Roland A. - PVD for microelectronics (779636), страница 20
Текст из файла (страница 20)
Hence, increasing effort by source designers has been directed at preventing arcs, suppressing them once they begin to form, and/orlimiting the delivered energy to extremely low levels ( < 1 mJ for a 10-kWprocess). As a result, modern PVD sources often include arc suppressioncircuitry within the source assembly housing or as an auxiliary piece ofhardware associated with the DC power supply. In spite of the importanceof arc formation and suppression during magnetron sputtering, many thinfilm scientists and engineers are unfamiliar with the topic. Also, much useful literature on this topic is found in product brochures, patents, and proceedings of industrial coating conferences that PVD workers are not likelyto have.
With this in mind, the reader is directed to refs. 4 . 1 1 - 4 . 1 7 for anintroduction to the field.Arcs can occur in a number of ways. One possibility is a bipolar arcfrom the cathode to some other part of the chamber that acts as an anode.For example, a low-impedance path can form between the target and agrounded surface in close proximity, such as a dark-space shield. This canoccur by electrical breakdown of the low-pressure gas in the interveningspace or by direct electrical conduction through a metallic flake thatbridges the target and shield.
This kind of arc can generally be preventedby proper spacing of the shield and periodic cleaning of the chamber.Similar bipolar arcs can form between the PVD shields and the substrate.Far more common is the unipolar arc, or microarc, whose onset and termination occur on one and the same electrode. It has been estimated thatmicroarcs account for 99.9% of the arc events in a properly designed PVDsystem. A microarc is initiated whenever a small region of target is able tosupply a sufficiently high current of electrons into the plasma, such that theresulting fields cause the plasma to locally collapse into a thread of ionsand electrons.
Microarcs can be caused by such things as inclusions or irregularities in the target surface that can act as field emission sources, localhot spots that lead to thermionic emission, and microbursts of trapped gasreleased from the target during sputtering. Microarcs can also occur wheninsulating material on the target surface charges up to voltages that exceedthe material's dielectric breakdown strength.
This electrical breakdowncan often initiate an arc and also produce negatively charged particles thatcan travel quite long distances, being accelerated by the electric fields atthe target. This could be the case if insulating hydrocarbon contaminationor native oxides were on the target surface, e.g., as the result of an im-THE PLANAR MAGNETRONproper target burn-in procedure (see Section 11.3). Alternatively, reactivegases could form insulating compounds on the target surface, such as ahigh partial pressure of 02 or H20 in Ar forming A1203 on lightly erodedparts of an A1 target.
It is instructive to consider the time to initiate such anarc under typical PVD conditions using elementary electrical arguments.Consider a small insulating region at the target surface of area A andthickness d that is undergoing Ar § bombardment with ion flux density J.This area acts like a microcapacitor with one plate being the metal targetand the other plate being the top surface of the insulator. Assuming thatthe dielectric constant of the insulator is e, its capacitance is then given byC = eeoA/d, where e0 = 8.8 x 10 -12 C 2/N - m 2.
As ions bombard thecapacitor, it builds up a positive surface charge after time t, given by Q =JAt. Since the voltage across a capacitor is V = Q/C, it is then easy toshow that an electric field of strength E = V/d develops after time t suchthatt =eeoEJ(4.1)The time to break down the insulating region is then given by Eq. (4.1)but with the breakdown field strength substituted for E. Using parametersappropriate to AI203 (e = 10, E = 108 V/m) and an ion current density of40 mA/cm 2 = 400 A/m 2, we calculate t = 2 msec.
This is a very short timeand is comparable to the time it takes for the plasma to extinguish fromelectron-ion recombination after shutting off the power supply.There are a number of hardware fixes that have been provided to address microarcing. These generally involve modifying the circuitry of theDC power supply to actively sense and then suppress arcs as soon as theyform. Arc formation is signaled by an abrupt drop in voltage across thecathode dark space. Once the arc is sensed, it is possible to suppress it byrapidly reversing the polarity of the target voltage and biasing it so that itis ~ 10-20 V more positive than the plasma potential.
This serves to attract electrons from the plasma to the target and quench the arc. Advancedarc suppression circuits are capable of reacting to cathode arcs in a few100 nsec and can suppress more than 2000 microarcs per second. As a result, the energy delivered to a given arc can be sufficiently low ( < 1 mJ)to avoid particulate generation and other target-related damage. For information on these and other methods of controlling arcs such as lowfrequency AC techniques, the use of RF alone or in combination with DCpower, dual magnetron sources, and unipolar pulsed DC magnetron sputtering, the reader is referred to refs. 4 .
1 1 - 4 . 1 7 .98R. POWELL AND S. M. ROSSNAGEL4.5 Low-Pressure SputteringThe efficient use of electrons in the DC magnetron allowed sputter deposition to be carried out at much lower pressure than had been possible withdiode sputtering. In fact, proper magnetron action requires that pressure below enough so that the electron mean free path associated with gas scattering is not significantly less than the electron gyratron radius. For representative DC voltages and magnetic field strengths (e.g., 500 V and 300 G),the gyration radius is calculated to be ~ 2 mm, with the result that operating pressure must be less than ~ 50 mTorr for efficient magnetron sputtering m this is 10-20 times lower than had been possible with diode sputtering. The reduced gas-phase scattering at these pressures providedbenefits such as improved sputtered flux directionality, a greater fractionof emitted material reaching the wafer, and retention of enough energy inthe incident adatoms to influence and control film morphology.In IC production, PVD films are typically deposited at operating pressure ~ 2-5 mTorr; however, there is a desire to reduce magnetron operating pressure to well below 1 mTorr to further reduce the effects ofgas-phase scattering and thereby improve directionality.
For example,long-throw sputtering must be carried out at source-to-substrate distancesthat are several times greater than in conventional PVD (e.g., 300 mm versus 100 mm), which requires a comparable increase in mean free path toprevent atom scattering. Hence, operating pressure in long-throw PVD isseveral times lower than in conventional PVD. In this regard, it should benoted that highly directional PVD methods based on ionized metal specieshave been developed that use much higher than conventional pressure( ~ 20 mTorr) to thermalize sputtered neutrals so they can be more efficiently ionized and directed at a biased wafer (see Chapter 9).
In this case,the ion acceleration occurs over such a short distance (the dark space) thatscattering is not an issue even though the chamber pressure is very high.That is, it is the product of distance and mean free path of the atom or ionthat is important and not the value of either quantity alone.Lower PVD operating pressures have already benefited conventionalprocesses such as collimation since this reduces the possibility that atomsexiting the collimator will scatter away from near-normal incidence beforereaching the wafer. Also, since the transmission of a collimator decreasesstrongly with its cell aspect ratio, reduced pressure has allowed lower aspect ratio collimators to be used to achieve the same directionality, with acorresponding increase in deposition rate. Whether reactive PVD processes such as collimated TiN will be possible at very low pressure is problematic since the flux of the reactive component may be insufficient toTHE PLANAR MAGNETRON99produce a stoichiometric compound in the allotted time.
For example,sputter deposition of a 200~ TiN film in 60 sec is equivalent to ~ 1 monolayer per second of TiN, or ~ 2 • 1015 Ti atoms/cm 2. Assuming the working gas is a 50/50 mixture of Ar/N 2 at 0.1 mTorr, the molecular incidencerate of N e at the wafer is about 1016 cme/sec, which is not much greaterthan the flux of Ti. In addition, the shields and collimators that becomecoated with Ti act as a getter-pump of chamber N e.
This serves to furtherreduce the amount of nitrogen available at the wafer surface to form TiN.Initiating and sustaining magnetron discharges at pressures < < 1 mTorris difficult due to the reduced number of Ar atoms available for ionization.This leads to reduced Ar § bombardment of the target and consequently decreased production of discharge-sustaining secondary electrons. On theother hand, the average electron energy may be somewhat greater at reduced pressure, which increases their cross section for Ar ionization andmitigates the reduction in Ar gas density. Greater magnetic field strengthcan also be applied to increase the electron mean free path and thereforethe probability of ionizing the working gas at low pressure; however, thishas its limits.
As described in Chapter 9, one way to reduce magnetronoperating pressure is to remove the working gas entirely and utilize selfsputtering of the metal to sustain the discharge. This method is well-suitedto materials such as Cu that have a high self-sputtering yield; however, itis not generally applicable to all materials of interest to IC fabrication.More generally, magnetron discharges can be initiated at quite low pressures and sustained at even lower pressure ( ~ 0.1 mTorr) if a sufficientsupply of low-energy electrons is provided.
This can be done by using anadditional source of electrons (e.g., injection of electrons from a hollowcathode electron source) or by preventing secondary electrons that werenot trapped in E • B drift orbits from leaving the plasma volume. In thelatter case, one method that has been successfully used is the so-calledbucking magnet in which a secondary, fixed ring of magnets is arranged atthe edge of the magnetron, at or behind the target plane. The magneticpoles in the fixed ring are aligned opposite to that of the primary magnetsin the rotating array (see Fig.
4.6), which creates a net magnetic field acting to redirect secondary electrons emitted at the edge of the target backtoward the plasma region.Other approaches to low-pressure magnetron operation are discussed inrefs. 4.18 and 4.19 and citations therein. Considering that many crossed-fielddevices already operate at pressures much less than 1 mTorr (e.g., sputter ionpumps and UHV ion gauges) and that high-vacuum planar magnetrons havebeen demonstrated [4.19], there is reason to expect that magnetrons designedfor sub-0.1 mTorr operation will be applied to future IC production.100R.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
