Electromigration reliability: Visual walkthrough — Hinglish deep dive

6.4.8 · D2 · Hardware › Power, Thermal & Reliability › Electromigration reliability

Hum assume karte hain ki tum sirf itna jaante ho: ek wire current carry karti hai, current moving charge hai, aur metal ek lattice mein baithe atoms se bana hai. Baaki sab hum draw karenge.

Step 1 — Current density $J$ kya hai, aur hum uski kyun parwah karte hain (current ki nahi)?

KYA HAI. Current $I$ yeh hai ki kitna charge per second flow karta hai. Lekin do wires same $I$ carry kar sakti hain aur bilkul alag behave kar sakti hain agar ek moti ho aur ek patli. Toh hum current ko wire ki cross-sectional area $A_{wire}$ se divide karte hain (woh "doorway" ka size jisme se charge squeeze hota hai):

$J = \frac{I}{A _{wire}}$

$I$ — wire ke through flow hone wala total charge per second (amps).
$A_{wire}$ — wire ki cross-section ki area (uski width $\times$ height).
$J$ — current density: flow kitna crowded hai. Units: amps per cm².

KYUN. Electromigration damage crowding ke baare mein hai, total flow ke nahi. Bahut saari cars ke saath ek wide highway shant rehti hai; wahi cars ek lane se force hone par chaos hai. Jaise chips shrink hoti hain, $A_{wire}$ shrink hoti hai lekin $I$ barely girta hai — toh $J$ shoot up karta hai. Yahi wajah hai ki $J$ , na ki $I$ , villain hai.

PICTURE. Same current, do doorways.

Step 2 — Ek metal atom par do forces

KYA HAI. Lattice mein baithe ek metal ion ko pick karo. Jab wire powered hoti hai, usse do pushes milte hain:

Direct field force $F_{field} = q E$ — electric field $E$ positive ion ko cathode ( $-$ terminal) ki taraf kheenchta hai.
Electron-wind force $F_{wind}$ — electrons ka ek dariya doosri taraf stream karta hai ( $+$ terminal, anode ki taraf). Har electron jo ion se takrata hai ek choti si thapki deta hai, jaise hawa ek jhande par.

$q$ — ek electron ka charge ( $1.6 \times 1 0^{- 19}$ C), hamare charge ki unit.
$E$ — electric field: wire ke saath voltage kitni steeply girta hai (volts per cm).

KYUN. Ek achhe metal mein bahut badi sankhya mein tez electrons hote hain aur woh tug-of-war jeet jaate hain. Atom par net push anode ki taraf hoti hai — wahi jis taraf electrons jaate hain. Yeh EM ka sabse zyada counter-intuitive fact hai (parent ke [!mistake] callout dekhein).

PICTURE. Ek ion, do arrows; wind arrow bada draw kiya hai kyunki woh dominate karta hai.

Step 3 — Force ko $J$ use karke rewrite karo (Ohm's law kaam aata hai)

KYA HAI. Hamare paas $F = Z^{*} q E$ hai, lekin $E$ (field) ek chip designer ko invisible hai. Jo woh control karte hain woh hai $J$ (current density). Local Ohm's law unhe connect karta hai:

$E = ρ J ⟹ F = Z^{*} q ρ J$

$ρ$ — resistivity: metal current se kitna ladta hai. Ek material constant.
$ρ J$ — $E$ ko replace karta hai. Bada current density → bada internal field → bada push.

KYUN. Yeh pivotal move hai. Ek tool (Ohm's law) enter karta hai kyunki yeh ek unmeasurable quantity ko us knob mein convert karta hai jise designers actually turn karte hain. Ab atom ki push directly $J$ ke saath scale karti hai — aur hum dekh chuke hain ki $J$ hi woh cheez hai jo chhoti chips mein explode karta hai.

PICTURE. Substitution ko same arrow par " $E$ " label ki jagah " $ρ J$ " swap karte hue dikhaya gaya hai.

Step 4 — Ek atom par force se bahut saare atoms ke flux tak

KYA HAI. Ek ion par force use slowly drift karati hai. Multiply karo ki kitne atoms hain aur woh kitni aasani se move karte hain, aur hume milta hai ek atomic flux — wire mein creep karte metal ka ek dariya:

$J_{atom} = \frac{N D}{k _{B} T} F = \frac{N D}{k _{B} T} Z^{*} q ρ J$

$N$ — unit volume per atoms ki sankhya (lattice kitna packed hai).
$D$ — diffusivity: ek atom kitni aasani se kisi neighbouring empty spot par hop kar sakta hai.
$k_{B}$ — Boltzmann's constant, temperature aur energy ke beech exchange rate.
$T$ — absolute temperature (kelvin).
$\frac{D}{k _{B} T}$ — Einstein relation: yeh "kitna mobile hai" ko "per unit force kitna drift" mein convert karta hai. Hum ise isliye use karte hain kyunki yeh exactly woh bridge hai force se speed tak diffusing particles ke liye.

KYUN. Failure ek atom ke baare mein nahi hai; yeh mass move karne ke baare mein hai. $J_{atom}$ per second slide hone wale metal ki amount hai. Lekin — crucial — flux akela harmless hai.

PICTURE. Metal atoms ki ek steady river sab same direction mein drift karti hui.

Step 5 — Kyun divergence maarta hai, flux nahi

KYA HAI. Agar utna hi metal jo kisi region mein flow karta hai in woh out bhi nikalta hai, toh kuch change nahi hota — mass in = mass out. Damage ke liye ek aisi jagah chahiye jahan flux change ho: ek via, ek blocking diffusion barrier, ek grain boundary. Wahan, zyada atoms aate hain than jaate hain (pile-up → hillock → short) ya zyada jaate hain than aate hain (depletion → void → open).

"Ek jagah par flux change hone" ke liye word hai divergence, likha jaata hai $\nabla \cdot J_{atom}$ .

$\nabla \cdot J_{atom} = 0$ — river untouched flow karti hai, wire survive karti hai.
$\nabla \cdot J_{atom} \neq = 0$ — mass accumulate ya drain hota hai → damage badhta hai.

KYUN. Yeh batata hai ki wires kahan break hoti hain (junctions aur barriers), yahi wajah hai ki real design rules vias aur line ends par obsess karte hain, na ki ek lambi straight wire ke beech par.

PICTURE. Do boxes: uniform flow (safe) vs. ek blocking wall jahan atoms ek taraf pile karte hain aur doosri taraf empty hote hain.

Step 6 — "Damage ki rate" ko "time to failure" mein convert karo

KYA HAI. Wire fail hoti hai jab ek critical amount of mass $Δ m_{crit}$ move ho chuki hoti hai. Agar damage ek steady rate se build hoti hai, toh

$time to failure = \frac{Δ m _{crit}}{rate} \propto \frac{1}{J _{atom}} \propto \frac{1}{J} e^{+ E_{a} / k_{B} T}$

Yahan do cheezein hui:

Rate ko invert karo. Tez damage → choti life. Toh lifetime hai $1/ rate$ . Yeh $J$ ko numerator se denominator mein flip karta hai.
$D$ ko unpack karo. Diffusivity follow karta hai $D = D_{0} e^{- E_{a} / k_{B} T}$ , jahan $E_{a}$ hai activation energy — woh energy hill jise ek atom ko hop karne ke liye climb karna hota hai. Kyunki $D$ rate ke numerator mein tha, $e^{- E_{a} / k_{B} T}$ lifetime mein $e^{+ E_{a} / k_{B} T}$ ban jaata hai.

KYUN. Positive exponent woh sanity check hai jise parent ne flag kiya: hotter wire → easier hops → faster damage → choti life. $e^{+ E_{a} / k_{B} T}$ jaise $T$ badhta hai decrease karta hai. ✔

PICTURE. Woh energy-hill jise ek atom ko climb karna hota hai, jaise heat use upar le jaane mein help karti hai woh shrink hoti hai.

Step 7 — Black's empirical fix: exponent $n$

KYA HAI. Hamare clean derivation ne lifetime $\propto J^{- 1}$ diya. Lekin experiments ne dikhaya ki current dependence aksar steeper hoti hai — ek power $n$ :

$MTTF = \frac{A}{J ^{n}} e^{E_{a} / k_{B} T}$

$MTTF$ — Median Time To Failure: woh time jis tak identical wires ki aadhi mar chuki hoti hain.
$A$ — ek constant jo geometry, $Z^{*}$ , $ρ$ , $N$ aur material details ko swallow karta hai.
$J^{n}$ — current density current exponent $n$ tak raise ki gayi.
$n$ — $\approx 1$ jab void nucleation life limit karti hai (hamare simple theory se match karta hai); $\approx 2$ jab void growth limit karta hai (ek growing void area bhi shrink karta hai, feedback deta hai).
$e^{E_{a} / k_{B} T}$ — Step 6 se Arrhenius temperature term.

KYUN. Simple physics $n \approx 1$ deta hai; reality feedback add karta hai (ek void wire ko shrink karta hai, $J$ locally raise karta hai, khud ko accelerate karta hai) jisse $n \approx 2$ milta hai. Galat $n$ use karne se lifetime orders of magnitude se mis-extrapolate ho jaati hai — yahi wajah hai ki parent ne iske liye poora ek [!mistake] devote kiya.

PICTURE. Do lifetime-vs- $J$ curves log axes par: slope $- 1$ (nucleation) vs slope $- 2$ (growth).

Step 8 — Degenerate case: short wires kabhi fail nahi hoti (Blech)

KYA HAI. Jaise atoms anode end par pile karte hain, woh pressure (compressive stress) build karte hain jo wind ke against back push karta hai. Agar wire short hai, toh yeh back-stress EM force ko balance kar leta hai before koi bhi void grow ho. Wire immortal hai. Condition ek product hai:

$(J \cdot L) < (J \cdot L)_{crit}$

$L$ — wire ki length.
$J \cdot L$ — Blech product, immortality decide karne wala single knob.

KYUN. Ek short line = ek stiff spring; pile-up push ko jaldi cancel kar deta hai. Ek long line = ek soft spring; wind jeet jaati hai. Yeh important degenerate case hai: haari poori "steady rate → failure" story break ho jaati hai jab geometry kaafi chhoti hoti hai. Short jumpers bade $J$ carry kar sakte hain for free.

PICTURE. Short wire (spring push rok deta hai) vs long wire (spring bahut weak, void grow hoti hai).

Ek-picture summary

Ise left se right padho: crowded current $J$ → atom par force $F = Z^{*} q ρ J$ → atomic flux $J_{atom}$ → barrier par divergence ek void build karta hai → wire ek time mein mar jaati hai $MTTF = A J^{- n} e^{E_{a} / k_{B} T}$ — jab tak woh kaafi short na ho ki immortal ho jaye.

Recall Poore walkthrough ki Feynman retelling

Ek hallway imagine karo jo metal marbles (wire ke atoms) se bhari hui hai. Usme se ek tez electric wind bhejo — woh wind electrons hain, aur itne saare hain, itni tez chalte hain, ki woh marbles ko hall ke neeche dhakelte hain. Wind ki crowdedness hai $J$ ; ek narrow hall use crowd karta hai aur zyada push karta hai. Har marble ko ek shove $F$ milta hai; hum field-pull aur wind-push ko ek number $Z^{*}$ mein roll karte hain taaki hum sirf ek arrow track karein. Multiply karo ki kitne marbles hain aur woh kitni aasani se roll karte hain, aur tumhe marbles ka ek steady river milta hai ( $J_{atom}$ ). Lekin — ek evenly flowing river kuch hurt nahi karti — trouble hai ek wall (ek barrier ya junction) jahan marbles ek taraf pile karte hain (ek bump jo next hallway ko touch karta hai = short) aur doosri taraf drain hote hain (ek hole = snap). Heat marbles ko floor ke chhote bumps zyada easily jump karne deti hai, toh woh faster move karte hain aur hall jaldi break hota hai — yahi hai $e^{+ E_{a} / k_{B} T}$ . Experiments kehte hain ki crowding ek-power se bhi zyada hurt karta hai, toh hum $J^{n}$ likhte hain $n$ ke saath jo 2 ke near hota hai jab ek growing hole khud par feed karta hai. Aur escape hatch: agar hallway sach mein short hai, toh marbles far end par jam ho jaate hain aur itni tez push back karte hain ki poora flow ruk jaata hai before koi hole form ho — woh wire hamesha ke liye jeeti hai.

Connections

Arrhenius reliability model — Step 6 mein paida hua $e^{E_{a} / k_{B} T}$ factor.
MTTF and FIT rates — statistically "median time to failure" ka matlab.
Joule heating & self-heating — woh $T$ raise karta hai jo Step 6 mein appear hota hai.
Design rules & current density limits — $J$ cap jo Steps 1 aur 7 justify karte hain.
Copper damascene process — Step 5 ke $A$ , $E_{a}$ , aur barrier walls set karta hai.
Stress migration — Step 8 ke Blech balance ke peeche back-stress cousin.
Interconnect RC delay — wahi shrinking wires, ek competing constraint.