DC and RF Magnetron Sputter System

1. Theory

• Definition: A sputter system bombards a target with ions from a glow discharge so that target atoms are knocked out and deposit on the substrate (Chen; SPECS).
• Sputter yield: Atoms ejected per incident ion, about 0.5 to 3 at 500 eV Ar${}^{+}$, setting the rate through $z/t=MS{j}_p/\rho N_{A}e$ (SPECS).
• Energetic arrival: Sputtered atoms leave at 10 to 50 eV, about 100 times evaporated atoms, and shadow less (Chen).
• Alloys: The target surface enriches in the low-yield element until the film matches the target (TU Wien).
• RF self-bias: With a blocking capacitor, mobile electrons charge the target negative, so insulating targets can be sputtered at 13.56 MHz (CNR-IFN).
• Reactive sputtering: Reactive gas poisons the target in an avalanche once inflow exceeds the getter rate, and the compound sputters less efficiently (De Gryse et al.).

2. Working Principle

• DC diode: Conducting targets at 1 to 10 mTorr, deposition at 1 to 10 nm/s (Chen).
• Magnetron: Magnets trap electrons above the target, intensifying the plasma and allowing operation down to $10^{-5}$ to $10^{-3}$ Torr (Chen; CNR-IFN).
• RF: A 13.56 MHz supply through a blocking capacitor; the large grounded area keeps the substrate from sputtering (CNR-IFN; NPTEL).
• Plasma diagnostics: Discharge density and potentials can be measured with the Langmuir Probe page methods (my reading).

3. Applications

• Interconnect and via metal: Sputtering’s smaller shadowing gives the step coverage that vias and multilayer metal need (Chen).
• Dielectric and oxide films: SiO${}_2$, ZnO and TiO${}_2$ by RF sputtering, or Al${}_2$O${}_3$ by reactive sputtering (Chen; De Gryse et al.).
• Alloy films: Composition carried over from the target (TU Wien).

Yield, Rate and Energy

Name	Formula	Related Formulae	Variables/Constants
Sputter yield	$S=\frac{\text{atoms ejected}}{\text{incident ion}}$	$S_{Al}=0.95,S_{Si}=0.55,S_W=0.6,S_{Au}=2.0(0.5\mathrm{keV}A{r}^{+})$	$\begin{array}{r}SPECS:perpendicularincidence,at1.0\mathrm{keV}\\ Al1.8,Si0.90,W1.0,Au3.5;DeGryse\\ etal.:Al0.5fromSRIM,A{l}_2{O}_{3}0.03\\ measured\end{array}$ $\begin{array}{r}(myreading:yieldroughlydoublesfrom\\ 0.5to1\mathrm{keV}formostlistedmetals)\end{array}$
Sputter rate	$\frac{z}{t}=\frac{MS{j}_p}{\rho N_{A}e}$	${\left(\frac{z}{t}\right)}_{Al}=630\mathring{\mathrm{A}}/min(500\mathrm{eV},1mA/c{\mathrm{m}}^2)$	$M:molarmass$ $\rho :density$ $\begin{array}{r}{j}_p:ioncurrentdensity;SPECS:Cu818,Au\\ 1553,Si{O}_{2}400,Si310\mathring{\mathrm{A}}/min\end{array}$ $\begin{array}{r}(computed:Alwithanassumed2.70g/c{\mathrm{m}}^3\\ gives591\mathring{\mathrm{A}}/min;invertingC{u}^{\prime }s818\mathring{\mathrm{A}}/min\\ withanassumed8.96g/c{\mathrm{m}}^{3}givesSabout\\ 1.9)\end{array}$
Sputtered-atom energy	$E_{sput}=10\text{to}50\mathrm{eV}$	$E_{sput}\approx 100E_{evap}$	$\begin{array}{r}Chen:ejectedatomsagainstevaporated\\ atoms;TUWien:targetsurfaceenrichesin\\ thelow-yieldcomponentuntilthedeposit\\ matchesthetarget\end{array}$ $\begin{array}{r}(computed:k_{B}Tatanassumed1500Kis0.13\\ \mathrm{eV})\end{array}$

DC and Magnetron Discharges

Name	Formula	Related Formulae	Variables/Constants
DC glow discharge	$p=1\text{to}10\mathrm{mTorr}$	$R_{dep}=1\text{to}10nm/s$	$\begin{array}{r}Chen:dischargestartedbyalargevoltage\\ acrossalow-pressuregap,reverse\\ polarityforback-sputtercleaning\end{array}$ $\begin{array}{r}(computed:1to10nm/sis10to100\\ angstrompersecond,thee-beamrangeon\\ theElectron-BeamEvaporatorpage)\end{array}$
Magnetron	$\vec{B}\text{traps electrons in spirals above the target}$	$p_{plasma}=10^{-5}\text{to}10^{-3}\mathrm{Torr}$	$\begin{array}{r}Chen:higherionizationletstheplasma\\ formatlowerpressure;CNR-IFN:field\\ concentratestheplasmaabovethetarget,\\ raisingratewithoutraisingpressure\end{array}$ $\begin{array}{r}(myreading:lowerpressurelengthensthe\\ meanfreepath,sosputteredatomskeep\\ theirenergy)\end{array}$
Step coverage	$\text{shadowing: evaporation large, sputtering small}$	$\text{conformal for vias, poor for lift-off}$	$\begin{array}{r}Chen:goodcoverageneededforviasand\\ multilayermetal,poorcoveragewantedfor\\ lift-off;ThermalEvaporatorpage\end{array}$ $\begin{array}{r}(myreading:gasscatteringatmTorr\\ pressuresbroadensthearrivalangles)\end{array}$

RF and Reactive Sputtering

Name	Formula	Related Formulae	Variables/Constants
RF self-bias	$f=13.56\mathrm{MHz},⟨I⟩=0$	$V_{RF}\sim 500\mathrm{V}$	$\begin{array}{r}CNR-IFN:blockingcapacitordevelopsthe\\ DCself-bias,mobileelectronschargethe\\ targetnegative,insulatingtargets\\ sputterbecausenetcurrentaveragesto\\ zero;Chen:RFforSi{O}_2,ZnO,Ti{O}_2\end{array}$ $\begin{array}{r}(myreading:theself-biasiswhatturns\\ theRFelectrodeintoasputtering\\ cathode)\end{array}$
Electrode-area law	$\frac{V_t}{V_d}={\left(\frac{A_d}{A_t}\right)}^4$	$A_d\gg A_t\Rightarrow V_d\text{small}$	$\begin{array}{r}{V}_t,V_d:sheathvoltagesattargetand\\ groundedelectrode\end{array}$ $\begin{array}{r}{A}_t,A_d:theirareas;NPTEL:fourth-power\\ dependence,groundedchamberandsubstrate\\ averylargeelectrodesolittlesubstrate\\ sputtering\end{array}$ $\begin{array}{r}(myreading:theformulaitselfwasan\\ imageontheNPTELpage;theexponent\\ followsthestatedfourth-power\\ dependence)\end{array}$
Reactive gas balance	$q_0=q_t+q_s+q_p$	$\frac{S_{Al}}{S_{A{l}_2{O}_3}}=\frac{0.5}{0.03}\approx 17$	$q_0:reactiveinflow$ $\begin{array}{r}{q}_t,q_s,q_p:consumptionattarget,\\ substrateandpump;DeGryseetal.:Berg\\ model,avalanchewheninflowexceedsthe\\ getterrate,onsetabout1\mathrm{sccm},0.2\mathrm{Pa}\\ Ar,224L/s\end{array}$ $(computed:0.5/0.03=16.7)$

Important Questions

Ques: What is the sputter yield, and how does it set the deposition rate?

It is the number of atoms ejected per incident ion. SPECS lists Ar${}^{+}$ yields at 0.5 $\mathrm{keV}$ of 0.95 for Al, 0.55 for Si and Ti, 0.6 for W, 2.0 for Au and 3.0 for Ag, roughly doubling at 1 $\mathrm{keV}$. Its rate formula is $z/t=MS{j}_p/(\rho N_{A}e)$. The same note lists measured rates at 500 $\mathrm{eV}$ and 1 $mA/c{\mathrm{m}}^2$: Al 630, Si 310, Cu 818, Au 1553 and SiO${}_2$ 400 $\mathring{\mathrm{A}}/min$. As a check, Al with an assumed density of 2.70 $g/c{\mathrm{m}}^3$ gives 591 $\mathring{\mathrm{A}}/min$ (computed), close to the measured value. Inverting Cu’s rate with an assumed 8.96 $g/c{\mathrm{m}}^3$ gives a yield of about 1.9 (computed). Yield, current density and the atomic volume together set the rate (my reading). Reference: SPECS GmbH, ‘Sputter info’ technical note: sputter rate z/t = M S j_p/(rho N_A e) with S the sputter yield in atoms per ion and j_p the ion current density; Ar+ yields at 0.5 and 1.0 keV of Al 0.95 and 1.8, Si 0.55 and 0.90, Ti 0.55 and 1.0, Ge 1.0 and 1.4, Ag 3.0 and 5.0, W 0.6 and 1.0, Au 2.0 and 3.5; measured rates for 500 eV Ar+ at 1 mA/cm2 of Al 630, Si 310, Ti 336, W 340, Cu 818, Au 1553 and SiO2 400 angstrom/min (fetch-tool summary).

Ques: How does sputtering differ from evaporation?

In atom energy, coverage and composition. Chen states that sputtered atoms leave at 10 to 50 $\mathrm{eV}$, about 100 times the energy of evaporated atoms, and that shadowing is small for sputtering but large for evaporation. Good coverage is needed for vias and multilayer metallization, while poor coverage is wanted for lift-off, so evaporation stays the lift-off tool (Chen; Thermal Evaporator page). Rates of 1 to 10 $nm/s$ match the 10 to 100 angstrom/s of electron-beam evaporation (Chen; Electron-Beam Evaporator page). TU Wien adds that a sputter target’s surface enriches in the low-yield component until the deposit matches the target. Alloys therefore transfer without the drift that vapour-pressure differences cause in a boat (my reading). Reference: T.-C. Chen, EE 527 Microfabrication Lecture 9, ‘Sputter deposition’, University of Washington, Winter 2014: glow discharge started by a large voltage across a low-pressure gas gap, chamber pressures 1 to 10 mTorr, ejected atoms escaping at 10 to 50 eV, about 100 times the energy of evaporated atoms, RF for dielectrics such as SiO2, ZnO, TiO2 and Si, magnetic field making electrons spiral and raising ionization so the plasma forms at 1e-5 to 1e-3 torr, typical sputtering rates 1 to 10 nm/s, reverse polarity back-sputtering as physical etching, shadowing large for evaporation and small for sputtering, good step coverage needed for vias and poor coverage wanted for lift-off (fetch-tool summary); TU Wien Institute of Solid State Physics, ‘Physik und Technologie Duenner Schichten / Thin Film Technology’, chapter 2: in sputtering the target surface enriches in the lower-yield component until the deposit matches the target composition (fetch-tool summary); Thermal Evaporator page of this vault; Electron-Beam Evaporator page of this vault.

Ques: What does the magnetron add to a sputter source?

A magnetic trap for electrons that raises the rate at lower pressure. Chen explains that the magnetic field makes electrons spiral, increasing ionization so that the plasma can form at $10^{-5}$ to $10^{-3}$ $\mathrm{Torr}$, against typical sputtering pressures of 1 to 10 $\mathrm{mTorr}$. CNR-IFN describes the field as concentrating and intensifying the plasma immediately above the target, raising ion bombardment and sputter rate without raising the operating pressure, and calls the magnetron the design of high-rate sources. De Gryse et al. model the eroded region as a racetrack of $10^{-2}$ ${\mathrm{m}}^2$ in their reactive magnetron. Lower pressure lengthens the mean free path, so sputtered atoms keep more of their 10 to 50 $\mathrm{eV}$, and the discharge can be checked with the methods of the Langmuir Probe page (my reading). Reference: T.-C. Chen, EE 527 Microfabrication Lecture 9, ‘Sputter deposition’, University of Washington, Winter 2014: glow discharge started by a large voltage across a low-pressure gas gap, chamber pressures 1 to 10 mTorr, ejected atoms escaping at 10 to 50 eV, about 100 times the energy of evaporated atoms, RF for dielectrics such as SiO2, ZnO, TiO2 and Si, magnetic field making electrons spiral and raising ionization so the plasma forms at 1e-5 to 1e-3 torr, typical sputtering rates 1 to 10 nm/s, reverse polarity back-sputtering as physical etching, shadowing large for evaporation and small for sputtering, good step coverage needed for vias and poor coverage wanted for lift-off (fetch-tool summary); CNR-IFN Trento, ‘RF sputtering principles’ facility page: RF makes sputtering of insulating targets possible because net current averages to zero each cycle, whereas an insulator on a DC cathode stops current; supply often fixed at 13.56 MHz; blocking capacitor develops the DC self-bias; negative charge builds on the target because electrons are more mobile than the heavier ions, so the target sits below plasma potential for most of the cycle; RF amplitude of order 500 V assumed; magnetic field concentrates the plasma above the target, raising ion bombardment and rate without raising pressure; magnetron as the high-rate source (fetch-tool summary); R. De Gryse, D. Depla and J. Haemers, ‘Aspects of the target voltage behavior in reactive sputtering’, 48th Annual Technical Conference Proceedings, Society of Vacuum Coaters (2005), p. 290: Berg model splitting reactive gas flow into target, substrate and pump paths summing to the inflow, gettering avalanche when reactive flow exceeds the getter rate giving abrupt changes in partial pressure, target voltage and deposition rate, compound sputtering less efficiently than the metal, poisoning also by ion implantation, Ar 0.2 Pa, 200 mA, 224 L/s, racetrack 1e-2 m2 and substrate plus wall 1 m2, sticking 0.03 (target) and 0.2 (substrate), Al yield 0.5 (SRIM) and Al2O3 0.03 (experimental), getter onset about 1 sccm (fetch-tool summary); Langmuir Probe page of this vault.

Ques: Why is RF needed for insulating targets, and where does the self-bias come from?

Because DC current cannot flow through an insulator. CNR-IFN explains that an insulator on a DC cathode stops the current, whereas with RF the net current averages to zero each cycle. A blocking capacitor lets negative charge build on the target, because electrons are far more mobile than the heavier ions, so the target sits below plasma potential for most of the cycle. This is the DC self-bias, with an RF amplitude assumed of order 500 $\mathrm{V}$ and a supply usually at 13.56 $\mathrm{MHz}$. NPTEL notes that both electrodes sputter in principle, with a fourth-power dependence on electrode area. The grounded chamber and substrate form a very large electrode, so little substrate sputtering occurs. Chen lists SiO${}_2$, ZnO and TiO${}_2$ as RF-sputtered materials. Reference: CNR-IFN Trento, ‘RF sputtering principles’ facility page: RF makes sputtering of insulating targets possible because net current averages to zero each cycle, whereas an insulator on a DC cathode stops current; supply often fixed at 13.56 MHz; blocking capacitor develops the DC self-bias; negative charge builds on the target because electrons are more mobile than the heavier ions, so the target sits below plasma potential for most of the cycle; RF amplitude of order 500 V assumed; magnetic field concentrates the plasma above the target, raising ion bombardment and rate without raising pressure; magnetron as the high-rate source (fetch-tool summary); NPTEL course 104103019, Module 4, Lecture 25 ‘Sputtering processes for thin film deposition’, page ‘Electrode-size effects’: both RF electrodes sputter, fourth-power dependence of the sheath-voltage ratio on electrode area so a large grounded electrode strongly raises the target sheath potential, substrate plus chamber forming a very large electrode so little substrate sputtering (fetch-tool summary); T.-C. Chen, EE 527 Microfabrication Lecture 9, ‘Sputter deposition’, University of Washington, Winter 2014: glow discharge started by a large voltage across a low-pressure gas gap, chamber pressures 1 to 10 mTorr, ejected atoms escaping at 10 to 50 eV, about 100 times the energy of evaporated atoms, RF for dielectrics such as SiO2, ZnO, TiO2 and Si, magnetic field making electrons spiral and raising ionization so the plasma forms at 1e-5 to 1e-3 torr, typical sputtering rates 1 to 10 nm/s, reverse polarity back-sputtering as physical etching, shadowing large for evaporation and small for sputtering, good step coverage needed for vias and poor coverage wanted for lift-off (fetch-tool summary).

Ques: Why does reactive sputtering show abrupt jumps in rate?

Because target poisoning runs away. De Gryse et al. summarise the Berg model, which splits the reactive gas inflow $q_0$ into consumption at the target, at the substrate and by the pump. While the inflow is below the getter rate, deposited metal consumes the gas. Above it, compound forms on the target, the getter rate falls, and an avalanche follows, with abrupt changes in partial pressure, target voltage and deposition rate. The compound sputters less efficiently: their Al${}_2$O${}_3$ yield of 0.03 against 0.5 for Al is about 17 times lower (computed). In their system the getter onset was about 1 $\mathrm{sccm}$ at 0.2 $\mathrm{Pa}$ argon. They add that ion implantation, not only chemisorption, drives poisoning. Reference: R. De Gryse, D. Depla and J. Haemers, ‘Aspects of the target voltage behavior in reactive sputtering’, 48th Annual Technical Conference Proceedings, Society of Vacuum Coaters (2005), p. 290: Berg model splitting reactive gas flow into target, substrate and pump paths summing to the inflow, gettering avalanche when reactive flow exceeds the getter rate giving abrupt changes in partial pressure, target voltage and deposition rate, compound sputtering less efficiently than the metal, poisoning also by ion implantation, Ar 0.2 Pa, 200 mA, 224 L/s, racetrack 1e-2 m2 and substrate plus wall 1 m2, sticking 0.03 (target) and 0.2 (substrate), Al yield 0.5 (SRIM) and Al2O3 0.03 (experimental), getter onset about 1 sccm (fetch-tool summary).

Ques: What is the sputter yield of aluminium, and which value is more likely?

About 1 at 500 $\mathrm{eV}$, with lower values at lower energy. SPECS lists 0.95 at 0.5 $\mathrm{keV}$ and 1.8 at 1.0 $\mathrm{keV}$ for perpendicular Ar${}^{+}$. De Gryse et al. use 0.5, taken from SRIM, in their reactive-sputtering model, without stating the ion energy in the summary available. Because the SPECS yields rise with energy, a yield of 0.5 implies ions below 500 $\mathrm{eV}$ (my reading). The SPECS measured rate of 630 $\mathring{\mathrm{A}}/min$ at 500 $\mathrm{eV}$ and 1 $mA/c{\mathrm{m}}^2$ is consistent with 0.95 (computed: 591 $\mathring{\mathrm{A}}/min$). The 0.95 value is therefore more likely correct for 500 $\mathrm{eV}$ ions. The 0.5 value is plausible for a lower-energy magnetron discharge. The ion energy must always be quoted with a yield (my reading). Reference: SPECS GmbH, ‘Sputter info’ technical note: sputter rate z/t = M S j_p/(rho N_A e) with S the sputter yield in atoms per ion and j_p the ion current density; Ar+ yields at 0.5 and 1.0 keV of Al 0.95 and 1.8, Si 0.55 and 0.90, Ti 0.55 and 1.0, Ge 1.0 and 1.4, Ag 3.0 and 5.0, W 0.6 and 1.0, Au 2.0 and 3.5; measured rates for 500 eV Ar+ at 1 mA/cm2 of Al 630, Si 310, Ti 336, W 340, Cu 818, Au 1553 and SiO2 400 angstrom/min (fetch-tool summary); R. De Gryse, D. Depla and J. Haemers, ‘Aspects of the target voltage behavior in reactive sputtering’, 48th Annual Technical Conference Proceedings, Society of Vacuum Coaters (2005), p. 290: Berg model splitting reactive gas flow into target, substrate and pump paths summing to the inflow, gettering avalanche when reactive flow exceeds the getter rate giving abrupt changes in partial pressure, target voltage and deposition rate, compound sputtering less efficiently than the metal, poisoning also by ion implantation, Ar 0.2 Pa, 200 mA, 224 L/s, racetrack 1e-2 m2 and substrate plus wall 1 m2, sticking 0.03 (target) and 0.2 (substrate), Al yield 0.5 (SRIM) and Al2O3 0.03 (experimental), getter onset about 1 sccm (fetch-tool summary).

References

1. T.-C. Chen, “Sputter deposition,” EE 527 Microfabrication, Lecture 9, University of Washington, Winter 2014. Available: https://faculty.washington.edu/tcchen/EE527Wi/Notes/EE 527 01-27 2014.pdf
2. SPECS GmbH, “Sputter info,” technical note. Available: https://www.specs-group.com/fileadmin/user_upload/products/technical-note/sputter-info.pdf
3. CNR-IFN Trento, “RF sputtering principles.” Available: https://www.tn.ifn.cnr.it/facilities/rf-sputtering-facility/rf-sputtering-principles
4. NPTEL course 104103019, Module 4, Lecture 25, “Sputtering processes for thin film deposition: electrode-size effects.” Available: https://archive.nptel.ac.in/content/storage2/courses/104103019/module4/lec25/16.html
5. R. De Gryse, D. Depla and J. Haemers, “Aspects of the target voltage behavior in reactive sputtering,” 48th Annual Technical Conference Proceedings, Society of Vacuum Coaters, 2005, p. 290. Available: https://www.svc.org/clientuploads/directory/resource_library/05_290.pdf
6. TU Wien Institute of Solid State Physics, “Physik und Technologie duenner Schichten (Thin film technology),” chapter 2. Available: https://static.ifp.tuwien.ac.at/homepages/Personen/duenne_schichten/pdf/t_p_tf_chapter2.pdf
7. Thermal Evaporator page of this vault.
8. Electron-Beam Evaporator page of this vault.
9. Langmuir Probe page of this vault.

Related pages in this vault: Thermal Evaporator, Electron-Beam Evaporator, Langmuir Probe, Optical Emission Spectroscopy Endpoint (OES), Residual Gas Analyzer (RGA), Electroplating (ECD) Cell, Parallel-Plate Reactive Ion Etcher (RIE), Ion Beam Milling System.