Setup: A surface has a finite density of reactive sites N0 (e.g. -OH groups per cm²). When precursor A flows, it reacts with an available site with some probability. Let θ(t) = fraction of sites already reacted.
The rate of new reactions is proportional to (precursor arriving) × (sites still free):
dtdθ=kP(1−θ)
where P is precursor partial pressure and k a rate constant. Why (1−θ)? Because a site already occupied cannot react again — the reaction runs out of fuel.
Solve this first-order ODE (separable):
∫0θ1−θ′dθ′=∫0tkPdt′⇒−ln(1−θ)=kPtθ(t)=1−e−kPt
Growth per cycle then follows:
GPC=Natoms/volθmaxN0(thickness added per full cycle)
Typical GPC ≈ 0.9–1.2 Å/cycle for Al₂O₃. Total thickness:
Imagine painting a wall but the paint only sticks to dry spots. You spray "paint A" — it sticks everywhere it can and then stops (no dry spots left). You wipe away the extra spray. Then you spray "paint B" that makes the wall dry again. Wipe again. Each A-then-B round adds exactly one thin coat, always the same thickness — even inside deep cracks — because the paint decides for itself when to stop. Want it thicker? Just do more rounds!
Dekho, ALD ka core idea bahut simple hai: film ko hum ek-ek atomic layer karke banate hain, ek hi baar mein nahi. Iske liye chemical reaction ko do halves mein tod dete hain. Pehle ek precursor (jaise TMA) daalte hain — ye surface ke sites (jaise -OH groups) ke saath react karta hai aur jaise hi ssite khatam, reaction apne aap ruk jaata hai. Isko bolte hain self-limiting. Phir purge karke extra gas nikaal dete hain, taaki dono precursor gas mein aapas mein na milein. Phir doosra precursor (paani) daalte hain, wo surface reset kar deta hai. Ek cycle = A, purge, B, purge.
Iska sabse bada faayda? Thickness bilkul cycle count se control hoti hai — 50 cycle chahiye toh ~50 Å (5 nm) mil jaata hai, ekdum accurate. Time zyada dene se film moti nahi hoti, kyunki reaction saturate hoke ruk chuki hai. Ye baat yaad rakhna warna galti ho jaati hai — log sochte hain "zyada pulse = zyada thick", but ALD mein aisa nahi hota.
Doosra bada faayda conformality hai. Agar bahut deep trench hai (DRAM capacitor jaisa, 40:1 aspect ratio), toh bhi ALD neeche tak same thickness deta hai, kyunki har surface apne aap saturate ho jaati hai. Bas dose thoda lamba karo taaki molecule andar tak diffuse ho jaayein. Yahi cheez modern transistors ke high-k gate dielectric (HfO2, Al2O3) aur memory fabrication mein ALD ko must-have bana deti hai.
Formula wala part: coverage θ(t)=1−e−kPt hota hai — exponential curve jo dheere-dheere 1 (ya θmax) tak pahunchti hai. Iska matlab last ke sites bharne mein time lagta hai, isliye recipe mein thoda over-dose karte hain guarantee ke liye. Bas itna samajh lo toh ALD clear hai.