Physical vapor deposition (PVD - sputtering): Question bank

Build-the-picture: four ideas the traps hinge on

Before the question bank, four small visual builds so every trap answer has a mental image behind it.

(A) Where $γ$ comes from — a two-ball collision

We have two book-keeping rules. Momentum (mass × velocity) can't appear or vanish, and for an elastic hit kinetic energy ( $\frac{1}{2} m v^{2}$ ) is also conserved. Writing the ion's final speed $v_{1}^{'}$ and the atom's final speed $v_{2}^{'}$ :

$M_{1} v = M_{1} v_{1}^{'} + M_{2} v_{2}^{'} \frac{1}{2} M_{1} v^{2} = \frac{1}{2} M_{1} v_{1}^{'2} + \frac{1}{2} M_{2} v_{2}^{'2}$

Solving these two together (standard 1-D elastic result) gives the struck atom's speed, and squaring it into its kinetic energy $E_{2} = \frac{1}{2} M_{2} v_{2}^{'2}$ divided by the incoming energy $E = \frac{1}{2} M_{1} v^{2}$ :

$γ = \frac{E _{2}}{E} = \frac{4 M _{1} M _{2}}{( M _{1} + M _{2} ) ^{2}}$

(B) Why $γ$ peaks at equal mass

Look at the hump: $γ$ climbs to exactly $1$ when $M_{1} = M_{2}$ and sags on both sides. A tiny ion off a huge atom bounces back (keeps its energy); a huge ion barely notices a tiny atom (keeps its energy). Perfect sharing needs a mass match — which is why heavy argon pairs well with mid-weight metals.

(C) Why a threshold energy exists

That is the whole origin of $E_{t h} = U_{s} / γ$ : below it the offered energy can't lift the atom over the wall, so yield is flat-zero; above it, extra energy frees more atoms and yield rises (until, far past threshold, ions bury too deep and it falls again).

(D) How $Y$ and $J$ combine into growth rate $R$ , and where pressure enters

The transport step depends on the mean free path $λ$ — the average straight-line distance an atom flies before hitting a gas molecule. From kinetic theory $λ \propto 1/ P$ : double the pressure, halve the free path.

When $λ$ is larger than the target–wafer gap, atoms fly ballistically (straight, energetic). When pressure rises and $λ$ drops below the gap, they go diffusive (zig-zag, arrive tired and scattered) → porous films. That single curve is why "more pressure = better" is a trap.

True or false — justify

Sputtering vaporizes the target by heating it.

False. Atoms leave by momentum transfer in a collision cascade (figure A); the target is actively cooled and stays solid. Only evaporation-PVD uses heat.

Argon is chosen mainly because it is cheap.

False (cheap is a bonus). The real reasons are that it is inert (won't react with target or film) and heavy (mass close to many metals, so

γ

sits near the hump peak in figure B).

A higher chamber pressure always gives a faster, better film.

False. More gas shortens the mean free path

λ \propto 1/ P

(figure D), so ejected atoms scatter, lose energy and arrive at random angles → porous, poorly-adhering films.

The energy-transfer fraction

γ = \frac{4 M _{1} M _{2}}{( M _{1} + M _{2} ) ^{2}}

is largest when the ion is much heavier than the target atom.

False.

γ

peaks at

M_{1} = M_{2}

(giving

γ = 1

, figure B); a large mass mismatch either way lowers it.

DC sputtering works for any target you can machine into a slab.

False. Insulating targets (SiO₂, Al₂O₃) charge up positive under DC and repel further ions, quenching the plasma; they need RF.

A magnetron raises the deposition rate by increasing the sputter yield

Y

False. It raises the ion flux

J

(denser plasma), not

Y

; since

R \propto Y \cdot J

, more

J

at the same

Y

gives more rate.

Below the threshold energy

E_{t h}

an ion still ejects atoms, just fewer of them.

False. Below

E_{t h}

the transferred energy

γ E

can't clear the escape gate

U_{s}

(figure C), so ideally zero atoms escape — it is a true threshold, not a slow taper.

RF sputtering exists so we can sputter faster than DC.

False. RF exists to neutralize charge build-up on insulators; for plain metals DC is simpler and perfectly fine.

Sputtering is a form of Chemical Vapor Deposition because both use gas.

False. The gas in sputtering (Ar) only supplies ions; the film atoms come physically from a solid target with no chemical reaction — that is exactly the PVD/CVD dividing line.

Reactive sputtering is still PVD even though a gas chemically reacts.

True, with care. The metal atoms are still ejected physically from a solid target (PVD), but a reactive gas (O₂ or N₂) is added so a compound like TiN or Al₂O₃ forms on the wafer — a hybrid where a chemical step rides on a physical ejection.

Spot the error

"We melt the tungsten target because tungsten's melting point is high, so sputtering it needs lots of heat."

The word melt is the error — high-melting metals sputter fine precisely because the process is mechanical (momentum, figure A), not thermal, so no melting is required.

"Sputter yield keeps climbing forever as we raise ion energy."

At very high energy the error appears — ions bury deep and deposit their energy too far below the surface to eject atoms, so yield saturates then falls (right side of figure C's yield curve).

"To sputter an insulator we just raise the DC voltage until the plasma stays on."

More DC voltage still can't stop positive charge accumulating on the insulator; the surface keeps charging until it repels ions — the fix is switching to RF, not raising voltage.

"Magnetron magnets trap the argon ions in the racetrack."

They trap electrons, not ions (ions are far too massive for the

E \times B

drift); trapped electrons then ionize more Ar.

"Higher

γ

means the ion loses less energy, so it's less efficient."

Reversed — higher

γ

means more of the ion's energy reaches the target atom, making ejection more efficient and lowering

E_{t h} = U_{s} / γ

"Since

E_{t h}

for Al is about 3.5 eV, we run the ions at 3.5 eV to sputter Al."

Running at threshold gives essentially no yield (figure C); in practice ions are accelerated to hundreds of eV, well into the linear-yield regime above threshold.

"In reactive sputtering we just flood the chamber with O₂ to speed up compound growth."

Too much reactive gas poisons the target — a compound layer forms on the target surface itself, its yield drops and the rate crashes; the reactive flow must be carefully throttled.

Why questions

Why does sputtering give better adhesion than a gentle deposition method?

Ejected atoms arrive with real kinetic energy, so they impinge hard and pack tightly into a dense, well-bonded film instead of settling loosely.

Why does the process need a vacuum first before backfilling argon?

To remove reactive gases (O₂, H₂O) that would contaminate the film, and to lengthen the mean free path so ions and ejected atoms travel without constant collisions.

Why does lowering pressure with a magnetron improve film quality even though there's "less gas"?

The dense trapped-electron plasma keeps ion flux high, so you don't need extra gas; the longer

λ

(figure D) lets atoms arrive straight and energetic → denser films. See Plasma Physics.

Why is momentum, not temperature, the controlling variable for whether an atom is ejected?

Ejection is a collision cascade — the ion's momentum threads through a few atomic layers and kicks a surface atom backward out; temperature never enters the escape condition

γ E > U_{s}

Why does

E_{t h}

depend on the masses of both ion and target, not just the binding energy?

Because only the fraction

γ

of the ion's energy actually reaches the atom, and

γ

itself is set by the mass ratio (figure B) — so the mass mismatch scales how much energy you must supply.

Why do we care about the surface binding energy

U_{s}

rather than the bulk cohesive energy?

Sputtering ejects surface atoms, so the barrier that matters is the energy holding an atom at the surface — that is precisely

U_{s}

(the gate height in figure C).

Why is sputtering the go-to for metal interconnects rather than CVD?

Many metals lack safe, volatile precursors for CVD, whereas sputtering pulls atoms straight off a solid metal slab with good adhesion and tunable stress. See also Thin Film Deposition.

Why do we say

R \propto Y \cdot J

rather than just

R \propto Y

Growth rate counts atoms/area/second, which is atoms-per-ion (

Y

) times ions-per-area-per-second (

J

); leaving out flux would miss the whole reason magnetrons deposit faster.

Edge cases

What happens to the sputter yield when the ion and target atom have equal mass (

M_{1} = M_{2}

γ

hits its maximum of

1

(figure B peak) — the ion can transfer all its energy in a head-on hit, giving the lowest possible threshold

E_{t h} = U_{s}

What happens if the target is a perfect insulator and you refuse to switch off DC?

Positive charge accumulates until the surface potential cancels the applied field, ions stop arriving, and the plasma extinguishes — deposition stops entirely.

What happens as chamber pressure approaches zero (no gas at all)?

No Ar means no ions can form, so there is nothing to bombard the target — sputtering cannot start; you need enough gas to sustain the glow discharge.

What happens to ejected atoms if pressure is pushed very high?

λ

shrinks below the target-to-wafer gap (figure D), so atoms scatter many times, arrive from random directions with little energy → this diffusive regime yields loose, porous films; see also Sputter Etching for the opposite (removal) use of bombardment.

At exactly the threshold energy

E = E_{t h}

, what is the ideal yield?

Essentially zero — the offered energy

γ E

just equals

U_{s}

, so an atom is marginally freed but with no leftover energy, meaning no net ejection in the idealized model.

What if you fire ions far below

E_{t h}

for a long time — does a slow film eventually build?

No — sub-threshold ions can't dislodge atoms at all, so long exposure just heats/implants the target without depositing anything.

What happens in reactive sputtering if the reactive gas flow is set just right?

The metal sputters cleanly and the compound forms only on the wafer — but the operating window is narrow, sitting on a hysteresis between "metallic mode" (fast, metal-rich) and "poisoned mode" (slow, compound-covered target).

Recall One-line self-test

Cover everything above: can you state why each "False" is false in one sentence, pointing at the right figure (A collision, B mass hump, C threshold gate, D pressure curve)? If yes, the concept traps won't catch you.