Roland A. - PVD for microelectronics (779636), страница 12
Текст из файла (страница 12)
A typical plasma is a partially ionized gas of Ar, Ne, Kr, or Xe, ora mixture of an inert gas and a chemically reactive gas such as oxygen ornitrogen. The ionization levels of these plasmas are low, and typically onlyone gas atom in 100-10,000 is ionized and the rest are neutral.Plasma technology is characterized by the use of multiple sets of unitsas well as misnamed usage of units. Depending on the author and the contemporary trend at the time of publication, plasma-related publicationsmay use SI units (kilograms, joules), cgs units (centimeters, grams, etc.),or other hybrid terms appropriate at the time. For the purpose of this book,we will use the terms most commonly used in the late 1990s, which unfortunately are a hybrid of various sets.
For energy terms, we will use eV(electron volts), in which 1 eV = 1.6 x 10 -19 joules. For densities, we willuse particles/cm 3. A gas density, for example, of 1 x 1015/cm 3 is equivalent to a chamber pressure of 30 mTorr or 4 pascals. We will use gas pressure units of mTorr, which are 10 -3 Torr, where 760 Torr equals one atmosphere. (The SI unit of pascal never really caught on in the late 1970sand early 1980s. The pascal pressure unit is equal to 7.5 mTorr. The unitmTorr has also been known in past as a micron.) For dimensions, we willuse centimeters, and for the various masses, we will use the atomic weightsin grams.
For calculations involving the masses of atoms or ions, we willtry to convert most equations into mass units of AMU (Ar -- 40 AMU,Xe = 131 AMU), which are easily found in the periodic table. Finally, fortemperatures, we will work with degrees K, although electron temperatures (described below) use an energy unit, eV.Even though it is somewhat tedious, we will try to restate the exact unitsas we go along to help those who might be searching for specific termsrather than reading the complete text. This is, perhaps, not the correct scientific approach to the purity of various systems of units, but it is consistent with the rather chaotic (and sloppy) usage within the various groupsworking with, publishing, and selling plasma technology.The most common method for producing a plasma for sputtering applications is to place a moderate voltage (hundreds to a few thousand voltsDC or RF) between two metal electrodes in a vacuum system (Fig.
3.1).Under the appropriate conditions of applied voltage and gas density, electrons may gain enough kinetic energy to ionize background gas atoms, andthe gas may break down and a plasma can be formed. The condition forbreakdown between two electrodes is a function of gas density; too low a52R. POWELL AND S. M.
ROSSNAGELFIG. 3.1Typical diode plasma system.density prohibits a cascade-like breakdown, and too high a pressure hastoo many gas atoms, which can damp down the plasma formation due totoo many collisions. This can be shown pictorially on a Paschen Curve(Fig. 3.2).For most sputtering applications, the cathode will serve as the source ofthe sputtered atoms: Ions from the plasma will bombard the cathode, alsoknown as the target, with sufficient energy to cause sputter emission ofcathode atoms. The typical parameters of a system like that in Fig. 3.1 area gas pressure in the 0.1 to 50 mTorr range, electrodes with dimensions ofa few to several tens of centimeters, and a separation of a few centimetersto perhaps 20 cm between the cathode and the sample.The vacuum systems used for sputtering usually have ultimate, or"base" pressures of below 10 -5 Torr (0.01 mTorr) and often as low as the10 -9 Torr range.
The chambers are constructed of steel or aluminum, andthe vacuum seals are either copper gaskets or viton o-rings (or a similarmaterial). Originally, most sputtering was done in diffusion-pumped chambers with a partially closed baffle between the diffusion pump and thechamber since the diffusion pump would not operate well at several mTorr.Most systems in current usage are pumped with either cryopumps or turbopumps, and many do not have pressure baffles between the sputteringchamber and the pump.
In this case, the plasma is formed and the cathodeis sputtered at the true chamber base pressure. If a baffle is used, it canoften degrade the base pressure 1 to 2 orders of magnitude. This last feature is often not documented well in many articles and trade publications.PLASMA SYSTEMS53FIG. 3.2 The breakdown voltage for plasma formation between two electrodes as a function of theproduct of the pressure and the electrode spacing. This curve is known generically as a PaschenCurve.3.1 Diode PlasmasThe two-electrode plasma system is known as a diode and is characterizedby a flux of ions that impact the cathode (the negative electrode) alongwith a flux of electrons that move toward the anode. The specific directionof the current flow leads to the term diode.
The bombardment of the cathode with ions causes the emission of secondary electrons from the cathode,which are then accelerated toward the anode and can gain enough energyto ionize more gas atoms. The flux of secondary electrons is the primarymeans of energy input into the plasma. From charge balance, it is necessary that each secondary electron create at least 1/gamma electron-ionpairs in the plasma, where gamma is the secondary electron yield.The secondary electron yield depends on both the bombarding ionspecies and the substrate.
The values can be quite low, and a sampling ofthis data is shown in Table 3.1 [3.1]. For a given material, the secondaryR. POWELL AND S. M. ROSSNAGEL54TABLE 3.1SECONDARY ELECTRON YIELDS FOR VARIOUS SEMICONDUCTOR-RELATEDMATERIALS [3.1 ].MetalAIAgCuMoMoTaPtWSi (100)Ni ( 111 )Ge ( 111 )Ar + (low E)Ar + (100 eV)0.120.010.030.10.020.30.30.050.050.070.120.0150.030.010.030.040.04Ar + (1 keV)Reference0.100.070.30.10.123.13.13.13.23.13.13.13.13.13.13.10.100.040.70.5electron yield is rather insensitive to the incoming ion's energy over abroad range from near 0 eV up to 1 keV or so (Fig.
3.3). The secondaryelectron emission process at these energies is an Auger-like process that isindependent of the ion's kinetic energy. At higher energies (above 1 keV orso), the secondary electron yield increases linearly with ion energy. In thisSecondary Electron Yield0.2Ar ions/0.15Ar neutrals0.10.05/500,sss.,s~ s,~ ~" ~" " ""1000 1500 2000 2500 3000Kinetic Energy (eV)FIG. 3.3 General energy dependence of the secondary electron yield for energies up to a few keV(adapted from [ 3.2 ] ).PLASMA SYSTEMS55energy regime, the yield is composed of the Auger component as well as akinetic component. A classic test of this is the comparison of the secondaryyields for ions and neutrals of the same species. This test, first done byMedved et al.
in 1963, shows the additive effect of these two componentsfor ions and the lack of the Auger process for neutrals [3.2].A quick, crude calculation from the secondary electron yield puts an estimate on the voltages needed to operate a discharge. If the secondary electron yield for Ar § on some cathode is 0.05, then each secondary electronis responsible (in some form) for the generation of 20 Ar ions. Since theionization potential for Ar is around 16 eV, the minimum energy necessaryfor the secondary electron to make 20 ions is 320 eV, requiring that theminimum discharge voltage be around 320 V.
The real situation in theplasma is immensely more complicated, and this simple estimate shouldnot be considered too binding. However, it does point out the inverse relationship between secondary electron yield and discharge voltage: If theyield of secondaries is high, the required discharge voltage is likely to besmaller than a similar case with a low secondary yield.Surface contamination can change the secondary electron yield. Thepresence of oxygen on the surface often increases the secondary electronyield by 20 to 100%.
Therefore, an oxidized cathode might have a loweroperating voltage than a clean cathode. This implies that when startingwith a dirty cathode, the operating or discharge voltage of a plasma systemmay be observed to increase with time until the oxided layer is sputteredoff the cathode. Conversely, if oxygen is added to a sputtering system (seeSection 3.8), the result may be to decrease the discharge voltage as thecathode becomes lightly oxidized.Many common diode plasma systems actually have only one obviouselectrode, which is the cathode, and the chamber walls function as theanode. Since the chamber walls are grounded for safety reasons, the cathode potential is then many hundreds of volts negative.Plasmas have specific regions of interest for sputtering applications. Thetwo most relevant regions are the bulk of the plasma and the sheath, ordark space, located between the cathode and the bulk of the plasma. Theplasma itself is a conductor, and as such its potential is fairly constantacross its width.
Virtually all of the voltage drop between the cathode andthe anode occurs in the thin dark space, or sheath, at the cathode.Current flow across the cathode sheath is limited by space charge effects, which relate to a self-shielding effect of a stream of charged particles. The current density across the sheath, j, the voltage across thesheath, V, and the sheath thickness, d, are related by the Child-LangmuirLaw (3.3):R. POWELLAND S.
M. ROSSNAGEL56j = ~4,n.d2(3.1)where M is the ion mass, V is the voltage, and d is the acceleration distance. The current density in most plasma systems is in the range ofmilliAmps per square centimeter (mA/cm2). This equation in a useful formbecomesJ= 5.5 X 10 - 5 .V1"5" d - 2 .W-o.5(3.2)where V is the voltage (in volts), d is the sheath width in cm, and W is theatomic weight of the ion in AMU.Example: The maximum, space-charge-limited current betweentwo grids spaced 0.2 cm apart with a voltage across them of1000 V for Argon (40 A M U ) is (d = 0.2, V = 1000, W = 40)j = 2.41 mA/cm 2. If Xe was used in place of Ar (W = 131), them a x i m u m current density drops to (d = 0.2, V - 1000, W 1 3 1 ) j = 1.33 mA/cm 2.For electrons, the conversion leads toJ\cm'/ = 2.3 •10 -3.
V !5. d -z(3.3)and the units are, again, volts and cm for V and d, respectively.By comparison, with the same voltage difference and d-spacing, themaximum electron current density (i.e., the space charge value) is ordersof magnitude larger than the ion current density value. This is evidenceof the much greater mobility of electrons in a plasma compared to theions. The electron space charge current is effectively ignored in mostcases and the fundamental limit is the ion current. Also, this means thatin a sheath at the edge of a p l a s m a - where the ions are passing from theplasma to the cathode and the electrons are moving from the cathode tothe p l a s m a - there are virtually no electrons in the sheath.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
