Doc #115 — Semiconductor Processing Roadmap

---

RECOMMENDED ACTIONS

Phase 1–3 (Years 0–7): Bridge and Foundation

1. Stockpile and extend the life of all existing electronic devices and components. This is the single highest-value action. Every year of extended electronics availability buys time for industrial development. (Doc #130: Device Life Extension)
2. Secure all semiconductor-grade materials currently in NZ — silicon wafers in university and industrial labs, electronic-grade chemicals, photoresist, clean room consumables. These are irreplaceable in the medium term and have near-zero peacetime visibility to requisition planners.
3. Identify and protect personnel with semiconductor fabrication knowledge — university researchers (particularly at the University of Canterbury, University of Auckland, and Victoria University of Wellington, which have or have had microelectronics research programmes), any NZ-based semiconductor industry workers (few, but not zero — packaging, testing, and design roles exist), and electronics engineers with relevant materials science backgrounds.²
4. Begin knowledge capture from semiconductor-knowledgeable personnel — document what they know about fabrication processes, equipment requirements, and materials while these people are available. This knowledge degrades as individuals retire, relocate, or die.
5. Preserve all existing semiconductor fabrication literature — textbooks, process manuals, equipment documentation. Print critical references during the printing window (Doc #5, Doc #29).

Phase 4–5 (Years 7–30): Discrete Components and Simple Semiconductor Processing

6. Commission the discrete transistor fabrication facility described in Doc #135. This is the foundation — without the ability to produce working transistors from local materials, no subsequent step is possible.
7. Develop germanium junction transistor capability (Doc #135, Part XI) — alloy junction transistors are more reliable and reproducible than point-contact devices, and the fabrication process teaches essential skills (controlled atmosphere heating, dopant diffusion, junction formation) that transfer directly to later silicon processing.
8. Begin silicon purification research. Even if germanium remains the primary semiconductor material for decades, understanding silicon chemistry is essential preparation for the eventual transition. The Siemens process (trichlorosilane distillation and chemical vapour deposition) should be studied and small-scale experiments attempted as chemical industry capability permits.
9. Develop precision optics capability (Doc #98: Glass Production, optical glass section) — photolithography requires lenses and mirrors of quality that NZ does not currently produce. Optical glass grinding and polishing is a prerequisite skill that takes years to develop.
10. Establish a photosensitive chemistry research programme — photoresists (the light-sensitive coatings used to pattern semiconductor surfaces) may be formulable from locally available materials (bitumen-based negative resists were used in early semiconductor fabrication), but bitumen resists offer substantially lower resolution and poorer process control than the synthetic resists used in 1960s fabrication.³ Developing a working resist with adequate resolution for even 25-micrometre features requires sustained experimental effort.

Phase 5–6 (Years 15–60): Planar Processing and Early Integration

11. Build a basic clean room facility — not the Class 1 environments of modern fabs, but a Class 10,000–100,000 environment (comparable to 1960s semiconductor fabrication) achievable with HEPA filtration, positive pressure, and disciplined gowning procedures. This is a significant construction project requiring glass fibre production for HEPA filters (Doc #98), HVAC fabrication (Doc #91), and non-shedding synthetic gown fabric that NZ does not currently produce (see section 4.4).⁴
12. Develop planar transistor fabrication — thermal oxidation of silicon, photolithographic patterning, dopant diffusion through oxide windows. This is the critical transition from discrete devices to the technology base for integrated circuits.
13. Produce simple integrated circuits — logic gates (NAND, NOR, NOT) containing 5–20 transistors on a single die. These are functionally equivalent to the 7400-series TTL logic that powered the computing revolution of the 1960s–1970s.
14. If trade with larger industrial regions has developed, negotiate for semiconductor imports and focus NZ effort on applications, packaging, and testing rather than wafer fabrication.

Phase 6–7 (Years 30–100+): Scaling

15. Expand integration scale progressively — from small-scale integration (SSI, tens of transistors) to medium-scale integration (MSI, hundreds), following the historical progression of the 1960s–1970s at whatever pace NZ’s industrial base and trade relationships permit.
16. Develop microprocessor capability — a single-chip computer comparable to the Intel 4004 (1971) or 8080 (1974), containing 2,000–5,000 transistors. This is the inflection point where computing becomes broadly accessible rather than a centralised national asset.

---

ECONOMIC JUSTIFICATION

The cost of no semiconductor capability

The economic case for semiconductor development is not comparable to the immediate, quantifiable cost-benefit analysis appropriate for projects like computer construction (Doc #135) or steel production (Doc #106). Semiconductor fabrication is infrastructure for infrastructure — it enables capabilities across every sector rather than solving a specific allocation problem.

The costs of not developing semiconductor capability are diffuse but compounding:

• Grid management degrades. NZ’s electrical grid relies on electronic control systems for load balancing, fault protection, and frequency regulation. Electromechanical substitutes exist for all of these functions (they were standard before the 1970s), but the transition from electronic to electromechanical control involves progressive loss of precision, response speed, and automation. Each relay-based substitution works but is less capable than the electronic system it replaces.⁵
• Telecommunications regresses. Without replacement components, the telecommunications network degrades from digital fibre to analogue copper to HF radio over a period of years to decades (Doc #128). Each step backward reduces bandwidth, reliability, and geographic coverage. A nation coordinating recovery across two main islands and hundreds of coastal communities pays a real coordination cost for every step of communications regression.
• Medical diagnostics narrow. Modern diagnostic equipment (ultrasound, patient monitors, laboratory analysers) depends on electronics. Without replacement, NZ’s medical system progressively loses diagnostic capability, reverting to clinical examination, basic laboratory chemistry, and mechanical instruments. The health cost is real but difficult to quantify — it manifests as missed diagnoses, delayed treatment, and higher mortality from conditions that are treatable if detected early.⁶
• Scientific instrumentation disappears. Measurement underpins every technical discipline. Electronic balances, pH meters, spectrophotometers, oscilloscopes, and data loggers are the tools of industrial chemistry, agriculture, public health, and engineering. Without them, NZ’s ability to do precise, reproducible science declines steadily.

Labour cost

The cumulative labour investment in semiconductor capability development is large but spread across decades:

Stage	Period	Estimated person-years	Prerequisite
Knowledge preservation and capture	Phase 1–3	5–10	Printing capability
Discrete transistor facility (Doc #135)	Phase 4–5	30–55	Chemistry, metallurgy, glasswork
Silicon purification research	Phase 4–5	10–20	Chemical industry
Clean room construction	Phase 5–6	15–30	HVAC, filtration, precision construction
Planar process development	Phase 5–6	20–50	All of the above
Simple IC fabrication	Phase 6	30–80	Planar processing, photolithography
Cumulative total	Decades 1–6	110–245

These figures are rough estimates based on the historical workforce requirements of early semiconductor facilities (Fairchild Semiconductor employed approximately 50–150 people in its first year of operation in 1957–1958 and grew to several hundred within two years; similar-scale operations in the 1960s typically employed 50–200 production workers plus 10–30 engineers and scientists).⁷ The numbers assume NZ is developing the capability from a base of discrete transistor fabrication, not from zero — the Doc #135 programme provides the essential foundation.

Comparison: trade versus local fabrication

If maritime trade with Australia or other regions develops on the timelines described in the Recovery Library baseline, NZ may be able to import semiconductors rather than manufacture them. Australia has a larger population (approximately 26 million pre-war), more diverse mineral resources, and a more extensive industrial base — though it too had minimal domestic semiconductor fabrication pre-war.⁸

The trade path is almost certainly faster and cheaper per device than local fabrication, but it creates dependency. The strategic question is whether NZ invests in the capability to manufacture semiconductors as insurance against trade disruption, or accepts semiconductor dependency on trading partners (as it does today) and invests the labour elsewhere. This is a genuine strategic choice, not a technical question, and this document cannot resolve it — it can only lay out what local fabrication requires.

The recommended approach: Invest in the prerequisite industries (chemical processing, glass, precision optics, metallurgy) that have broad value regardless of whether semiconductors are manufactured locally, and maintain semiconductor fabrication knowledge and small-scale experimental capability. Defer the decision on full-scale local fabrication until the trade situation, industrial base, and competing demands are clearer — likely a Phase 5–6 decision point.

---

4. WHY SEMICONDUCTORS ARE DIFFICULT

4.1 The purity problem

The fundamental challenge of semiconductor fabrication is purity. A working transistor requires semiconductor material with impurity concentrations measured in parts per billion — a level of chemical purity that has no parallel in any other industrial process NZ is likely to attempt.

To put this in context: metallurgical-grade silicon (the starting material for semiconductor processing) is approximately 98–99% pure. This is the output of an electric arc furnace reducing quartz sand with carbon — a process comparable in complexity to steelmaking. But metallurgical-grade silicon is useless for electronics. Electronic-grade silicon must be 99.9999999% pure (nine nines) — a factor of roughly one billion times purer.⁹

The historical process for achieving this purity — the Siemens process, developed in the 1950s and still used today — involves:

1. Crushing and grinding metallurgical-grade silicon into fine particles.
2. Reacting with hydrogen chloride gas (HCl) at 300–350°C to form trichlorosilane (SiHCl₃), a liquid that boils at 31.8°C.
3. Fractional distillation of trichlorosilane — repeated distillation cycles, each removing a fraction of the impurities. This is where the extreme purity is achieved, exploiting the different boiling points of trichlorosilane and its impurity-bearing analogues. Multiple distillation columns in series are required.¹⁰
4. Chemical vapour deposition (CVD) — the purified trichlorosilane is decomposed on a heated silicon rod at approximately 1,100°C in a hydrogen atmosphere, depositing ultra-pure polycrystalline silicon. This is the Siemens reactor — a bell jar containing an inverted U-shaped silicon rod electrically heated to white-hot, with trichlorosilane gas flowing over it.
5. Crystal growth — the polycrystalline silicon is melted (at 1,414°C) and a single crystal is pulled from the melt using the Czochralski method (slowly withdrawing a rotating seed crystal from a crucible of molten silicon). The single crystal boule is then sliced into thin wafers.

Each of these steps requires specific industrial capability that NZ does not currently possess and would need to develop over years to decades. The distillation requires large quantities of high-purity HCl (which requires a chlor-alkali industry — electrolysis of salt water). The CVD reactor requires precise temperature control, hydrogen gas supply, and materials that can withstand 1,100°C in a corrosive atmosphere. The Czochralski puller requires a vacuum-capable furnace, a rotating mechanism with very low vibration, and a crucible material (typically high-purity quartz) that does not contaminate the melt.

4.2 The germanium alternative

Doc #135 specifies germanium rather than silicon for the first generation of transistors, and this choice remains relevant for early semiconductor development. Germanium’s lower melting point (938°C vs. 1,414°C for silicon) and the lower boiling point of its purification intermediate (germanium tetrachloride, GeCl₄, boils at 84°C — making fractional distillation significantly easier) mean that germanium devices can be fabricated with less extreme infrastructure than silicon requires.¹¹

However, germanium has fundamental limitations that prevented its use in integrated circuits historically and would do so again:

• Germanium oxide (GeO₂) is water-soluble, which means it cannot form the stable, protective surface oxide layer that is essential to planar transistor processing and integrated circuit fabrication. Silicon dioxide (SiO₂) — naturally stable, electrically insulating, chemically resistant, and growable by thermal oxidation (though the process requires furnaces capable of sustained 900–1,200°C operation in a controlled atmosphere) — is the enabling material for the entire planar process and everything that followed it.¹²
• Germanium has a smaller bandgap (0.67 eV vs. 1.12 eV for silicon), making germanium devices more sensitive to temperature — leakage current increases rapidly above about 75°C, limiting operating conditions.
• Germanium is far rarer than silicon. Silicon constitutes approximately 28% of the Earth’s crust by weight; germanium is measured in parts per million. NZ’s germanium sources (principally from sphalerite deposits) are limited and uncertain in concentration (Doc #135).

The practical consequence: germanium is the right material for discrete transistors and the first generation of computing (Doc #135), but integrated circuits require silicon. The transition from germanium to silicon is itself a major industrial development step.

4.3 The photolithography problem

Integrated circuits are manufactured by patterning thin films on a semiconductor surface — depositing a layer of material (oxide, metal, polysilicon), coating it with a light-sensitive photoresist, exposing the resist through a patterned mask, developing the resist to create a pattern, and etching the underlying layer through the pattern openings. This process is repeated multiple times (typically 5–15 mask layers for circuits of 1960s–1970s complexity) to build up the three-dimensional structure of transistors, resistors, capacitors, and interconnections on a single chip.¹³

Each step in this process has requirements that compound:

• Photoresist: Early photoresists (1950s–1960s) were based on dichromated gelatin or Kodak Thin Film Resist (KTFR), a cyclised poly(cis-isoprene) rubber with a bisazide photoactivator.¹⁴ These materials are not available in NZ and would need to be synthesised. Simpler alternatives exist — bitumen (Judean bitumen was the photoresist used in Niépce’s earliest photographic experiments in the 1820s, and bitumen-based resists were used in early semiconductor work) — but resolution and process control are inferior. Developing a working photoresist chemistry from locally available materials is a significant research programme.
• Masks: The pattern to be transferred to the wafer must first be created on a mask — a glass plate with an opaque pattern (typically chromium on glass). Creating masks requires either a precision mechanical ruling engine (to cut the pattern at large scale for photographic reduction) or a photographic process with very high resolution. The minimum feature size achievable is directly limited by mask quality.
• Alignment: Each mask layer must be aligned to the previous layers with accuracy comparable to the minimum feature size. For 1960s-era circuits (features of 10–25 micrometres), alignment accuracy of 2–5 micrometres is required. This demands a contact or proximity aligner — essentially a precision microscope combined with a mechanical stage — that is within NZ’s manufacturing capability given a functioning machine shop and optical glass production, but is a precision instrument requiring careful design and construction.¹⁵
• Etching: Wet chemical etching (hydrofluoric acid for SiO₂, various acid mixtures for metals) was standard in the 1960s and is within reach of a functioning chemical industry. The acids involved are hazardous (HF is particularly dangerous — it penetrates skin and causes deep tissue burns with delayed onset).¹⁶
• Clean environment: Particulate contamination at any step can destroy a circuit. A dust particle larger than the minimum feature size that lands on the wafer during photoresist coating, exposure, or etching creates a defect. For 10–25 micrometre features, this means particles above approximately 5 micrometres must be controlled — achievable with HEPA filtration and basic clean room discipline, but requiring purpose-built facilities.

4.4 The clean room

A semiconductor clean room is not a sterile room in the medical sense. It is a room with controlled airborne particle counts, achieved through:

• HEPA filters (High Efficiency Particulate Air) — fibreglass filter media that capture 99.97% of particles 0.3 micrometres and larger. Manufacturing HEPA filters requires producing fine glass fibres and assembling them into a pleated filter medium in a sealed frame. This is within the capability of a glass production industry (Doc #98) but is a specialised product that would need development.
• Positive air pressure — clean air is forced into the room at slightly higher pressure than the surrounding environment, so that any air leakage flows outward (carrying particles out) rather than inward.
• Laminar airflow — air moves in a uniform, unidirectional pattern (typically downward from ceiling to floor), sweeping particles away from work surfaces rather than recirculating them.
• Gowning — workers wear full-body suits, gloves, masks, and booties that contain the millions of skin particles and fibres that the human body sheds per hour. The gown fabric must be non-shedding synthetic material — NZ’s wool and harakeke, while excellent for many purposes, are particle-generating fibres unsuitable for clean room use. Woven nylon or polyester is required, and NZ does not produce synthetic fibres.¹⁷
• Temperature and humidity control — photoresist behaviour is sensitive to both. Typical clean room conditions are 20–22°C and 40–45% relative humidity, maintained to tight tolerances.

The clean room standard required for 1960s-era semiconductor fabrication is approximately Class 10,000 (no more than 10,000 particles ≥0.5 micrometres per cubic foot of air). For comparison, a typical office environment is Class 500,000–1,000,000, and a modern semiconductor fab operates at Class 1 or better. Class 10,000 is achievable with HEPA filtration, basic airflow management, and disciplined procedures — it does not require the extraordinarily expensive environmental control systems of a modern fab.¹⁸

---

5. THE STAGED DEVELOPMENT PATH

5.1 Stage 0: The electronics bridge (Phase 1–4, Years 0–15)

NZ enters the recovery with an estimated several million electronic devices — computers, phones, radios, medical instruments, industrial controllers, vehicles with electronic systems.¹⁹ These devices are the semiconductor capability for the first decade or more. The bridge strategy (Doc #130) extends their useful life through:

• Centralised repair and component harvesting from failed devices
• Environmental management (temperature, humidity, dust) to slow degradation
• Prioritisation of electronic capability for highest-value applications
• Progressive substitution of electromechanical alternatives (relays, mechanical controls, analogue instruments) where feasible

The bridge ends when stockpiled devices and salvageable components are exhausted. This is not a cliff edge — different classes of electronics fail at different rates (simple analogue circuits last longer than complex digital ones; military-specification components outlast consumer-grade ones), so capability degrades progressively over perhaps 10–25 years depending on the category.²⁰

5.2 Stage 1: Discrete transistors (Phase 4–5, Years 7–30)

This is the Doc #135 programme: germanium point-contact transistors, then alloy junction transistors, fabricated from locally extracted and purified germanium. The endpoint is the ability to produce individual transistors with sufficient consistency and yield to build computers, radio equipment, and simple instrumentation.

What this gives NZ: Computers (Doc #134), HF radio transmitters and receivers with semiconductor amplification (more reliable, lower-power, and more compact than vacuum tube equivalents, though vacuum tubes can handle higher power levels per device), basic electronic instrumentation (amplifiers, oscillators, voltage regulators), and the industrial knowledge base for semiconductor materials processing.

What this does not give NZ: Integrated circuits, digital telecommunications, complex instrumentation, or anything approaching modern electronic capability. The gap between a discrete-transistor computer containing hundreds to low thousands of transistors and a modern microprocessor (containing billions of transistors) is a factor of roughly one million to one billion in circuit complexity.²¹

Prerequisites assumed met: Chemistry lab with acid production capability (Doc #113), glassblowing for distillation apparatus (Doc #98), wire drawing (Doc #70), machine shop (Doc #91), electrical grid.

5.3 Stage 2: Silicon purification and planar processing (Phase 5–6, Years 15–60)

The transition from germanium to silicon is the critical step toward integrated circuits. It requires:

Silicon purification chain:

1. Quartz sand to metallurgical-grade silicon — reduction with carbon in an electric arc furnace at approximately 1,800–2,000°C. NZ has high-purity quartz sand (Parengarenga Harbour deposit — see Doc #106) and electric power. The furnace technology is similar to that used in steelmaking (Doc #106) but requires graphite or carbon electrodes and a different operating regime. This step produces 98–99% pure silicon.²²
2. Metallurgical-grade to trichlorosilane — reaction with HCl gas at 300–350°C in a fluidised bed reactor. This requires an HCl supply from the chlor-alkali industry (electrolysis of brine — see Doc #103: Salt Production for the salt supply chain; the electrolysis step itself requires platinum or graphite electrodes and significant electrical power).²³
3. Trichlorosilane purification — fractional distillation in a column of sufficient height and efficiency to achieve the required purity. Multiple passes may be needed. The distillation column must itself be extremely clean — glass or high-purity quartz-lined construction.
4. Siemens CVD reactor — deposition of polycrystalline silicon from purified trichlorosilane. The reactor is a sealed bell jar (quartz or stainless steel) containing a silicon rod heated to approximately 1,100°C by passing electric current through it, with trichlorosilane and hydrogen gas flowing through the chamber. The deposited polysilicon builds up on the rod over hours to days.²⁴
5. Crystal growth — melting the polysilicon in a quartz crucible at 1,414°C and pulling a single crystal using the Czochralski method. The pull rate, rotation speed, and temperature gradients must be precisely controlled to produce a single crystal without defects. This is a significant precision engineering challenge.
6. Wafer slicing and polishing — the crystal boule is sliced into thin wafers (approximately 300–500 micrometres thick for early processes) using a diamond wire saw or internal-diameter saw, then polished to a mirror finish using progressively finer abrasives and chemical-mechanical polishing.

Thermal oxidation: Once silicon wafers are available, the first processing step is thermal oxidation — growing a thin layer of silicon dioxide by heating the wafer to 900–1,200°C in an oxygen or steam atmosphere. This oxide layer is the foundation of all subsequent processing. The furnace must be extremely clean (quartz tube, high-purity gas supply) to avoid contaminating the oxide.

Planar processing: The combination of thermal oxidation, photolithography, and selective etching that enables integrated circuits. A transistor is formed by: (1) growing an oxide, (2) patterning and etching windows in the oxide, (3) introducing dopant atoms (phosphorus for n-type, boron for p-type) through the windows by high-temperature diffusion, (4) growing another oxide, (5) patterning contact windows, and (6) depositing metal contacts. Each step must be controlled to tolerances measured in micrometres and fractions of micrometres.

Estimate: The development of this full capability from the Doc #135 foundation is a 15–40 year programme depending on the rate of industrial development, the availability of trade for critical materials and equipment, and the number of skilled people NZ can dedicate to the effort. This estimate is based on the historical timeline: the Siemens process was developed in 1954–1957; planar processing was invented at Fairchild in 1959; the first commercial integrated circuits appeared in 1961–1962. That 5–8 year arc from concept to product occurred in the United States with massive Cold War funding, dozens of PhD-level researchers, existing chemical and materials industries, and the knowledge that these things were physically possible. NZ would be working with a smaller team and a less developed industrial base, but with the advantage of knowing exactly what it is trying to achieve and how to get there.

5.4 Stage 3: Simple integrated circuits (Phase 6, Years 30–60)

The first integrated circuits are arrays of a few transistors and resistors on a single silicon chip, performing basic logic functions. The historical equivalents are:

• SSI (Small-Scale Integration): 1–10 logic gates per chip, 10–100 transistors total. Examples: 7400-series TTL logic (quad NAND gate = 4 gates, approximately 16 transistors).²⁵ These devices, produced in quantity, enable construction of computers from standardised building blocks rather than individual transistors — reducing assembly time by roughly an order of magnitude, improving reliability, and enabling more complex designs.
• MSI (Medium-Scale Integration): 10–100 logic gates per chip, 100–1,000 transistors. Examples: counters, shift registers, arithmetic logic units, small memory arrays. These enable single-board computers comparable to the minicomputers of the late 1960s.

The manufacturing requirements for SSI/MSI are demanding but well within what a 1960s-level industrial base could support:

• Feature size: 10–25 micrometres (achievable with contact lithography using UV light from a mercury lamp)
• Wafer size: 25–50 mm diameter (small by modern standards, but adequate)
• Number of mask levels: 5–8
• Clean room class: 10,000 (achievable with HEPA filtration)
• Yield: early production will be low (perhaps 10–30% of dice functional), improving with experience²⁶

5.5 Stage 4: Microprocessors (Phase 7, Years 50–100+)

A microprocessor — a complete central processing unit on a single chip — requires LSI (Large-Scale Integration), with thousands to tens of thousands of transistors on a single die. The Intel 4004 (1971) contained 2,300 transistors with a 10-micrometre feature size. The Intel 8080 (1974) contained 4,500 transistors at 6 micrometres.²⁷

Achieving this level of integration requires:

• Improved photolithography (projection alignment rather than contact, with better resolution)
• More precise process control (thinner oxides, shallower junctions, tighter dimensional tolerances)
• Higher-purity materials throughout the process chain
• More sophisticated circuit design tools (which themselves require computing capability — creating a productive feedback loop where earlier-generation computers assist in designing the next generation)
• Larger clean room facilities with better environmental control

This is the point at which the self-reinforcing nature of semiconductor development becomes important: better semiconductors enable better computers, which enable better design tools, which enable better semiconductors. The historical semiconductor industry crossed this threshold in the early 1970s. A recovering NZ (or more likely, a recovering region with a larger industrial base and population with which NZ trades) might reach this point somewhere in the range of 50–100+ years after the event, but this estimate is genuinely speculative. It depends on too many variables — population growth, trade development, competing priorities, institutional stability, access to materials — to narrow further.

---

6. NZ-SPECIFIC CONSIDERATIONS

6.1 Raw materials

Silicon (as quartz): NZ has abundant high-purity quartz sand at the Parengarenga Harbour deposit in Northland. This sand is approximately 96–99% SiO₂ and is suitable as a starting material for metallurgical-grade silicon production.²⁸ The deposit is large (tens of millions of tonnes) and the sand is accessible by surface mining. Transport from Parengarenga to a processing site is the main logistical challenge — the deposit is remote, accessible by unsealed road and coastal shipping.

Germanium: Available as a trace element in sphalerite (zinc sulfide) deposits. NZ’s sphalerite resources are limited and germanium concentration is variable (see Doc #135). Adequate for discrete transistor fabrication at small scale; likely insufficient for a large semiconductor industry.

Dopants: Phosphorus (n-type dopant) is available from NZ rock phosphate deposits (the same sources used for agricultural fertiliser). Boron (p-type dopant) is more problematic — NZ has no known borate mineral deposits. Boron could potentially be obtained from geothermal fluids (some NZ geothermal waters contain boron at concentrations of 10–50 mg/L) or from trade with Australia or Turkey (the world’s major borate producers).²⁹

Hydrochloric acid: Required for trichlorosilane production. Produced by reacting salt (NaCl) with sulfuric acid (Doc #113), or by direct synthesis from hydrogen and chlorine gas (products of brine electrolysis). NZ has salt (Doc #103) and a renewable electricity supply for electrolysis.

Hydrogen: Required for the Siemens process and various semiconductor processing steps. Produced by water electrolysis using NZ’s renewable grid. NZ has an abundance of the two inputs required: water and electricity.

Copper, aluminium, and gold: Used for interconnections on integrated circuits. Copper and aluminium are available (Doc #109). Gold is produced in NZ (historically from the Coromandel and West Coast goldfields, with small-scale production continuing). Gold wire bonding for chip packaging is the standard interconnection method and requires very thin gold wire — a wire drawing capability (Doc #70) adapted for precious metals.³⁰

6.2 Energy

Semiconductor fabrication is energy-intensive. The Siemens CVD process alone consumes approximately 100–200 kWh per kilogram of polysilicon produced.³¹ An electric arc furnace for metallurgical-grade silicon requires approximately 11–13 kWh per kilogram.³² Crystal growth, thermal oxidation, and clean room HVAC add further demand.

NZ’s renewable electrical grid is a genuine advantage here. A small semiconductor research and production facility might consume 1–5 MW of continuous power — significant but manageable within NZ’s grid capacity (approximately 9,500–10,000 MW of installed generation capacity pre-war, with effective capacity lower due to hydro variability and plant availability).³³

6.3 Population and workforce

Semiconductor fabrication requires a concentration of specialised skills that is disproportionate to NZ’s population. A minimal integrated circuit fabrication facility (producing simple SSI/MSI circuits at low volume) might require:

• 5–10 process engineers with deep knowledge of semiconductor physics and chemistry
• 20–50 production technicians trained in clean room operations
• 5–10 equipment maintenance technicians (mechanical, electrical, vacuum systems)
• 3–5 circuit designers
• 5–10 chemical processing workers (silicon purification, resist preparation, etchant formulation)
• Supporting roles: quality control, administration, training

Total: approximately 50–100 skilled workers dedicated to semiconductor production. In a nation of 4–5 million, this is a small number — but these are among the most highly trained technical workers in the economy, and every one of them is a person not doing something else. The opportunity cost is real and must be weighed against the alternative of importing semiconductors through trade.³⁴

6.4 NZ’s realistic role

The honest assessment: NZ is more likely to be a consumer and adapter of semiconductors than a manufacturer. The historical semiconductor industry has always been concentrated in regions with large populations, extensive industrial infrastructure, and massive capital investment — first the United States, then Japan, South Korea, Taiwan, and China. Even pre-war, NZ had no semiconductor wafer fabrication and minimal packaging or testing.

NZ’s most productive semiconductor-related roles are likely to be:

1. Knowledge preservation and dissemination — NZ’s AI facility and printing capability can produce and distribute semiconductor fabrication documentation to other surviving regions
2. Circuit design — designing integrated circuits for specific applications requires computing capability and engineering skill but not fabrication infrastructure
3. Testing and packaging — testing fabricated circuits and packaging them for use requires less infrastructure than fabrication
4. Application development — designing systems that use whatever semiconductors are available (from trade or local production) to solve NZ-specific problems

If NZ does pursue wafer fabrication, it should be at the scale of a research and development facility producing small quantities of simple circuits, not a production fab competing with larger regions. The value would be in maintaining the knowledge and capability as strategic insurance, not in cost-competitive manufacturing.

---

CRITICAL UNCERTAINTIES

Uncertainty	Range	Impact if Worse	Impact if Better
Trade availability of semiconductors	May never develop; may develop within decades	NZ must develop local capability or accept permanent electronics regression	NZ imports chips, focuses effort elsewhere
Achievable silicon purity with NZ-built equipment	Unknown — may be limited to 99.99% (four nines) initially	Devices are low-performance; useful for some applications but not complex ICs	Faster path to working ICs
NZ boron availability	Geothermal extraction feasibility unknown	Must trade for boron or develop alternative dopant strategies	P-type doping achievable locally
Clean room gown material	NZ does not produce synthetic fibres	Must trade for gown material or accept higher particle counts	Clean room operational
Prerequisite industry development timeline	15–50 years to build full chemical/metallurgical base	IC fabrication delayed proportionally	Earlier start possible
Workforce availability	50–100 specialist workers needed	Competition with other critical industries for talent	Sufficient workforce if training pipeline established early
Photoresist development	No local source; must synthesise or trade	Circuit patterning quality limited	Adequate feature resolution for SSI/MSI

---

CROSS-REFERENCES

Document	Relationship
Doc #135: Computer Construction	Primary prerequisite. Establishes discrete transistor fabrication, germanium processing, and the computing capability that semiconductor development depends on. The bridge from Doc #135’s discrete transistors to this document’s integrated circuits is the central technical narrative.
Doc #106: NZ Steel Glenbrook	Steel for equipment construction; electric arc furnace experience relevant to silicon smelting.
Doc #91: Machine Shop Operations	Precision fabrication of semiconductor processing equipment — furnaces, jigs, alignment tools, clean room components.
Doc #98: Glass Production	Quartz tubes for furnaces, laboratory glassware for chemical processing, optical glass for photolithography lenses.
Doc #113: Sulfuric Acid	Acid production enables the chemical industry that semiconductor fabrication depends on.
Doc #70: Wire and Cable Production	Copper wire for equipment, interconnections, and electrical supply. Fine wire drawing for chip bonding.
Doc #103: Salt Production	Salt for chlor-alkali industry producing HCl and NaOH — essential feedstocks.
Doc #102: Charcoal Production	Carbon source for metallurgical-grade silicon production.
Doc #128: HF Radio Network	Radio systems that benefit from discrete transistor and eventually IC availability.
Doc #130: Device Life Extension	The bridge strategy that buys time for semiconductor development.
Doc #134: Computing Self-Sufficiency Roadmap	Broader computing strategy within which semiconductor development sits.
Doc #157: Accelerated Trade Training	Training pipeline for the specialist workforce semiconductor fabrication requires.
Doc #162: University and Research Reorientation	Research resource allocation — semiconductor development competes with other priorities.

---

FOOTNOTES

---

1. NZ’s population and industrial constraints relative to semiconductor fabrication: NZ’s pre-war population of approximately 5 million is roughly comparable to the population of Singapore (5.9 million) or Ireland (5.1 million), neither of which developed domestic wafer fabrication despite significantly larger industrial bases and more extensive international supply chain integration. The minimum efficient scale for even a research-grade fab historically required hundreds of specialists and a supporting industrial ecosystem that nations of this size have generally imported rather than built. See: general semiconductor industry analysis; Lécuyer (note 4).↩︎

2. NZ semiconductor-related academic research: The University of Canterbury’s Electrical and Computer Engineering department has conducted microelectronics and MEMS research. The University of Auckland’s Department of Electrical, Computer, and Software Engineering has expertise in electronic design. Victoria University of Wellington’s School of Engineering and Computer Science also has relevant programmes. The MacDiarmid Institute for Advanced Materials and Nanotechnology, a national Centre of Research Excellence, has conducted research on semiconductor materials. None of these constitute a semiconductor fabrication capability, but the researchers and graduate students represent NZ’s closest domestic knowledge base. See: MacDiarmid Institute, https://www.macdiarmid.ac.nz/; university department websites.↩︎

3. Bitumen photoresist limitations: Bitumen (asphalt) resists, while historically significant (Niépce used bitumen of Judea for the first photographic experiments in the 1820s), have substantially poorer resolution than synthetic resists. Bitumen resists are typically limited to feature sizes above approximately 50–100 micrometres due to poor edge definition, slow exposure response, and difficulty controlling film thickness uniformly. The KTFR synthetic resists used in 1960s semiconductor fabrication achieved 5–10 micrometre resolution. The performance gap between locally formulable bitumen resists and synthetic resists is therefore roughly one order of magnitude in feature size. See: Moreau (note 11); Thompson et al. (note 11).↩︎

4. Clean room classification: Federal Standard 209E (US, now replaced by ISO 14644) classified clean rooms by the maximum number of particles ≥0.5 micrometres per cubic foot of air. Class 10,000 (ISO 7) permits 10,000 such particles per cubic foot. Early semiconductor fabs (1960s) operated at approximately Class 10,000. Modern fabs operate at Class 1 (ISO 3) or better. The jump from Class 10,000 to Class 1 represents approximately four orders of magnitude improvement and required decades of engineering development. See: ISO 14644-1:2015; Whyte (note 14).↩︎

5. Grid control transition from electronic to electromechanical: NZ’s power grid operated with electromechanical relays and analogue controls before the transition to digital systems (SCADA — Supervisory Control and Data Acquisition) from the 1980s onward. Electromechanical protection relays (overcurrent, distance, differential) are well-proven technology and NZ’s grid was designed around them. The regression would reduce monitoring granularity, response speed, and remote control capability, but would not prevent grid operation. See: Transpower NZ; general power systems engineering literature (e.g., Glover, Sarma, and Overbye, “Power Systems Analysis and Design”).↩︎

6. Medical diagnostic capability without electronics: Pre-electronic medicine relied on clinical examination, auscultation (stethoscope), manual blood pressure measurement (sphygmomanometer), optical microscopy, and basic chemical laboratory tests (urinalysis, blood chemistry by manual methods). These provide significant diagnostic capability but miss conditions detectable by imaging (ultrasound, X-ray) and advanced laboratory analysis (complete blood count by automated analyser, blood gas analysis, immunoassays). See: general medical history literature.↩︎

7. Fairchild Semiconductor early workforce: Fairchild Semiconductor was founded in 1957 by eight engineers (the “traitorous eight”) with financial backing from Fairchild Camera and Instrument. The company grew rapidly — from a handful of employees to several hundred within 2–3 years and several thousand by the mid-1960s. Early semiconductor firms (Texas Instruments, Fairchild, Motorola, RCA) each employed hundreds to thousands of workers in fabrication, with production workers outnumbering engineers by approximately 5:1 to 10:1. See: Lécuyer, C., “Making Silicon Valley: Innovation and the Growth of High Tech, 1930–1970,” MIT Press, 2006; Riordan, M. and Hoddeson, L., “Crystal Fire: The Birth of the Information Age,” W.W. Norton, 1997.↩︎

8. Australia’s semiconductor industry: Pre-war Australia had minimal semiconductor wafer fabrication. Some design, testing, and packaging operations existed, but virtually all wafer fabrication for the Australian market occurred offshore (primarily in Asia). Australia does have relevant research capability (CSIRO, university programmes) and a more extensive chemical and mining industry than NZ. See: Australian Semiconductor Industry Association; CSIRO manufacturing research.↩︎

9. Silicon purity requirements: Metallurgical-grade silicon is 98–99% pure. Solar-grade silicon requires approximately 99.9999% (six nines). Electronic-grade silicon for integrated circuits requires 99.9999999% (nine nines) or better. Each additional nine of purity is exponentially harder to achieve. See: Luque, A. and Hegedus, S. (eds.), “Handbook of Photovoltaic Science and Engineering,” Wiley; Green, M.A., “Solar Cells: Operating Principles, Technology and System Applications.”↩︎

10. The Siemens process: Developed by Siemens AG in Germany in the 1950s, this process remains the dominant method for producing high-purity polycrystalline silicon as of the 2020s. Alternative processes (fluidised bed reactor deposition, upgraded metallurgical-grade silicon) have been developed for lower-cost solar-grade silicon but do not achieve electronic-grade purity. See: Ceccaroli, B. and Lohne, O., “Solar Grade Silicon Feedstock,” in “Handbook of Photovoltaic Science and Engineering”; Dietl, J. et al., “Crystals: Growth, Properties, and Applications,” Springer.↩︎

11. Germanium vs. silicon processing temperatures: Germanium melts at 938.25°C; silicon at 1,414°C. Germanium tetrachloride (GeCl₄) boils at 84°C; silicon tetrachloride (SiCl₄) boils at 57.6°C, but trichlorosilane (SiHCl₃, the preferred purification intermediate) boils at 31.8°C. The lower temperatures for germanium processing translate directly to simpler furnace requirements, less demanding refractory materials, and lower energy consumption. See: Sze, S.M. and Ng, K.K., “Physics of Semiconductor Devices,” Wiley.↩︎

12. Silicon dioxide as enabling material: The importance of SiO₂ to semiconductor technology is difficult to overstate. A thermally grown SiO₂ layer on silicon is: (a) an excellent electrical insulator (breakdown strength ~10 MV/cm), (b) a diffusion barrier for common dopants (phosphorus, boron can be selectively diffused through windows in the oxide while the oxide protects other areas), (c) chemically stable and resistant to most processing chemicals, (d) selectively patterned by hydrofluoric acid etching (HF dissolves SiO₂ but not silicon), and (e) forms a nearly perfect interface with the silicon surface. No other semiconductor has an equivalently useful native oxide. This is the fundamental reason silicon dominates electronics. See: Deal, B.E. and Grove, A.S., “General Relationship for the Thermal Oxidation of Silicon,” Journal of Applied Physics, 36(12), 1965.↩︎

13. Integrated circuit fabrication process overview: The planar process was invented by Jean Hoerni at Fairchild Semiconductor in 1959. The first monolithic integrated circuit was demonstrated by Robert Noyce (Fairchild) in 1959 and Jack Kilby (Texas Instruments) in 1958. The basic process of oxidation, photolithography, etching, and diffusion remained the core of IC fabrication through the 1960s–1980s, with progressive refinements in feature size, layer count, and process complexity. See: Riordan and Hoddeson (note 4); Mead, C. and Conway, L., “Introduction to VLSI Systems,” Addison-Wesley, 1980.↩︎

14. Early photoresist materials: Kodak Thin Film Resist (KTFR), introduced in the late 1950s, was the first widely used photoresist for semiconductor fabrication. It is a negative-acting resist based on cyclised polyisoprene with a bisazide crosslinker. When exposed to UV light, the bisazide crosslinks the polymer, making it insoluble in the developer. Dichromated gelatin was used earlier. Bitumen (asphalt) resists were used in the earliest photolithographic experiments (Niépce, 1826) and in some early semiconductor work. See: Moreau, W.M., “Semiconductor Lithography: Principles, Practices, and Materials,” Springer; Thompson, L.F. et al. (eds.), “Introduction to Microlithography,” ACS.↩︎

15. Contact aligner specifications: A contact aligner for 1960s-era semiconductor fabrication consists of: a UV light source (mercury lamp at 365 nm or 405 nm), a mask holder, a wafer chuck, a microscope for alignment, and mechanical stage controls for X, Y, and theta positioning. Feature sizes of 10–25 micrometres require alignment accuracy of approximately 2–5 micrometres. These are achievable with precision mechanical components — lead screws, flexure stages, and micrometer adjustments — within the capability of a well-equipped machine shop. See: general semiconductor equipment literature; Jaeger, R.C., “Introduction to Microelectronic Fabrication,” Prentice Hall.↩︎

16. Hydrofluoric acid hazards: HF is uniquely dangerous among common acids. It penetrates skin rapidly (the fluoride ion is very small and mobile), causes deep tissue and bone damage, and can cause fatal systemic poisoning through skin absorption of relatively small quantities (as little as 2.5% body surface area exposure to concentrated HF can be lethal). Burns may not be immediately painful, delaying treatment. Treatment requires calcium gluconate. Anyone working with HF must be specifically trained in its hazards. See: Kirkpatrick, J.J.R. et al., “Hydrofluoric acid burns: a review,” Burns, 21(7), 1995.↩︎

17. Clean room gowning materials: Standard clean room garments are made from continuous-filament polyester (typically Dacron or equivalent) woven in a pattern that minimises particle shedding. Natural fibres (wool, cotton, linen, harakeke) are unsuitable because they shed short fibres continuously. NZ produces no synthetic fibres domestically. Clean room gowns would need to be either stockpiled from pre-war supplies, imported via trade, or potentially fabricated from NZ-produced nylon if a caprolactam synthesis pathway is developed (a significant chemical industry challenge in itself). See: IEST (Institute of Environmental Sciences and Technology) clean room standards; Whyte, W., “Cleanroom Technology: Fundamentals of Design, Testing and Operation,” Wiley.↩︎

18. Clean room classification: Federal Standard 209E (US, now replaced by ISO 14644) classified clean rooms by the maximum number of particles ≥0.5 micrometres per cubic foot of air. Class 10,000 (ISO 7) permits 10,000 such particles per cubic foot. Early semiconductor fabs (1960s) operated at approximately Class 10,000. Modern fabs operate at Class 1 (ISO 3) or better. The jump from Class 10,000 to Class 1 represents approximately four orders of magnitude improvement and required decades of engineering development. See: ISO 14644-1:2015; Whyte (note 14).↩︎

19. NZ electronic device inventory: As of the 2020s, NZ had approximately 5–6 million smartphones, 2–3 million personal computers, and millions of additional electronic devices across industrial, medical, transport, and household applications. The total number of discrete electronic devices is difficult to estimate precisely but is likely in the range of 15–30 million units of various types. See: NZ Telecommunications Forum; Stats NZ household surveys; general industry estimates.↩︎

20. Electronics degradation rates: Electrolytic capacitors (common failure point in power supplies and many circuits) have typical lifespans of 5–15 years depending on temperature and quality. Semiconductor devices themselves can last decades under benign conditions but are sensitive to electrostatic discharge and power surges. Displays (LCD, LED, CRT) degrade over 5–20 years. Connectors and solder joints degrade from thermal cycling and oxidation. Military-specification components are designed for longer life (25+ years) but are a small fraction of NZ’s electronics inventory. See: general reliability engineering literature; MIL-HDBK-217 (Military Handbook: Reliability Prediction of Electronic Equipment).↩︎

21. Complexity gap between discrete-transistor and modern computing: A discrete-transistor computer of the type described in Doc #135 might contain 500–2,000 transistors. A modern microprocessor (as of the 2020s) contains 10–100 billion transistors. The ratio is approximately 10⁶ to 10⁸ — a factor of one million to one hundred million. This gap represents roughly 60–70 years of semiconductor scaling and cannot be bridged in a single generation. See: Intel corporate history; Moore, G.E., “Cramming More Components onto Integrated Circuits,” Electronics, 38(8), 1965.↩︎

22. Metallurgical-grade silicon production: The carbothermic reduction of quartz (SiO₂ + 2C → Si + 2CO) occurs in a submerged-arc electric furnace at approximately 1,800–2,000°C. Global production of metallurgical-grade silicon is approximately 2–3 million tonnes per year (as of 2020s), primarily in China. A small furnace producing tens to hundreds of tonnes per year would be adequate for NZ’s semiconductor research needs. See: Schei, A. et al., “Production of High Silicon Alloys,” Tapir; general metallurgical engineering literature.↩︎

23. Chlor-alkali process: Electrolysis of sodium chloride brine produces chlorine gas (Cl₂), sodium hydroxide (NaOH), and hydrogen gas (H₂). The chlorine can be combined with hydrogen to produce hydrochloric acid (HCl). Modern chlor-alkali plants use membrane cell technology, but the older mercury cell and diaphragm cell processes are simpler to construct. This is a mature industrial process with over a century of commercial operation. NZ has salt from seawater evaporation or rock salt deposits, and electricity from its renewable grid. See: O’Brien, T.F. et al., “Handbook of Chlor-Alkali Technology,” Springer.↩︎

24. Siemens CVD reactor details: A Siemens reactor typically produces 50–150 kg of polysilicon per batch over 60–120 hours. The silicon rod (seed rod) is heated to approximately 1,050–1,150°C by passing electric current through it. Trichlorosilane (diluted in hydrogen at a molar ratio of approximately 1:10) flows over the heated rod and decomposes, depositing silicon. The process is energy-intensive — approximately 100–200 kWh per kg of polysilicon, with most of the energy going to maintaining the rod temperature and heating the gas mixture. See: Ceccaroli and Lohne (note 7); Dietl et al. (note 7).↩︎

25. 7400-series TTL logic: The Texas Instruments SN7400 (quad 2-input NAND gate), introduced in 1964, contains 4 NAND gates with approximately 16 transistors total. The 7400 series became the standard building-block logic family for digital systems from the mid-1960s through the 1980s. See: Texas Instruments, “TTL Data Book”; Lancaster, D., “TTL Cookbook,” Howard W. Sams, 1974.↩︎

26. Semiconductor yield: Yield (the fraction of fabricated dice that function correctly) depends on defect density, die area, process maturity, and design complexity. Early integrated circuit production in the 1960s typically achieved yields of 5–30% for SSI circuits, improving to 30–70% as processes matured. Yield is directly related to defect density and die area — smaller dice and cleaner processes yield more. See: Van Zant, P., “Microchip Fabrication: A Practical Guide to Semiconductor Processing,” McGraw-Hill.↩︎

27. Early microprocessor transistor counts: Intel 4004 (1971): 2,300 transistors, 10 µm feature size, 12 mm² die area. Intel 8008 (1972): 3,500 transistors, 10 µm. Intel 8080 (1974): 4,500 transistors, 6 µm. These devices were fabricated using PMOS or NMOS processes with 5–7 mask layers. See: Intel corporate history; Faggin, F. et al., “The History of the 4004,” IEEE Micro, 16(6), 1996.↩︎

28. Parengarenga Harbour silica sand: The deposit at the northern tip of the North Island is one of NZ’s most significant mineral resources. The sand is high-purity silica (approximately 96–99% SiO₂, with the balance primarily alumina and iron oxide), suitable for glass production and potentially for metallurgical-grade silicon after beneficiation. The deposit contains an estimated 20–40+ million tonnes. It has been mined intermittently for glass production and export. See: GNS Science mineral resources; Crown Minerals; NZ Petroleum and Minerals.↩︎

29. Boron in NZ geothermal fluids: Several NZ geothermal fields contain boron at concentrations of 10–50 mg/L in geothermal waters (notably Wairakei, Kawerau, and Ohaaki). Boron extraction from geothermal brines is practised in some locations globally (notably at the Salton Sea, California). The concentrations in NZ geothermal waters are lower than optimal for extraction but potentially viable if demand is small (semiconductor doping requires very small quantities of boron). See: Ármannsson, H., “The fluid geochemistry of Icelandic high temperature geothermal areas,” Applied Geochemistry; NZ geothermal chemistry literature.↩︎

30. NZ gold production: NZ has a long history of gold mining, primarily from the Coromandel Peninsula (Hauraki goldfield) and the West Coast of the South Island. As of the 2020s, NZ produces approximately 5–10 tonnes of gold per year (the Waihi mine in the Coromandel being the largest producer). Gold for semiconductor wire bonding requires very high purity (99.99% or better), achievable through electrolytic refining — a process within NZ’s capability given electricity and a basic chemical lab. See: NZ Petroleum and Minerals; Crown Minerals; Waihi Gold (OceanaGold). https://www.nzpam.govt.nz/↩︎

31. Siemens process energy consumption: Estimates of energy consumption for the Siemens CVD process range from approximately 80–250 kWh per kg of polysilicon, depending on reactor design, batch size, and operating conditions. The most commonly cited figure is approximately 120–150 kWh/kg. See: Ceccaroli and Lohne (note 7); Müller, A. et al., “Silicon for Photovoltaics,” in Fraas, L.M. and Partain, L.D. (eds.), “Solar Cells and Their Applications,” Wiley.↩︎

32. Metallurgical-grade silicon energy consumption: Approximately 11–13 kWh per kg of silicon metal, in a submerged-arc furnace. This makes silicon smelting energy-intensive but not exceptionally so by industrial standards — comparable to aluminium smelting (approximately 13–15 kWh/kg). See: Schei et al. (note 17); USGS Mineral Commodity Summaries, Silicon.↩︎

33. NZ electricity generation capacity: As of the 2020s, NZ’s installed electricity generation capacity is approximately 9,500–10,000 MW, with effective capacity somewhat lower due to hydro variability and plant availability. Peak demand is approximately 6,500–7,000 MW. Under recovery conditions, demand would be significantly lower, and the surplus could support industrial development. See: MBIE, “NZ Energy Data Tables”; Electricity Authority. https://www.ea.govt.nz/↩︎

34. Workforce requirements for semiconductor fabrication: The estimates in this document are based on the historical size of early semiconductor operations (1960s), scaled for low-volume production. A modern semiconductor fab employs thousands of workers, but modern fabs produce billions of transistors per chip — the scale is incomparable. A facility producing simple SSI/MSI circuits at volumes of thousands to tens of thousands of dice per year (sufficient for NZ’s domestic needs) would operate at a scale closer to a university research fab or a specialised low-volume foundry. See: Lécuyer (note 4); Van Zant (note 20).↩︎