• Vacancy Diffusion: Host solvent atoms or substitutional solute atoms migrate from their normal lattice sites into adjacent unoccupied vacant sites. The process requires two coordinated events: a vacancy must first be present on an adjacent neighbor site, and the migrating atom must acquire sufficient thermal vibrational energy to squeeze through surrounding atoms into that vacancy. Operates for self-diffusion and substitutional solutes (e.g., Cu, Ni, Zn).
• Interstitial Diffusion: Small solute atoms (e.g., H, C, N, O) migrate from one interstitial void (interstice) to an adjacent unoccupied interstitial void. Because host atoms occupy lattice sites and solute concentrations are small, the vast majority of adjacent interstitial voids are empty and immediately available.

The total activation energy for vacancy-mediated diffusion consists of the sum of two distinct thermodynamic energies:

$$Q_{\text{substitutional}} = \Delta H_f + \Delta H_m$$

where $\Delta H_f$ is the enthalpy to create an adjacent vacancy, and $\Delta H_m$ is the enthalpy of atomic migration.

For interstitial diffusion, neighboring interstitial sites are naturally empty. No prior defect creation step is required ($\Delta H_f = 0$). Furthermore, interstitial atoms are much smaller than host atoms, so squeezing through interstitial bottlenecks requires significantly less elastic lattice distortion:

$$Q_{\text{interstitial}} = \Delta H_m \ll (\Delta H_f + \Delta H_m)$$

Consequently, interstitial diffusion proceeds with less than half the activation energy of substitutional diffusion.

$$\frac{D_{\text{C}}}{D_{\text{Fe}}} = \frac{1.5 \times 10^{-11} \text{ m}^2/\text{s}}{2.0 \times 10^{-16} \text{ m}^2/\text{s}} = \mathbf{75,000 \approx 7.5 \times 10^4}$$

Carbon diffuses roughly 75,000 times faster through the iron lattice than iron atoms themselves! This enormous kinetic disparity enables rapid industrial case-hardening of steel (carburizing gears in hours at $900^\circ\text{C}$) while keeping the underlying iron grain structure intact.

• Self-Diffusion: Atomic migration occurring in pure elemental crystals or homogeneous alloys in the absence of a chemical composition gradient. Atoms wander randomly by Brownian thermal motion; no net macroscopic mass transport occurs.
• Interdiffusion (Chemical / Impurity Diffusion): Atomic migration occurring in a concentration gradient (e.g., across a Cu-Ni diffusion couple). Driven by chemical potential gradients, a net directional flux of solute and solvent atoms occurs to homogenize the system.

To measure self-diffusion in pure nickel, a sub-micron film of radioactive isotope (e.g., $^{63}\text{Ni}$) is electroplated onto a polished pure nickel disc. The sample is annealed at temperature $T$ for time $t$. After cooling, thin parallel surface layers (a few micrometers thick) are sequentially shaved off using a precision lathe or microtome, and the radioactive activity of each layer is measured with a scintillation counter. Fitting the specific radioactivity profile $A(x)$ to the thin-film Gaussian solution:

$$A(x, t) = \frac{A_0}{\sqrt{\pi D t}} \exp\left(-\frac{x^2}{4Dt}\right)$$

plotting $\ln A$ versus $x^2$ gives a straight line whose slope $-\frac{1}{4Dt}$ yields the true self-diffusion coefficient $D^*$.

The true fundamental thermodynamic driving force for all diffusion is the gradient of chemical potential $\nabla\mu_i$, not the concentration gradient $\nabla C_i$:

$$J_i = -M_i C_i \frac{\partial\mu_i}{\partial x}$$

For ideal or dilute solutions, $\mu_i = \mu_i^0 + RT \ln C_i$, giving $\frac{\partial\mu_i}{\partial x} = \frac{RT}{C_i}\frac{\partial C_i}{\partial x}$, which reduces to Fick's first law $J = -D \frac{\partial C}{\partial x}$ ("downhill" diffusion from high to low $C$).

However, inside a thermodynamic miscibility gap (where $\frac{\partial^2 G}{\partial C^2} < 0$), the chemical potential gradient has the opposite sign of the concentration gradient. Atoms spontaneously diffuse from regions of low concentration toward regions of high concentration (uphill diffusion) during spinodal decomposition, phase-separating into A-rich and B-rich modulated nanostructures without nucleation.

Ernest Kirkendall and Alice Smigelskas wrapped fine, chemically inert, insoluble molybdenum wires around a block of 70-30 brass ($70\text{ wt}\%\text{ Cu}, 30\text{ wt}\%\text{ Zn}$). They electroplated a thick layer of pure copper over the assembly, embedding the molybdenum wire markers directly at the original weld interface between brass and pure copper. The diffusion couple was then annealed at $785^\circ\text{C}$ for several weeks.

Upon cross-sectioning and microscopic inspection, Kirkendall discovered that the inert molybdenum markers had moved inward toward the brass core. The distance between the marker wires on opposite sides had measurably decreased.

1. Unequal Intrinsic Diffusion Rates ($D_{\text{Zn}} > D_{\text{Cu}}$): Zinc atoms (lower melting point, weaker bonds) diffuse out of the brass into the pure copper faster than copper atoms diffuse into the brass ($J_{\text{Zn}} > J_{\text{Cu}}$).
2. Refutation of Direct Exchange: If diffusion occurred by direct pairwise exchange of adjacent atoms or ring rotation, every atom of Zn moving right would require exactly one atom of Cu moving left ($J_{\text{Zn}} = J_{\text{Cu}}$). Net mass flux across the interface would be zero, and markers could never move.
3. Confirmation of Vacancy Mechanism: Because each atomic jump requires exchanging with a vacancy, the net outward atomic flux ($J_{\text{Zn}} - J_{\text{Cu}}$) must be balanced by an equal and opposite net flux of vacancies diffusing into the brass: $$J_v = -(J_{\text{Zn}} - J_{\text{Cu}})$$ As excess vacancies flow into the brass, lattice planes on the brass side are destroyed by vacancy annihilation (climb of edge dislocations). On the copper side, extra lattice planes are created by vacancy emission. The entire atomic lattice shifts rigidly relative to the inert markers at velocity $v = (D_{\text{Zn}} - D_{\text{Cu}})\frac{dX_{\text{Zn}}}{dx}$.

When the influx of vacancies into the faster-diffusing side (brass) exceeds the rate at which edge dislocations can climb and absorb them, the vacancy population becomes supersaturated. The excess vacancies condense and coalesce into macroscopic microscopic voids, termed Kirkendall voids (porosity). This causes embrittlement and delamination in microelectronic solder joints (e.g., Cu-Sn and Ni-Au interfaces in semiconductor packaging).

• Marker Velocity Equation: $$v = (D_A - D_B) \frac{\partial X_A}{\partial x} = -(D_A - D_B) \frac{\partial X_B}{\partial x}$$ where $v$ is the velocity of the inert marker frame relative to the fixed laboratory frame.
• Chemical Interdiffusion Coefficient: $$\tilde{D} = X_B D_A + X_A D_B$$ For non-ideal solutions, Darken introduced the thermodynamic factor $\Phi$: $$\tilde{D} = (X_B D_A + X_A D_B) \left(1 + \frac{\partial \ln \gamma_A}{\partial \ln X_A}\right)$$

Given $X_{\text{Cu}} = 0.60, X_{\text{Ni}} = 0.40$, $D_{\text{Cu}} = 4.0 \times 10^{-14} \text{ m}^2/\text{s}$, $D_{\text{Ni}} = 1.0 \times 10^{-14} \text{ m}^2/\text{s}$, and $\frac{\partial X_{\text{Ni}}}{\partial x} = 2.5 \times 10^3 \text{ m}^{-1}$ (meaning $\frac{\partial X_{\text{Cu}}}{\partial x} = -2.5 \times 10^3 \text{ m}^{-1}$):

• Interdiffusion Coefficient $\tilde{D}$: $$\tilde{D} = X_{\text{Ni}} D_{\text{Cu}} + X_{\text{Cu}} D_{\text{Ni}} = (0.40 \times 4.0 \times 10^{-14}) + (0.60 \times 1.0 \times 10^{-14})$$ $$\tilde{D} = 1.60 \times 10^{-14} + 0.60 \times 10^{-14} = \mathbf{2.20 \times 10^{-14} \text{ m}^2/\text{s}}$$
• Marker Velocity $v$: $$v = (D_{\text{Cu}} - D_{\text{Ni}}) \left(-\frac{\partial X_{\text{Cu}}}{\partial x}\right) = (D_{\text{Cu}} - D_{\text{Ni}}) \frac{\partial X_{\text{Ni}}}{\partial x}$$ $$v = (4.0 \times 10^{-14} - 1.0 \times 10^{-14} \text{ m}^2/\text{s}) \times (2.5 \times 10^3 \text{ m}^{-1})$$ $$v = (3.0 \times 10^{-14}) \times (2.5 \times 10^3) = \mathbf{7.50 \times 10^{-11} \text{ m/s}}$$

  Convert to $\text{nm/s}$ ($1 \text{ m} = 10^9 \text{ nm}$):

  $$v = (7.50 \times 10^{-11} \text{ m/s}) \times 10^9 \text{ nm/m} = \mathbf{0.075 \text{ nm/s}}$$

  Convert to $\mu\text{m/day}$ ($1 \text{ day} = 86,400 \text{ s}$, $1 \text{ m} = 10^6 \mu\text{m}$):

  $$v = (7.50 \times 10^{-11} \text{ m/s}) \times (86,400 \text{ s/day}) \times 10^6 \mu\text{m/m} = \mathbf{6.48 \mu\text{m/day}}$$

$$J = -D \frac{dC}{dx}$$

• $J$: Diffusion Flux. The mass (or number of atoms) traversing perpendicular to a unit cross-sectional area per unit time. SI units: $\mathbf{\text{kg}/(\text{m}^2\cdot\text{s})}$ or $\mathbf{\text{atoms}/(\text{m}^2\cdot\text{s})}$.
• $D$: Diffusion Coefficient (Diffusivity). The kinetic rate constant for atomic mobility in the host lattice. SI units: $\mathbf{\text{m}^2/\text{s}}$.
• $C$: Concentration. The mass (or number of atoms) of solute per unit volume of host. SI units: $\mathbf{\text{kg/m}^3}$ or $\mathbf{\text{atoms/m}^3}$.
• $\frac{dC}{dx}$: Concentration Gradient. The spatial rate of change of concentration along direction $x$. SI units: $\frac{\text{kg/m}^3}{\text{m}} = \mathbf{\text{kg/m}^4}$ or $\mathbf{\text{atoms/m}^4}$.

Diffusion is a passive, spontaneous transport process driven down a concentration gradient. If concentration decreases with increasing distance ($x$), the spatial slope $\frac{dC}{dx}$ is negative. The negative sign ensures that the resultant net flux $J$ is a positive quantity pointing in the direction of lower concentration:

$$\frac{dC}{dx} < 0 \implies J = -D \left(-\left|\frac{dC}{dx}\right|\right) > 0 \quad (\text{Net mass flows from High to Low } C)$$

Let $M_{\text{solute}}$ be the molar mass ($\text{kg/mol}$) and $N_A = 6.022 \times 10^{23} \text{ atoms/mol}$ be Avogadro's number:

$$J_{\text{at}} = \frac{J_m \cdot N_A}{M_{\text{solute}}} \quad \left[\frac{\text{atoms}}{\text{m}^2\cdot\text{s}}\right]$$ $$J_m = \frac{J_{\text{at}} \cdot M_{\text{solute}}}{N_A} \quad \left[\frac{\text{kg}}{\text{m}^2\cdot\text{s}}\right]$$

Under steady-state conditions in a flat planar membrane, the concentration profile is linear:

$$\frac{dC}{dx} = \frac{C_B - C_A}{L - 0} = \frac{0.40 \text{ kg/m}^3 - 2.40 \text{ kg/m}^3}{0.20 \times 10^{-3} \text{ m}} = \frac{-2.00 \text{ kg/m}^3}{2.0 \times 10^{-4} \text{ m}} = \mathbf{-10,000 \text{ kg/m}^4}$$

Using Fick's First Law ($J = -D \frac{dC}{dx}$):

$$J = -(1.6 \times 10^{-9} \text{ m}^2/\text{s}) \times (-10,000 \text{ kg/m}^4) = \mathbf{1.60 \times 10^{-5} \text{ kg}/(\text{m}^2\cdot\text{s})}$$

Calculate atomic flux using $M_{\text{H}} = 1.008 \times 10^{-3} \text{ kg/mol}$:

$$J_{\text{at}} = \frac{J \cdot N_A}{M_{\text{H}}} = \frac{(1.60 \times 10^{-5} \text{ kg/m}^2\cdot\text{s}) \times (6.022 \times 10^{23} \text{ atoms/mol})}{1.008 \times 10^{-3} \text{ kg/mol}} = \mathbf{9.56 \times 10^{21} \text{ atoms}/(\text{m}^2\cdot\text{s})}$$

The mass flow rate $\dot{m}$ through area $A$ is:

$$\dot{m} = J \times A = (1.60 \times 10^{-5} \text{ kg/m}^2\cdot\text{s}) \times (0.15 \text{ m}^2) = 2.40 \times 10^{-6} \text{ kg/s}$$

In one hour ($t = 3600 \text{ s}$):

$$m_{\text{hr}} = \dot{m} \times 3600 \text{ s} = (2.40 \times 10^{-6} \text{ kg/s}) \times 3600 \text{ s} = \mathbf{8.64 \times 10^{-3} \text{ kg/hr} = 8.64 \text{ g/hr}}$$

Since flux is inversely proportional to thickness ($J \propto 1/L$):

$$\frac{L_1}{L_2} = \frac{0.20 \text{ mm}}{0.05 \text{ mm}} = 4.0$$ $$m_{\text{hr, new}} = 4.0 \times 8.64 \text{ g/hr} = \mathbf{34.56 \text{ g/hr}}$$

Reducing membrane thickness by a factor of 4 quadruples the purification throughput without increasing temperature.

The linear concentration gradient is:

$$\frac{dC}{dx} = \frac{C_2 - C_1}{L} = \frac{0.10 - 1.20 \text{ kg/m}^3}{5.0 \times 10^{-3} \text{ m}} = \frac{-1.10}{5.0 \times 10^{-3}} = -220 \text{ kg/m}^4$$

Using Fick's First Law:

$$J = -D \frac{dC}{dx} = -(4.2 \times 10^{-11} \text{ m}^2/\text{s}) \times (-220 \text{ kg/m}^4) = \mathbf{9.24 \times 10^{-9} \text{ kg}/(\text{m}^2\cdot\text{s})}$$

For steady-state 1D diffusion, concentration varies linearly with position:

$$C(x) = C_1 + \left(\frac{dC}{dx}\right) x = 1.20 \text{ kg/m}^3 + (-220 \text{ kg/m}^4) \times (2.0 \times 10^{-3} \text{ m})$$ $$C(2.0 \text{ mm}) = 1.20 - 0.44 = \mathbf{0.76 \text{ kg/m}^3}$$

$$m_{\text{day}} = J \times A \times t = (9.24 \times 10^{-9} \text{ kg/m}^2\cdot\text{s}) \times (0.80 \text{ m}^2) \times (86,400 \text{ s})$$ $$m_{\text{day}} = (7.392 \times 10^{-9} \text{ kg/s}) \times 86,400 \text{ s} = 6.387 \times 10^{-4} \text{ kg} = \mathbf{0.639 \text{ g/day}}$$

Using $C [\text{kg/m}^3] = \left(\frac{\text{wt}\%}{100}\right) \times \rho_{\text{steel}}$:

$$C_1 = \left(\frac{1.15}{100}\right) \times 7800 \text{ kg/m}^3 = \mathbf{89.70 \text{ kg/m}^3}$$ $$C_2 = \left(\frac{0.25}{100}\right) \times 7800 \text{ kg/m}^3 = \mathbf{19.50 \text{ kg/m}^3}$$

Thickness $L = 1.5 \text{ mm} = 1.5 \times 10^{-3} \text{ m}$:

$$\frac{dC}{dx} = \frac{C_2 - C_1}{L} = \frac{19.50 - 89.70 \text{ kg/m}^3}{1.5 \times 10^{-3} \text{ m}} = \frac{-70.20}{1.5 \times 10^{-3}} = -46,800 \text{ kg/m}^4$$ $$J = -D \frac{dC}{dx} = -(3.0 \times 10^{-11} \text{ m}^2/\text{s}) \times (-46,800 \text{ kg/m}^4) = \mathbf{1.404 \times 10^{-6} \text{ kg}/(\text{m}^2\cdot\text{s})}$$

Mass flow rate through $1.0 \text{ cm}^2$:

$$\dot{m} = J \times A = (1.404 \times 10^{-6} \text{ kg/m}^2\cdot\text{s}) \times (1.0 \times 10^{-4} \text{ m}^2) = 1.404 \times 10^{-10} \text{ kg/s}$$

Convert to carbon atoms per second ($M_{\text{C}} = 0.012011 \text{ kg/mol}$):

$$\dot{N} = \frac{\dot{m} \cdot N_A}{M_{\text{C}}} = \frac{(1.404 \times 10^{-10} \text{ kg/s}) \times (6.022 \times 10^{23} \text{ atoms/mol})}{0.012011 \text{ kg/mol}} = \mathbf{7.04 \times 10^{15} \text{ atoms/s}}$$

Consider a pipe of length $L$. At radius $r$, the cylindrical surface area through which diffusion occurs is $A(r) = 2\pi r L$.

By Fick's first law, radial diffusion flux is $J(r) = -D \frac{dC}{dr}$.

The total mass flow rate $Q$ across the cylindrical shell is:

$$Q = J(r) \cdot A(r) = -D \left(\frac{dC}{dr}\right) (2\pi r L) = -2\pi D L r \frac{dC}{dr}$$

In steady state with no internal sources or sinks, conservation of mass requires $Q = \text{constant}$ (independent of $r$):

$$r \frac{dC}{dr} = -\frac{Q}{2\pi D L} = \text{constant } K$$

Separating variables:

$$dC = K \frac{dr}{r} \implies C(r) = K \ln r + C_0$$

Apply boundary conditions:

• At $r = r_1$: $C(r_1) = C_1 = K \ln r_1 + C_0$
• At $r = r_2$: $C(r_2) = C_2 = K \ln r_2 + C_0$

Subtracting the two equations:

$$C_2 - C_1 = K (\ln r_2 - \ln r_1) = K \ln\left(\frac{r_2}{r_1}\right) \implies K = \frac{C_2 - C_1}{\ln(r_2 / r_1)}$$

Substituting $K$ and solving for $C(r)$:

$$C(r) - C_1 = K (\ln r - \ln r_1) = K \ln\left(\frac{r}{r_1}\right) = (C_2 - C_1) \frac{\ln(r / r_1)}{\ln(r_2 / r_1)}$$ $$C(r) = C_1 + (C_2 - C_1) \frac{\ln(r / r_1)}{\ln(r_2 / r_1)}$$

From Part (a), $Q = -2\pi D L K$. Substituting $K = \frac{C_2 - C_1}{\ln(r_2 / r_1)}$:

$$Q = -2\pi D L \left[\frac{C_2 - C_1}{\ln(r_2 / r_1)}\right] = \frac{2\pi D L (C_1 - C_2)}{\ln(r_2 / r_1)}$$

Dividing by length $L$:

$$\frac{Q}{L} = \frac{2\pi D (C_1 - C_2)}{\ln(r_2 / r_1)}$$

$$\ln\left(\frac{r_2}{r_1}\right) = \ln\left(\frac{4.6}{4.0}\right) = \ln(1.15) = 0.13976$$ $$\ln\left(\frac{r_{\text{mid}}}{r_1}\right) = \ln\left(\frac{4.3}{4.0}\right) = \ln(1.075) = 0.07232$$ $$C(r_{\text{mid}}) = C_1 + (C_2 - C_1)\frac{\ln(r_{\text{mid}}/r_1)}{\ln(r_2/r_1)} = 0.50 + (0.05 - 0.50) \times \left(\frac{0.07232}{0.13976}\right)$$ $$C(r_{\text{mid}}) = 0.50 - 0.45 \times (0.51746) = 0.50 - 0.2329 = \mathbf{0.2671 \text{ kg/m}^3}$$

The simple arithmetic mean is $\frac{0.50 + 0.05}{2} = 0.2750 \text{ kg/m}^3$. The logarithmic profile is slightly lower ($0.2671$) because flux is concentrated over smaller areas at smaller radii.

$$\frac{Q}{L} = \frac{2\pi D (C_1 - C_2)}{\ln(r_2 / r_1)} = \frac{2\pi \times (5.0 \times 10^{-10} \text{ m}^2/\text{s}) \times (0.50 - 0.05 \text{ kg/m}^3)}{0.13976}$$ $$\frac{Q}{L} = \frac{3.1416 \times 10^{-9} \times 0.45}{0.13976} = \frac{1.4137 \times 10^{-9}}{0.13976} = 1.0115 \times 10^{-8} \text{ kg}/(\text{m}\cdot\text{s})$$

Convert to $\text{mg}/(\text{m}\cdot\text{day})$ ($1 \text{ day} = 86,400 \text{ s}$, $1 \text{ kg} = 10^6 \text{ mg}$):

$$\frac{Q}{L} = (1.0115 \times 10^{-8} \text{ kg/s}) \times (86,400 \text{ s/day}) \times 10^6 \text{ mg/kg} = \mathbf{873.9 \text{ mg}/(\text{m}\cdot\text{day})}$$

Using $J(r) = \frac{Q/L}{2\pi r}$:

• At $r_1 = 4.0 \times 10^{-3} \text{ m}$: $$J(r_1) = \frac{1.0115 \times 10^{-8}}{2\pi \times (4.0 \times 10^{-3})} = \frac{1.0115 \times 10^{-8}}{0.02513} = \mathbf{4.025 \times 10^{-7} \text{ kg}/(\text{m}^2\cdot\text{s})}$$
• At $r_2 = 4.6 \times 10^{-3} \text{ m}$: $$J(r_2) = \frac{1.0115 \times 10^{-8}}{2\pi \times (4.6 \times 10^{-3})} = \frac{1.0115 \times 10^{-8}}{0.02890} = \mathbf{3.500 \times 10^{-7} \text{ kg}/(\text{m}^2\cdot\text{s})}$$

Physical Explanation: To maintain constant total mass flow rate $Q$, flux times area must be constant ($J_1 A_1 = J_2 A_2$). Because the outer area is $15\%$ larger than the inner area ($A_2 / A_1 = 4.6 / 4.0 = 1.15$), the outer flux is reduced by exactly $1/1.15 = 0.87\times$.

Consider an elemental volume $\Delta V = A \Delta x$ bounded by planes at $x$ and $x + \Delta x$ over time interval $\Delta t$:

• Mass of diffusing solute entering through face at $x$: $m_{\text{in}} = J(x) \cdot A \cdot \Delta t$
• Mass of diffusing solute leaving through face at $x + \Delta x$: $m_{\text{out}} = J(x + \Delta x) \cdot A \cdot \Delta t$
• Net accumulation of solute inside the slice: $\Delta m = m_{\text{in}} - m_{\text{out}} = [J(x) - J(x + \Delta x)] A \Delta t$

The change in average concentration inside the volume is $\Delta C = \frac{\Delta m}{\Delta V} = \frac{\Delta m}{A \Delta x}$:

$$\Delta C = \frac{[J(x) - J(x + \Delta x)] A \Delta t}{A \Delta x} = -\frac{J(x + \Delta x) - J(x)}{\Delta x} \Delta t$$

Dividing by $\Delta t$ and taking the limit as $\Delta x \to 0$ and $\Delta t \to 0$:

$$\lim_{\Delta t \to 0} \frac{\Delta C}{\Delta t} = -\lim_{\Delta x \to 0} \frac{J(x + \Delta x) - J(x)}{\Delta x} \implies \frac{\partial C}{\partial t} = -\frac{\partial J}{\partial x}$$

From Fick's First Law, $J = -D \frac{\partial C}{\partial x}$. Substituting this flux expression into the continuity equation:

$$\frac{\partial C}{\partial t} = -\frac{\partial}{\partial x}\left(-D \frac{\partial C}{\partial x}\right) = \frac{\partial}{\partial x}\left(D \frac{\partial C}{\partial x}\right)$$

If the diffusion coefficient $D$ is constant (independent of $x$ and $C$):

$$\frac{\partial C}{\partial t} = D \frac{\partial}{\partial x}\left(\frac{\partial C}{\partial x}\right) = D \frac{\partial^2 C}{\partial x^2}$$

This is Fick's Second Law, mathematically identical to Fourier's heat conduction equation.

Diffusivity depends strongly on concentration when:

1. High Solute Concentrations: In concentrated solid solutions (e.g., carbon $> 0.5 \text{ wt}\%$ in austenite, or Cu-Ni couples), dissolved solute alters the host lattice parameter and local bond stiffness, accelerating jump frequencies. In austenite, $D_{\text{C}}$ increases by a factor of 4 as carbon concentration rises from $0.1\%$ to $1.4\%$.
2. Non-Ideal Thermodynamic Solutions: When activity coefficients vary strongly with concentration, Darken's thermodynamic factor $\left(1 + \frac{\partial \ln \gamma}{\partial \ln X}\right)$ causes $D$ to be highly concentration-dependent. (In such cases, the non-linear equation must be solved using the Boltzmann-Matano graphical method).

1. Initial Condition: Before diffusion begins ($t = 0$), the solute is uniformly distributed throughout the solid at bulk concentration $C_0$: $$C(x, 0) = C_0 \quad \text{for } 0 \le x < \infty$$
2. Surface Boundary Condition: For all times $t > 0$, the surface at $x = 0$ is held at an instantaneous, constant composition $C_s$ (maintained by an active gas or liquid atmosphere): $$C(0, t) = C_s \quad \text{for } t > 0$$
3. Far-Field Boundary Condition: Deep in the interior ($x \to \infty$), the solid is unaffected by surface diffusion, maintaining the original composition: $$C(\infty, t) = C_0 \quad \text{for all } t$$

$$\frac{C(x,t) - C_0}{C_s - C_0} = 1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right) = \text{erfc}\left(\frac{x}{2\sqrt{Dt}}\right)$$

• $C(x,t)$: Solute concentration at depth $x$ after diffusion time $t$.
• $C_0$: Initial uniform bulk concentration of the solid before treatment.
• $C_s$: Constant surface concentration maintained at the interface.
• $x$: Perpendicular depth into the solid ($\text{m}$).
• $D$: Diffusion coefficient ($\text{m}^2/\text{s}$).
• $t$: Diffusion time ($\text{s}$).

By defining the Boltzmann similarity variable $\eta = \frac{x}{2\sqrt{Dt}}$, the partial differential equation $\frac{\partial C}{\partial t} = D \frac{\partial^2 C}{\partial x^2}$ is converted into an ordinary differential equation $\frac{d^2 C}{d\eta^2} + 2\eta \frac{dC}{d\eta} = 0$. Physically, $\eta$ scales the spatial coordinate $x$ by the characteristic diffusion penetration distance $2\sqrt{Dt}$, collapsing all time-dependent concentration profiles onto a single universal curve.

A finite component of half-thickness $L$ (or sheet of thickness $2L$ exposed on both sides) behaves as a semi-infinite solid as long as diffusing solute has not penetrated to the center of the part ($x = L$). Because the error function approaches zero ($1 - \text{erf}(z) < 0.005$) for $z \ge 2.0$, the diffusion field penetrates to $x_{\max} \approx 4\sqrt{Dt}$. Therefore, the semi-infinite model is strictly valid as long as:

$$L > 4\sqrt{Dt} \implies t < \frac{L^2}{16 D}$$

$$\text{erf}(z) = \frac{2}{\sqrt{\pi}} \int_0^z e^{-\eta^2} d\eta$$

• $\text{erf}(0) = \frac{2}{\sqrt{\pi}} \int_0^0 e^{-\eta^2} d\eta = \mathbf{0}$
• $\text{erf}(\infty) = \frac{2}{\sqrt{\pi}} \left(\frac{\sqrt{\pi}}{2}\right) = \mathbf{1.0}$
• $\text{erf}(-z) = -\text{erf}(z)$ (odd function).
• For small arguments ($z \ll 1$): $\text{erf}(z) \approx \frac{2}{\sqrt{\pi}} z \approx 1.1284 z$.
• Complementary Error Function: $\text{erfc}(z) = 1 - \text{erf}(z) = \frac{2}{\sqrt{\pi}} \int_z^\infty e^{-\eta^2} d\eta$.

• Finding $\text{erf}(0.425)$ between $z_1 = 0.40$ and $z_2 = 0.45$: $$\text{erf}(0.40) = 0.4284, \quad \text{erf}(0.45) = 0.4755$$ $$\frac{\text{erf}(0.425) - 0.4284}{0.425 - 0.40} = \frac{0.4755 - 0.4284}{0.45 - 0.40}$$ $$\text{erf}(0.425) = 0.4284 + \left(\frac{0.025}{0.050}\right) \times (0.4755 - 0.4284)$$ $$\text{erf}(0.425) = 0.4284 + 0.5 \times (0.0471) = 0.4284 + 0.02355 = \mathbf{0.45195 \approx 0.4520}$$
• Finding $z$ for $\text{erf}(z) = 0.5000$: Looking at the table: $\text{erf}(0.45) = 0.4755$ and $\text{erf}(0.50) = 0.5205$. $$\frac{z - 0.45}{0.5000 - 0.4755} = \frac{0.50 - 0.45}{0.5205 - 0.4755}$$ $$z - 0.45 = \frac{0.0245}{0.0450} \times (0.05) = 0.5444 \times 0.05 = 0.02722$$ $$z = 0.45 + 0.02722 = \mathbf{0.4772}$$

From the semi-infinite solution, the argument of the error function is $z = \frac{x}{2\sqrt{Dt}}$. Significant composition change occurs when $z \approx 1$, which defines the characteristic penetration distance $\bar{x} \approx 2\sqrt{Dt}$. Because distance scales with $\sqrt{t}$, the penetration velocity is:

$$v = \frac{d\bar{x}}{dt} = \frac{d(2\sqrt{Dt})}{dt} = \sqrt{\frac{D}{t}} = \frac{2D}{\bar{x}}$$

As the case depth increases, the penetration speed decays inversely with depth ($v \propto 1/\bar{x}$). To double the depth, the diffusion time must be quadrupled ($4\times$).

For fixed initial concentration $C_0$ and surface concentration $C_s$, achieving a specific target concentration $C_x$ means the left side of the error function solution is a constant:

$$\frac{C_x - C_0}{C_s - C_0} = \text{const} \implies 1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right) = \text{const} \implies \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right) = \text{const}$$

Because the error function is strictly monotonic, its argument must be constant:

$$\frac{x}{2\sqrt{Dt}} = \text{const} \implies \frac{x^2}{Dt} = \text{const}$$

At constant temperature, $D$ is constant, giving the scaling rule:

$$\frac{x_1^2}{t_1} = \frac{x_2^2}{t_2} \iff \frac{x_1}{\sqrt{t_1}} = \frac{x_2}{\sqrt{t_2}}$$

• Time to achieve $x_2 = 1.20 \text{ mm}$ (doubling case depth): $$\frac{t_2}{t_1} = \left(\frac{x_2}{x_1}\right)^2 = \left(\frac{1.20 \text{ mm}}{0.60 \text{ mm}}\right)^2 = (2.0)^2 = 4.0$$ $$t_2 = 4.0 \times t_1 = 4.0 \times 2.0 \text{ hr} = \mathbf{8.0 \text{ hours}}$$

  Doubling the case depth requires 4 times the furnace processing time!

• Case depth after $t_3 = 8.0 \text{ hours}$: $$x_3 = x_1 \sqrt{\frac{t_3}{t_1}} = 0.60 \text{ mm} \times \sqrt{\frac{8.0}{2.0}} = 0.60 \times \sqrt{4.0} = 0.60 \times 2 = \mathbf{1.20 \text{ mm}}$$

Given parameters: $C_0 = 0.20 \text{ wt}\%$, $C_s = 1.20 \text{ wt}\%$, $C(x,t) = 0.60 \text{ wt}\%$, $x = 0.75 \text{ mm} = 7.5 \times 10^{-4} \text{ m}$, $D = 1.28 \times 10^{-11} \text{ m}^2/\text{s}$.

Substitute into the error function equation:

$$\frac{C(x,t) - C_0}{C_s - C_0} = 1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)$$ $$\frac{0.60 - 0.20}{1.20 - 0.20} = \frac{0.40}{1.00} = 0.4000 = 1 - \text{erf}(z) \implies \text{erf}(z) = 1.0000 - 0.4000 = 0.6000$$

From the error function table in Problem 13:

$$\text{erf}(0.55) = 0.5633, \quad \text{erf}(0.60) = 0.6039$$

Interpolate for $\text{erf}(z) = 0.6000$:

$$\frac{z - 0.55}{0.6000 - 0.5633} = \frac{0.60 - 0.55}{0.6039 - 0.5633} \implies \frac{z - 0.55}{0.0367} = \frac{0.05}{0.0406} = 1.2315$$ $$z - 0.55 = 1.2315 \times 0.0367 = 0.0452 \implies z = \mathbf{0.5952}$$

Now equate $z = \frac{x}{2\sqrt{Dt}}$:

$$2\sqrt{Dt} = \frac{x}{z} = \frac{7.5 \times 10^{-4} \text{ m}}{0.5952} = 1.2601 \times 10^{-3} \text{ m}$$ $$\sqrt{Dt} = 6.3004 \times 10^{-4} \text{ m} \implies Dt = (6.3004 \times 10^{-4})^2 = 3.9695 \times 10^{-7} \text{ m}^2$$ $$t = \frac{3.9695 \times 10^{-7} \text{ m}^2}{1.28 \times 10^{-11} \text{ m}^2/\text{s}} = 31,012 \text{ seconds}$$

Convert to hours:

$$t = \frac{31,012 \text{ s}}{3600 \text{ s/hr}} = \mathbf{8.614 \text{ hours} = 8 \text{ hours and 37 minutes}}$$

$$z_2 = \frac{x_2}{2\sqrt{Dt}} = \frac{3.5 \times 10^{-4} \text{ m}}{1.2601 \times 10^{-3} \text{ m}} = 0.2777$$

Interpolate error function for $z_2 = 0.2777$ ($\text{erf}(0.25) = 0.2763, \text{erf}(0.30) = 0.3286$):

$$\text{erf}(0.2777) = 0.2763 + \left(\frac{0.2777 - 0.25}{0.05}\right) \times (0.3286 - 0.2763) = 0.2763 + 0.554 \times (0.0523) = 0.3053$$ $$\frac{C(0.35 \text{ mm}) - 0.20}{1.20 - 0.20} = 1 - 0.3053 = 0.6947$$ $$C(0.35 \text{ mm}) = 0.20 + 0.6947(1.00) = \mathbf{0.895 \text{ wt}\% \approx 0.90 \text{ wt}\%}$$

As proved in Chapter 4, the maximum interstitial solubility of carbon in BCC ferrite ($\alpha$-Fe) is only $0.022 \text{ wt}\%$. It is impossible to achieve a high-carbon hard surface ($0.8\text{--}1.2\text{ wt}\%$) in ferrite because iron carbide (cementite) would immediately precipitate as an impenetrable crust on the outer surface, blocking further carbon absorption. In FCC austenite ($\gamma$-Fe), the maximum solubility is $2.14 \text{ wt}\%$, allowing large quantities of carbon to dissolve in solid solution and subsequently quench into ultra-hard martensite.

Evaluate concentration ratio:

$$\frac{C_x - C_0}{C_s - C_0} = \frac{0.45 - 0.20}{1.00 - 0.20} = \frac{0.25}{0.80} = 0.3125$$ $$\frac{C_x - C_0}{C_s - C_0} = 1 - \text{erf}(z) \implies \text{erf}(z) = 1 - 0.3125 = 0.6875$$

From the error function table: $\text{erf}(0.70) = 0.6778$ and $\text{erf}(0.75) = 0.7112$.

Interpolate for $z$:

$$\frac{z - 0.70}{0.6875 - 0.6778} = \frac{0.75 - 0.70}{0.7112 - 0.6778} \implies \frac{z - 0.70}{0.0097} = \frac{0.05}{0.0334} = 1.497$$ $$z - 0.70 = 1.497 \times 0.0097 = 0.0145 \implies z = \mathbf{0.7145}$$

Compute case depth $x_1$ ($t_1 = 4.0 \times 3600 = 14,400 \text{ s}$):

$$x_1 = 2 z \sqrt{D t_1} = 2 \times 0.7145 \times \sqrt{(1.20 \times 10^{-11} \text{ m}^2/\text{s}) \times (14,400 \text{ s})}$$ $$D t_1 = 1.728 \times 10^{-7} \text{ m}^2 \implies \sqrt{D t_1} = 4.157 \times 10^{-4} \text{ m}$$ $$x_1 = 2 \times 0.7145 \times (4.157 \times 10^{-4} \text{ m}) = 5.940 \times 10^{-4} \text{ m} = \mathbf{0.594 \text{ mm}}$$

$$D t_2 = (1.20 \times 10^{-11}) \times (32,400) = 3.888 \times 10^{-7} \text{ m}^2 \implies \sqrt{D t_2} = 6.235 \times 10^{-4} \text{ m}$$ $$x_2 = 2 z \sqrt{D t_2} = 2 \times 0.7145 \times (6.235 \times 10^{-4} \text{ m}) = 8.910 \times 10^{-4} \text{ m} = \mathbf{0.891 \text{ mm}}$$

$$\frac{x_2}{x_1} = \frac{0.891 \text{ mm}}{0.594 \text{ mm}} = \mathbf{1.500}$$ $$\sqrt{\frac{t_2}{t_1}} = \sqrt{\frac{9.0 \text{ hr}}{4.0 \text{ hr}}} = \sqrt{2.25} = \mathbf{1.500}$$

The calculated depth ratio matches the theoretical square-root scaling law with exact precision.

• At $T_1 = 900^\circ\text{C} = 1173 \text{ K}$: $$R T_1 = (8.314 \text{ J/(mol}\cdot\text{K)}) \times (1173 \text{ K}) = 9752.3 \text{ J/mol} = 9.7523 \text{ kJ/mol}$$ $$D_1 = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{9.7523}\right) = (2.3 \times 10^{-5}) \exp(-15.1759)$$ $$D_1 = (2.3 \times 10^{-5}) \times (2.5656 \times 10^{-7}) = \mathbf{5.901 \times 10^{-12} \text{ m}^2/\text{s}}$$
• At $T_2 = 1000^\circ\text{C} = 1273 \text{ K}$: $$R T_2 = (8.314) \times (1273) = 10583.7 \text{ J/mol} = 10.5837 \text{ kJ/mol}$$ $$D_2 = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{10.5837}\right) = (2.3 \times 10^{-5}) \exp(-13.9838)$$ $$D_2 = (2.3 \times 10^{-5}) \times (8.4510 \times 10^{-7}) = \mathbf{1.9437 \times 10^{-11} \text{ m}^2/\text{s}}$$

Diffusivity ratio:

$$\frac{D_2}{D_1} = \frac{1.9437 \times 10^{-11}}{5.901 \times 10^{-12}} = \mathbf{3.294}$$

Increasing temperature by just $100^\circ\text{C}$ speeds up carbon mobility by more than $3.3\times$!

For an identical concentration profile, the argument $z = \frac{x}{2\sqrt{Dt}}$ is constant for any depth $x$, which requires:

$$D_1 t_1 = D_2 t_2 \implies t_2 = t_1 \left(\frac{D_1}{D_2}\right)$$ $$t_2 = 10.0 \text{ hr} \times \left(\frac{1}{3.294}\right) = \mathbf{3.036 \text{ hours} = 3 \text{ hours and 2 minutes}}$$

Percentage time savings:

$$\% \text{ Time Saved} = \frac{10.0 - 3.036}{10.0} \times 100\% = \mathbf{69.64\% \approx 70\%}$$

The furnace cycle is reduced from 10 hours down to just 3 hours.

At $1000^\circ\text{C}$, the high thermal energy causes significant prior austenite grain growth (grain coarsening). Coarse austenite grains transform upon quenching into coarse martensite packets, which drastically reduces the impact toughness and fatigue resistance of the core steel. (To mitigate this, microalloyed steels with fine niobium/titanium carbide precipitates are used to pin grain boundaries and prevent grain growth).

The standard semi-infinite solution holds identically for outward diffusion (depletion), where $C_s < C_0$:

$$\frac{C(x,t) - C_0}{C_s - C_0} = 1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right) \implies C(x,t) = C_0 - (C_0 - C_s)\left[1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)\right]$$ $$C(x,t) = C_s + (C_0 - C_s) \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)$$

$$2\sqrt{Dt} = 2 \sqrt{(5.9 \times 10^{-12} \text{ m}^2/\text{s}) \times (10,800 \text{ s})} = 2 \sqrt{6.372 \times 10^{-8} \text{ m}^2} = 2 \times (2.524 \times 10^{-4} \text{ m}) = 5.048 \times 10^{-4} \text{ m}$$ $$z = \frac{x}{2\sqrt{Dt}} = \frac{2.0 \times 10^{-4} \text{ m}}{5.048 \times 10^{-4} \text{ m}} = \mathbf{0.3962}$$

From the error function table: $\text{erf}(0.35) = 0.3794$ and $\text{erf}(0.40) = 0.4284$.

$$\text{erf}(0.3962) = 0.3794 + \left(\frac{0.3962 - 0.35}{0.05}\right) \times (0.4284 - 0.3794) = 0.3794 + 0.924 \times (0.0490) = 0.4247$$ $$C(0.20 \text{ mm}) = 0.05 + (0.80 - 0.05) \times (0.4247) = 0.05 + 0.75 \times 0.4247 = 0.05 + 0.3185 = \mathbf{0.369 \text{ wt}\%}$$

$$\frac{C_x - C_0}{C_s - C_0} = \frac{0.65 - 0.80}{0.05 - 0.80} = \frac{-0.15}{-0.75} = 0.2000$$ $$1 - \text{erf}(z) = 0.2000 \implies \text{erf}(z) = 0.8000$$

From table: $\text{erf}(0.90) = 0.7969$ and $\text{erf}(0.95) \approx 0.8209$.

$$\text{erf}(z) = 0.8000 \implies z \approx \mathbf{0.9065}$$ $$x = 2 z \sqrt{Dt} = 0.9065 \times (5.048 \times 10^{-4} \text{ m}) = 4.576 \times 10^{-4} \text{ m} = \mathbf{0.458 \text{ mm}}$$

A minimum machining depth of at least $0.46 \text{ mm}$ must be ground off the spring steel bar to remove the weakened, decarburized outer layer and restore full fatigue resistance.

Because the gas phase maintains a constant surface concentration $C_s$ during the entire pre-deposition step, the profile is given by the complementary error function ($\text{erfc}$):

$$C(x,t) = C_s \text{erfc}\left(\frac{x}{2\sqrt{Dt}}\right) = C_s \left[1 - \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)\right]$$

(Since $C_s = 5 \times 10^{19} \gg C_0 = 10^{15}$, the background concentration $C_0$ is completely negligible in the denominator).

The metallurgical $p\text{-}n$ junction occurs at that precise depth $x_j$ where the concentration of diffused acceptor dopant atoms ($C_{\text{boron}}$) exactly equals the background donor dopant concentration ($C_0$):

$$C(x_j, t) = C_0 \implies C_s \text{erfc}\left(\frac{x_j}{2\sqrt{Dt}}\right) = C_0 \implies \text{erfc}\left(\frac{x_j}{2\sqrt{Dt}}\right) = \frac{C_0}{C_s}$$

Diffusion time: $t = 30.0 \times 60 = 1800 \text{ s}$.

$$2\sqrt{Dt} = 2 \sqrt{(1.40 \times 10^{-13} \text{ cm}^2/\text{s}) \times (1800 \text{ s})} = 2 \sqrt{2.52 \times 10^{-10} \text{ cm}^2} = 2 \times (1.5875 \times 10^{-5} \text{ cm}) = 3.175 \times 10^{-5} \text{ cm}$$

Evaluate concentration ratio:

$$\text{erfc}(z_j) = \frac{C_0}{C_s} = \frac{1.0 \times 10^{15} \text{ cm}^{-3}}{5.0 \times 10^{19} \text{ cm}^{-3}} = 2.0 \times 10^{-5}$$

For very small $\text{erfc}$ values ($\text{erfc}(z) \ll 1$), we use the asymptotic expansion $\text{erfc}(z) \approx \frac{e^{-z^2}}{z\sqrt{\pi}}$ or standard semiconductor tables:

• For $\text{erfc}(z) = 2.0 \times 10^{-5} \implies z_j \approx \mathbf{3.045}$

Now solve for junction depth $x_j$:

$$x_j = z_j \times (2\sqrt{Dt}) = 3.045 \times (3.175 \times 10^{-5} \text{ cm}) = 9.668 \times 10^{-5} \text{ cm}$$

Convert to micrometers ($1 \text{ cm} = 10,000 \mu\text{m}$):

$$x_j = 9.668 \times 10^{-5} \times 10,000 = \mathbf{0.967 \mu\text{m}}$$

Integrating the pre-deposition complementary error function profile over all $x$ yields the total incorporated dose $Q$ ($\text{atoms/cm}^2$):

$$Q = \int_0^\infty C_s \text{erfc}\left(\frac{x}{2\sqrt{D_{\text{pre}} t_{\text{pre}}}}\right) dx = \frac{2}{\sqrt{\pi}} C_s \sqrt{D_{\text{pre}} t_{\text{pre}}}$$

During drive-in, the wafer surface is sealed with an oxidizing ambient ($\text{SiO}_2$ capping layer), so no dopant enters or escapes (zero-flux boundary condition $\left.\frac{\partial C}{\partial x}\right|_{x=0} = 0$). Symmetrical reflection of the infinite Green's function yields the Gaussian half-space profile:

$$C(x, t) = \frac{Q}{\sqrt{\pi D t}} \exp\left(-\frac{x^2}{4Dt}\right)$$

Given $Q = 1.80 \times 10^{15} \text{ atoms/cm}^2, D = 4.0 \times 10^{-13} \text{ cm}^2/\text{s}, t = 7200 \text{ s}$:

$$D t = (4.0 \times 10^{-13} \text{ cm}^2/\text{s}) \times (7200 \text{ s}) = 2.88 \times 10^{-9} \text{ cm}^2$$ $$\sqrt{\pi D t} = \sqrt{\pi \times (2.88 \times 10^{-9} \text{ cm}^2)} = \sqrt{9.0478 \times 10^{-9}} = 9.512 \times 10^{-5} \text{ cm}$$

• Surface Concentration ($x = 0$): $$C(0, t) = \frac{Q}{\sqrt{\pi D t}} = \frac{1.80 \times 10^{15} \text{ cm}^{-2}}{9.512 \times 10^{-5} \text{ cm}} = \mathbf{1.892 \times 10^{19} \text{ atoms/cm}^3}$$ Notice that surface concentration has dropped from $5 \times 10^{19}$ to $1.89 \times 10^{19}$ as dopant atoms diffuse inward!
• Final Junction Depth $x_j$ (where $C(x_j) = C_0 = 1.0 \times 10^{15} \text{ cm}^{-3}$): $$C_0 = C(0, t) \exp\left(-\frac{x_j^2}{4Dt}\right) \implies \exp\left(-\frac{x_j^2}{4Dt}\right) = \frac{C_0}{C(0, t)}$$ $$\exp\left(\frac{x_j^2}{4Dt}\right) = \frac{C(0, t)}{C_0} = \frac{1.892 \times 10^{19}}{1.0 \times 10^{15}} = 18,920$$ $$\frac{x_j^2}{4Dt} = \ln(18,920) = 9.848$$ $$x_j^2 = 9.848 \times (4Dt) = 9.848 \times (4 \times 2.88 \times 10^{-9} \text{ cm}^2) = 1.1345 \times 10^{-7} \text{ cm}^2$$ $$x_j = \sqrt{1.1345 \times 10^{-7} \text{ cm}^2} = 3.368 \times 10^{-4} \text{ cm} = \mathbf{3.37 \mu\text{m}}$$

For an infinite couple with initial conditions $C(x, 0) = C_1$ for $x < 0$ and $C(x, 0) = C_2$ for $x > 0$:

$$C(x,t) = \left(\frac{C_1 + C_2}{2}\right) - \left(\frac{C_1 - C_2}{2}\right) \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)$$

Equivalently written:

$$\frac{C(x,t) - C_1}{C_2 - C_1} = \frac{1}{2}\left[1 + \text{erf}\left(\frac{x}{2\sqrt{Dt}}\right)\right]$$

At the weld plane $x = 0$ for any time $t > 0$:

$$z = \frac{0}{2\sqrt{Dt}} = 0 \implies \text{erf}(0) = 0$$ $$C(0, t) = \frac{C_1 + C_2}{2} - \left(\frac{C_1 - C_2}{2}\right)(0) \equiv \mathbf{\frac{C_1 + C_2}{2}}$$

The composition at the original interface remains pinned at the exact arithmetic mean of the two initial terminal concentrations for all times.

Given $C_1 = 100 \text{ at}\%$ (for $x < 0$), $C_2 = 70 \text{ at}\%$ (for $x > 0$), $D = 2.0 \times 10^{-14} \text{ m}^2/\text{s}$:

$$C(0) = \frac{100 + 70}{2} = \mathbf{85.0 \text{ at}\% \text{ Cu}}$$ $$2\sqrt{Dt} = 2 \sqrt{(2.0 \times 10^{-14}) \times (172,800)} = 2 \sqrt{3.456 \times 10^{-9} \text{ m}^2} = 2 \times (5.879 \times 10^{-5} \text{ m}) = 1.1758 \times 10^{-4} \text{ m} = 117.58 \mu\text{m}$$

• At $x = +10 \mu\text{m}$ (into the alloy): $$z = \frac{+10 \mu\text{m}}{117.58 \mu\text{m}} = +0.08505$$

  Using $\text{erf}(z) \approx \frac{2}{\sqrt{\pi}} z \approx 1.1284 \times 0.08505 = 0.0960$:

  $$C(+10 \mu\text{m}) = 85.0 - \left(\frac{100 - 70}{2}\right)(0.0960) = 85.0 - 15(0.0960) = 85.0 - 1.44 = \mathbf{83.56 \text{ at}\% \text{ Cu}}$$
• At $x = -10 \mu\text{m}$ (into the pure copper): $$z = -0.08505 \implies \text{erf}(-z) = -0.0960$$ $$C(-10 \mu\text{m}) = 85.0 - 15(-0.0960) = 85.0 + 1.44 = \mathbf{86.44 \text{ at}\% \text{ Cu}}$$

$$\frac{C_x - C_0}{C_s - C_0} = \frac{5.0 - 0.0}{20.0 - 0.0} = \frac{5.0}{20.0} = 0.2500$$

$$\frac{C_x - C_0}{C_s - C_0} = 1 - \text{erf}(z) = 0.2500 \implies \text{erf}(z) = 1.0000 - 0.2500 = 0.7500$$

From the error function table in Problem 13:

$$\text{erf}(0.80) = 0.7421, \quad \text{erf}(0.85) = 0.7707$$

Linear interpolation for $\text{erf}(z) = 0.7500$:

$$\frac{z - 0.80}{0.7500 - 0.7421} = \frac{0.85 - 0.80}{0.7707 - 0.7421} \implies \frac{z - 0.80}{0.0079} = \frac{0.05}{0.0286} = 1.7483$$ $$z - 0.80 = 1.7483 \times 0.0079 = 0.0138 \implies z = \mathbf{0.8138}$$

Diffusion time: $t = 16.0 \text{ hr} = 16.0 \times 3600 \text{ s} = 57,600 \text{ s}$.

Depth: $x = 0.40 \text{ mm} = 4.0 \times 10^{-4} \text{ m}$.

From $z = \frac{x}{2\sqrt{Dt}}$:

$$\sqrt{Dt} = \frac{x}{2 z} = \frac{4.0 \times 10^{-4} \text{ m}}{2 \times 0.8138} = \frac{4.0 \times 10^{-4} \text{ m}}{1.6276} = 2.4576 \times 10^{-4} \text{ m}$$

Squaring both sides:

$$Dt = (2.4576 \times 10^{-4} \text{ m})^2 = 6.0398 \times 10^{-8} \text{ m}^2$$ $$D = \frac{6.0398 \times 10^{-8} \text{ m}^2}{57,600 \text{ s}} = \mathbf{1.049 \times 10^{-12} \text{ m}^2/\text{s}}$$

In $\text{cm}^2/\text{s}$ ($1 \text{ m}^2 = 10^4 \text{ cm}^2$):

$$D = 1.049 \times 10^{-12} \times 10^4 = \mathbf{1.05 \times 10^{-8} \text{ cm}^2/\text{s}}$$

Evaluate concentration ratio:

$$\frac{C_x - C_0}{C_s - C_0} = \frac{0.10 - 0.005}{0.40 - 0.005} = \frac{0.095}{0.395} = 0.2405$$ $$\text{erf}(z) = 1.0000 - 0.2405 = 0.7595$$

From table: $\text{erf}(0.80) = 0.7421, \text{erf}(0.85) = 0.7707$.

$$z - 0.80 = \frac{0.7595 - 0.7421}{0.7707 - 0.7421} \times (0.05) = \frac{0.0174}{0.0286} \times 0.05 = 0.6084 \times 0.05 = 0.0304 \implies z = \mathbf{0.8304}$$

Compute penetration depth $x$:

$$2\sqrt{Dt} = 2 \sqrt{(2.50 \times 10^{-12} \text{ m}^2/\text{s}) \times (43,200 \text{ s})} = 2 \sqrt{1.080 \times 10^{-7} \text{ m}^2} = 2 \times (3.2863 \times 10^{-4} \text{ m}) = 6.573 \times 10^{-4} \text{ m}$$ $$x = z \times (2\sqrt{Dt}) = 0.8304 \times (6.573 \times 10^{-4} \text{ m}) = 5.458 \times 10^{-4} \text{ m} = \mathbf{0.546 \text{ mm}}$$

Carburizing is performed at $900\text{--}950^\circ\text{C}$ in the FCC austenite regime and requires aggressive liquid oil/water quenching to transform austenite into martensite, generating massive thermal and transformation stresses that warp gears. Nitriding is performed at $525^\circ\text{C}$ entirely within the ferritic temperature window (below the $727^\circ\text{C}$ eutectoid). No phase transformation or quenching is involved; parts are slowly furnace-cooled, resulting in near-zero dimensional distortion.

In Nitralloy steels, dissolved Aluminum, Chromium, and Vanadium react with the diffusing nitrogen to precipitate coherent, ultra-fine sub-nanometer alloy nitrides ($\text{AlN}, \text{CrN}, \text{VN}$) within the ferrite grains. These ultra-dense coherent nanoprecipitates create massive dislocation glide resistance via the Orowan bypass mechanism, driving surface hardness to extreme values of $1000\text{--}1200 \text{ HV}$ (70 HRC), providing exceptional wear and anti-scuffing properties.

Substitutional dopants in silicon migrate via an interstitialcy mechanism (kick-out and vacancy-assisted exchange). Phosphorus ($r_{\text{cov}} = 0.106 \text{ nm}$) is relatively small and forms mobile dopant-interstitial pairs ($(\text{P}-I)$) that migrate with low activation energy ($Q \approx 3.66 \text{ eV}$). Arsenic ($r_{\text{cov}} = 0.122 \text{ nm}$) has an atomic radius virtually identical to silicon ($0.118 \text{ nm}$); its large mass and size inhibit interstitialcy kick-out, causing it to migrate strictly via slower neutral vacancy exchanges ($Q \approx 4.05 \text{ eV}$).

• For Phosphorus ($D_{\text{P}} = 6.5 \times 10^{-14} \text{ cm}^2/\text{s}$): $$2\sqrt{D_{\text{P}} t} = 2 \sqrt{(6.5 \times 10^{-14}) \times (3600)} = 2 \sqrt{2.34 \times 10^{-10} \text{ cm}^2} = 2 \times (1.5297 \times 10^{-5} \text{ cm}) = 3.059 \times 10^{-5} \text{ cm} = \mathbf{0.306 \mu\text{m} = 306 \text{ nm}}$$
• For Arsenic ($D_{\text{As}} = 2.0 \times 10^{-15} \text{ cm}^2/\text{s}$): $$2\sqrt{D_{\text{As}} t} = 2 \sqrt{(2.0 \times 10^{-15}) \times (3600)} = 2 \sqrt{7.20 \times 10^{-12} \text{ cm}^2} = 2 \times (2.683 \times 10^{-6} \text{ cm}) = 5.367 \times 10^{-6} \text{ cm} = \mathbf{0.0537 \mu\text{m} = 53.7 \text{ nm}}$$

In advanced nanometer CMOS transistors, source and drain junctions must have extremely shallow depths ($x_j < 20\text{--}50 \text{ nm}$) to prevent drain-induced barrier lowering (short-channel effects). Phosphorus diffuses so rapidly ($306 \text{ nm/hr}$) that during the thermal budget required to activate implants, it diffuses deeply into the channel, causing catastrophic punch-through leakage. Arsenic's extremely slow diffusivity ($54 \text{ nm/hr}$, nearly $6\times$ shorter penetration) allows the formation of razor-sharp, ultra-shallow, steep dopant profiles that maintain strict short-channel control.

From the semi-infinite solution:

$$C(x,t) = C_0 + (C_s - C_0)\left[1 - \frac{2}{\sqrt{\pi}} \int_0^{\frac{x}{2\sqrt{Dt}}} e^{-\eta^2} d\eta\right]$$

Differentiating with respect to $x$ using Leibniz's rule:

$$\frac{\partial C}{\partial x} = -(C_s - C_0) \frac{2}{\sqrt{\pi}} \exp\left(-\frac{x^2}{4Dt}\right) \frac{\partial}{\partial x}\left(\frac{x}{2\sqrt{Dt}}\right) = -\frac{C_s - C_0}{\sqrt{\pi D t}} \exp\left(-\frac{x^2}{4Dt}\right)$$

Evaluating at the surface $x = 0$ ($\exp(0) = 1$):

$$\left.\frac{\partial C}{\partial x}\right|_{x=0} = -\frac{C_s - C_0}{\sqrt{\pi D t}}$$

Applying Fick's first law $J_0 = -D \left.\frac{\partial C}{\partial x}\right|_{x=0}$:

$$J_0(t) = -D \left(-\frac{C_s - C_0}{\sqrt{\pi D t}}\right) = (C_s - C_0) \frac{D}{\sqrt{\pi D t}} = \mathbf{(C_s - C_0) \sqrt{\frac{D}{\pi t}}}$$

At $t \to 0$, the concentration step is infinitely steep ($\partial C/\partial x \to -\infty$), yielding infinite initial flux. As a saturated boundary layer forms, the diffusion barrier thickens, causing flux to decay as $1/\sqrt{t}$.

The total mass absorbed per unit surface area up to time $t$ is the time integral of surface flux:

$$M_t = \int_0^t J_0(t') dt' = \int_0^t (C_s - C_0) \sqrt{\frac{D}{\pi}} (t')^{-1/2} dt' = (C_s - C_0) \sqrt{\frac{D}{\pi}} \left[ \frac{(t')^{1/2}}{1/2} \right]_0^t$$ $$M_t = 2 (C_s - C_0) \sqrt{\frac{Dt}{\pi}}$$

Given $\Delta C = C_s - C_0 = 93.6 - 15.6 = 78.0 \text{ kg/m}^3$, $D = 1.28 \times 10^{-11} \text{ m}^2/\text{s}$:

• Instantaneous Surface Flux at $t = 21,600 \text{ s}$: $$\sqrt{\frac{D}{\pi t}} = \sqrt{\frac{1.28 \times 10^{-11}}{\pi \times 21,600}} = \sqrt{\frac{1.28 \times 10^{-11}}{67,858}} = \sqrt{1.8863 \times 10^{-16}} = 1.3734 \times 10^{-8} \text{ m/s}$$ $$J_0 = (78.0 \text{ kg/m}^3) \times (1.3734 \times 10^{-8} \text{ m/s}) = 1.0713 \times 10^{-6} \text{ kg}/(\text{m}^2\cdot\text{s})$$

  Convert to $\text{g}/(\text{m}^2\cdot\text{hr})$ ($1 \text{ kg} = 1000 \text{ g}, 1 \text{ hr} = 3600 \text{ s}$):

  $$J_0 = (1.0713 \times 10^{-6}) \times 1000 \times 3600 = \mathbf{3.857 \text{ g}/(\text{m}^2\cdot\text{hr})}$$
• Total Mass Absorbed by Billet: $$M_t = 2 (78.0 \text{ kg/m}^3) \sqrt{\frac{(1.28 \times 10^{-11}) \times (21,600)}{\pi}} = 156.0 \times \sqrt{\frac{2.7648 \times 10^{-7}}{\pi}} = 156.0 \times \sqrt{8.8006 \times 10^{-8}}$$ $$M_t = 156.0 \times (2.9666 \times 10^{-4} \text{ m}) = 0.04628 \text{ kg/m}^2 = 46.28 \text{ g/m}^2$$

  Total mass for entire area $A = 2.5 \text{ m}^2$:

  $$m_{\text{total}} = M_t \times A = (0.04628 \text{ kg/m}^2) \times (2.5 \text{ m}^2) = \mathbf{0.1157 \text{ kg} = 115.7 \text{ grams}}$$

$$D = D_0 \exp\left(-\frac{Q_d}{R T}\right)$$

• $D$: Diffusion coefficient at absolute temperature $T$ ($\mathbf{\text{m}^2/\text{s}}$).
• $D_0$: Temperature-independent pre-exponential factor ($\mathbf{\text{m}^2/\text{s}}$).
• $Q_d$: Activation energy for diffusion ($\mathbf{\text{J/mol}}$ or $\mathbf{\text{kJ/mol}}$ when using gas constant $R = 8.314 \text{ J/(mol}\cdot\text{K)}$; or $\mathbf{\text{eV/atom}}$ when using Boltzmann's constant $k_B = 8.617 \times 10^{-5} \text{ eV/K}$).
• $T$: Absolute temperature in Kelvin ($\mathbf{\text{K}}$).

• Pre-exponential Factor $D_0$: From random walk theory, $D = \frac{1}{6} \alpha z a^2 \Gamma$, where $a$ is jump distance, $z$ is coordination number, and $\Gamma$ is successful jump frequency. Transition state theory expresses $D_0$ as: $$D_0 = \frac{1}{6} \alpha z a^2 \nu_0 \exp\left(\frac{\Delta S}{R}\right)$$ where $\nu_0$ is the Debye vibrational frequency ($\sim 10^{13} \text{ s}^{-1}$) and $\Delta S$ is the activation entropy due to lattice distortion during the atomic jump.
• Activation Energy $Q_d$: The potential energy barrier that an atom must overcome to execute a successful jump:
  • For substitutional / self-diffusion: $Q_d = \Delta H_f + \Delta H_m$ (energy to form a vacancy plus energy to move into it).
  • For interstitial diffusion: $Q_d = \Delta H_m$ (energy to migrate through interstitial bottlenecks; vacancy formation energy is zero).

Taking the natural logarithm of the Arrhenius equation:

$$\ln D = \ln D_0 - \left(\frac{Q_d}{R}\right) \frac{1}{T}$$

This matches the standard linear form $y = m x + c$, where $y = \ln D$ and $x = \frac{1}{T}$:

• Slope: $m = -\frac{Q_d}{R} \implies Q_d = -R \times \text{slope}$.
• $y$-Intercept: $c = \ln D_0 \implies D_0 = e^c$.

$$\ln D_1 = \ln D_0 - \frac{Q_d}{R T_1}, \quad \ln D_2 = \ln D_0 - \frac{Q_d}{R T_2}$$

Subtracting the first equation from the second:

$$\ln\left(\frac{D_2}{D_1}\right) = -\frac{Q_d}{R} \left(\frac{1}{T_2} - \frac{1}{T_1}\right) = \frac{Q_d}{R} \left(\frac{T_2 - T_1}{T_1 T_2}\right)$$ $$Q_d = \frac{R \ln(D_2 / D_1)}{\frac{1}{T_1} - \frac{1}{T_2}} = \frac{R T_1 T_2}{T_2 - T_1} \ln\left(\frac{D_2}{D_1}\right)$$

$$\frac{D_2}{D_1} = \frac{6.80 \times 10^{-14}}{5.40 \times 10^{-16}} = 125.926 \implies \ln(125.926) = 4.8357$$ $$\frac{1}{T_1} - \frac{1}{T_2} = \frac{1}{1000} - \frac{1}{1200} = 1.000 \times 10^{-3} - 8.3333 \times 10^{-4} = 1.6667 \times 10^{-4} \text{ K}^{-1}$$ $$Q_d = \frac{(8.314 \text{ J/(mol}\cdot\text{K)}) \times 4.8357}{1.6667 \times 10^{-4} \text{ K}^{-1}} = 241,250 \text{ J/mol} = \mathbf{241.3 \text{ kJ/mol}}$$

Convert to $\text{eV/atom}$ ($1 \text{ eV/atom} = 96.485 \text{ kJ/mol}$):

$$Q_d = \frac{241.3 \text{ kJ/mol}}{96.485 \text{ kJ/(mol}\cdot\text{eV)}} = \mathbf{2.50 \text{ eV/atom}}$$

$$D_0 = D_1 \exp\left(\frac{Q_d}{R T_1}\right)$$ $$\frac{Q_d}{R T_1} = \frac{241,250 \text{ J/mol}}{(8.314 \text{ J/(mol}\cdot\text{K)}) \times (1000 \text{ K})} = \frac{241,250}{8314} = 29.0173$$ $$D_0 = (5.40 \times 10^{-16} \text{ m}^2/\text{s}) \times \exp(29.0173) = (5.40 \times 10^{-16}) \times (4.0098 \times 10^{12}) = \mathbf{2.165 \times 10^{-3} \text{ m}^2/\text{s} \approx 2.2 \times 10^{-3} \text{ m}^2/\text{s}}$$

$$\frac{Q_d}{R T_3} = \frac{241,250}{8.314 \times 1300} = \frac{241,250}{10,808.2} = 22.321$$ $$D_3 = (2.165 \times 10^{-3} \text{ m}^2/\text{s}) \times \exp(-22.321) = (2.165 \times 10^{-3}) \times (2.0235 \times 10^{-10}) = \mathbf{4.38 \times 10^{-13} \text{ m}^2/\text{s}}$$

Let $x_i = 1/T_i$ and $y_i = \ln D_i$ for $N = 4$ points:

• $\sum x = 8.177 + 7.559 + 7.027 + 6.566 = 29.329 \times 10^{-4} \text{ K}^{-1} \implies \bar{x} = 7.3323 \times 10^{-4}$
• $\sum y = -36.228 - 33.951 - 32.012 - 30.380 = -132.571 \implies \bar{y} = -33.1428$
• $\sum (x_i - \bar{x})^2 = (0.8447)^2 + (0.2267)^2 + (-0.3053)^2 + (-0.7663)^2 = (0.7135 + 0.0514 + 0.0932 + 0.5872) \times 10^{-8} = 1.4453 \times 10^{-8}$
• $\sum (x_i - \bar{x})(y_i - \bar{y}) = (0.8447)(-3.085) + (0.2267)(-0.808) + (-0.3053)(1.131) + (-0.7663)(2.763) = (-2.606 - 0.183 - 0.345 - 2.117) \times 10^{-4} = -5.251 \times 10^{-4}$

The best-fit slope is:

$$m = \frac{\sum (x_i - \bar{x})(y_i - \bar{y})}{\sum (x_i - \bar{x})^2} = \frac{-5.251 \times 10^{-4}}{1.4453 \times 10^{-8}} = \mathbf{-36,331 \text{ K}}$$

The correlation coefficient is $r^2 = 0.9998$, confirming near-perfect Arrhenius linearity.

Since slope $m = -\frac{Q_d}{R}$:

$$Q_d = -R \cdot m = -(8.314 \text{ J/(mol}\cdot\text{K)}) \times (-36,331 \text{ K}) = 302,056 \text{ J/mol} = \mathbf{302.1 \text{ kJ/mol}}$$

The $y$-intercept is:

$$c = \bar{y} - m \bar{x} = -33.1428 - (-36,331) \times (7.3323 \times 10^{-4}) = -33.1428 + 26.6399 = \mathbf{-6.5029}$$ $$D_0 = \exp(c) = \exp(-6.5029) = \mathbf{1.50 \times 10^{-3} \text{ m}^2/\text{s}}$$

$$D(T) = (1.50 \times 10^{-3} \text{ m}^2/\text{s}) \exp\left(-\frac{302.1 \text{ kJ/mol}}{R T}\right)$$

In a 3D random walk with step length $\lambda$ and jump rate $\Gamma$:

$$\langle r^2 \rangle = \Gamma t \lambda^2 = 6 D t \implies D = \frac{1}{6} \Gamma \lambda^2 \implies \Gamma = \frac{6 D}{\lambda^2}$$

In FCC, nearest neighbors lie along the face diagonal $\frac{a}{2}\langle 110\rangle$:

$$\lambda = \frac{a\sqrt{2}}{2} = \frac{0.3615 \text{ nm} \times 1.4142}{2} = \mathbf{0.2556 \text{ nm} = 2.556 \times 10^{-10} \text{ m}}$$

$$\lambda^2 = (2.556 \times 10^{-10} \text{ m})^2 = 6.533 \times 10^{-20} \text{ m}^2$$ $$\Gamma = \frac{6 \times (2.0 \times 10^{-16} \text{ m}^2/\text{s})}{6.533 \times 10^{-20} \text{ m}^2} = \frac{1.20 \times 10^{-15}}{6.533 \times 10^{-20}} = \mathbf{18,368 \text{ jumps/s} \approx 1.84 \times 10^4 \text{ jumps/s}}$$

The average residence time is:

$$\tau = \frac{1}{\Gamma} = \frac{1}{18,368 \text{ s}^{-1}} = \mathbf{5.44 \times 10^{-5} \text{ seconds} = 54.4 \mu\text{s}}$$

At $800^\circ\text{C}$, a copper atom stays on its lattice site for about 54 microseconds before jumping to an adjacent vacancy.

$$\text{Success Fraction} = \frac{\Gamma}{\nu_0} = \frac{1.84 \times 10^4 \text{ s}^{-1}}{1.0 \times 10^{13} \text{ s}^{-1}} = \mathbf{1.84 \times 10^{-9}}$$

Only about 2 out of every billion atomic vibrations possess sufficient thermal energy and an adjacent vacancy to successfully execute a lattice jump!

The vector connecting an edge center $(a/2, 0, 0)$ to the adjacent body center $(a/2, a/2, a/2)$ is $(0, a/2, a/2)$. The distance is:

$$\lambda = \sqrt{0^2 + \left(\frac{a}{2}\right)^2 + \left(\frac{a}{2}\right)^2} = \frac{a\sqrt{2}}{2} = \frac{a}{\sqrt{2}}$$ $$\lambda = \frac{0.364 \text{ nm}}{\sqrt{2}} = \mathbf{0.2574 \text{ nm} = 2.574 \times 10^{-10} \text{ m}}$$

$$\lambda^2 = (2.574 \times 10^{-10} \text{ m})^2 = 6.625 \times 10^{-20} \text{ m}^2$$ $$\Gamma = \frac{6 D}{\lambda^2} = \frac{6 \times (2.0 \times 10^{-11} \text{ m}^2/\text{s})}{6.625 \times 10^{-20} \text{ m}^2} = \frac{1.20 \times 10^{-10}}{6.625 \times 10^{-20}} = \mathbf{1.81 \times 10^9 \text{ jumps/s}}$$

A dissolved carbon atom jumps nearly 2 billion times every second at $1000^\circ\text{C}$! Its residence time is less than one nanosecond ($\tau \approx 0.55 \text{ ns}$).

$$\frac{\Gamma_{\text{interstitial}}}{\Gamma_{\text{substitutional}}} = \frac{1.81 \times 10^9 \text{ s}^{-1}}{1.84 \times 10^4 \text{ s}^{-1}} \approx \mathbf{100,000 \times}$$

Interstitial carbon executes jumps 100,000 times more frequently than substitutional atoms at comparable temperatures because all adjacent interstitial sites are vacant and migration barriers are low.

$R T_1 = (8.314 \text{ J/(mol}\cdot\text{K)}) \times (1223 \text{ K}) = 10,168 \text{ J/mol} = 10.168 \text{ kJ/mol}$.

• Carbon Diffusivity: $$D_{\text{C}} = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{10.168}\right) = (2.3 \times 10^{-5}) \exp(-14.555) = \mathbf{1.107 \times 10^{-11} \text{ m}^2/\text{s}}$$
• Iron Self-Diffusivity: $$D_{\text{Fe}} = (6.5 \times 10^{-5}) \exp\left(-\frac{284.0}{10.168}\right) = (6.5 \times 10^{-5}) \exp(-27.931) = \mathbf{4.814 \times 10^{-17} \text{ m}^2/\text{s}}$$

Ratio at $950^\circ\text{C}$:

$$\frac{D_{\text{C}}}{D_{\text{Fe}}} = \frac{1.107 \times 10^{-11}}{4.814 \times 10^{-17}} = \mathbf{230,000 \approx 2.3 \times 10^5}$$

At $950^\circ\text{C}$, carbon diffuses 230,000 times faster than iron!

$R T_2 = (8.314) \times (1423) = 11,831 \text{ J/mol} = 11.831 \text{ kJ/mol}$.

• Carbon Diffusivity: $$D_{\text{C}} = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{11.831}\right) = (2.3 \times 10^{-5}) \exp(-12.509) = \mathbf{8.455 \times 10^{-11} \text{ m}^2/\text{s}}$$
• Iron Self-Diffusivity: $$D_{\text{Fe}} = (6.5 \times 10^{-5}) \exp\left(-\frac{284.0}{11.831}\right) = (6.5 \times 10^{-5}) \exp(-24.005) = \mathbf{2.430 \times 10^{-15} \text{ m}^2/\text{s}}$$

Ratio at $1150^\circ\text{C}$:

$$\frac{D_{\text{C}}}{D_{\text{Fe}}} = \frac{8.455 \times 10^{-11}}{2.430 \times 10^{-15}} = \mathbf{34,800 \approx 3.5 \times 10^4}$$

Physical Explanation: Because iron self-diffusion has a much higher activation energy ($284 \text{ kJ/mol}$ vs $148 \text{ kJ/mol}$), its Arrhenius slope is much steeper. As temperature increases, the process with the higher activation energy accelerates much more rapidly, shrinking the kinetic disparity from $230,000\times$ at $950^\circ\text{C}$ down to $35,000\times$ at $1150^\circ\text{C}$.

$R T = (8.314 \text{ J/(mol}\cdot\text{K)}) \times (1185 \text{ K}) = 9852.1 \text{ J/mol} = 9.8521 \text{ kJ/mol}$.

• In BCC $\alpha$-Fe at $912^\circ\text{C}$: $$\frac{Q_d}{R T} = \frac{251.0}{9.8521} = 25.4768$$ $$D_\alpha = (2.0 \times 10^{-4} \text{ m}^2/\text{s}) \times \exp(-25.4768) = (2.0 \times 10^{-4}) \times (8.621 \times 10^{-12}) = \mathbf{1.724 \times 10^{-15} \text{ m}^2/\text{s}}$$
• In FCC $\gamma$-Fe at $912^\circ\text{C}$: $$\frac{Q_d}{R T} = \frac{284.0}{9.8521} = 28.8263$$ $$D_\gamma = (6.5 \times 10^{-5} \text{ m}^2/\text{s}) \times \exp(-28.8263) = (6.5 \times 10^{-5}) \times (3.026 \times 10^{-13}) = \mathbf{1.967 \times 10^{-17} \text{ m}^2/\text{s}}$$

$$\frac{D_\alpha(\text{BCC})}{D_\gamma(\text{FCC})} = \frac{1.724 \times 10^{-15} \text{ m}^2/\text{s}}{1.967 \times 10^{-17} \text{ m}^2/\text{s}} = \mathbf{87.6 \approx 88}$$

Upon crossing the $912^\circ\text{C}$ transformation line from BCC into FCC, the self-diffusion coefficient drops discontinuously by a factor of nearly 90 times!

1. Atomic Packing Factor: BCC is an open, non-close-packed crystal structure with $\text{APF} = 0.68$, leaving $32\%$ free volume. FCC is close-packed with $\text{APF} = 0.74$, leaving only $26\%$ free volume.
2. Coordination Number and Bond Strength: In BCC, an atom has only 8 nearest neighbors ($CN = 8$); in FCC, it has 12 nearest neighbors ($CN = 12$). Squeezing an atom between fewer neighbors requires severing or stretching fewer coordinating bonds.
3. Saddle-Point Bottleneck: During a vacancy jump, the migrating atom must push through a constriction formed by coordinating atoms (the saddle-point configuration). In the open BCC lattice, the window between atoms is wider, requiring substantially less elastic strain work ($\Delta H_m$ is lower). Consequently, both activation enthalpy for vacancy formation ($\Delta H_f$) and migration ($\Delta H_m$) are significantly lower in BCC, yielding $Q_{\text{BCC}} = 251 \text{ kJ/mol} < Q_{\text{FCC}} = 284 \text{ kJ/mol}$.

$R T = (8.314) \times (1185) = 9852.1 \text{ J/mol} = 9.8521 \text{ kJ/mol}$.

• In BCC Ferrite ($\alpha$-Fe): $$D_{\text{C}}(\alpha) = (6.2 \times 10^{-7}) \exp\left(-\frac{80.0}{9.8521}\right) = (6.2 \times 10^{-7}) \exp(-8.120) = \mathbf{1.841 \times 10^{-10} \text{ m}^2/\text{s}}$$
• In FCC Austenite ($\gamma$-Fe): $$D_{\text{C}}(\gamma) = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{9.8521}\right) = (2.3 \times 10^{-5}) \exp(-15.022) = \mathbf{7.014 \times 10^{-12} \text{ m}^2/\text{s}}$$

Ratio at $912^\circ\text{C}$:

$$\frac{D_{\text{C}}(\alpha)}{D_{\text{C}}(\gamma)} = \frac{1.841 \times 10^{-10}}{7.014 \times 10^{-12}} = \mathbf{26.25 \approx 26}$$

Even though carbon solubility is 100 times smaller in ferrite, its diffusion rate in ferrite is 26 times faster!

$R T = (8.314) \times (873) = 7258.1 \text{ J/mol} = 7.2581 \text{ kJ/mol}$.

• $D_{\text{C}}(\alpha) = (6.2 \times 10^{-7}) \exp\left(-\frac{80.0}{7.2581}\right) = (6.2 \times 10^{-7}) \exp(-11.022) = \mathbf{1.027 \times 10^{-11} \text{ m}^2/\text{s}}$
• $D_{\text{C}}(\gamma) = (2.3 \times 10^{-5}) \exp\left(-\frac{148.0}{7.2581}\right) = (2.3 \times 10^{-5}) \exp(-20.391) = \mathbf{3.238 \times 10^{-14} \text{ m}^2/\text{s}}$

Ratio at $600^\circ\text{C}$:

$$\frac{D_{\text{C}}(\alpha)}{D_{\text{C}}(\gamma)} = \frac{1.027 \times 10^{-11}}{3.238 \times 10^{-14}} = \mathbf{317.2 \approx 317}$$

Physical Reason: Because the activation energy in austenite ($148 \text{ kJ/mol}$) is nearly double that in ferrite ($80 \text{ kJ/mol}$), lowering the temperature freezes carbon diffusion in austenite far more aggressively, widening the ratio from $26\times$ at $912^\circ\text{C}$ to over $317\times$ at $600^\circ\text{C}$. This ultra-fast carbon diffusion in ferrite governs the formation kinetics of pearlite and bainite.

$$D_{\text{surface}} > D_{\text{grain boundary}} > D_{\text{dislocation pipe}} \gg D_{\text{lattice / volume}}$$

• External Surfaces: Atoms are bonded only to the layer below; open 2D space offers minimal geometric resistance.
• Grain Boundaries: Disordered transition zone containing excess free volume, dilated channels, and reduced bond coordination.
• Dislocation Cores: 1D pipe containing an extra half-plane termination with high core strain and dilated vacancy-like channels.
• Bulk Lattice: Perfect close-packed crystal; atoms are constrained by full coordination numbers ($CN = 12$).

$$Q_{\text{surface}} < Q_{\text{grain boundary}} < Q_{\text{dislocation pipe}} < Q_{\text{lattice}}$$

Typical empirical metallurgical rules of thumb:

• $Q_{\text{surface}} \approx 0.25\text{ to }0.35 \times Q_L$
• $Q_{\text{grain boundary}} \approx 0.40\text{ to }0.60 \times Q_L \approx \mathbf{0.5 Q_L}$
• $Q_{\text{dislocation pipe}} \approx 0.60\text{ to }0.70 \times Q_L$
• $Q_{\text{lattice}} \equiv Q_L$

The total mass transport is the product of diffusivity and the cross-sectional area fraction of each path:

$$J_{\text{total}} \propto A_L D_L + A_{\text{gb}} D_{\text{gb}} + A_{\text{pipe}} D_{\text{pipe}}$$

In a typical polycrystalline metal with grain size $d \approx 50 \mu\text{m}$ and grain boundary width $\delta \approx 0.5 \text{ nm}$:

$$\frac{A_{\text{gb}}}{A_L} \approx \frac{2\delta}{d} = \frac{2 \times 0.5 \times 10^{-9} \text{ m}}{50 \times 10^{-6} \text{ m}} = 2.0 \times 10^{-5}$$

Over $99.998\%$ of the cross-sectional area consists of bulk crystal lattice! At high temperatures ($T > 0.7 T_m$), thermal energy is sufficient to activate lattice jumps easily ($\exp(-Q_L/RT)$ increases dramatically). The vast geometric area of the bulk crystal completely overwhelms the higher intrinsic diffusivity of the tiny, nanometer-thin boundary network.

Fisher modeled a grain boundary as a semi-infinite uniform isotropic slab of narrow width $\delta \approx 0.5 \text{ nm}$ oriented perpendicular to the external surface ($y$-axis along the boundary, $x$-axis perpendicular to it). The boundary possesses an exceptionally high diffusion coefficient $D_{\text{gb}} \gg D_L$. Solute diffuses rapidly down the boundary "conduit" along $y$. As solute accumulates in the boundary, it simultaneously diffuses laterally out of the boundary slab into the adjacent low-diffusivity bulk grains along $x$ (lateral leakage).

Because atomic migration is orders of magnitude faster down the boundary conduit ($D_{\text{gb}} \gg D_L$), solute penetrates to depths along the boundary that are far beyond the reach of normal bulk lattice diffusion. At any given depth, solute leaks slowly into adjacent grain interiors, decaying rapidly with lateral distance $|x|$. The resulting curves of constant concentration $C(x,y) = \text{const}$ form deep, acute, V-shaped isoconcentration wedges or "penetration fingers" pointing deep into the material along each grain boundary.

$$\frac{D_{\text{gb}}}{D_L} = \frac{1.0 \times 10^{-11} \text{ m}^2/\text{s}}{1.0 \times 10^{-17} \text{ m}^2/\text{s}} = \mathbf{1,000,000 = 10^6}$$

At $500^\circ\text{C}$, atoms travel one million times faster along grain boundaries than through the grain interiors! In the bulk lattice, the diffusion distance in $t = 10 \text{ hours}$ ($36,000 \text{ s}$) is only $2\sqrt{D_L t} = 2\sqrt{3.6 \times 10^{-13}} \approx 1.2 \mu\text{m}$. Along the grain boundaries, the penetration distance is $2\sqrt{D_{\text{gb}} t} = 2\sqrt{3.6 \times 10^{-7}} \approx 1200 \mu\text{m} = \mathbf{1.2 \text{ mm}}$!

This colossal kinetic disparity explains why oxygen, molten metals, and aggressive environmental agents rapidly penetrate deep along grain boundaries at moderate temperatures, causing catastrophic stress-corrosion cracking, liquid metal embrittlement (e.g., liquid gallium penetrating aluminum grain boundaries), and intergranular oxidation.

Hart's equation models diffusion when atoms sample both lattice and dislocation conduits frequently (random mixing over time $t$ where $\sqrt{D_L t} \gg$ dislocation spacing):

$$D_{\text{eff}} = f D_{\text{pipe}} + (1 - f) D_L$$

Let $A_{\text{core}} = \pi r_0^2 \approx \pi (0.47 \times 10^{-9} \text{ m})^2 \approx 7.0 \times 10^{-19} \text{ m}^2$. In a unit cross-sectional area ($1 \text{ m}^2$), there are $\rho_d$ dislocation cores. The total area occupied by dislocation cores is:

$$f = \frac{A_{\text{cores}}}{A_{\text{total}}} = \rho_d \cdot A_{\text{core}} = \rho_d (\pi r_0^2)$$

$$f = (2.0 \times 10^{15} \text{ m}^{-2}) \times (7.0 \times 10^{-19} \text{ m}^2) = \mathbf{1.40 \times 10^{-3} \approx 0.14\%}$$

Dislocation cores occupy about 0.14% of the crystal volume.

Given $D_L = 1.0 \times 10^{-20} \text{ m}^2/\text{s}$ and $D_{\text{pipe}} = 5.0 \times 10^{-14} \text{ m}^2/\text{s}$:

• Dislocation pipe contribution: $$f D_{\text{pipe}} = (1.40 \times 10^{-3}) \times (5.0 \times 10^{-14} \text{ m}^2/\text{s}) = \mathbf{7.00 \times 10^{-17} \text{ m}^2/\text{s}}$$
• Lattice contribution: $$(1 - f) D_L \approx 1 \times (1.0 \times 10^{-20} \text{ m}^2/\text{s}) = \mathbf{1.00 \times 10^{-20} \text{ m}^2/\text{s}}$$

Total effective diffusivity:

$$D_{\text{eff}} = 7.00 \times 10^{-17} + 1.00 \times 10^{-20} \approx \mathbf{7.00 \times 10^{-17} \text{ m}^2/\text{s}}$$

Enhancement ratio:

$$\frac{D_{\text{eff}}}{D_L} = \frac{7.00 \times 10^{-17}}{1.00 \times 10^{-20}} = \mathbf{7,000 \times}$$

Severe cold deformation multiplies the effective atomic mobility by 7,000 times, enabling rapid low-temperature strain-aging and solute segregation.

• Regime A (High Temperature, Long Times: $\sqrt{D_L t} \gg d$):
  • Physics: Atoms diffuse so far during time $t$ that their diffusion distance greatly exceeds the grain diameter $d$. Every atom samples multiple grain boundaries and grain interiors. The diffusion fields of adjacent boundaries overlap completely, yielding smooth planar isoconcentration contours.
  • Measured Parameter: Follows Fick's law with an effective homogenized diffusivity: $$D_{\text{eff}} = D_L + \left(\frac{2\delta}{d}\right) D_{\text{gb}}$$
• Regime B (Intermediate Temperature: $\delta \ll \sqrt{D_L t} \ll d$):
  • Physics: Fast diffusion occurs down grain boundaries, accompanied by lateral leakage into the immediately adjacent grain interiors. Because $\sqrt{D_L t} \ll d$, the leakage fields from adjacent boundaries do not overlap. Profiles exhibit the classic Fisher V-shaped wedges.
  • Measured Parameter: Penetration plots of $\ln C$ versus depth $y^{6/5}$ yield the triple product $s \delta D_{\text{gb}}$ (via the Le Claire / Whipple-Suzuoka equation). One cannot separate $D_{\text{gb}}$ from boundary width $\delta$ without independent assumptions.
• Regime C (Low Temperature, Short Times: $\sqrt{D_L t} \ll \delta$):
  • Physics: Bulk lattice diffusion is completely frozen ($\sqrt{D_L t} \approx 0$). Zero lateral leakage occurs into grain interiors. Diffusion takes place strictly and exclusively within the two-dimensional grain boundary network, with grains acting as completely inert, impermeable blocks.
  • Measured Parameter: Standard Fickian Gaussian analysis yields the true, uncoupled grain boundary diffusivity $D_{\text{gb}}$ directly!

• Regime A: High homologous temperatures ($T > 0.7\text{--}0.8 T_m$), where bulk lattice diffusion is rapid.
• Regime B: Intermediate temperatures ($0.4 T_m < T < 0.7 T_m$), typical of industrial heat treatments and creep.
• Regime C: Low cryogenic to ambient temperatures ($T < 0.3\text{--}0.4 T_m$), typical of thin films and room-temperature aging.

Model grains as cubes or spheres of diameter $d$. The crystalline core has dimension $(d - \delta)$. The volume of the grain interior is $V_{\text{core}} = (d - \delta)^3$, and total grain volume is $V_{\text{total}} = d^3$:

$$f_{\text{gb}} = \frac{V_{\text{total}} - V_{\text{core}}}{V_{\text{total}}} = \frac{d^3 - (d - \delta)^3}{d^3} = 1 - \left(1 - \frac{\delta}{d}\right)^3$$

Using the binomial expansion for $\delta \ll d$:

$$\left(1 - \frac{\delta}{d}\right)^3 \approx 1 - \frac{3\delta}{d} \implies f_{\text{gb}} \approx \mathbf{\frac{3\delta}{d}}$$

• Coarse-Grained ($d = 50 \mu\text{m} = 5.0 \times 10^{-5} \text{ m}$): $$f_{\text{gb}} = \frac{3 \times (1.0 \times 10^{-9} \text{ m})}{5.0 \times 10^{-5} \text{ m}} = 6.0 \times 10^{-5} = \mathbf{0.006\%}$$
• Fine-Grained ($d = 5 \mu\text{m} = 5.0 \times 10^{-6} \text{ m}$): $$f_{\text{gb}} = \frac{3 \times 1.0 \times 10^{-9}}{5.0 \times 10^{-6}} = 6.0 \times 10^{-4} = \mathbf{0.06\%}$$
• Nanocrystalline ($d = 10 \text{ nm} = 1.0 \times 10^{-8} \text{ m}$): $$f_{\text{gb}} = 1 - \left(\frac{10 - 1}{10}\right)^3 = 1 - (0.9)^3 = 1 - 0.729 = \mathbf{0.271 = 27.1\%}$$

  In a $10 \text{ nm}$ nanocrystal, over one-quarter of all atoms reside inside grain boundaries and triple junctions!

Given $D_L = 1.0 \times 10^{-23} \text{ m}^2/\text{s}$ and $D_{\text{gb}} = 1.0 \times 10^{-14} \text{ m}^2/\text{s}$:

• For Coarse-Grained Metal ($f_{\text{gb}} = 6.0 \times 10^{-5}$): $$D_{\text{eff}} = (6.0 \times 10^{-5}) \times (1.0 \times 10^{-14}) + (1) \times (1.0 \times 10^{-23}) = 6.0 \times 10^{-19} + 1.0 \times 10^{-23} = \mathbf{6.0 \times 10^{-19} \text{ m}^2/\text{s}}$$
• For Nanocrystalline Metal ($f_{\text{gb}} = 0.271$): $$D_{\text{eff}} = (0.271) \times (1.0 \times 10^{-14}) + (0.729) \times (1.0 \times 10^{-23}) = \mathbf{2.71 \times 10^{-15} \text{ m}^2/\text{s}}$$

Acceleration factor:

$$\frac{D_{\text{eff}}(\text{nano})}{D_{\text{eff}}(\text{coarse})} = \frac{2.71 \times 10^{-15}}{6.0 \times 10^{-19}} = \mathbf{4,517 \approx 4,500 \times}$$

Nanocrystalline copper diffuses 4,500 times faster than conventional copper, enabling low-temperature sintering and fast diffusion bonding, but creating severe thermal instability (rapid grain coarsening at modest temperatures).

Cations are smaller than anions and generally possess much higher individual diffusion coefficients ($D_+ \gg D_-$). If cations diffused freely down a concentration gradient ahead of anions, positive charge would accumulate at the leading edge, leaving negative charge behind. To prevent macroscopic charge separation and uphold strict electroneutrality, an internal electrostatic potential gradient (electric field $E = -\nabla\phi$) is established. This field retards the faster-moving cations and accelerates the slower-moving anions, forcing both ions to migrate at a single, coupled mutual rate.

The total flux of each ion under combined concentration gradient and electric field $E$ is given by the Nernst-Planck equation:

$$J_+ = -D_+ \frac{dC_+}{dx} + \frac{z_+ e D_+ C_+}{k_B T} E$$ $$J_- = -D_- \frac{dC_-}{dx} - \frac{z_- e D_- C_-}{k_B T} E$$

To prevent electrical current flow ($I = e(z_+ J_+ - z_- J_-) = 0$), electroneutrality requires $z_+ J_+ = z_- J_-$. Solving for the electric field $E$ and substituting back gives the coupled ambipolar flux:

$$J = -D_{\text{amb}} \frac{dC}{dx} \quad \text{where} \quad D_{\text{amb}} = \frac{(z_+ + z_-) D_+ D_-}{z_+ D_+ + z_- D_-}$$

For symmetric valence $z_+ = z_- = 2$:

$$D_{\text{amb}} = \frac{(2 + 2) D_+ D_-}{2 D_+ + 2 D_-} = \frac{4 D_+ D_-}{2(D_+ + D_-)} = \frac{2 D_+ D_-}{D_+ + D_-}$$

Since $D_{\text{Mg}} = 10^{-14} \text{ m}^2/\text{s} \gg D_{\text{O}} = 10^{-17} \text{ m}^2/\text{s}$:

$$D_+ + D_- \approx D_+ \implies D_{\text{amb}} \approx \frac{2 D_+ D_-}{D_+} = 2 D_-$$ $$D_{\text{amb}} = \frac{2 \times (1.0 \times 10^{-14}) \times (1.0 \times 10^{-17})}{1.0 \times 10^{-14} + 1.0 \times 10^{-17}} = \frac{2.0 \times 10^{-31}}{1.001 \times 10^{-14}} = \mathbf{1.998 \times 10^{-17} \text{ m}^2/\text{s} \approx 2 D_{\text{O}^{2-}}}$$

Macroscopic mass transport processes (sintering, creep, grain boundary sliding) require stoichiometric formula units ($\text{Mg}^{2+} + \text{O}^{2-}$) to move together to build or remove lattice planes. You cannot sinter $\text{MgO}$ by moving only magnesium; oxygen ions must move as well. Because the ambipolar diffusion coefficient is pinned to the slower species ($D_{\text{amb}} \approx 2 D_{\text{O}}$), the slower-diffusing oxygen ion ($\text{O}^{2-}$) is the strict rate-controlling bottleneck for all high-temperature deformation and sintering of $\text{MgO}$.

$$P = D \cdot S$$

• $P$: Permeability Coefficient. Measures total steady-state gas transmission per unit pressure gradient. SI units: $\mathbf{\frac{\text{m}^3\text{(STP)}\cdot\text{m}}{\text{m}^2\cdot\text{s}\cdot\text{Pa}} = \frac{\text{m}^2}{\text{s}\cdot\text{Pa}} \text{ or } \frac{\text{kg}\cdot\text{m}}{\text{m}^2\cdot\text{s}\cdot\text{Pa}}}$.
• $D$: Diffusivity. Kinetic mobility of gas molecules hopping between polymer free-volume cavities ($\mathbf{\text{m}^2/\text{s}}$).
• $S$: Solubility Coefficient. Thermodynamic partition coefficient relating dissolved gas concentration to external partial pressure via Henry's law: $C = S \cdot p$ ($\mathbf{\text{kg}/(\text{m}^3\cdot\text{Pa})}$).

• Amorphous Regions: Polymer chains are randomly coiled with loose, disordered packing and substantial fractional free volume ($V_f \sim 2.5\text{--}10\%$). Gas molecules easily squeeze through fluctuating free-volume gaps.
• Crystalline Lamellae: Polymer chains are folded into tightly packed, ordered crystallites with negligible free volume. Gas molecules cannot penetrate crystalline lamellae ($D_{\text{cryst}} \approx 0$).
• Tortuosity Effect: Crystalline lamellae act as impenetrable obstacles, forcing diffusing gas molecules to follow long, winding, convoluted paths around crystallites. Permeability drops steeply as crystallinity fraction $X_c$ increases: $$P = P_{\text{amorphous}} \left(\frac{1 - X_c}{\tau}\right)$$

Given $L = 0.40 \text{ mm} = 4.0 \times 10^{-4} \text{ m}, A = 0.050 \text{ m}^2, \Delta p = 3.5 \times 10^5 \text{ Pa}$, and $P = 1.20 \times 10^{-17} \text{ m}^2/(\text{s}\cdot\text{Pa})$:

Volumetric flux of gas (at standard temperature and pressure STP, $V_{\text{mol}} = 0.022414 \text{ m}^3/\text{mol}$):

$$J_v = \frac{P \Delta p}{L} = \frac{(1.20 \times 10^{-17} \text{ m}^2/\text{s}\cdot\text{Pa}) \times (3.5 \times 10^5 \text{ Pa})}{4.0 \times 10^{-4} \text{ m}} = \frac{4.20 \times 10^{-12}}{4.0 \times 10^{-4}} = 1.050 \times 10^{-8} \text{ m}^3\text{(STP)}/(\text{m}^2\cdot\text{s})$$

Total volumetric gas flow rate through bottle area $A = 0.050 \text{ m}^2$:

$$\dot{V} = J_v \times A = (1.050 \times 10^{-8}) \times 0.050 = 5.250 \times 10^{-10} \text{ m}^3\text{(STP)/s}$$

Molar flow rate ($\dot{n} = \dot{V} / V_{\text{mol}}$):

$$\dot{n} = \frac{5.250 \times 10^{-10} \text{ m}^3/\text{s}}{0.022414 \text{ m}^3/\text{mol}} = 2.3423 \times 10^{-8} \text{ mol/s}$$

Mass loss rate in grams per second ($M_{\text{CO}_2} = 44.01 \text{ g/mol}$):

$$\dot{m} = \dot{n} \times 44.01 \text{ g/mol} = (2.3423 \times 10^{-8}) \times 44.01 = 1.0309 \times 10^{-6} \text{ g/s}$$

Loss in one month ($t = 30 \text{ days} = 30 \times 86,400 = 2,592,000 \text{ s}$):

$$m_{\text{month}} = (1.0309 \times 10^{-6} \text{ g/s}) \times (2,592,000 \text{ s}) = \mathbf{2.672 \text{ g/month}}$$

A typical 2-liter soda bottle contains about $8 \text{ g}$ of $\text{CO}_2$. Losing $\sim 2.7 \text{ g/month}$ explains why plastic soda bottles go flat after 2 to 3 months of shelf storage, whereas glass and aluminum cans maintain carbonation indefinitely.

Fine powders possess an immense total surface area $A_{\text{powder}}$. The fundamental thermodynamic driving force for sintering is the reduction in total interfacial excess Gibbs free energy achieved by replacing high-energy solid-vapor external surfaces ($\gamma_{sv} \approx 1.0\text{--}2.0 \text{ J/m}^2$) with lower-energy grain boundaries ($\gamma_{\text{gb}} \approx 0.3\text{--}0.6 \text{ J/m}^2$) and ultimately eliminating pores entirely:

$$\Delta G_{\text{sintering}} = \Delta (\gamma A) = \gamma_{\text{gb}} A_{\text{gb}} - \gamma_{sv} A_{sv} < 0$$

The neck growth kinetics depend on the atomic pathway connecting the source of matter to the neck sink:

• Volume / Lattice Diffusion from Boundary ($n = 5, m = 3$): Matter diffuses from the interparticle grain boundary through the bulk crystal to the neck surface ($x^5 \propto t$).
• Grain Boundary Diffusion ($n = 6, m = 4$): Matter diffuses along the interparticle grain boundary core to the neck ($x^6 \propto t$).
• Surface Diffusion ($n = 7, m = 4$): Atoms migrate from the convex particle surface along the exterior surface into the concave neck ($x^7 \propto t$).

• Densifying Mechanisms (Grain Boundary & Volume Diffusion from Boundary): The source of matter is the grain boundary between the two particle centers. Removing atoms from the interparticle contact zone allows the centers of the two particles to approach each other ($\Delta L / L_0 < 0$). This brings about macroscopic shrinkage, pore elimination, and bulk density increase.
• Non-Densifying Mechanisms (Surface Diffusion & Evaporation-Condensation): The source of matter is the convex outer surface of the particles. Matter merely redistributes from the convex particle surface to the concave neck. The centers of the two spheres do not move closer together ($\Delta L = 0$); the compact gains mechanical neck strength, but the total volume of pores remains unchanged and bulk density does not increase.

From Fick's second law: $\frac{\partial C}{\partial t} = D \frac{\partial^2 C}{\partial x^2}$.

Assume a separation of variables solution: $C(x,t) - \bar{C} = X(x) \cdot T(t) = \Delta C_0 \sin\left(\frac{2\pi x}{\lambda}\right) e^{-t/\tau}$.

Compute spatial derivative:

$$\frac{\partial^2 C}{\partial x^2} = -\left(\frac{2\pi}{\lambda}\right)^2 \Delta C_0 \sin\left(\frac{2\pi x}{\lambda}\right) e^{-t/\tau} = -\frac{4\pi^2}{\lambda^2} [C(x,t) - \bar{C}]$$

Compute time derivative:

$$\frac{\partial C}{\partial t} = -\frac{1}{\tau} [C(x,t) - \bar{C}]$$

Equating $\frac{\partial C}{\partial t} = D \frac{\partial^2 C}{\partial x^2}$:

$$-\frac{1}{\tau} [C - \bar{C}] = -D \frac{4\pi^2}{\lambda^2} [C - \bar{C}] \implies \frac{1}{\tau} = \frac{4\pi^2 D}{\lambda^2} \implies \tau_{\text{hom}} = \frac{\lambda^2}{4\pi^2 D}$$

$$\lambda^2 = (5.0 \times 10^{-5} \text{ m})^2 = 2.50 \times 10^{-9} \text{ m}^2$$ $$\tau_{\text{hom}} = \frac{2.50 \times 10^{-9} \text{ m}^2}{4\pi^2 \times (2.0 \times 10^{-14} \text{ m}^2/\text{s})} = \frac{2.50 \times 10^{-9}}{7.8957 \times 10^{-13}} = 3166.3 \text{ seconds}$$ $$\tau_{\text{hom}} = \frac{3166.3 \text{ s}}{3600 \text{ s/hr}} = \mathbf{0.8795 \text{ hours} \approx 52.8 \text{ minutes}}$$

$$\frac{\Delta C(t)}{\Delta C_0} = \exp\left(-\frac{t}{\tau_{\text{hom}}}\right) = 0.05 \implies -\frac{t}{\tau_{\text{hom}}} = \ln(0.05) = -2.9957$$ $$t = 2.9957 \times \tau_{\text{hom}} \approx 3.0 \times \tau_{\text{hom}}$$ $$t = 2.9957 \times (0.8795 \text{ hr}) = \mathbf{2.635 \text{ hours} \approx 2 \text{ hours and 38 minutes}}$$

Because homogenization time scales quadratically with dendrite arm spacing ($t \propto \lambda^2$):

$$\frac{t_2}{t_1} = \left(\frac{\lambda_2}{\lambda_1}\right)^2 = \left(\frac{200 \mu\text{m}}{50 \mu\text{m}}\right)^2 = 4^2 = \mathbf{16.0}$$ $$t_2 = 16.0 \times 2.635 \text{ hr} = \mathbf{42.16 \text{ hours}}$$

A $4\times$ increase in dendrite spacing requires multiplying the soaking furnace time by 16 times (nearly 2 full days of soaking at $1200^\circ\text{C}$), demonstrating why fine as-cast dendrite structures are essential in continuous casting.

1. The dissolved oxygen permeability in solvent $A$ must exceed the solute diffusion product: $N_O^{(s)} D_O \gg N_B^{(0)} D_B$ (oxygen diffuses inward much faster than solute $B$ can diffuse outward).
2. The oxide $BO_v$ must have high thermodynamic stability (very negative free energy of formation $\Delta G_f^\circ$).
3. The solute concentration $N_B^{(0)}$ must be dilute (below the critical transition concentration $N_B^* \sim 15\text{--}25 \text{ at}\%$) so that discrete isolated precipitates form rather than a continuous protective external passivation scale.

Given $N_O^{(s)} D_O = 1.20 \times 10^{-14} \text{ m}^2/\text{s}, v = 1.5$ (for $\text{AlO}_{1.5} \equiv \text{Al}_2\text{O}_3$), $N_{\text{Al}}^{(0)} = 0.0050$:

$$v N_{\text{Al}}^{(0)} = 1.5 \times 0.0050 = 0.0075$$ $$\frac{2 N_O^{(s)} D_O}{v N_{\text{Al}}^{(0)}} = \frac{2 \times (1.20 \times 10^{-14} \text{ m}^2/\text{s})}{0.0075} = \frac{2.40 \times 10^{-14}}{0.0075} = 3.20 \times 10^{-12} \text{ m}^2/\text{s}$$

Compute oxidation front depth $X(t)$:

$$X^2 = (3.20 \times 10^{-12} \text{ m}^2/\text{s}) \times (14,400 \text{ s}) = 4.608 \times 10^{-8} \text{ m}^2$$ $$X = \sqrt{4.608 \times 10^{-8} \text{ m}^2} = 2.1466 \times 10^{-4} \text{ m} = \mathbf{0.2147 \text{ mm} = 214.7 \mu\text{m}}$$

Hydrogen is the smallest atom ($r_{\text{H}} \approx 0.037 \text{ nm}$), residing in the open tetrahedral interstitial sites of BCC ferrite with an ultra-low migration barrier ($Q_m \approx 7\text{ kJ/mol}$). However, microstructural defects (non-metallic inclusions, coherent carbide interfaces, grain boundaries, and the tensile hydrostatic stress fields of dislocation cores) act as deep potential energy wells (traps) with binding energies $E_b \approx 20\text{--}80 \text{ kJ/mol}$. Trapped hydrogen atoms cannot jump until thermally activated out of the trap, reducing apparent room-temperature diffusivity by several orders of magnitude.

At $190\text{--}200^\circ\text{C}$ ($463\text{--}473 \text{ K}$), thermal energy $k_B T$ is sufficient to overcome the trap binding energy ($E_b$), de-trapping hydrogen atoms back into mobile interstitial solid solution. Once de-trapped, hydrogen diffuses rapidly through the matrix to the external surface and desorbs into the atmosphere as harmless $\text{H}_2$ gas before it can segregate to stress concentration points to nucleate microcracks.

$$R^2 = (6.0 \times 10^{-3} \text{ m})^2 = 3.60 \times 10^{-5} \text{ m}^2$$ $$t_{\text{degas}} \approx \frac{R^2}{4 D_{\text{eff}}} = \frac{3.60 \times 10^{-5} \text{ m}^2}{4 \times (2.50 \times 10^{-10} \text{ m}^2/\text{s})} = \frac{3.60 \times 10^{-5}}{1.00 \times 10^{-9}} = 36,000 \text{ seconds}$$

Convert to hours:

$$t_{\text{degas}} = \frac{36,000 \text{ s}}{3600 \text{ s/hr}} = \mathbf{10.0 \text{ hours}}$$

Standard aerospace specifications mandate baking for 18 to 24 hours, providing a generous safety factor to ensure complete hydrogen exhaustion.

Copper is an ultra-fast interstitial diffuser in silicon. If copper directly contacts silicon during processing or operation ($100\text{--}400^\circ\text{C}$):

1. Copper reacts aggressively with silicon at low temperatures ($200^\circ\text{C}$) to form resistive Copper Silicide ($\text{Cu}_3\text{Si}$), causing colossal volumetric expansion that shatters dielectric layers.
2. Dissolved copper atoms introduce deep-level electronic trap states right in the center of the silicon energy bandgap ($E_v + 0.52 \text{ eV}$). These states act as lethal generation-recombination centers, annihilating transistor minority carrier lifetimes and producing catastrophic leakage currents that destroy the microprocessor.

• Ultra-low copper diffusivity ($D < 10^{-18} \text{ cm}^2/\text{s}$).
• Thermodynamically inert and immiscible with both Cu and Si (no intermetallic formation).
• Amorphous or nanocrystalline microstructure (absence of high-diffusivity grain boundaries).
• Low electrical resistivity and excellent adhesion.

$$x^2 = (3.0 \times 10^{-7} \text{ cm})^2 = 9.0 \times 10^{-14} \text{ cm}^2$$

• In Bare Silicon ($D = 1.0 \times 10^{-9} \text{ cm}^2/\text{s}$): $$t_{\text{Si}} \approx \frac{x^2}{2 D} = \frac{9.0 \times 10^{-14} \text{ cm}^2}{2 \times (1.0 \times 10^{-9} \text{ cm}^2/\text{s})} = \mathbf{4.5 \times 10^{-5} \text{ seconds} = 45 \mu\text{s}}$$ Without a barrier, copper penetrates the entire layer in 45 microseconds!
• In Amorphous $\text{TaN}$ Barrier ($D = 1.0 \times 10^{-18} \text{ cm}^2/\text{s}$): $$t_{\text{TaN}} \approx \frac{9.0 \times 10^{-14} \text{ cm}^2}{2 \times (1.0 \times 10^{-18} \text{ cm}^2/\text{s})} = 45,000 \text{ seconds} = \mathbf{12.5 \text{ hours}}$$ At normal chip operating temperatures ($75^\circ\text{C}$), $D_{\text{TaN}}$ plunges to $< 10^{-25} \text{ cm}^2/\text{s}$, extending barrier breakthrough lifetime to over 100 years.

For an intermetallic layer to grow, copper and tin atoms must diffuse across the existing intermetallic layer of thickness $x$. By Fick's first law, the flux of reactants arriving at the growth interface is inversely proportional to layer thickness: $J \propto 1/x$. Since the growth velocity is proportional to flux:

$$\frac{dx}{dt} \propto J \implies \frac{dx}{dt} = \frac{k}{2x} \implies 2x dx = k dt \implies x^2 = k t \implies x(t) = \sqrt{k t}$$

• After $t_1 = 100 \text{ hours} = 360,000 \text{ s}$: $$x_1 = \sqrt{(4.0 \times 10^{-16} \text{ m}^2/\text{s}) \times (360,000 \text{ s})} = \sqrt{1.440 \times 10^{-10} \text{ m}^2} = 1.20 \times 10^{-5} \text{ m} = \mathbf{12.0 \mu\text{m}}$$
• After $t_2 = 1,000 \text{ hours} = 3,600,000 \text{ s}$: $$x_2 = \sqrt{(4.0 \times 10^{-16} \text{ m}^2/\text{s}) \times (3,600,000 \text{ s})} = \sqrt{1.440 \times 10^{-9} \text{ m}^2} = 3.795 \times 10^{-5} \text{ m} = \mathbf{37.95 \mu\text{m}}$$

Intermetallic compounds ($\text{Cu}_6\text{Sn}_5, \text{Cu}_3\text{Sn}$) are brittle intermetallic phases with covalent-ionic bonding characteristics, high hardness, and very low fracture toughness ($K_{Ic} \approx 1\text{--}2 \text{ MPa}\sqrt{\text{m}}$). Furthermore, Kirkendall microvoids coalesce at the $\text{Cu}_3\text{Sn}/\text{Cu}$ interface due to unequal diffusion rates. When the layer thickens beyond $\sim 5 \mu\text{m}$, thermal mismatch stresses and Kirkendall void planes cause catastrophic brittle cleavage failure along the IMC interface during mobile phone drop-impact tests.

1. Melting of Interlayer: The assembly is heated to a bonding temperature $T_B$ above the melting point of the thin interlayer foil ($25\text{--}50 \mu\text{m}$ thick) but well below the melting point of the parent base superalloy. The interlayer melts and wets the joint interface.
2. Base Metal Dissolution: The liquid dissolves a thin layer of the solid substrate until thermodynamic solid-liquid phase equilibrium is established at the interface.
3. Isothermal Solidification: The melting point depressant (boron) diffuses out of the liquid joint into the vast solid base metal. As the liquid loses boron, its liquidus temperature rises above $T_B$, causing the solid-liquid interface to advance inward and solidify the entire joint at constant temperature.
4. Solid-State Homogenization: A post-bond heat treatment diffuses residual boron deeply into the matrix, eliminating brittle borides and matching parent single-crystal properties.

Nickel melts at $1455^\circ\text{C}$. Adding small interstitial boron atoms ($2\text{--}3\text{ wt}\%$) forms a low-melting eutectic at $1040^\circ\text{C}$. This allows the joint to liquefy at $T_B \approx 1150^\circ\text{C}$ without damaging the surrounding parent superalloy ($T_m \approx 1350^\circ\text{C}$). Furthermore, because boron is an interstitial atom, it diffuses rapidly out of the liquid into the solid base metal.

Solidification normally requires cooling ($\Delta T < 0$). In TLP bonding, solidification is driven chemically by compositional change rather than temperature change. As boron diffuses away into the solid, the liquid becomes boron-depleted pure nickel-base alloy. Looking at the phase diagram, shifting composition to lower solute increases the melting point above the furnace temperature $T_B$. The joint solidifies completely under strictly isothermal conditions.

The rate-controlling step is the solid-state diffusion of boron into the base metal. Following Fick's second law for a moving boundary, the distance the solid-liquid interface moves scales with $\sqrt{Dt}$:

$$w = 2 \gamma \sqrt{D t_{\text{iso}}} \implies t_{\text{iso}} = \frac{w^2}{4 \gamma^2 D} \propto \frac{w^2}{D}$$

where $w$ is the maximum half-width of the liquid zone and $\gamma$ is a dimensionless growth parameter determined by phase diagram partition coefficients. Halving the joint gap reduces the required furnace time by a factor of 4.

• Direct Electrostatic Force ($\mathbf{F}_{\text{direct}} = Z e \mathbf{E}$): The electrostatic attraction pulling the positively charged metal ion toward the negative cathode.
• Electron Wind Force ($\mathbf{F}_{\text{wind}} = -e n_e \sigma_{\text{tr}} \lambda_e \mathbf{E}$): High-velocity conduction electrons colliding with activated metal ions at the saddle-point of their diffusion jump impart ballistic momentum, pushing the metal ions toward the positive anode.
• In metals with high electron conductivities, the electron wind force is $10\text{--}100$ times stronger than the direct electrostatic force! Metal atoms migrate in the direction of electron flow (from cathode to anode), leaving voids at the cathode and building hillocks/whiskers at the anode.

The net effective charge is defined as $Z^* = Z_{\text{direct}} + Z_{\text{wind}}$ (typically negative, $Z^* \approx -1\text{ to }-10$):

$$J_{\text{EM}} = \frac{C D}{k_B T} Z^* e E = \frac{C D}{k_B T} Z^* e (\rho_e j)$$

In 1969, J. R. Black formulated the standard reliability lifetime equation:

$$\text{MTTF} = \frac{A}{j^n} \exp\left(\frac{E_a}{k_B T}\right) \quad (n \approx 1.5\text{--}2.0)$$

Given $\text{MTTF}_1 = 10 \text{ years}$, $j_1 = 1.0 \times 10^6 \text{ A/cm}^2, T_1 = 378 \text{ K}$, $j_2 = 1.5 \times 10^6 \text{ A/cm}^2, T_2 = 398 \text{ K}$, $E_a = 0.90 \text{ eV}$, $n = 2$:

$$\frac{\text{MTTF}_2}{\text{MTTF}_1} = \left(\frac{j_1}{j_2}\right)^n \exp\left[\frac{E_a}{k_B}\left(\frac{1}{T_2} - \frac{1}{T_1}\right)\right]$$

• Current density factor: $$\left(\frac{j_1}{j_2}\right)^2 = \left(\frac{1.0}{1.5}\right)^2 = \left(\frac{2}{3}\right)^2 = \mathbf{0.4444}$$
• Thermal factor: $$\frac{1}{T_2} - \frac{1}{T_1} = \frac{1}{398} - \frac{1}{378} = 2.5126 \times 10^{-3} - 2.6455 \times 10^{-3} = -1.329 \times 10^{-4} \text{ K}^{-1}$$ $$\frac{E_a}{k_B} \Delta(1/T) = \frac{0.90 \text{ eV}}{8.617 \times 10^{-5} \text{ eV/K}} \times (-1.329 \times 10^{-4} \text{ K}^{-1}) = 10,444.5 \times (-1.329 \times 10^{-4}) = -1.388$$ $$\exp(-1.388) = \mathbf{0.2496}$$

Combined lifetime ratio:

$$\frac{\text{MTTF}_2}{\text{MTTF}_1} = 0.4444 \times 0.2496 = \mathbf{0.1109}$$ $$\text{MTTF}_2 = 10 \text{ years} \times 0.1109 = \mathbf{1.11 \text{ years}}$$

A $50\%$ increase in current density combined with a $20^\circ\text{C}$ temperature rise slashes the microprocessor lifetime from 10 years down to just 1.1 years (an $89\%$ reduction in reliability!).

$$J_{\text{TM}} = -\frac{C D Q^*}{R T^2} \frac{dT}{dx}$$

The heat of transport $Q^*$ represents the net thermal energy carried by an atom per mole during an atomic jump, minus the enthalpy of the atom on its lattice site:

• If $Q^* > 0$ (Positive Heat of Transport): The diffusing atom transports excess heat forward. Energetically, jumping toward the colder side lowers free energy: atoms migrate down the thermal gradient from HOT to COLD.
• If $Q^* < 0$ (Negative Heat of Transport): Atoms migrate up the thermal gradient from COLD to HOT (common for interstitial carbon and nitrogen in certain refractory metals).

In operating $\text{UO}_2$ nuclear reactor fuel rods, the thermal gradient is immense ($\sim 5 \times 10^5 \text{ K/m}$). Pores and bubbles migrate rapidly up the thermal gradient toward the hot center via an evaporation-condensation mechanism (vaporizing at the hot side of the pore and condensing on the cold side). As millions of pores migrate inward, they coalesce at the cylinder axis, forming a hollow, continuous central void tunnel surrounded by columnar grain structures.

In high-power semiconductor packaging, Joule heating concentrates at tiny contact corners where current enters the solder bump, generating steep localized temperature gradients ($\nabla T > 1000^\circ\text{C/cm}$ over a $50 \mu\text{m}$ bump). Under this driving force, tin and lead atoms migrate rapidly from the hot silicon chip side toward the colder organic substrate side. This leaves behind a dense band of coalesced vacancy voids at the hot interface, culminating in sudden electrical open-circuit failure.

Given $C_0 = 0.20 \text{ wt}\%, C_s = 1.20 \text{ wt}\%, D_1 = 1.28 \times 10^{-11} \text{ m}^2/\text{s}, x = 1.00 \text{ mm} = 1.0 \times 10^{-3} \text{ m}$:

$$2\sqrt{D_1 t_1} = 2 \sqrt{(1.28 \times 10^{-11} \text{ m}^2/\text{s}) \times (14,400 \text{ s})} = 2 \sqrt{1.8432 \times 10^{-7} \text{ m}^2} = 2 \times (4.2932 \times 10^{-4} \text{ m}) = 8.5865 \times 10^{-4} \text{ m}$$ $$z = \frac{x}{2\sqrt{D_1 t_1}} = \frac{1.0 \times 10^{-3} \text{ m}}{8.5865 \times 10^{-4} \text{ m}} = \mathbf{1.1646}$$

Using standard error function values: $\text{erf}(1.10) = 0.8802, \text{erf}(1.20) = 0.9103$:

$$\text{erf}(1.1646) = 0.8802 + \left(\frac{1.1646 - 1.10}{0.10}\right) \times (0.9103 - 0.8802) = 0.8802 + 0.646 \times (0.0301) = 0.8802 + 0.0194 = \mathbf{0.8996}$$ $$\frac{C(1.00 \text{ mm}) - 0.20}{1.20 - 0.20} = 1 - \text{erf}(z) = 1 - 0.8996 = 0.1004$$ $$C(1.00 \text{ mm}) = 0.20 + 0.1004(1.00) = \mathbf{0.300 \text{ wt}\%}$$

Specification Check: At $1.00 \text{ mm}$, the carbon content is only $0.30 \text{ wt}\%$, which falls short of the required $0.40 \text{ wt}\%$. The specification is not met by Stage 1 alone; additional diffuse time is mandatory.

Using the integrated mass uptake formula from Problem 25:

$$\Delta C = (1.20 - 0.20 \text{ wt}\%) \times 7800 \text{ kg/m}^3 = 78.0 \text{ kg/m}^3$$ $$M_t = 2 \Delta C \sqrt{\frac{D_1 t_1}{\pi}} = 2 \times (78.0) \times \sqrt{\frac{1.8432 \times 10^{-7} \text{ m}^2}{\pi}} = 156.0 \times \sqrt{5.867 \times 10^{-8}} = 156.0 \times (2.422 \times 10^{-4} \text{ m})$$ $$M_t = 0.03779 \text{ kg/m}^2 = \mathbf{37.79 \text{ g/m}^2}$$

If the gear were quenched directly with $1.20 \text{ wt}\%$ surface carbon:

1. High dissolved carbon depresses the martensite finish temperature ($M_f$) far below room temperature ($M_f \approx -50^\circ\text{C}$). Upon oil quenching to $25^\circ\text{C}$, the transformation cannot complete, leaving $30\text{--}50\%$ soft, ductile retained austenite at the surface. Retained austenite degrades surface hardness from 62 HRC down to $< 52 \text{ HRC}$ and causes catastrophic contact fatigue spalling.
2. Excessive carbon causes continuous proeutectoid cementite ($\text{Fe}_3\text{C}$) networks to precipitate along prior austenite grain boundaries during cooling, leading to severe embrittlement.
3. The Diffuse stage lowers surface carbon to the eutectoid composition ($0.80\text{--}0.85\text{ wt}\%$), raising $M_s/M_f$ to ensure $100\%$ transformation to hard plate martensite while driving carbon deeper to satisfy the $1.00 \text{ mm}$ case depth specification.

During the rapid quench, carbon diffusion is kinetically frozen in the bulk lattice, trapping carbon into octahedral interstitial sites to force the diffusionless FCC $\to$ BCT martensite transformation. However, in the lower-carbon core ($0.20 \text{ wt}\%$, high $M_s \approx 420^\circ\text{C}$), the transformation begins earlier, forming high-toughness lath martensite and bainite. The resulting composite component possesses a wear-resistant, high-hardness case ($62 \text{ HRC}$) under compressive residual stress, supported by a ductile, fracture-resistant core ($32 \text{ HRC}$).