Bkm

# BKM: Best Known Method in Semiconductor Foundry Operations

A Best Known Method (BKM) is the currently preferred, documented, and validated way to execute a defined semiconductor manufacturing, equipment, metrology, maintenance, or engineering task within a controlled scope. A BKM captures the method that has produced the strongest repeatable result using the evidence available at that time.

The scope matters. A BKM is not a universal recipe and not an informal suggestion. It is tied to conditions such as technology node, product family, material stack, process module, tool platform, chamber configuration, hardware revision, software release, consumable set, measurement system, factory, and qualification state. A method that is best for one chamber or device integration may be invalid for another.

In a professional foundry, the word known means supported by traceable engineering evidence, while best means preferred against explicit technical and business criteria. Those criteria can include yield, defectivity, process capability, matching, reliability, cycle time, uptime, cost per wafer, safety, and environmental performance.

---

## 1. Why Foundries Use BKMs

Semiconductor manufacturing contains thousands of interacting process and equipment variables. Small differences in chamber condition, hardware, incoming material, queue time, recipe implementation, metrology, or maintenance can shift critical dimensions, film properties, overlay, profile, defects, electrical behavior, and final yield.

A BKM converts successful engineering learning into a controlled, repeatable method. Its main purposes are to:

• reduce lot-to-lot, chamber-to-chamber, tool-to-tool, and fab-to-fab variation;
• shorten process development and technology-transfer cycles;
• preserve learning from experiments, excursions, and root-cause investigations;
• establish a qualified starting point for recipes, setup, maintenance, and troubleshooting;
• train engineers and technicians using one controlled method;
• improve equipment uptime, process capability, yield, and cost of ownership;
• prevent obsolete or locally improvised practices from silently becoming production norms.

A BKM is therefore both an engineering artifact and a knowledge-control mechanism.

---

## 2. BKM, Best Practice, SOP, and Process of Record

These terms overlap but are not interchangeable.

A BKM may become part of an SOP or POR after qualification and approval. A POR can also lag a newer BKM while the new method completes reliability, integration, customer, or manufacturing qualification. Conversely, a production POR may remain the correct released method even when an experimental BKM gives better results in development.

The safe rule is: BKM identifies the strongest currently supported method; release governance determines where it may be used.

---

## 3. Major Types of Foundry BKM

### Process BKM

A process BKM defines the preferred method for a unit process or integrated module. Examples include plasma etch, atomic layer deposition, CVD, PVD, ion implantation, anneal, CMP, clean, lithography, track processing, electroplating, wafer bonding, and advanced packaging.

It can specify the allowed material stack, precondition, recipe structure, setpoints, endpoint method, queue-time limits, chamber state, wafer handling, metrology, acceptance criteria, and recovery response.

### Equipment BKM

An equipment BKM standardizes tool setup and operation. It may cover chamber configuration, hardware part numbers, firmware and software revisions, calibration, matching, qualification wafers, sensor checks, fault handling, and startup or shutdown sequences.

The objective is not only recipe repeatability. It is maintaining a known equipment state capable of executing the process correctly.

### Preventive-maintenance BKM

A maintenance BKM defines the safest and most repeatable way to inspect, clean, replace, rebuild, calibrate, and return equipment to service. It includes parts, consumables, torque or alignment requirements, contamination controls, post-maintenance seasoning, qualification tests, and release criteria.

Maintenance BKMs are especially important because an apparently successful intervention can still change chamber matching, particles, process drift, or equipment reliability.

### Metrology and inspection BKM

A metrology BKM controls how a characteristic is measured. It can define sampling, target selection, alignment, focus, recipe parameters, model version, calibration, reference material, filtering, data quality checks, and gauge capability.

KLA has described BKM-based optimization for pattern-recognition models on semiconductor profilers. The principle is broader: production decisions are only as reliable as the repeatability and reproducibility of the measurement method.

### Troubleshooting and excursion BKM

A troubleshooting BKM provides a disciplined diagnostic sequence for known failure signatures. It identifies required evidence, non-invasive checks, safe isolation steps, decision points, escalation criteria, and recovery qualification. It should support root-cause analysis rather than encourage untracked parameter changes.

### Virtual-process and simulation BKM

A virtual BKM can define the calibrated process steps and model settings used for predictive manufacturing or 3D process simulation. Lam Research describes SEMulator3D workflows that use a BKM process flow with design layout and search criteria to identify potential shorts, opens, excess material, and other structural failures before wafer experimentation.

---

## 4. Minimum Content of a Professional BKM

A usable BKM should answer the following questions without relying on tribal knowledge.

### Identity and scope

• unique document or method identifier;
• title, owner, approvers, revision, and effective date;
• fab, area, module, technology, product, and layer scope;
• approved tool platforms, tool IDs, chambers, and hardware configurations;
• explicit exclusions and conditions where the BKM must not be used.

### Preconditions

• incoming material and prior-process requirements;
• tool health, chamber state, calibration, and qualification status;
• required software, firmware, recipe, model, library, and document revisions;
• environmental, facility, chemical, gas, and consumable conditions;
• queue-time, storage, handling, and contamination requirements.

### Method

• ordered steps with controlled parameters and units;
• allowable ranges and parameters that must remain fixed;
• decision points, hold points, and human approvals;
• wafer sampling and monitor strategy;
• required safety and equipment-protection controls;
• abnormal-event, abort, recovery, and escalation instructions.

### Verification

• required measurements and inspection points;
• acceptance limits, control limits, and capability targets;
• tool, chamber, and measurement-system matching requirements;
• electrical, structural, reliability, yield, and defectivity criteria;
• required number of wafers, lots, runs, and time window;
• comparison baseline and statistical analysis method.

### Records and traceability

• recipe and parameter file hashes or controlled revision IDs;
• equipment state, chamber history, parts, and consumable lots;
• wafer and lot genealogy;
• raw data and analysis location;
• deviations, approvals, qualification evidence, and linked change records.

---

## 5. How a BKM Is Established

### Step 1: Define the problem and metric

The team defines the scope and success criteria. Examples include lowering killer-defect density, improving etch-profile uniformity, increasing chamber mean time between cleans, reducing overlay residuals, improving measurement success rate, or reducing recovery time without compromising yield.

### Step 2: Establish a trustworthy baseline

The existing method is measured using controlled material, stable equipment, qualified metrology, and enough repetitions to separate real improvement from random variation. Baseline data should include distribution, not only an average.

### Step 3: Generate and test candidates

Candidate methods are evaluated through designed experiments, split lots, simulation, fault-tree analysis, or structured troubleshooting. Confounding variables must be controlled. The study should examine interactions with upstream and downstream processes, not only the local module result.

### Step 4: Validate repeatability and transferability

The preferred candidate is repeated across time, operators, lots, chambers, and tools as appropriate. Tool-to-tool matching, measurement-system capability, wafer position, product mix, and chamber age should be considered. A method is not a foundry BKM merely because one expert achieved one strong result.

### Step 5: Qualify risk

The team checks defectivity, reliability, integration, equipment protection, EHS, cost, capacity, and supply-chain effects. For production use, qualification may include electrical test, reliability stress, change notification, customer approval, and a controlled ramp.

### Step 6: Review and release

Process, equipment, integration, yield, metrology, manufacturing, quality, reliability, and EHS owners review the evidence relevant to their boundary. The released BKM receives an owner, revision, scope, training plan, effective date, and rollback method.

### Step 7: Monitor and improve

Post-release monitoring determines whether the expected benefit persists. Control charts, fault detection, equipment-state data, yield, defects, maintenance history, and audit results can trigger review. A BKM is a controlled current state, not a permanent claim.

---

## 6. Evidence Required to Call a Method “Best”

The evidence depends on the application, but professional qualification commonly examines:

• mean, variation, confidence interval, and outlier behavior;
• repeatability and reproducibility of the measurement system;
• process capability such as Cp, Cpk, Pp, or Ppk where appropriate;
• across-wafer, wafer-to-wafer, lot-to-lot, chamber-to-chamber, and tool-to-tool variation;
• defect density, excursion rate, false-call rate, and nuisance rate;
• electrical performance, parametric yield, functional yield, and reliability;
• throughput, cycle time, availability, utilization, and recovery time;
• consumable life, parts usage, cost per wafer, and cost per good die;
• worker safety, chemical risk, equipment protection, waste, emissions, and energy.

No single metric automatically defines the BKM. The selected method must satisfy hard constraints and show the best controlled tradeoff for its intended scope.

---

## 7. Tool-to-Tool and Fab-to-Fab Transfer

Copying parameter values is not the same as transferring a BKM. Nominally identical tools can differ because of chamber geometry, component age, sensor calibration, RF delivery, gas conductance, thermal behavior, endpoint response, software revision, facilities, and local maintenance history.

A professional transfer separates:

• invariants: outputs or physical conditions that must be preserved;
• implementation variables: tool-specific settings used to achieve those outputs;
• matching metrics: measurements that demonstrate equivalence;
• local qualifications: evidence required before release at the receiving site.

For a plasma process, preserving ion energy, radical balance, profile, selectivity, uniformity, damage, and defect performance may matter more than forcing every numerical setpoint to match. The receiving tool may require a locally tuned recipe inside the BKM’s approved transfer rules.

---

## 8. BKM Governance and Change Control

Every released BKM needs a named technical owner and a controlled lifecycle:

Draft
  -> engineering review
  -> qualification
  -> approval and release
  -> training and controlled deployment
  -> performance monitoring
  -> revision, supersession, or retirement

A BKM review should be triggered by significant changes such as:

• new technology node, product, layer, or material stack;
• tool, chamber, hardware, sensor, firmware, or software change;
• recipe, chemical, gas, target, consumable, or supplier change;
• metrology recipe, model, calibration, or sampling change;
• repeated excursion, drift, yield loss, reliability issue, or audit finding;
• improved method demonstrated by new evidence;
• transfer to another tool set, production line, or fab.

Revision history must explain what changed, why it changed, the supporting evidence, affected scope, required requalification, and rollback plan. Obsolete copies must be removed from points of use so that two revisions cannot quietly operate as competing standards.

---

## 9. Common Failure Modes

### Calling a preference a BKM

An expert opinion can start an experiment, but it is not a BKM until the method and benefit are documented and validated.

### Ignoring scope

A method qualified on one product, chamber, or film stack can fail elsewhere. Scope, exclusions, and transfer conditions are mandatory.

### Optimizing a local metric

Faster etch rate, fewer inspection nuisance events, or longer maintenance interval can damage profile, sensitivity, downstream integration, reliability, or yield. BKM evaluation must include system-level constraints.

### Losing the equipment state

A recipe without hardware, chamber, software, calibration, consumable, and seasoning state is incomplete. The same recipe name does not guarantee the same process.

### Weak measurement capability

If the metrology method is unstable or biased, apparent improvement may be measurement noise. Gauge qualification and reference correlation are part of the evidence.

### No retirement mechanism

An old BKM can become a source of variation after tools, materials, or integration change. Effective dates, periodic review, supersession, and archival controls prevent stale methods from remaining active.

---

## 10. Digital BKM Systems

A modern BKM can be represented as controlled, machine-readable knowledge linked to recipe management, equipment engineering, manufacturing execution, statistical process control, fault detection, maintenance, metrology, yield, and document-control systems.

Useful capabilities include:

• revision and approval workflow;
• access control and electronic signatures;
• parameter-schema and unit validation;
• automatic comparison of deployed versus approved configuration;
• links to qualification datasets and analysis notebooks;
• tool and chamber applicability rules;
• training and certification records;
• drift monitoring and automatic review triggers;
• search by failure signature, tool state, product, module, or process layer;
• audit-ready history and controlled rollback.

AI can help retrieve relevant BKMs, compare tool states, summarize evidence, detect deviations, or propose experiments. It should not independently declare a new BKM or release a production change. Foundry owners must review the evidence and authorize the controlled revision.

---

## 11. Key Takeaway

In semiconductor foundry operations, a Best Known Method is the best currently validated and controlled method for a specific engineering scope. It turns experimental and operational learning into repeatable execution across people, shifts, chambers, tools, and fabs.

A credible BKM has five properties:

1. Scoped to defined products, processes, equipment, materials, and revisions.
2. Evidence-backed by repeatable technical and manufacturing results.
3. Controlled through ownership, approval, training, and document revision.
4. Transferable through explicit invariants, matching metrics, and local qualification.
5. Living because new evidence, equipment changes, and process drift can replace it.

The professional distinction is simple: a best practice is advice; a BKM is controlled foundry knowledge supported by evidence.

---

## References

• Applied Materials, semiconductor equipment and process documentation describing Best-Known-Method libraries for inspection-recipe creation and BKM tracking for installed-base productivity.
• KLA, application guidance describing a Best Known Method for semiconductor profiler pattern-recognition model optimization.
• Lam Research, semiconductor process and equipment materials describing BKM processing, troubleshooting, maintenance, and SEMulator3D BKM process flows.
• SEMI, semiconductor manufacturing guidance and standards materials using best-known-method concepts for controlled work and EHS methodology.