20 %.

3. Breakthrough monitoring

Pressure drop alone is misleading. The real indicator is breakthrough (outlet / inlet concentration ratio). Replace when breakthrough exceeds ~10 %, not at 50 %.

Practical Cost Control

A representative TFT-LCD Gen-6 fab used a quarterly AMC monitoring + filter rotation plan to extend chemical-filter life from 6 months to 9 months, cutting annual consumable cost by ~30 %.

Rotation logic:

  1. 1Track "install date + accumulated runtime" per filter
  2. 2Decide replacement on measured breakthrough, not calendar
  3. 3Filters approaching end-of-life but still functional — step them down to less-sensitive zones (non-critical processes)

AMC control is not "install more filters." It is "install the right filter in the right place, and swap it at the right time." Any single tactic falls short — source control, path filtration, monitoring, and rotation all run together.