After you have patterned photoresist onto a wafer, etching is how you remove the material that is NOT protected . The resist acts like a stencil; etching is the "spray paint" that only marks where the stencil has holes.
The whole game is a fight between two goals:
Selectivity — eat the film you want, spare the mask and the layer underneath.
Anisotropy — cut straight down , not sideways, so tiny features keep their shape.
Wet etching (liquid chemistry) is cheap and gentle but etches sideways too. Dry/plasma etching (reactive gas) can cut vertically, which is why every modern chip uses it.
A subtractive fabrication step that removes exposed thin-film material through a patterned mask, transferring the mask pattern into the underlying film. Two families:
Wet etching : wafer dipped in a liquid etchant that chemically dissolves the film.
Dry etching : wafer exposed to a plasma of reactive ions/radicals in vacuum.
Definition Two figures of merit
Selectivity S = R film R mask/underlayer S = \dfrac{R_{\text{film}}}{R_{\text{mask/underlayer}}} S = R mask/underlayer R film — ratio of etch rates. High S S S = you barely touch what you want to keep.
Anisotropy A = 1 − R lateral R vertical A = 1 - \dfrac{R_{\text{lateral}}}{R_{\text{vertical}}} A = 1 − R vertical R lateral — measures how vertical the sidewall is.
Picture a mask opening of width w w w , over a film of thickness t t t . The etch removes material both down (rate R v R_v R v ) and sideways under the mask (rate R l R_l R l ).
To clear the film we etch for time τ \tau τ such that the vertical cut reaches the bottom:
R v τ = t ⇒ τ = t R v R_v\,\tau = t \quad\Rightarrow\quad \tau = \frac{t}{R_v} R v τ = t ⇒ τ = R v t
In that same time, the etch also crept sideways by an amount called the undercut u u u :
u = R l τ = R l ⋅ t R v = R l R v t u = R_l\,\tau = R_l\cdot\frac{t}{R_v} = \frac{R_l}{R_v}\,t u = R l τ = R l ⋅ R v t = R v R l t
Why this step? Undercut is set by the ratio of rates, scaled by film thickness — thicker films always undercut more for the same anisotropy.
Now define anisotropy A = 1 − R l / R v A = 1 - R_l/R_v A = 1 − R l / R v . Substitute R l / R v = 1 − A R_l/R_v = 1-A R l / R v = 1 − A :
u = ( 1 − A ) t \boxed{u = (1-A)\,t} u = ( 1 − A ) t
Perfectly isotropic (R l = R v R_l=R_v R l = R v ): A = 0 ⇒ u = t A=0 \Rightarrow u = t A = 0 ⇒ u = t . You lose a full film-thickness of width on each side!
Perfectly anisotropic (R l = 0 R_l=0 R l = 0 ): A = 1 ⇒ u = 0 A=1 \Rightarrow u = 0 A = 1 ⇒ u = 0 . Straight walls.
Intuition Feynman version of isotropy
A liquid molecule doesn't know which way is "down". Once the etchant touches the film through the mask hole, it reacts equally in every direction it can reach — including sideways under the resist. So R l ≈ R v R_l \approx R_v R l ≈ R v , giving A ≈ 0 A\approx 0 A ≈ 0 and rounded, undercut sidewalls.
Mechanism (3 steps, purely chemical):
Reactant diffuses to the surface.
Surface reaction dissolves the film (e.g. SiO 2 + 6 HF → H 2 SiF 6 + 2 H 2 O \text{SiO}_2 + 6\,\text{HF} \to \text{H}_2\text{SiF}_6 + 2\,\text{H}_2\text{O} SiO 2 + 6 HF → H 2 SiF 6 + 2 H 2 O ).
Products diffuse away.
Because it's chemical, wet etching is very selective (you pick chemistry that ignores the mask) but isotropic and rate-limited by diffusion/temperature (Arrhenius: R ∝ e − E a / k T R \propto e^{-E_a/kT} R ∝ e − E a / k T ).
Intuition The directional trick
In a plasma, ions are accelerated by an electric field toward the wafer — they arrive nearly vertical . Two things happen:
Physical sputtering : ions physically knock atoms off (directional, like sandblasting straight down).
Reactive Ion Etching (RIE) : reactive radicals chemically etch, but the vertical ion bombardment is what makes the reaction proceed at the bottom of trenches, while sidewalls (shielded by passivation films) barely react.
Net effect: R v ≫ R l ⇒ A → 1 R_v \gg R_l \Rightarrow A \to 1 R v ≫ R l ⇒ A → 1 . Straight walls, tiny features possible.
Property
Wet
Dry / Plasma
Mechanism
Chemical (liquid)
Chemical + physical (ions)
Anisotropy A A A
~0 (isotropic)
high (near 1 with RIE)
Selectivity
very high
moderate
Feature size
large only (undercut)
sub-micron ✔
Cost / throughput
cheap, batch
expensive vacuum tools
Damage
none
ion-induced surface damage
Waste
chemical disposal
toxic gases
Worked example 1 — Undercut of an oxide film
A 0.5 μ m 0.5\,\mu m 0.5 μ m SiO₂ film is wet-etched (isotropic, A = 0 A=0 A = 0 ). What is the undercut per side?
u = ( 1 − A ) t = ( 1 − 0 ) × 0.5 = 0.5 μ m u=(1-A)t=(1-0)\times0.5=0.5\,\mu m u = ( 1 − A ) t = ( 1 − 0 ) × 0.5 = 0.5 μ m .
Why this step? A = 0 A=0 A = 0 means lateral = vertical rate, so the mask is undercut by exactly one film thickness on each side. A designed 1 μ m 1\,\mu m 1 μ m line ends up 1 + 2 ( 0.5 ) = 2 μ m 1+2(0.5)=2\,\mu m 1 + 2 ( 0.5 ) = 2 μ m wide — hopeless for dense circuits.
Worked example 2 — Partially anisotropic dry etch
A 1 μ m 1\,\mu m 1 μ m film is etched with R v = 100 R_v=100 R v = 100 nm/min and R l = 10 R_l=10 R l = 10 nm/min.
A = 1 − 10 / 100 = 0.9 A = 1 - 10/100 = 0.9 A = 1 − 10/100 = 0.9 . Undercut u = ( 1 − 0.9 ) ( 1 μ m ) = 0.1 μ m u=(1-0.9)(1\,\mu m)=0.1\,\mu m u = ( 1 − 0.9 ) ( 1 μ m ) = 0.1 μ m per side.
Why this step? Even a "good" 90%-anisotropic etch still loses 100 nm sideways — critical for defining exactly how small your features can be.
Worked example 3 — Selectivity and overetch
Film 500 nm, mask etch rate 20 nm/min, film etch rate 400 nm/min. We overetch 20% (i.e. etch 1.2 × 1.2\times 1.2 × the time to clear film).
Time to clear: 500 / 400 = 1.25 500/400=1.25 500/400 = 1.25 min. With overetch: 1.5 1.5 1.5 min.
Mask lost: 20 × 1.5 = 30 20\times1.5=30 20 × 1.5 = 30 nm. Selectivity S = 400 / 20 = 20 S=400/20=20 S = 400/20 = 20 .
Why this step? Overetch is needed because films aren't perfectly uniform; high S S S ensures the mask/underlayer survives the extra time.
Common mistake "Higher selectivity means straighter walls."
Why it feels right: both sound like "quality" of the etch. The fix: selectivity and anisotropy are independent . Selectivity = ratio of etch rates between different materials . Anisotropy = ratio of vertical vs lateral rate in the same material . Wet etch can be super-selective yet totally isotropic.
Common mistake "Plasma etching is always faster than wet etching."
Why it feels right: "high-tech = fast." The fix: dry etch rates are often slower ; you use it for precision (anisotropy) , not speed. Wet etch can strip thick layers very quickly in a batch.
Common mistake "Undercut depends only on the etch chemistry."
Why it feels right: chemistry sets R l / R v R_l/R_v R l / R v . The fix: u = ( 1 − A ) t u=(1-A)t u = ( 1 − A ) t — undercut also scales with film thickness . A thick film undercuts more even at the same anisotropy.
Common mistake "RIE is purely physical sputtering."
Why it feels right: ions are physical. The fix: RIE = reactive ion etching — it's a synergy of chemical radicals + directional ion bombardment. The synergy etches faster than either alone.
Recall Explain to a 12-year-old (Feynman)
Imagine you tape a stencil onto a cake and you want to carve out the uncovered parts.
Wet etching is like dipping the cake in acid syrup — it melts the cake everywhere it touches, including sideways under the tape , so your carved shape gets fat and blurry.
Dry/plasma etching is like a super-fine sandblaster shooting straight down — it only digs where you aim, so the walls stay sharp and thin.
That straight-down trick is why your phone can have billions of tiny transistors.
Mnemonic Remember the trade-off
"Wet is Wide, Dry is Direct."
Wet → W ide (isotropic, undercut). Dry → D irectional (anisotropic). And RIE = Ram + React (ions ram down, radicals react ).
What is the purpose of the etching step? To remove exposed film material through a patterned mask, transferring the mask pattern into the film.
Define selectivity in etching. Ratio of etch rate of the target film to that of the mask/underlayer,
S = R f i l m / R m a s k S=R_{film}/R_{mask} S = R f i l m / R ma s k .
Define anisotropy. A = 1 − R l / R v A=1-R_l/R_v A = 1 − R l / R v ; measures how vertical the etch is (1 = perfectly vertical, 0 = isotropic).
Formula for undercut per side. u = ( 1 − A ) t u=(1-A)t u = ( 1 − A ) t where
t t t is film thickness.
Why is wet etching isotropic? The liquid etchant reacts equally in all directions (
R l ≈ R v R_l\approx R_v R l ≈ R v ), so it undercuts the mask.
Which etch type is highly selective but poor for small features? Wet etching.
Which etch enables sub-micron features and why? Dry/plasma (RIE) — vertical ion bombardment gives high anisotropy.
What does RIE combine? Chemical etching by radicals + directional physical ion bombardment (synergy).
For isotropic etch (A=0), undercut equals? The full film thickness
t t t on each side.
Why do we overetch? To clear film despite thickness/rate non-uniformity; requires high selectivity so mask/underlayer survives.
Temperature dependence of wet etch rate? Arrhenius:
R ∝ e − E a / k T R\propto e^{-E_a/kT} R ∝ e − E a / k T .
Main drawbacks of dry etching? Ion-induced surface damage, toxic gases, expensive vacuum equipment, lower selectivity.
Photolithography — creates the mask that etching transfers.
Photoresist — the etch-protecting stencil; selectivity is measured against it.
Thin-Film Deposition — puts down the film that etching later removes.
Plasma Physics — sheath fields accelerate ions vertically (source of anisotropy).
Critical Dimension & Feature Scaling — undercut sets the minimum line width.
Arrhenius Equation — governs temperature dependence of wet etch rate.
exposes film through holes
Patterned photoresist mask
Etching removes exposed film
Selectivity S = Rfilm / Rmask
Anisotropy A = 1 - Rl / Rv
Wet etching liquid etchant
Min feature needs s greater than 2t
Intuition Hinglish mein samjho
Etching ka matlab hai — photoresist mask lagane ke baad, jo material khula (exposed) hai use hataa dena. Mask ek stencil ki tarah kaam karta hai, aur etchant sirf hole wali jagah material ko kha jaata hai. Do main goals hote hain: selectivity (sirf film hate, mask aur neeche wala layer bache) aur anisotropy (seedha neeche kate, side me na phaile).
Wet etching me wafer ko liquid chemical me dubaate hain. Liquid ko pata nahi ki "neeche" kaunsi direction hai, isliye wo har taraf barabar react karta hai — matlab side me bhi, mask ke neeche. Isko kehte hain isotropic (A ≈ 0 A\approx0 A ≈ 0 ). Formula yaad rakho: undercut u = ( 1 − A ) t u=(1-A)t u = ( 1 − A ) t . Agar A = 0 A=0 A = 0 , to u = t u=t u = t — poori film-thickness jitna side me phail jaata hai! Isliye wet etch chhote (sub-micron) features ke liye bekaar hai, lekin sasta aur bahut selective hota hai.
Dry/plasma etching (RIE) me vacuum chamber me reactive gas ka plasma banate hain. Electric field ions ko wafer ki taraf seedha neeche accelerate karta hai. To etching mostly vertical hoti hai (R v ≫ R l R_v \gg R_l R v ≫ R l ), anisotropy A → 1 A\to1 A → 1 , aur walls straight rehti hain. Yahi wajah hai ki aaj ke chips me billions transistors fit ho paate hain. Trade-off: dry etch mehenga hai, thoda slow ho sakta hai, aur surface ko ion damage ho sakta hai.
Yaad rakhne ka trick: "Wet is Wide, Dry is Direct." Aur RIE = ions ram down + radicals react — dono ka synergy. Exam me sabse important cheez: selectivity aur anisotropy alag-alag concepts hain, confuse mat karna.