The present invention relates to industrial and other processes in which large volumes of gases must be pumped at pressures as low as 1-10 milliTorr. Industrial processes in this category include, for example, various types of plasma processing, such as plasma enhanced chemical vapor deposition, plasma mediated etching of surfaces, and other types of surface modification processes.
In many such processes employing plasmas, it is generally considered by those skilled in the art to be advantageous to generate the processing plasmas in suitable mixtures of gases maintained at pressures as low as 1-10 milliTorr. The purity and composition of the gas can best be controlled if the flow rate of fresh gas into the processing chamber is high relative to the processing rate. However, existing vacuum pumping technology can provide only limited throughput of gas in this pressure range.
The pumping speed of widely used turbomolecular vacuum pumps, for example, generally decreases rapidly with increasing pressure at pressures above roughly 1 milliTorr. Robust, cost-effective systems for achieving high-speed pumping in the pressure range from 1-10 milliTorr have not been developed to date.
Furthermore, many gases involved in industrial processing of, for example, VLSI systems are toxic or hazardous and must be isolated and controlled with great care. It would be advantageous if the toxic or hazardous gases to be exhausted from the processing chambers could be converted into less toxic or hazardous forms through dissociation of the molecules of such gases. This process is often referred to as pyrolyzation of toxic or hazardous gases and has been investigated for many years, particularly in connection with toxic gases used by the military.
Conventional vacuum pumping technology utilizes one of two fundamental mechanisms: (1) increasing the momentum of gas molecules in a preferred direction and exhausting the gas molecules through a valve or baffle structure which inhibits the reverse flow of gas; or (2) condensing the gas to be pumped on special surfaces. The first mechanism is usually implemented through some type of piston, blower, or rapidly moving vanes which impart directed momentum to the gas from rapidly moving mechanical structures or streams of pumping molecules, such as mercury or readily condensable pumping oils. The second mechanism is commonly used in systems with low to moderate throughput requirements. In the range of pressures where industrial plasma processes are carried out (1-100 milliTorr), turbomolecular pumps are used almost universally as the first stage of a compound pumping system designed to pump large fluxes (“throughputs”) of process gases.
Turbomolecular pumps impart directed momentum to gas molecules through collisions with rapidly spinning discs. This mechanism is most effective at sufficiently low gas pressures that the mean-free-path of the molecules is larger than the dimensions of the pumping structures. The resulting limit on maximum gas throughput that can be achieved with turbomolecular pumps is a drawback in the plasma processing industry where substantial throughputs of reactant gases are needed to prevent the buildup of reaction products to concentrations that would be deleterious to the process. Further, high-speed turbomolecular pumps are necessarily complex and expensive devices, in which large angular momentum is stored. Moreover, many plasma processes yield solid and/or corrosive byproducts that can be potentially damaging to these pumps.
The ability to pump at 3 to 5 times the presently available pumping speed has been shown (using smaller substrates) to enhance the reliability of the process and performance of etch and deposition processes. The present limit of pumping speed in a properly designed system is determined both by the speed of the pump and the conductance of the chamber to the pump inlet. At present the capability of turbomolecular pumps is limited to 5500 liters per second; although the use of the largest available turbomolecular pumps is further limited by the cost of the large pumps and the expected lack of reliability of a pump this large. The cost of a turbomolecular pump of this size exceeds $80,000, which, at $15.5/(liter/sec), is considerably higher in cost per unit of pumping speed than the $30,000, or $10.6/(liter/sec), for a 3300-liter per second pump. Thus, the cost per unit pumping speed for the larger pumps is 50% greater than for the smaller pumps.
Even so, the 5500 liters/second (1/sec) pump is smaller than would be optimum for processing 200 mm wafers. Extension to 300 mm wafers significantly exacerbates the problem of providing adequate pumping speed. The required pumping speed scales most closely with the area of the substrate. The scaling of process equipment to from 200 mm to 300 mm wafers requires at least a 2.25× increase in pumping speed. The available increase from 3300 liter/second to 5500 liter/second is only an increase of 1.67×. This leaves the 300 mm systems with pumping options that are grossly inadequate.
The problem of gas handling in 300 mm systems is complicated even further by the fact that it is the pumping speed at the wafer (substrate) that is most important. Providing pumping speed at a location remote from the wafer means that the gas atoms must be conducted to the pump inlet through some transitional structure. In the transition region there are invariably reductions in conductance, which is typically measured in liters/sec. Computer programs can predict the overall pumping speed when the gas atoms are in the laminar or molecular flow regimes. In most processes of commercial interest, however, the flow is characterized as transition flow and the computer models are less reliable in predicting performance. The conductance between the wafer and the pump in most designs provides a loss of at least 50%-75% in effective pumping speed.
In addition, the effective handling of gas flow must account for the gas species that must be removed from the system but spend much time attached to the processing chamber walls. A similar problem occurs with any barely volatile species that may result from the process itself, for example, multiple carbon species that are polymerized either by the electrons or photons of the plasma. It is easy for the plasma electron or photon flux to affix these fragments of molecules to the walls. In the same way these materials are subsequently released from the walls, perhaps as a different species. The same processes of attachment, synthesis, decomposition and evolution occur at the substrate but with the difference that the substrate is of different material and has the additional energy flux of the ion bombardment that is used, for example, to promote the etch process. The substrate is a source of organic material and silicon from the etch process. If this flow of new material ceases when the wafer bias is removed the reactor tends to stabilize in that the active groups that are volatile become combined into the degenerate brown film that is often seen in plasma processing systems. This material is highly cross-linked and very stable with respect to thermal desorption or plasma bombardment.
The quality of the process results depends in considerable measure upon the presence or absence of weakly volatile materials such as, for example, in the use of side-wall passivation to provide straight walls in high aspect ratio channels. This passivation results from the deposition of material that originated as photoresist and injected gas and is modified by the electron collisions as a free molecule in the plasma volume and may undergo multiple changes while adsorbed on surfaces. This material is then redeposited on the wafer surface specifically in the freshly etched surface of the feature being etched. The deposited material having experienced multiple interactions with the plasma is particularly resistant to decomposition by the ions and chemically reactive molecules that provide the flux that etches the wafer. Probably much of the process dynamics that allow selectivity between organic surfaces and surfaces that are more inorganic (the plasma tends to spread molecular species to all surfaces in the plasma) are affected by changes in the thin volume of the wall that is in contact with the plasma. For these reasons, if the residence time for these species in the process chamber could be reduced significantly, perhaps shorter that some accommodation time (the time for them to find this stable chemical site), then etch chemistry in the chamber could be controlled and optimized.
For some time there has been a growing appreciation of the possible benefits of using plasmas as the active element in vacuum pumping technologies; for example, plasmas can pump a wide range of gasses, including hydrogen and helium, with equally high efficiencies. Plasma vacuum pumps can be highly tolerant of solid or corrosive process by-products. As described herein, these benefits result from the generic underlying properties of plasma vacuum pumps, in which three-dimensional flow of the neutral gas to be pumped is transformed into one-dimensional flow of a magnetized plasma which can be magnetically compressed and guided through suitable baffle structures. Neutral gas is composed of neutral, i.e. non-ionized, atoms and molecules. Momentum can be imparted to the plasma through various electromagnetic interactions and, in turn, to the neutral gas through collisions between energetic charged particles and neutral gas molecules.
These potential benefits have not yet been fully realized in practice for a number of technical reasons relating to efficient generation of plasma, the creation of a magnetic field suitable for both the plasma generation and the necessary channeling of the plasma flow, and simple and effective mechanisms for driving the plasma flow at pressures in the range of importance to plasma processing applications. This last technical difficulty is exacerbated by the plasma's ability to shield its interior from low-frequency external electric fields, together with the complex atomic and molecular processes that become important in the pressure range of interest.
Another type of plasma vacuum pump is disclosed in International Application No. PCT/US99/12827, filed on Jun. 29, 1999, entitled PLASMA VACUUM PUMPING CELL, the entire disclosure of which is incorporated herein by reference, and pending U.S. Provisional Application No. 60/114,453, filed on Dec. 30, 1998, entitled PLASMA VACUUM PUMP, the entire disclosure of which is incorporated herein by reference. This pump utilizes the plasma excited in a processing system by a high-density plasma source, such as a plasma utilizing electron cyclotron resonance (ECR), an inductively coupled plasma (ICP), or an electrostatically shielded radio frequency (ESRF) plasma, to pump gas out of the system. This type of pumping cell must be designed and built according to the source plasma properties of the system in which it is to be installed. It can not be used as a stand-alone vacuum pump.
According to the invention, a stand-alone plasma vacuum pump for pumping gas from a low-pressure inlet to high-pressure outlet combines an electron cyclotron resonance effect with a specially shaped permanent magnet field to propel ions from the inlet to the outlet. The plasma vacuum pump includes a housing enclosing a pumping region located between the inlet and the outlet, a plurality of permanent magnet assemblies providing magnetic fields that extend in the pumping region between the inlet and the outlet, the magnetic fields providing at least one magnetic flux channel for guiding and confining a plasma, and a source of microwave power coupled into the flux channel to heat plasma electrons, ionize the gas, and create forces that propel plasma ions in a direction from the inlet toward the outlet. The pump is further constructed to promote the flow of plasma and electrically neutral gas molecules toward the outlet while impeding flow of neutral gas molecules in the direction from the outlet toward the inlet.
A plasma vacuum pump according to the invention is especially well suited for pyrolyzing effluent gases, in that all gases to be pumped must pass through a region populated by electrons with enough energy and density to ionize the gas molecules which enter the region with high efficiency. The resulting molecular ions will generally dissociate into the constituent atomic species in times that are less than their transit time through the pump.
A pump according to the invention employs four separate and distinct plasma technologies in a novel configuration. The first is electron cyclotron heating technology that provides efficient ionization of the neutral gas to be pumped. The second is plasma confinement technology that confines a plasma which results from electron cyclotron heating within the pumping duct by creating at least one magnetic flux channel having the cross sectional shape of an asymmetric magnetic mirror. The third is a momentum-transfer-pumping technology that drives the electron cyclotron heated plasma along the magnetic flux channel. The fourth is baffle technology that allows unrestricted outward-directed plasma flow but impedes the back-stream of thermalized neutral gas molecules.
According to one novel aspect of the present invention, a single longitudinally extended plasma region is generated in a magnetic flux channel within a rectangular pumping duct. Alternatively, 4, 8, or more longitudinally extended plasma regions are each generated in a respective flux channel within a cylindrical enclosure. In this form of construction, the channels are spaced apart around the axis of the cylindrical enclosure, each channel extends along that axis, and each channel is formed to pump gas in a radial direction. This cylindrical configuration provides a very large pumping surface area to accommodate large gas loads. In addition, neutral gas molecules can enter the enclosure and travel parallel to the axis of the enclosure in the space or spaces between the plasma regions. As the gas molecules pass through the electron cyclotron resonance region, most of the incident molecules will be ionized by energetic electrons which accumulate within this region as the result of electron cyclotron heating. Once the molecules are ionized they can escape the plasma pump essentially over only a small solid angle at the outlet end of each flux channel.
Electron cyclotron heating serves two essential functions in the invention. It provides a flexible and efficient means for heating electrons to energies around 100 eV, at which energy inelastic electron collisions with gas molecules have their greatest probability of ionizing the molecules. In addition, electron cyclotron heating in the magnetic field configuration of the invention creates internal space-charge electric fields that accelerate ions along the magnetic lines of force.
According to a second novel aspect of the present invention, each flux channel is defined by a very efficient magnetic field configuration created using permanent magnets. The magnetic field plays three important roles in the plasma pump mechanism: (1) it delimits an electron cyclotron resonance (ECR) zone for effective electron heating and plasma production, (2) it provides a diverging, expanding, magnetic flux tube or channel within which there are diverging magnetic lines of force, or flux lines, or a decreasing magnetic field strength, and within which the kinetic energy associated with the orbiting movement of electrons in a direction normal to the magnetic field lines is converted to kinetic energy associated with electron movement parallel to the magnetic field lines, and (3) it guides the plasma flow from the ECR resonance region, through the momentum-transfer pumping ducts and a baffle, and into a fore-pump region. All three roles are accomplished with minimal amounts of permanent magnet materials.
A magnetic flux tube is generally defined as a region in space whose transverse surfaces are everywhere parallel to magnetic lines of force. A comprehensive discussion of magnetic flux tubes is given, for example, in “The Structure of Magnetic Fields”, by A. I. Morozov and L. S. Solov'ev, in Reviews of Plasma Physics, Vol. 2, pages 9-11, Ed. M. A. Leontovich, Consultants Bureau, New York, 1966. In the rectangular embodiment of the invention magnetic lines of force lie in planes. In Cartesian coordinates these planes can be chosen to be planes on which the coordinate z is constant, so that the magnetic field has only x- and y-components. The magnetic flux tubes or channels extend the length of the magnetic structure in the z-direction. In the cylindrical embodiment the magnetic lines of force lie in planes on which the axial coordinate, z, is constant. The magnetic field has components in the radial, r, and azimuthal directions. The magnetic flux tubes or channels extend the length of the magnetic structure in the axial, z-direction.
In an ECR resonance zone, a high-density plasma with a density of 1012-1013 ions/cm3 is generated by whistler waves launched from high magnetic field regions near a microwave antenna. As shown by R. A. Dandl and G. E. Guest (“On the Low-Pressure Mode Transition in Electron Cyclotron Heated Plasmas”, J. Vac. Sci. Technol. A9(6), November/December 1991, pp.3119-3125) the density produced is approximately proportional to the microwave power coupled into the plasma region.
This plasma contains primary and secondary electrons. The primary resonantly-heated electrons may have energies approaching 100 eV. These electrons are highly ionizing, and are capable of ionizing a large flux of gas molecules, provided adequate electron cyclotron heating power is provided, and thus achieving a high pumping speed. The temperature of the secondary electrons resulting from impact ionization of incident gas molecules is in the range of 3-10 eV, depending on the heating power and operating pressure.
According to a further novel aspect of the present invention, the plasma ions are accelerated collectively by an internal space-charge electric field in the region of diverging magnetic field lines downstream from the ECR zone.
According to a still further novel aspect of the present invention, directed neutral gas flow in each flux channel is strongly enhanced in that the neutral gas atoms gain directed momentum rapidly from resonant charge-exchange and other collisions with the plasma ions. In this way, parallel momentum is continuously transferred to the neutral molecules, while the newly charge-exchanged ions quickly gain directed energy from the collective electric field. Plasma ions may subsequently recombine with plasma electrons at the flux channel exit, where backward flow of neutral molecules is impeded. Since the plasma readily flows along the converging magnetic lines of force, it can be guided through a suitably restricted exit orifice; i.e., the geometry of the pumping ducts may be made to coincide with the magnetic field lines. Using criteria familiar in the art (see, for example, Scientific Foundations of Vacuum Technique, Second Edition, by Saul Dushman, Ed. J. M. Lafferty, John Wiley and Sons, New York, (1962) Chapter 2) the dimensions of the exit orifice can be chosen to restrict the flow of gas from the high-pressure outlet side of the orifice to the low-pressure inlet side. In this way the backward flow rate of gas through the orifice can be made smaller than the forward flow rate of plasma and entrained gas.
The construction and performance of the pumping ducts will be discussed in more detail below.
Ωe=2πfe=eB/m
Here e and m are the electric charge and mass of the electron, respectively. In a magnetic field of 875 Gauss, for example, the electron gyrofrequency, fe, equals 2.45 GHz, the frequency of the microwave power used in many commercial applications.
If microwave power of some particular frequency, fμ, is radiated into a region in space where the magnetic field strength has the value at which the electron gyrofrequency equals the frequency of the microwave power, electrons will undergo continuous acceleration transverse to the magnetic field for as long as they remain in this “resonance region”. The resonant value of the magnetic field strength is given by
Bres=fμ(m/2πe)
As is know in the art, this resonant acceleration can readily increase the electron kinetic energy by more than 100 eV before electrons move out of the resonance region. Such resonantly heated electrons can thus have energies which are optimum for ionizing any gas molecules with which they may collide.
Provided the magnetic field strength varies sufficiently gradually in space, the electron motion can be further characterized by the magnetic moment associated with its (transverse) gyration. The gyrating electron constitutes a microscopic current loop whose current is je=efe. Moreover, if the electron velocity transverse to the magnetic field has the magnitude v⊥, then the radius of the gyration is given by ρe=v⊥/Ωe. The magnetic moment associated with the microscopic electron current loop is the product of the current and the area of the loop:
μe=jeπρe2=W⊥/B
Here W⊥ is the electron kinetic energy associated with motion transverse to the magnetic field. In a spatially varying magnetic field, electrons will experience a force along the magnetic field given by −μe∇B, where ∇B is the gradient of the magnitude of the magnetic field strength. Because this force is antiparallel to the gradient, it will accelerate electrons along the magnetic lines of force toward regions of lower field strength.
The flux channel has a substantially constant flux density along each line perpendicular to the plane of
Lines L are not intended to depict the quantitative variations in flux density, but only the qualitative nature and direction of those variations. However, as is illustrated, it is desired that the flux density at plane d be less than that at plane a.
The microwave power source 2 is disposed to introduce high frequency microwave energy into the flux channel to produce high energy electrons and ionize gas molecules in an electron cyclotron resonance zone ECR between planes a and b in which electrons undergo gyration at the resonance frequency, experience a force along the weakening magnetic field and generate an internal space-charge field to accelerate ions along the magnetic field, as discussed earlier. The portion of the flux channel between planes a and b will constitute the preferred location of the inlet end of the flux channel. However gas to be pumped can also be introduced into the portion between planes b and c.
The internal space-charge electric field arises due to the fact that the electrons experience the −μe∇B force along the magnetic field between planes a and c, where the flux density is decreasing, thus “separating” from the ions. Consequently, the electrons “drag” the ions through the flux channel by electric forces. This is known in the art.
The plasma ions are thus accelerated collectively by the internal space-charge electric field to drift downstream with the electrons, which are constrained to flow parallel to the magnetic field. Plasma drift energies of tens of electron volts are expected from this acceleration mechanism
In the flux channel, downstream of the ECR resonance zone, supra-thermal electrons directly heated in the ECR zone convert part of their energy perpendicular to the magnetic field lines into energy parallel thereto as those electrons move down to plane c. Perpendicular energy herein refers to the kinetic energy associated with the electron velocities perpendicular to the magnetic field lines, i.e.,
Wperpendicular (or W⊥)=½mv2 perpendicular (or v⊥2)
Parallel energy is herein referred to the kinetic energy associated with electron velocities in a direction parallel with the magnetic field lines; i.e.
Wparallel (or W∥)=½mv2parallel (or v2∥)
The electrons and ions continue to move through the magnetic field portion between planes c and d to enter the outlet portion between planes d and e.
Although the flux density in the flux channel has a positive gradient between planes c and d, the electrons and ions have sufficient parallel kinetic energy to move past plane d. This movement is permitted by the fact that the gradient between planes c and d is smaller in magnitude than the gradient between the ECR and plane c.
The electrons and ions then enter the portion between planes d and e, which constitutes the outlet end of the flux channel, where they will recombine into neutral gas and be acted on by a fore-pump. Electrons downstream of plane e see an increasing magnetic field in the flux channel, between planes d and e, which impedes their movement back into the flux channel. The movement of gas in this direction is further impeded by the effective narrowing of the channel at plane d, which reduces the channel cross section as seen from the outlet end of the channel.
In addition, neutral molecules are displaced along the channel by collisions with positive ions travelling down the flux channel.
As shown, magnets 12 and 14 are longer than magnets 16 and 18, respectively. This serves to create the field pattern described above. Alternatively, or in addition, magnets 12 and 14 can be made magnetically stronger than magnets 16 and 18, respectively.
In a pumping region according to the invention, source 2 (
This pump is constituted by a rectangular enclosure formed by pole pieces 20 and two side walls 30 (not shown in
Slotted waveguide 27 can be loaded with a dielectric material such as Teflon® in order to reduce the size of the component. Furthermore, as indicated in
As shown in
The embodiment shown in
A simplified perspective view of a stand-alone plasma vacuum pump according to a cylindrical embodiment of the invention is shown for illustration in
A center septum structure 60 extends along the axis of enclosure 50 between plate 52 and outlet opening 58. Structure 60 separates the plasma and gas flows out of the four axially extending pumping ducts or channels and provides surfaces on which the plasma ions can recombine with plasma electrons to form neutral gas. Structure 60 thus divides the central conduit into four quadrants to prevent plasma and gas from one of the four pumping regions from streaming into adjacent pumping regions.
Microwave antennas 70 are used for plasma formation and heating in each of the four pumping ducts. According to one embodiment of the invention, each antenna is formed by modifying a standard 4-port slot-coupled hybrid coupler to include a tunable impedance matching front end section. Antennas of this general type have been used successfully in plasma sources described in U.S. Pat. Nos. 5,133,826, 5,203,960, and 5,370,765 issued to Raphael A. Dandl. Modifications for adapting the microwave couplers and antennas for particular embodiments are known to those skilled in the art of microwave systems.
Each microwave antenna 70 radiates, for example, 2.45 GHz microwave power into a respective pumping region within enclosure 50 through a respective one of four separate quartz microwave windows 72. This microwave power may be supplied by one or more commercially available sources coupled to antenna 70. The number of sources and the means for distributing the power from individual sources may be dictated by the total power required and the availability of suitable sources. Generally speaking, the pumping speed will be proportional to microwave power, so high-speed pumps may require multiple sources of microwave power. An example of a 2.45 GHz microwave source that is commercially available is an ASTEX 1500 (Maximum output power of 1500 W@ 2.45 GHz). Other sources include an ASTEX S1000i (1 kW) and an ASTEX S700i (700 W).
The pump further includes four axially extending outer magnet assemblies mounted at the cylindrical wall of enclosure 50, and four axially extending inner magnet assemblies spaced radially inwardly from the outer magnet assemblies. These magnet assemblies produce four magnetic fields each similar to the magnetic field shown in
The magnetic flux in the air gaps between adjacent pairs of inner and outer magnets forms flux channels that are specifically shaped to function as plasma pumping channels 138. Thus, in the illustrated cylindrical embodiment there are four such pumping channels. The more constricted portions at the outlet ends of the magnetic flux channels pass through suitably designed orifices which form baffles which, as described earlier herein, impede flow of ions back into the flux channels. In the cylindrical embodiment, a set of baffles represented by slats are shown in dashed lines for a single quadrant in
The strength and dimensions of magnets 130 and 134 are selected to generate an electron cyclotron resonance surface 140 (
An example of the magnetic flux surfaces created by the magnetic structures is shown in FIG. 11. The lines appearing in
As shown particularly in
The pump inlet of the cylindrical stand-alone plasma vacuum pump shown in
According to a further feature of the embodiment shown in
It will be noted that in a pump according to the invention gas to be pumped can enter each duct in any direction perpendicular to the pumping direction. In the embodiment of
The multi-channel plasma configuration provides a large plasma surface for ionizing gas molecules which enter the plasma pump. Thus this pump is capable of pumping a large gas load. The 4-channel pump can be designed to have a total plasma surface area, i.e., the total area over which gas that has passed through inlet openings 54 can enter the four pumping ducts, of greater than 1000 cm2 and is capable of pumping a gas load of 10 Torr liter/sec or higher. The gas entering the pump through inlet openings 54 is first ionized by the fast electrons heated by microwave power launched from the antennas in the high magnetic field regions. The plasma parameters that can be obtained in this way are in the ranges of: ni=1012-1013 cm−3; Te=3-10 eV, and degree of ionization=1-10%, where ni is the plasma density and Te is the secondary electron temperature.
The plasma is produced in magnetic flux channels 138 that guide the plasma flow to the outlet end of each flux channel while back streaming of the gas is impeded by orifices bounding the flux channels. The plasma throughput resulting from the free flow of plasma along the magnetic fields is about 10-30 Torr·liter/sec.
The present invention employs collective plasma acceleration to further enhance the plasma throughput by a factor of 3 to 5. The plasma acceleration results from the microwave heating dynamics and the magnetic field design. It does not require additional hardware components.
Finally, the magnetic field configuration downstream of plane d in
Details on the theory of the operation of the stand-alone plasma vacuum pump are described below. Specifically, there will be presented a description of: the plasma vacuum pump requirements; the magnetic field designs and calculations; the formation and heating of the over-dense plasma using high-field launched whistler wave heating at a microwave frequency of 2.45 GHz; the collective plasma acceleration mechanism employed in the present invention; and the momentum transfer pumping mechanism.
Plasma Vacuum Pump Requirements:
Plasma Surface Area:
For a given gas throughput (“load”) Qo the plasma pump must provide a pumping speed So large enough to maintain the desired operating pressure P. The required pumping speed is given by:
So=Qo/P
The pumping speed is a volume flow rate having the units of volume per unit time, i.e. liters per second. The operating pressure is assumed to be the operating pressure of the pumping ducts and the pumping speed is the volume flow rate through the total summation of the “outer” magnetic field surfaces; these surfaces are referred to later as the vacuum - plasma interface.
For a plasma vacuum pump, gas molecules must be ionized before they are removed or pumped out of the system. Since the gas molecules enter the plasma with a characteristic speed vo, the plasma surface area must be large enough to allow the gas to flow into the plasma at a rate corresponding to the desired pumping speed. The required plasma surface area, A, is given by
So=¼vo·A (2)
where So is the desired pumping speed, A is the total surface area of all surfaces of the vacuum-plasma interface and vo is the gas velocity normal to these surfaces integrated on these surfaces. In the embodiment of
Ion Throughput:
The ion throughput is given by the ion flow equation:
Qion=¼ni·viAp (3)
Here, Qion is the total ion throughput through all of the pumping ducts, ni is the ion density, vi is the ion velocity, and Ap is the total cross sectional area of all of the channels through which the plasma flows. Here the factor ¼ is the three-dimensional effect, as in the case of neutral gas flow. Because the ion density and the ion velocity are both limited within narrow parameter ranges, the cross-section area must be large enough to produce a large ion throughput. For a typical plasma density of ni=5×1012 cm−3 and Te=5 eV (the ion velocity is vi≈3×105 cm/sec for Ar), to achieve the ion throughput, namely, Qion≈10 Torr liter/sec 3×1020 particle/sec, the required cross section area is quite large:
Ap=4Qion/(ni·vi)=1000 cm2 (4)
Differential Pressure:
The balance between the gas load and the throughput of the pump will determine the differential pressure between the pump inlet and the pump outlet:
Qo=Qpump+C(Pin−Pout) (5)
where Qpump is the throughput generated by the pump, C is the pump conductance, Pin is the inlet pressure, and Pout is the outlet pressure. The compression ratio of the pump is defined as Pout/Pin. The differential pressure and the compression ratio can be expressed as
(Pin−Pout)=(Qo−Qpump)/C (6)
and
Pout/Pin=(Qpump−Qo)/CPin+1 (7)
From Equation (7), it becomes clear that Qpump must be greater than Qo in order to produce a compression ratio greater than one. In the present invention, several enhancements are employed to enhance the pumping throughput. Symbolically
Qpump=βQion (8)
Here β is the enhancement factor: making β≧100 is possible with the enhancement mechanisms implemented, as described in detail in the following sections.
Magnetic Field Configuration:
The magnetic field described herein for implementing the present invention was designed using the 2-dimensional SUPERFISH-PANDIRA code and the 3-dimensional ANSYS/Multiphysics code v5.4. Designs for rectangular plasma pumps and cylindrical plasma pumps with 4, 6, and 8 plasma channels have been generated. The 4-channel configuration is shown in
In each quadrant of the 4-channel pump, an auxiliary pair of rectangular magnets 134 and a pole piece 136 are used near the center region to optimize the local mirror fields in the following manner:
The magnetic flux channels 138 extend from the local magnetic field maximum Bmax2 towards the cylindrical axis of enclosure 50 in order that they can guide the plasma to flow into the pumping ducts in a one-dimensional manner.
Much of the radially inwardly directed magnetic flux is connected to magnets 134 and then returned to outer magnets 130. This further improves the magnetic field strength in the mirror region. The mirror region is, in general, the region extending radially between the two maxima in the magnetic field, i.e. the region between the two B-field maxima, Bmax1 and Bmax2 in
The ECR resonance (Bres=875 gauss for 2.45 GHz microwave frequency) surfaces 140 are positioned several inches inwardly from outer magnets 130, allowing outer magnets 130 to be located outside of the vacuum vessel.
The magnetic field strength, Bmax1, in the gap between each pair of outer magnets 130 is significantly higher than the resonant field, Bmax1/Bres=1.2, in order that the wave-guide antenna located here can excite whistler waves from the high field side for “high-field launch” heating. The magnetic field strength, Bmax1, is at the peak of the curve in
The resonant surfaces (ECR) are found at the location where Bres=875 gauss, in the region of radially inwardly decreasing magnetic field strength between the location of Bmax1 and Bmin pointing radially to the pumping ducts. The magnitude of field gradient is optimized for improved electron heating.
The flux channels are converging downstream, from Bmin to Bmax2. to guide the plasma to the outlets of the pumping ducts and into the quadrants formed by structure 60. Gas is then pumped along these quadrants to outlet 58 by a pump (not shown, coupled to outlet 58. The local field maximum, Bmax2, however, is made smaller than the resonant field strength.
The field minimum, Bmin, in the mirror region is significantly lower then the resonant field strength to allow efficient plasma acceleration.
Formation and Heating of Over-Dense Plasmas:
As discussed by, for example, B. H. Quon and R. A. Dandl, “Preferential electron-cyclotron heating of hot electrons and formation of overdense plasma”, in Physics of Fluids B1, (10), October 1989; and G. E. Guest, M. E. Fetzer, and R. A. Dandl, “Whistler-wave electron cyclotron heating in uniform and nonuniform magnetic fields”, in Physics of Fluids B2(6), June 1990, experience in electron cyclotron heating plasma technology has proven that strong wave absorption occurs and over-dense plasma is generated when whistler waves are launched toward the resonant zones from locations of higher magnetic field strength: B>Bres, where Bres=2πfμ/(m/e). Here fμ is the applied microwave frequency, and e and m are the electron charge and mass, respectively. The whistler waves propagate into the plasma with
E(z,t)=E⊥(z)cos(ωt−kz cos θ) (9)
In this expression, k is the magnitude of the wave-number vector, E is the microwave electric field strength with component E⊥ perpendicular to the static magnetic field, z is the distance along a magnetic field line, and k=2π/λ is the magnitude of the propagation vector of the waves whose wavelength and (angular) frequency are λ and ω, respectively.
The whistler wave is the right-hand circularly-polarized mode described by the dispersion relation
n2=1−αω)/(ω−Ω cos θ) (10)
where n2=(kc/ω)2 is the square of the refractive index, α=(ωpe/ω), and Ω=eB/m, θ is the angle between the wave vector, k, and the magnetic field, B, c is the speed of light, e is the electron charge, m is the electron mass, ωpe is the electron plasma frequency and ω is the angular frequency of the wave. For small θ, the power absorbed by the plasma is expected on the basis of theory to be given by:
Pabs=Po[1−exp(−∫2kidz)] (11)
ki≈∥ωpe2ω/(2c2k2ve)exp{−[(ω−ωce)/kve]2}|| (12)
where Pabs is the absorbed power, Po is the incident wave power, and ki is imaginary part of the wave number, ωce=eB/m is the cyclotron frequency and ve=(2Te/m)½ is the electron thermal speed.
With microwave power launched at ωce/ω=1.2-1.5 from a matched antenna or a wave-guide horn with a quarter-wave window, the whistler wave will propagate mainly along the magnetic field lines, with a corresponding reduction in wavelength (e.g. increasing k) and strong attenuation (e.g. increasing ki) by plasma absorption. Wave absorption occurs where the Doppler-shifted resonance condition
ω−ωce=k·v=kv cos θ (13)
is satisfied. Electrons with |v∥≦2ve are able to resonate with the wave. For Te≈6 eV, the resonant zone covers magnetic field strengths in the interval 810≦B≦940 G, or
ωce/ω=0.92-1.08.
Notice that in the high field side, v∥ is negative; i.e., the heated electrons are moving toward the antenna. These heated electrons are reflected by the higher magnetic field at the front surface of the antenna.
The typical plasma density generated by whistler waves launched in the high-field region is in the range of 1-3×1012 cm−3 without antenna tuning. To achieve still higher plasma density, a special coupling network and a low-impedance antenna will be provided. Standard 4-port, side-wall hybrid couplers, wave-guide transitions with low impedance horns and quarter-wave quartz windows will be provided to achieve higher plasma density. Wave-guides filled with alumina will be used to reduce the dimensions of the network when necessary. Such a coupling network and antenna can be obtained by modifying available hardware in ways that are already known in the art.
Plasma Acceleration Techniques:
Electron cyclotron heated electrons gain kinetic energy in motion perpendicular to the magnetic field lines of force:
W⊥=½mv∥2 (14)
As discussed, for example, by David J. Rose and Melville Clark, Jr., in Plasmas and Controlled Fusion, The M.I.T. Press and John Wiley&Sons (1961) especially Chapter 10, the motion of electrons in the space beyond the resonant interaction zone can be described in terms of their total energy, ε, and magnetic moment, μ:
ε=W⊥+W∥−eφ (15)
μ=W⊥/B (16)
Here φ is any electrostatic potential that may be present, and
W∥=½mv∥2 (17)
Outside of the resonant interaction zones, ε and μ remain constant due to conservation of energy and the magnetic moment, if the magnetic field does not vary too rapidly in space.
The electrons experience a force parallel to the magnetic field, F∥:
F∥=−μ∇B=−(W⊥,res/Bres)(dB/ds∥) (18)
Thus, in the absence of electrostatic fields, the “parallel” kinetic energy of the electrons at any point beyond the resonant surface is given by
For a magnetic flux channel extending radially as shown in
Equation (19) signifies that electrons are accelerated radially inwardly parallel to a spatially decreasing magnetic field in the direction antiparallel to the gradient. Since heated electrons are being accelerated but ions are not, charge separation will result and an electrostatic potential, φ, will build up. This potential will retard the electrons and accelerate the ions until both species of particles have the same “ambipolar” parallel drift velocity.
For the electrons,
since W⊥=μB=W⊥,resB/Bres. For singly-charged ions of charge e,
where the subscript e denotes electron.
Adding the ion equation (21) and the electron equation (20) yields
W∥,e+W∥,i=Wres(1−B/Bres) (22)
Where the subscript i denotes ion.
In equilibrium, since v∥,=v∥,e=va, the “ambipolar” speed,
½mva2+½Mva2=Wres(1−B/Bres) (23)
or
va=[2Wres(1−B/Bres)/(M+m)]½ (24)
where M is the ion mass.
It is important to ensure that the plasma is not reflected or significantly retarded by the magnetic field as the plasma approaches the local region of high magnetic field near the outlet ducts. The condition for the plasma to flow through the local field maximum, Bmax, can be expressed as
½Mva2≈Wres(1−Bmax/Bres) (25)
where the ion kinetic energy is given by equation (25). Thus the local field maximum should not be higher then the resonant field.
Momentum Transfer Pumping Mechanism
The moving plasma ions are subject not only to inertial force and field forces, but also to stochastic effects that cause them to be diffused or to be impeded in their motion. Momentum transfer collisions can induce these effects in the presence of an electric field. In this case, the plasma-associated throughput is modified to include a term proportional to the ion mobility, μ:
Qp=ni·vi·Ap+E·μμ·ni·Ap (26)
Here E is the electric field. Denoting vd=Eμ, where vd is the ion drift velocity, the ion throughput can be expressed as:
Qp=ni(vi+vd)Ap=ni·vi(1+vd/vi)Ap (27)
If (1+vd/vi) is replaced by βμ, then
Qp=βμni·vi·Ap (28)
With large E at high pressure, the value of βμ can be large.
The pumping throughput is actually higher when the neutral drift velocity is included. In fact, most of the neutral molecules in the system under consideration are also drifting with nearly the same drift velocity, owing to the ion-neutral momentum transfer collisions. Thus the total pumping throughput is:
Here no is the neutral gas number density and βt is the enhancement factor for the ion and neutral flows. Experimental testing in a single plasma duct has confirmed a large enhancement using the momentum transfer pumping technique according to the present invention.
The method and apparatus according to the invention can also be used for pyrolyzing effluent gas by operating the pump in the manner described above and supplying effluent gas to the pump inlet. A stand-alone plasma vacuum pump according to the invention is capable of pumping any constituents in gaseous form as long as the ionization of the gasses leads to a predominantly electro-positive plasma.
While the description above refers to particular embodiments of the present invention, it will be understood that many modifications may be made without departing from the spirit thereof. The accompanying claims are intended to cover such modifications as would fall within the true scope and spirit of the present invention.
The presently disclosed embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims, rather than the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
This is a continuation of International Application No. PCT/US01/11111, which was filed on Apr. 6, 2001, and also claims benefit of U.S. application No. 60/196,920, which was filed Apr. 13, 2000, the contents of both of which are incorporated herein in their entirety.
Number | Name | Date | Kind |
---|---|---|---|
3555331 | Chuan | Jan 1971 | A |
3994625 | Welch | Nov 1976 | A |
4401054 | Matsuo et al. | Aug 1983 | A |
4512868 | Fujimura et al. | Apr 1985 | A |
4641060 | Dandl | Feb 1987 | A |
4745337 | Pichot et al. | May 1988 | A |
4883570 | Efthimion et al. | Nov 1989 | A |
4902934 | Miyamura et al. | Feb 1990 | A |
4933650 | Okamoto | Jun 1990 | A |
5024716 | Sato | Jun 1991 | A |
5133826 | Dandl | Jul 1992 | A |
5165861 | Jahns | Nov 1992 | A |
5198725 | Chen et al. | Mar 1993 | A |
5203960 | Dandl | Apr 1993 | A |
5240381 | Nagai et al. | Aug 1993 | A |
5256036 | Cole | Oct 1993 | A |
5296714 | Treglio | Mar 1994 | A |
5370765 | Dandl | Dec 1994 | A |
5453125 | Krogh | Sep 1995 | A |
5754008 | Wartski et al. | May 1998 | A |
5762814 | Ohara et al. | Jun 1998 | A |
5824607 | Trow et al. | Oct 1998 | A |
5855686 | Rust | Jan 1999 | A |
5948168 | Shan et al. | Sep 1999 | A |
5975855 | Ensberg et al. | Nov 1999 | A |
5985091 | Suzuki | Nov 1999 | A |
6422825 | Dandl et al. | Jul 2002 | B2 |
6559601 | Johnson et al. | May 2003 | B1 |
Number | Date | Country | |
---|---|---|---|
20030122492 A1 | Jul 2003 | US |
Number | Date | Country | |
---|---|---|---|
60196920 | Apr 2000 | US |
Number | Date | Country | |
---|---|---|---|
Parent | PCTUS01/11111 | Apr 2001 | US |
Child | 10268970 | US |