RF power in semiconductor processing is not a dial that makes plasma "stronger"; it is an impedance-matching problem in which a 50 $\Omega$ generator must deliver energy into a plasma whose impedance is typically 2–10 $\Omega$ resistive with a reactive component that drifts as chemistry, pressure, and density change. Without a matching network, a 5 $\Omega$ plasma load presents a reflection coefficient $|\Gamma| = |(Z_L - Z_0)/(Z_L + Z_0)| = 0.818$, and the fraction of power actually delivered is $1 - |\Gamma|^2 = 33.1\%$ — two-thirds of the generator's output bounces back. Add 50 $\Omega$ of uncompensated reactance (a match that has drifted out of tune) and efficiency falls to 18.1%. The matching network's job is to present the conjugate of the plasma impedance to the generator, driving $\Gamma$ to zero and delivered power to 100% of forward power. This is not a support system; it is the mechanism that determines how much energy the plasma receives.

The matching network is an L-section whose component values are set entirely by the plasma resistance and the operating frequency. For a plasma resistance $R_p < Z_0$, the required quality factor is $Q = \sqrt{Z_0/R_p - 1}$, the series reactance is $X_s = Q \cdot R_p$ (an inductor), and the shunt reactance is $X_p = Z_0/Q$ (a capacitor). At 13.56 MHz with $R_p = 5$ $\Omega$: $Q = 3.0$, $L$ = 176 nH, $C$ = 704 pF. At $R_p = 2$ $\Omega$ (denser plasma, lower resistance): $Q = 4.9$, $L = 115$ nH, $C = 1{,}150$ pF. The capacitors in a production match are motor-driven vacuum variables that re-tune in 0.1–1 s; the inductor is fixed. The speed of re-tuning determines how quickly the system recovers from a plasma impedance shift — which happens every time gas composition, pressure, or power changes.

Frequency determines ion energy, not etch rate, and the boundary is the ion plasma frequency $f_{pi}$. For argon at a density of $10^{10}$ cm$^{-3}$, $f_{pi} = (1/2\pi)\sqrt{n_e e^2/(m_i \varepsilon_0)} = 3.33$ MHz. Below $f_{pi}$, ions respond to the instantaneous sheath electric field and arrive at the wafer with an energy spread equal to the full RF voltage swing: at 400 kHz the ion energy distribution function (IEDF) spans 300 eV, producing broadband bombardment that damages the sidewall and erodes the mask. Above $f_{pi}$, ions cannot follow the oscillating field and see only the time-averaged sheath voltage: at 13.56 MHz the energy spread narrows to 18.1 eV, and at 60 MHz to 0.92 eV — nearly monoenergetic bombardment. This is the entire reason dual-frequency etch chambers exist: a high-frequency source (60 MHz or 27.12 MHz) generates plasma density while a low-frequency bias (400 kHz or 2 MHz) independently controls ion energy. Mixing the two roles on a single frequency forces a trade-off between density and bombardment energy that advanced nodes cannot afford.

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 760 470" font-family="Georgia,serif" font-size="11">
<rect x="0" y="0" width="760" height="470" fill="#0d1117"/>
<line x1="65" y1="415.0" x2="350" y2="415.0" stroke="#21262d" stroke-width="0.5"/>
<text x="60" y="415.0" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">0</text>
<line x1="65" y1="312.1" x2="350" y2="312.1" stroke="#21262d" stroke-width="0.5"/>
<text x="60" y="312.1" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">100</text>
<line x1="65" y1="209.3" x2="350" y2="209.3" stroke="#21262d" stroke-width="0.5"/>
<text x="60" y="209.3" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">200</text>
<line x1="65" y1="106.4" x2="350" y2="106.4" stroke="#21262d" stroke-width="0.5"/>
<text x="60" y="106.4" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">300</text>
<line x1="181.5" y1="55" x2="181.5" y2="415" stroke="#d2a8ff" stroke-width="1.5" stroke-dasharray="6,3"/>
<text x="184.5" y="67" fill="#d2a8ff" font-size="10">f_ion = 3.3 MHz</text>
<line x1="65" y1="415" x2="350" y2="415" stroke="#8b98a5" stroke-width="1.5"/>
<line x1="65" y1="55" x2="65" y2="415" stroke="#8b98a5" stroke-width="1.5"/>
<text x="76.6" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">0.4</text>
<text x="156.3" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">2</text>
<text x="251.1" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">13.56</text>
<text x="324.7" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">60</text>
<text x="207.5" y="450" text-anchor="middle" fill="#e6edf3" font-size="11">RF frequency (MHz)</text>
<text x="25" y="235.0" text-anchor="middle" fill="#e6edf3" font-size="11" transform="rotate(-90,25,235.0)">Ion energy spread (eV)</text>
<text x="207.5" y="35" text-anchor="middle" fill="#e6edf3" font-size="12" font-weight="bold">IEDF narrows above ion plasma frequency</text>
<polyline points="62.4,106.4 87.7,106.4 111.0,106.4 122.0,106.4 142.1,106.4 156.3,106.4 176.4,106.4 190.6,201.5 201.7,278.3 218.3,345.3 236.0,380.8 251.1,396.4 270.3,406.5 290.4,411.2 304.6,412.9 324.7,414.1 339.0,414.5 350.0,414.7" fill="none" stroke="#3fb950" stroke-width="2.5"/>
<circle cx="76.6" cy="106.4" r="4" fill="#3fb950"/>
<text x="76.6" y="96.4" text-anchor="middle" fill="#3fb950" font-size="10">300 eV</text>
<circle cx="156.3" cy="106.4" r="4" fill="#3fb950"/>
<circle cx="251.1" cy="396.4" r="4" fill="#3fb950"/>
<text x="251.1" y="386.4" text-anchor="middle" fill="#3fb950" font-size="10">18 eV</text>
<circle cx="285.4" cy="410.4" r="4" fill="#3fb950"/>
<circle cx="324.7" cy="414.1" r="4" fill="#3fb950"/>
<text x="324.7" y="404.1" text-anchor="middle" fill="#3fb950" font-size="10">1 eV</text>
<text x="122.0" y="363.57142857142856" text-anchor="middle" fill="#d2a8ff" font-size="10" font-style="italic">Ions follow field</text>
<text x="304.6348390113877" y="363.57142857142856" text-anchor="middle" fill="#d2a8ff" font-size="10" font-style="italic">Ions see average</text>
<line x1="410" y1="370.0" x2="730" y2="370.0" stroke="#21262d" stroke-width="0.5"/>
<text x="405" y="370.0" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">10^9</text>
<line x1="410" y1="280.0" x2="730" y2="280.0" stroke="#21262d" stroke-width="0.5"/>
<text x="405" y="280.0" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">10^10</text>
<line x1="410" y1="190.0" x2="730" y2="190.0" stroke="#21262d" stroke-width="0.5"/>
<text x="405" y="190.0" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">10^11</text>
<line x1="410" y1="100.0" x2="730" y2="100.0" stroke="#21262d" stroke-width="0.5"/>
<text x="405" y="100.0" text-anchor="end" dominant-baseline="middle" fill="#e6edf3" font-size="10">10^12</text>
<line x1="410" y1="415" x2="730" y2="415" stroke="#8b98a5" stroke-width="1.5"/>
<line x1="410" y1="55" x2="410" y2="415" stroke="#8b98a5" stroke-width="1.5"/>
<text x="435.5" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">50</text>
<text x="474.0" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">100</text>
<text x="563.5" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">500</text>
<text x="602.0" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">1K</text>
<text x="691.5" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">5K</text>
<text x="730.0" y="430" text-anchor="middle" fill="#e6edf3" font-size="10">10K</text>
<text x="570.0" y="450" text-anchor="middle" fill="#e6edf3" font-size="11">RF power (W)</text>
<text x="365" y="235.0" text-anchor="middle" fill="#e6edf3" font-size="11" transform="rotate(-90,365,235.0)">Plasma density (cm^-3)</text>
<text x="570.0" y="35" text-anchor="middle" fill="#e6edf3" font-size="12" font-weight="bold">ICP delivers 40x the density of CCP</text>
<polyline points="435.5,253.2 474.0,226.1 512.5,199.0 563.5,163.2 602.0,136.1 640.5,109.0 691.5,73.2 730.0,46.1" fill="none" stroke="#58a6ff" stroke-width="2.5"/>
<circle cx="435.5" cy="253.2" r="3" fill="#58a6ff"/>
<circle cx="474.0" cy="226.1" r="3" fill="#58a6ff"/>
<circle cx="512.5" cy="199.0" r="3" fill="#58a6ff"/>
<circle cx="563.5" cy="163.2" r="3" fill="#58a6ff"/>
<circle cx="602.0" cy="136.1" r="3" fill="#58a6ff"/>
<circle cx="640.5" cy="109.0" r="3" fill="#58a6ff"/>
<circle cx="691.5" cy="73.2" r="3" fill="#58a6ff"/>
<circle cx="730.0" cy="46.1" r="3" fill="#58a6ff"/>
<polyline points="435.5,386.3 474.0,370.0 512.5,353.7 563.5,332.3 602.0,316.0 640.5,299.7 691.5,278.3 730.0,262.0" fill="none" stroke="#f85149" stroke-width="2.5"/>
<circle cx="435.5" cy="386.3" r="3" fill="#f85149"/>
<circle cx="474.0" cy="370.0" r="3" fill="#f85149"/>
<circle cx="512.5" cy="353.7" r="3" fill="#f85149"/>
<circle cx="563.5" cy="332.3" r="3" fill="#f85149"/>
<circle cx="602.0" cy="316.0" r="3" fill="#f85149"/>
<circle cx="640.5" cy="299.7" r="3" fill="#f85149"/>
<circle cx="691.5" cy="278.3" r="3" fill="#f85149"/>
<circle cx="730.0" cy="262.0" r="3" fill="#f85149"/>
<line x1="425" y1="70" x2="445" y2="70" stroke="#58a6ff" stroke-width="2.5"/>
<text x="450" y="74" fill="#58a6ff" font-size="10">ICP (n_e ~ P)</text>
<line x1="425" y1="88" x2="445" y2="88" stroke="#f85149" stroke-width="2.5"/>
<text x="450" y="92" fill="#f85149" font-size="10">CCP (n_e ~ P^0.6)</text>
<line x1="602.0" y1="141.1" x2="602.0" y2="311.0" stroke="#6e7681" stroke-width="1" stroke-dasharray="3,2"/>
<text x="607.0" y="226.0" fill="#6e7681" font-size="10">100x</text>
</svg>

ICP and CCP are not two names for the same thing; they differ by a factor of 40 in plasma density at the same power and scale with different exponents. In an inductively coupled plasma (ICP), the RF coil couples energy through a magnetic field, and the steady-state density follows $n_e \approx P_{\mathrm{abs}} / (A_{\mathrm{loss}} \cdot v_B \cdot \varepsilon_c)$ where $\varepsilon_c \approx 70$ eV is the energy cost per ion-electron pair and $v_B$ is the Bohm velocity. The relationship is linear: at 100 W, $n_e = 3.97 \times 10^{10}$ cm$^{-3}$; at 1 kW, $3.97 \times 10^{11}$ cm$^{-3}$. In a capacitively coupled plasma (CCP), the RF field couples through the sheath and the scaling is sublinear ($n_e \propto P^{0.6}$): at 100 W, $n_e = 1.00 \times 10^{9}$ cm$^{-3}$; at 1 kW, $3.98 \times 10^{9}$ cm$^{-3}$. The ratio at 1 kW is 100:1. The skin depth — $\delta = c/\omega_{pe}$ — determines which regime applies: at $10^{10}$ cm$^{-3}$, $\delta = 5.32$ cm, larger than the electrode gap, so the RF field fills the chamber (CCP). At $10^{12}$ cm$^{-3}$, $\delta = 0.053$ cm, and the field is confined to a thin layer (ICP). The plasma frequency at $10^{10}$ cm$^{-3}$ is 0.90 GHz, well above the 13.56 MHz excitation, confirming that the plasma is overdense and the RF cannot propagate through it — it can only couple at the surface.

The self-bias voltage in a CCP is not a design parameter; it is a consequence of the area asymmetry between the powered and grounded electrodes. Because electrons are much more mobile than ions, during each RF cycle the powered electrode collects excess electrons and charges negatively. In steady state, $V_{dc} \approx -(V_{pp}/2) \cdot (A_g/A_p)^q$ where $q \approx 1.5$ (Koenig-Maissel). For symmetric electrodes ($A_g/A_p = 1$) at $V_{pp} = 600$ V, $V_{dc} = -300$ V. At an area ratio of 3, $V_{dc} = -600$ V (clamped at $-V_{pp}$). This is why production CCP chambers use a small powered electrode and a large grounded chamber wall — the asymmetry concentrates the voltage drop at the wafer, maximising ion energy without increasing the total RF power. Conversely, an ICP source is inherently symmetric (the coil does not collect current), which is why ICP provides high density with low self-bias: the ion energy in an ICP is set by a separate bias supply, decoupled from the density-generating source power.

Of 1,000 W leaving the RF generator, only 951 W reaches the plasma, and most of that goes into heating the chamber walls, not etching the wafer. The matching network dissipates 30 W (3%) in resistive losses through its inductor and capacitor contacts. The cable and vacuum feedthrough dissipate another 19 W (2%). Of the 951 W absorbed by the plasma, 570 W (60%) accelerates ions into surfaces — predominantly the chamber walls and the wafer — producing the physical bombardment that drives anisotropic etch. Another 143 W (15%) heats neutral gas molecules through elastic collisions, 95 W (10%) is radiated as VUV photons, and 143 W (15%) is carried to the walls by electrons. The wafer receives roughly 70% of the ion power: about 399 W of thermal load from a 1 kW setpoint. This is why electrostatic chuck backside helium cooling is not optional — without it, a 300 mm wafer under 1 kW of bias power would rise by several hundred degrees in seconds.

Pulsed RF replaces the CW plasma's fixed electron temperature with a two-state cycle that separates radical generation from surface reaction. During the ON phase ($T_e \approx 3.0$ eV), the plasma generates radicals and ions at the CW rate. During the OFF phase, electron temperature drops to approximately 0.3 eV within 5 $\mu$s (the electron energy relaxation time in argon), ion bombardment ceases, and only thermal radicals reach the surface. At 10% duty cycle with 1,000 W peak power, the time-averaged power is 100 W and the time-averaged electron temperature is 0.57 eV; at 50% duty cycle, 500 W and 1.65 eV. The process consequence is selectivity: during the OFF phase, the etch proceeds isotropically via radical chemistry alone, which preferentially attacks the target material (whose bonds the radicals were chosen to break) over the mask or underlayer. Pulsed RF at 20–50% duty cycle is now standard in high-aspect-ratio dielectric etch, where CW operation produces notching, bowing, and mask erosion that pulsing eliminates by giving the accumulated charge on insulating surfaces time to dissipate between bursts.

The 13.56 MHz standard is not an optimal frequency for plasma processing; it is the ISM-band allocation that avoids interfering with communications, and every other standard frequency in the industry is a harmonic or subharmonic of it. 27.12 MHz is the second harmonic; 2 MHz and 400 kHz are chosen for low-frequency bias because they sit below the ion plasma frequency of typical process plasmas. 60 MHz (used in advanced CCP sources) is not a harmonic of 13.56 MHz — it was selected because higher source frequency at fixed power produces higher plasma density (the electron heating efficiency scales with $\omega$ in the ohmic regime) while pushing the bias function to a separate, lower frequency. The trend toward VHF sources (60–200 MHz) is ultimately a density argument: at constant power, $n_e \propto \omega^{0.5}$ in the stochastic heating regime, so doubling the source frequency gains 41% in density. But VHF introduces standing-wave non-uniformity across the electrode — at 60 MHz the free-space wavelength is 5 m and the electrode-sheath-plasma structure compresses it to roughly 1.5 m, comparable to the 300 mm wafer diameter, creating a centre-to-edge density variation that requires phase-shifted multi-zone feeding or Gaussian-profile electrode shaping to correct.

Through the lens of the matching network, RF power is a transmission-line problem, not a plasma-physics problem, and every process excursion that changes the plasma's complex impedance is also a reflected-power excursion that the match must chase. A recipe transition from high-pressure polysilicon etch ($R_p \approx 8$ $\Omega$, $X_p \approx -30$ $\Omega$) to low-pressure oxide etch ($R_p \approx 3$ $\Omega$, $X_p \approx -80$ $\Omega$) swings $|\Gamma|$ from 0.72 to 0.93 if the match does not re-tune — reflected power jumps from 52% to 86% of forward power. The 0.1–1 s re-tune window of a motor-driven match is the limiting factor in step transitions. Solid-state matches (frequency-tuning or electronically switched impedance networks) reduce this to microseconds, and are increasingly adopted for pulsed-RF processes where the plasma impedance oscillates at the pulse repetition rate and a motor-driven match cannot follow.