Patent Application
20250079128
The present invention relates generally to plasma processing techniques, and, in particular embodiments, to systems and methods for processing substrates with plasma modulated by DC magnetic fields.
An integrated circuit (IC) is a monolithic structure comprising a network of electronic components in a semiconductor substrate. An array of IC units is formed on each substrate as a batch of substrates gets processed through a series of patterning levels. At each level, a patterned layer of various materials are formed using deposition, photolithography, and etch processes. Typically, a minimum half-pitch of an array of trenches or holes fabricated at a patterning level defines its patterning capability. About every two years, a new IC node is introduced having more advanced technology that enables doubling the component density to reduce cost per unit and, concurrently, enhance performance and functionality. The doubling results from shrinking minimum feature sizes and deploying three dimensional (3D) devices, a combination that requires fabricating high aspect ratio structures with controlled low dimensional variations.
A first step in patterning a layer is printing, in which a pattern of actinic radiation is transferred to a photoresist layer coated on the substrate. In a patterning process, referred to as single patterning, printing is followed by etching to transfer the resist pattern to an optional hard mask and thence to a target layer. The patterning capability of single patterning is the resolution limit of the optics, which is proportional to λ/NA, where λ is the radiation wavelength and NA is numerical aperture. The minimum half-pitch was reduced from 1 micron to 0.1 micron in two decades, during which λ was reduced from 436 nm (ultraviolet (UV) mercury lamp) to 193 nm (deep-UV (DUV) ArF excimer laser). A shorter λ=13.5 nm (extreme UV (EUV) tin plasma) lithography system took almost another two decades to develop. In the interim, scaling was enabled by using techniques such as optical proximity correction (OPC), phase shift photomask, immersion lithography (193-i), and multiple patterning, but at a higher processing cost and new patterning errors such as pitch walking. Recently, 13.5 nm EUV is being used for cost-effective manufacturing at the sub-10 nm nodes, where lateral features less than 20 nm appear frequently in IC designs, thus pushing the aspect ratios of the respective vertical structures to a range requiring special etch processes, generally referred to as high aspect ratio contact (HARC) etch.
Many of the deposition and etch steps in forming 3D devices (e.g., nanosheet transistors and vertical NAND (V-NAND) memory) are plasma processes, some of which, such as the HARC etch, may need special plasma properties to provide, for example, a highly collimated ion flux to the substrate. Further innovations are desired to gain enhanced control over plasma properties such as plasma density, electron temperature, radical flux, ion flux, and ion angle and energy distributions to meet the challenge of patterning high aspect ratio features at a nanoscale pitch with high-yield for volume manufacturing.
A method for plasma processing a substrate, where the method includes generating a plasma in a plasma chamber within which the substrate is held during processing, where generating the plasma includes: flowing a discharge gas through the plasma chamber; coupling a radio frequency (RF) source signal to a first RF electrode, where the coupling ionizes the discharge gas; and coupling a bias signal to a second RF electrode, the bias signal being a periodic series of bias pulses, each period having a bias-ON time and a bias-OFF time, where a bias voltage waveform is applied during the bias-ON time; generating a pulsed DC magnetic field in the plasma chamber, by coupling a magnetizing signal to an electromagnet, the magnetizing signal being a periodic series of current pulses; and prior to coupling the magnetizing signal, synchronizing the periodic series of current pulses with the bias signal to flow a DC magnetizing current during the bias-ON time.
A system for plasma processing, where the system includes a plasma chamber; a substrate holder configured to hold a substrate in the plasma chamber; a gas flow system configured to flow a discharge gas through the plasma chamber; a first radio frequency (RF) electrode; a second RF electrode, where the first RF electrode and the second RF electrode are configured to cooperatively generate a plasma in the plasma chamber; an electromagnet configured to generate a pulsed DC magnetic field in the plasma chamber, where the pulsed DC magnetic field is configured to modulate an electron temperature profile of the plasma; a first electrical circuit configured to output an RF source signal; a second electrical circuit configured to output a bias signal; a third electrical circuit configured to output a magnetizing signal; and a controller configured to send control signals to the first, second, and third electrical circuits, to adjust and synchronize the RF source signal, the bias signal, and the magnetizing signal.
A method for plasma processing a substrate, where the method includes flowing a discharge gas through a chamber; coupling a radio frequency (RF) source signal to a first RF electrode; and coupling a bias signal to a second RF electrode, the bias signal being a periodic series of bias pulses, each period being a sum of a bias-ON time and a bias-OFF time, where a bias voltage waveform is applied during the bias-ON time; generating a pulsed DC magnetic field in the chamber, by coupling a magnetizing signal to an electromagnet, the magnetizing signal being a periodic series of current pulses; and prior to coupling the magnetizing signal, synchronizing the periodic series of current pulses with the bias signal to flow a DC magnetizing current during the bias-ON time.
For a more complete understanding of the present invention, and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawings, in which:
This disclosure describes embodiments of methods for processing a substrate with plasma modulated by a pulsed DC magnetic field in a plasma chamber coupled to a gas flow system to flow a discharge gas at a low pressure through the chamber. The plasma may be generated by coupling radio frequency (RF) power to ionize the gas to ignite and sustain a glow discharge plasma, and the pulsed DC magnetic field may be generated by an electromagnet magnetized with pulsed DC current. The electric and magnetic fields in the chamber may be configured to reduce an effective electron temperature (Te) of the plasma in a region of the chamber above the substrate, as explained in further detail below. An effect of reducing Te is to alter ion velocities such that variances of ion angle and ion energy get reduced in a joint distribution of the angle and energy of ions in an ion flux incident on the substrate. Thus, with the embodiments of methods described in this disclosure, a highly collimated flux of ions with precisely controlled kinetic energy may be made available to perform plasma processing.
An ion flux with very small spread in ion angle and energy is advantageous for an anisotropic reactive ion etch (RIE) process, especially when the RIE process is for etching high aspect ratio features, commonly referred to as a high aspect ratio contact (HARC) etch. With lateral dimensions of the electronic components in ICs being scaled more rapidly than vertical dimensions, the aspect ratios of vertical structures, at most patterning levels, are on an upward trend. For example, in the back end of line (BEOL), widths of holes (e.g., for contacts and vias) and trenches (e.g., for metal lines) in interlayer dielectric (ILD) layers have shrunk more than the ILD thicknesses and trench depths. In the front end of line (FEOL), the advent of 3D field-effect transistors (FET) (e.g., FinFET, nanosheet FET, forksheet FET, and complementary FET (CFET)) and 3D memory (e.g., the vertical NAND (V-NAND) memory) has, likewise, led to vertical structures with aspect ratios increasing with miniaturization. Self-aligned quadruple patterning (SAQP) with 193 nm immersion (193-i) lithography was used at the 10 nm node to fabricate features at a minimum half-pitch of about 25 nm. But, IC designs targeting the sub-10 nm nodes include features in several patterning layers, where the minimum half-pitch is less than 20 nm. For these, the more cost effective 13.5 nm EUV single exposure lithography is used to print the pattern.
After printing, the pattern is transferred to a hard mask layer using the resist as an etch mask, and the hard mask layer is used as the etch mask to pattern vertical openings (e.g., via holes) in a target layer (e.g., an ILD layer) using anisotropic RIE processes. For the dimensions at the sub-10 nm nodes, commonly used vertical structures such as contacts and vias may have aspect ratios exceeding 10; and the highest aspect ratios may be in a range of 50 to 100. It is non-trivial to provide a HARC etch for fabricating nanoscale width openings having such high aspect ratios. In addition, in order to achieve high yield and reliability, the process may have to meet stringent specifications for sidewall profile control and local critical dimension uniformity (LCDU). As explained below, to help the HARC etch meet these requirements, the plasma processing equipment and method may need to generate a vertically downward ion flux to the substrate, where the ion distribution is controlled to have very narrow spreads in ion angle and energy. For specificity, a zero degree ion angle is the vertically downward direction of a velocity of an ion when it strikes a major surface of the substrate in a direction normal to the surface.
During HARC etch processing, it is undesirable to have energetic ions colliding with sidewalls exposed by the etching. Such collisions may cause undesired effects such as line edge roughness (LER) and undercut. Furthermore, inelastic collisions and ion absorption at the sidewall may randomly reduce availability of ions with high vertically directed energy at the bottom of a partially etched opening to cause a correspondingly random variation in etch rate. As the bottom of a partially etched feature progresses deeper during anisotropic etching, the aspect ratio of a partially etched feature increases, thus squeezing a tolerance for fluctuation in ion angles around zero degree. Clearly, a reduced spread in the angle distribution provides an advantage for a HARC etch process by reducing the probability of an incoming ion colliding with the sidewall.
Ions within the plasma in the chamber collide with neighboring particles, which are mostly neutral particles. In general, ion angle and kinetic energy are both affected by random collisions. Hence, the variance in ion energy is related to the variance in ion angle. It is noted that, in this disclosure, plasma processing refers to direct plasma processing, so the ion angle and energy of interest are that of ions just prior to striking the surface of the substrate being processed, after accelerating across a plasma sheath that is formed adjacent above the surface of the substrate. The sheath is a narrow positively charged region of high electric field directed toward the surface that forms along the periphery adjacent to a boundary surface, whereas bulk of the glow discharge plasma, referred to as bulk plasma, is a charge-neutral region, (i.e., there is negligible difference between positive and negative charge densities). Between the bulk plasma and the sheath, is a quasi-neutral transition region, referred to as the pre-sheath. The ion velocity distributions in these three regions and the effect of the electron temperature (Te) on the velocity distribution of ions in the ion flux incident on the surface from the sheath are described below.
The glow discharge plasma in the chamber comprises two species of charged particles, ions and free electrons, along with a majority of neutral particles. For simplicity, in the discussions in this disclosure, we lump together all the neutral particles as a single third species, referred to as neutrals. Furthermore, we assume the ions are a single species having one unit of positive electronic charge. These simplifications are for clarity of the description and not to be construed as limiting the scope of the invention.
The particles are distributed over a wide range of energies. The energy distribution of the particles of each species may be approximated by a Maxwellian distribution function, which is proportional to exp (−E/kBT), where, E denotes the total energy (kinetic and potential energy) of a particle, kB is Boltzmann's constant, and T denotes the effective temperature, which represents an average kinetic energy (equal to 3/2 kBT) of a particle of the species. Note that, in the Maxwellian approximation, the kinetic energy is distributed isotropically; hence, the directed component of kinetic energy (directed vertically downwards) is not included in 3/2 kBT. For example, ions in the highly collimated ion flux impinging on the surface have a large directed component of kinetic energy, but an effective ion temperature in the sheath represents only the much smaller random isotropic component of the ion kinetic energy.
The RF power sustaining the glow discharge increases the average kinetic energy of the gaseous mixture in the chamber beyond the thermal equilibrium value given by an ambient temperature. The excess kinetic energy is split unevenly among the three species because of unequal energy absorption and distribution of via collisions. Thus, while in thermal equilibrium, T is equal to an ambient temperature, in non-equilibrium, the steady state may result in unequal effective temperatures for each of the species. The discharge plasma is considered not to be in thermal equilibrium because of a high rate at which electric potential energy from the externally induced RF electric field gets converted to kinetic energy of the free electrons in the plasma. In general, the steady state effective temperature of a species is the net result of the rates at which the particles of that species gain and lose kinetic energy during an initial transient that settles to the steady state of the plasma.
The electrons being extremely light compared to ions (typically, having a mass ratio of several thousand), have extremely high mobility. Due to the high electron mobility, the free electrons acquire kinetic energy at a high rate from the RF electric fields in the bulk plasma (much higher than that for ions). Furthermore, although free electrons collide elastically with the other species (i.e. the heavier ions and neutrals), very little momentum (hence kinetic energy) is transferred, again primarily because of the large mass ratio. With the rate of kinetic energy gained being much larger than the rate of kinetic energy lost, the average kinetic energy of free electrons rapidly rises to a steady state value corresponding to Te of the order of 10,000 K or 1 eV (i.e., roughly 30 to 200 times the thermal equilibrium value). The energetic free electrons cause further ionization by inelastic collisions with neutrals, thus generating the ion-electron pairs that sustain the bulk plasma and the steady state fluxes of ions and free electrons to the boundary surfaces, which include the chamber walls and the major surface of the substrate.
Neutrals, on the other hand, do not absorb energy from electromagnetic (EM) fields. Besides, as mentioned above, collisions of neutrals with high energy free electrons may be inelastic, where, instead of increasing the kinetic energy of the neutrals, kinetic energy is transferred from the free electrons to ionize the neutrals. With negligible increase in their kinetic energy, the effective temperature of the neutrals, remains at the ambient temperature (e.g., about 300 K or 0.026 eV). In contrast, on average, the kinetic energy of ions is increased slightly through collisions with the more energetic electrons, while a fraction of the gain is lost, mostly in collisions with neutrals. The net result is that the effective temperature of the ions (Ti) is slightly higher than that of the neutrals (which is the ambient temperature). However, as explained above, the increase in Ti is smaller than the increase in Te. Typically, in the bulk plasma, Ti is a few hundred degrees higher than the ambient (e.g., about 350 K to 900 K or 0.03 eV to 0.8 eV).
As described above, the initial transient after plasma ignition rapidly settles to the steady state glow discharge plasma. In the description, we have divided the steady state into three regions, depending on the motion of the ions. Being scattered by collisions, the ions may move with velocities distributed in all directions. At any location, the ion velocity distribution is approximated by a displaced Maxwellian function, which comprises an isotropic part superposed on an anisotropic part. The isotropic part is a Maxwellian with Ti as its temperature, representing the random motion of the ions at that location. The anisotropic part accounts for the ion motion directed perpendicular to the surface that averages to a net directed component of ion velocity, Vd, which gives rise to the ion flux at that location. In other words, at any location, the average kinetic energy per ion may be partitioned into a random component equal to 3/2 kBTi and a directed component equal to ½ Mion Vd2, where Mion is the mass of one ion.
Farthest from the surface is the bulk plasma having the highest densities of ions and free electrons, a negligible net charge and electric field, as well as a high collision rate, mostly with neutrals. As a result, the steady state velocity distributions of ions and free electrons in bulk plasma are almost perfectly isotropic, implying that Vd is almost zero and the average kinetic energy per ion may be almost entirely attributed to the random component. Although Vd is low in the charge-neutral bulk plasma, there is a slowly moving ion flux (and free electron flux) “drifting” toward the surface. As an ion approaches the vicinity of the substrate, it enters the narrow sheath adjacent to the surface of the substrate. As mentioned above, the plasma sheath is the narrow region of high electric field, where the field is directed to rapidly accelerate the positively charged ion to a high velocity toward the surface.
Prior to entering the sheath, the ions emerging from the bulk plasma pass through the pre-sheath. As mentioned above, the pre-sheath is the region where properties of the steady state plasma transition from their values in bulk plasma to pronouncedly different values at the edge of the sheath. Outside the bulk plasma, the ion and free electron densities reduce progressively toward the surface, with the free electron density dropping faster than the ion density. This results in a net positive charge density profile that starts with a negligibly low value close to the charge-neutral bulk plasma but increases progressively toward the surface through the pre-sheath to reach much higher values in the sheath. The pre-sheath region may be considered to be quasi-neutral, in comparison to the high charge densities inside the sheath. The direction of the electric field increases the ion velocities preferentially toward the surface, thus increasing the anisotropy of the ion distribution function. Indeed, Vd starts to increase from almost zero near the bulk plasma and, within a short distance into the pre-sheath, the directed component of kinetic energy (½ Mion Vd2) becomes equal the random component of ion kinetic energy (3/2 kBTi) and increases further to far exceed 3/2 kBTi in most of the pre-sheath. In addition, Ti itself increases in the pre-sheath because increasing Vd also increases the average kinetic energy of the ions that get scattered in ion-neutral collisions in the pre-sheath region to add to the random motion of ions. However, as explained further below, there may be no further increase in Ti inside the sheath, i.e., Ti reaches its highest value at the edge of the sheath.
The sheath adjacent to the surface of the substrate is formed during the initial transient, when a rapid out-diffusion of free electrons from the plasma to the substrate charges the substrate negatively. The transients settle to where the negative charge of the substrate is balanced by positive space charge in the sheath and an associated electric field profile that attracts ions and repels free electrons. This establishes the steady state comprising a potential profile and ion and free electron fluxes that satisfy the boundary conditions for the surface, for example the applied voltage if the substrate is biased or a net zero current requirement if the substrate is floating. As known to persons skilled in the art, in order to obtain a stable steady state sheath, the average directed velocity of ions entering the sheath needs to exceed a minimum value, referred to as the Bohm velocity. The quasi-neutral pre-sheath (described above) is the region where Vd increases till it equals the Bohm velocity at a location defined to be the sheath edge. The pre-sheath being quasi-neutral, most of the positive charge and potential drop occurs the narrow space between the sheath edge and the surface of the substrate, for which the electric field reaches very high values in the sheath.
Consider an ion in the pre-sheath arriving at the sheath edge, where it enters the sheath adjacent above the substrate with a random velocity, Vo. Depending on its magnitude and direction, Vo may be resolved into a parallel component, Vx0, and a perpendicular component, Vy0, (parallel and perpendicular to the surface of the substrate). On average, at this interface (the sheath edge), the parallel and perpendicular components of ion velocity are similar in magnitude. Upon entering the sheath, the ion is accelerated toward the surface by the sheath electric field to acquire high kinetic energy. The sheath being a narrow region of high electric field, the ion is swept across the sheath, typically with a transit time much shorter than the mean free time between collisions. In the absence of collisions, the increase in ion velocity in the sheath is directed parallel to the field, i.e., perpendicular to the surface. In other words, if the velocity with which the ion exits the sheath and impinges on the surface is V1 then its perpendicular component, Vy1, is increased such that Vy1>>Vy0, but its parallel component, Vx1, is unchanged, i.e., Vx1=Vx0. In addition, since the magnitudes of Vx0 and Vy0 are similar, generally, Vy1>>Vx0.
Note that increasing ion speed in the direction of the field reduces a ratio of the parallel component to the perpendicular component. This, in turn, reduces the ion angle, i.e., brings it closer to the vertical. When the ion enters the sheath at the sheath edge, the ion angle, θo, =tan−1 (Vx0/Vy0) and, by the time it strikes the surface, the ion angle, θ1, =tan−1 (Vx1/Vy1). Since, Vx1=Vx0 and Vy1>>Vy0, it follows that θ1>>θo. Furthermore, it can be shown from the inequalities discussed above that the reduction in ion angle (θo−θ1) is larger for an ion that enters the sheath with a bigger angle, θo. Thus, the acceleration in the sheath tightens the ion angle distribution, making the ion flux more directional. The sheath electric field may be increased by applying, for example, a negative substrate bias. However, the increased average energy with which the surface is bombarded may have some undesirable effects such as surface damage, surface roughness, and heating.
As mentioned above, reducing the ion incidence angle is important, especially in the etching of high aspect ratio features. The ion angle may be reduced either by increasing the vertical component of ion velocity (Vy1) by increasing the sheath electric field with substrate bias or by reducing the horizontal component of ion velocity (Vx0) in the pre-sheath. The kinetic energy associated with Vx0 originates in random collisions; hence, Vx0 may be reduced by reducing the random component of ion kinetic energy, i.e., by reducing Ti. In the absence of randomizing collisions in the sheath (as described above), the random component of ion kinetic energy remains unchanged from that of the ions in the pre-sheath region near the sheath edge. With no further increase in the random component of ion kinetic energy, there is no further increase in Ti inside the sheath, i.e., Ti reaches its highest value at the edge of the sheath.
As discussed above, the ion temperature (Ti) refers to the randomized average energy of ions, randomized by collisions with ions in the charge-neutral bulk plasma and the quasi-neutral pre-sheath. Thus, Ti is the net result of kinetic energy gained by ions through collisions, (e.g., elastic collisions with electrons) and kinetic energy lost by ions through collisions (e.g., collisions with neutrals). An upper limit for Ti may be determined by considering an energy balance condition in the steady state, where ion energy lost by collisions (mostly with neutrals) is balanced, on average, by ion energy gained by collisions (mostly with high energy electrons). This relates the ion temperature, Ti, (which attains its highest value at the sheath edge) to the electron temperature, Te, As known to persons skilled in the art, an approximate mathematical expression using the theory of plasma physics may be derived, wherein Ti ∝Te2/3. From this analysis, it is apparent that a reduction of Ti in the bulk plasma and the pre-sheath may be achieved by reducing the electron temperature there. As explained above, such a condition would improve the ion incidence angle at the substrate being processed by achieving a more collimated ion flux incident on the surface.
Embodiments of the present disclosure use a pulsed DC magnetic field in the chamber to lower Te in the plasma locally above the substrate in order to reduce the spreads in ion angle and energy of the ions in the ion flux used to process the substrate. While it is desirable to reduce Te of the plasma above the substrate in order to achieve a better controlled and highly collimated ion flux incident on the substrate, a high Te may still be desirable for efficiently generating ions and radicals to sustain a desired plasma density in the plasma chamber. Various embodiments of a plasma processing system and method, described in this disclosure, provide the advantage of locally modulating Te during processing, using the pulsed DC magnetic field. As explained in further detail below, a timing of the pulsed DC magnetic field.
The DC magnetic field being a pulsed DC magnetic field, the profile of electron temperature changes with time. A first profile is produced during a time when the DC magnetic field is present and a second profile is produced during a time when the DC magnetic field is absent. An average Te in the central region of the plasma chamber 110 is a first Te (Te1) for the first profile, and a second Te (Te2) for the second profile, where Te2 is greater than or equal to Te1. The ratio, Te2/Te1, increases if the DC magnetizing current is increased.
In the example embodiment in
In some embodiments, a portion of the walls of the chamber 110 may comprise a conductor (e.g., aluminum) coated with an insulator (e.g., yttria), where the conductor is, typically, coupled to a reference potential, commonly referred to as ground.
As mentioned above, RF source power and pulsed bias power may be coupled to the gas in the chamber 110 to generate plasma. An RF source signal may be coupled to the first RF electrode 106 and the bias signal may be coupled to a second RF electrode 108.
In the embodiments described in this disclosure (e.g., system 100 in
It is noted that a person skilled in the art may adapt an embodiment having the ICP configuration is used (e.g., system 100 in
The plasma may be sustained by RF power received in an upper portion of the chamber 110 vertically below the first RF electrode 106. However, the plasma extends horizontally to a sidewall 116 and vertically from the ceiling to the substrate holder 104. A dashed rectangle is used in
One advantage of configuring the chamber 110 to generate plasma with power coupled from two independent sources (RF source power and bias power) is that plasma density and the vertical potential and electric field profiles may be adjusted independently. Generally, the RF source signal is used to adjust the plasma density and the bias signal is used to adjust the potential drop across the sheath and the pre-sheath.
A magnetizing signal comprising a periodic series of current pulses may be coupled to the electromagnet (e.g., the planar electromagnet 112A) to generate the pulsed DC magnetic field.
In some embodiments (e.g., the system 100), the first RF electrode 106 is an antenna shaped like a planar coil, positioned over a central portion of a ceiling of the chamber, as illustrated in
In some embodiments, the chamber 110 has conductive material in a portion of the walls (as mentioned above). That portion excludes the dielectric window 114. For example, an embodiment may have a sidewall 116 of the chamber 110, where the sidewall 116 includes a conductive material (e.g., aluminum) coated with an insulator (e.g., yttria) coupled to ground to form a ground plane around the sides of the plasma chamber 110.
In some embodiments, the second RF electrode 108 is a part of the upper portion of the multipurpose pedestal, which is the substrate holder 104 in the system 100. In the embodiment illustrated in
In some embodiments, the electromagnet for generating the pulsed DC magnetic field in the chamber 110 may be a planar electromagnet 112A. The planar electromagnet 112A, illustrated schematically in
The flow of current, in the planar electromagnet 112A and the solenoidal electromagnet 112B, is axisymmetric around the central axis of the chamber 110, thus the pulsed DC magnetic field produced by the axisymmetric current flow is also axisymmetric. Because of the large inner diameter (larger than the diameter of the substrate 102) of the coil of the planar electromagnet 112A and the coil of the solenoidal electromagnet 112B, the DC magnetizing current flows near the sidewall 116 of the chamber 110. Considering that the strength of a DC magnetic field generated by a current in a wire decreases with increasing distance from the wire, the strength of the pulsed DC magnetic field in the chamber 110 is high near the sidewall 116 and reduces with reducing radial distance from the central axis. Accordingly, on average, the magnetic field strength is low in a central region of the plasma chamber 110 above the substrate 102 relative to the higher average field strength in a contiguous edge region surrounding the central region. Note that the planar electromagnet 112A of the system 100 (in
A DC magnetic field forces mobile charged particles to rotate around magnetic field lines resulting in a spiral motion of free electrons along the field lines of the DC magnetic field in the chamber 110. When a spiraling free electron reaches the sheath then the electric field in the sheath reflect the free electron spiraling back into the bulk plasma. The stronger the DC magnetic field, the smaller is the radius of the spiral. The spiraling motion tends to confine the free electrons along the magnetic field lines. The average values of Te in the chamber 110 change with time because the DC magnetic field is a pulsed DC magnetic field.
The RF source signal, the bias signal, and the magnetizing signal are output by a first electrical circuit 120, a second electrical circuit 130, and a third electrical circuit 140, respectively, as illustrated in
The first electrical circuit 120 may comprise an RF oscillator to generate a continuous wave (CW) RF signal and a power amplifier configured to provide sufficient power to the plasma. In a CW RF signal, an RF sinusoidal voltage waveform is applied continuously. In some embodiments, the RF source signal is the CW RF signal. Generally, the first electrical circuit 120 includes an impedance matcher between the power amplifier and the first RF electrode 106 for efficient power transfer from the first electrical circuit 120 to the first RF electrode 106.
While in some embodiments, the RF source signal is a CW RF signal, in some other embodiments the RF source signal is a periodic series of source pulses. Thus, a chopper circuit may be included to allow the first electrical circuit 120 to be configured to output the RF source signal as a CW RF signal or as a periodic series of source pulses. In the example embodiments, i.e. the system 100 in
The bias signal output by the second electrical circuit 130 is the periodic series of bias pulses (mentioned above), where each period is a sum of a bias-ON time and a bias-OFF time, and for each bias pulse, the bias voltage waveform is applied only during the bias-ON time. The general architecture of the second electrical circuit 130 is similar to that of the first electrical circuit 120 in that both circuits include a signal source, a chopper circuit, a power amplifier, and an impedance matcher. The differences between the first electrical circuit 120 and the second electrical circuit 130 depend on differences between the bias voltage waveform and the RF source voltage waveform.
In some embodiments, the periodic series of bias pulses, which is the bias signal, is a periodic series of DC pulses. In each DC pulse, the bias voltage waveform that is applied during the bias-ON time is a constant DC voltage level. Hence, the periodic series of DC pulses may be generated by “chopping” a continuous constant DC voltage level output from a DC voltage source. The polarity of the DC voltage level is such that the vertically downward component of the electric field in the sheath above the substrate 102 in the chamber 110 is increased.
In some other embodiments, the periodic series of bias pulses is a periodic series of DC-burst pulses. In each DC-burst pulse, the bias voltage waveform applied during the bias-ON time is a plurality of DC pulses. In other words, if a periodic series of DC pulses were sent through another chopper then each DC pulse of the periodic series of DC pulses would be converted to a DC-burst pulse. Hence, a second chopper circuit may be included to allow the second electrical circuit 130 to be configured to output the bias signal as a periodic series of DC pulses or as a periodic series of DC-burst pulses.
As discussed above, one embodiment of the second electrical circuit 130 that can be configured to output the bias signal as a periodic series of DC pulses or as a periodic series of DC-burst pulses comprises a DC voltage source, a first chopper circuit configured to generate a periodic series of DC pulses cooperatively with the DC voltage source, a second chopper circuit configured to generate a periodic series of DC-burst pulses from the periodic series of DC-burst pulses generated by the first chopper circuit, a power amplifier, and an impedance matcher. The second electrical circuit 130 may be configured by the controller 150 with an appropriate control signal that enables the first chopper and the second chopper, depending on which bias voltage waveform is selected to be applied during the bias-ON time.
In yet some other embodiments, the periodic series of bias pulses is a periodic series of RF pulses. In such embodiments, the second electrical circuit 130 comprises an RF oscillator, a chopper circuit, a power amplifier, and an impedance matcher, similar to the first electrical circuit 120, described above.
The magnetizing signal output by the third electrical circuit 140 is a periodic series of current pulses. In order to generate the periodic series of current pulses, the third electrical circuit 140 includes a DC current source and a chopper circuit. The chopper circuit generates current pulses cooperatively with the DC current source, when enabled by a control signal from the controller 150. The controller 150 may synchronize a timing of the control signal to the third electrical circuit 140 with a timing of the control signal to the second electrical circuit 130 to synchronize a timing of the current pulses with a timing of the bias pulses.
The controller 150 may comprise, for example, a processor and a memory. Instructions may be stored in the memory which, when executed in the processor, generate control signals for the first electrical circuit 120 (to select a source signal), the second electrical circuit 130 (to select a bias signal), and the third electrical circuit 140 (to select a magnetizing signal). The control signals from the controller 150 may also be used to synchronize the pulsed signal waveforms generated in the first electrical circuit 120, the second electrical circuit 130, and the third electrical circuit 140 in order to synchronously apply the source signal, the bias signal, the magnetizing signal, and the reference potential to generate the plasma in the chamber 110, as needed for the plasma process.
In the embodiments of plasma processing systems described in this disclosure, prior to coupling the magnetizing signal to the electromagnet (e.g., the planar electromagnet 112A or the solenoidal electromagnet 112B), the magnetizing signal is synchronized with the bias signal to flow the DC magnetizing current only during the bias-ON time. The synchronization ensures that the electromagnet is magnetized during the bias-ON time and demagnetized during the bias-OFF time. Generally, applying bias power increases the potential drop and the vertical electric field in the sheath, thus increasing the vertically directed velocity (Vy) with which the ions impinge on the substrate 102, when the bias power is present. Because of the increased kinetic energy of the incident ions, a major portion of the anisotropic etching occurs during the bias-ON time. During the bias-OFF time, the processing progresses with an ion flux having a lower kinetic energy and with radicals diffusing out from the plasma reacting with the substrate 102.
By synchronizing the current pulses with the bias pulses, the magnetic field is turned on only when the bias power is coupled to the plasma in the chamber 110 (i.e., the bias-ON time). This timing of the pulsed DC magnetic field is utilized in the embodiments of this invention to lower Te when anisotropic etching is enhanced with bias power, and allow Te to be returned to its high value after the bias power is removed. Between successive bias pulses (i.e., the bias-OFF time), the substrate 102 remains exposed to the ion flux having a lower kinetic energy and a diffusive radical flux from the plasma, which continue to modify the surface. Returning Te to the higher value with the RF source power restores the plasma density to an initial value. This technique of lowering Te (hence, providing a highly collimated ion flux) during a period of enhanced anisotropic etching along with the raising Te (hence, providing a high plasma density) in at least a part of the time between successive periods of high anisotropic etching is advantageous for etching high aspect ratio features.
During the time the DC magnetizing current is flowing (i.e., the bias-ON time), a magnetic field profile is established in the chamber 110. As mentioned above, the electromagnet (i.e., the electromagnet 112A for system 100 or the electromagnet 112B for system 101) is configured such that the average magnetic field strength is low in a central region of the plasma chamber 110 above the substrate 102 relative to the higher average field strength in a contiguous edge region surrounding the central region. Although the DC magnetic field profile in the chamber 110 of the system 100 (in
Being charged particles, the free electrons and ions, moving in random directions in the bulk plasma, experience a force due to the DC magnetic field. The force is directly proportional to a product of the particle velocity, charge, and the DC magnetic field strength. On average, the massive ions move slowly relative to the much lighter free electrons in the bulk plasma. Thus, on average, compared to the free electrons, the ions experience a much weaker force due to the DC magnetic field. Moreover, the acceleration due to a force is inversely proportional to mass. Thus, the impact of the DC magnetic field on ion motion may be ignored.
Consider a free electron in the bulk plasma in the presence of the DC magnetic field. According to the theory of electromagnetism, the force acting on the free electron due to the DC magnetic field (commonly referred to as the Lorentz force) is directed perpendicular to both the direction of its velocity and the direction of the local magnetic field. This causes the free electron to gyrate in a spiral trajectory along a magnetic field line of the DC magnetic field. A radius of the spiral is commonly referred to as the Larmor radius. In general, for a particle having a charge q and mass m, moving at a velocity v in a DC magnetic field of strength B, the Larmor radius, r, is given by r=(mv)/(qB). In other words, each particle is confined within a region that reduces with increasing magnetic field strength and reducing particle mass.
Accordingly, application of the external pulsed DC magnetic field confines the lighter mass free electrons very effectively, and a similar effect on the heavier ions may be ignored. The DC magnetic field does not change the kinetic energy of the free electron. However, as known to persons skilled in the art, the magnetic confinement of the free electrons affects collisions with electrons such that, the transport of electrons via collisions (represented by a diffusion coefficient) and the transport of the random component of electron kinetic energy (represented by a thermal conductivity coefficient), is reduced in the direction along the magnetic field lines (i.e., perpendicular to the direction of the pulsed DC magnetic field). In the example embodiments (system 100 and system 101), the magnetic field lines in the plasma pass roughly from the top to the bottom of the chamber 110, consistent with the configuration of electromagnet 112A in system 100 (see
A general method 200 for achieving a time-varying Te in the plasma above the substrate using the pulsed DC magnetic field is described below with reference to a flowchart illustrated in
Referring to
A controlled gas flow may be maintained using a gas flow system coupled to the chamber 110 of the example embodiment of a plasma processing system 100, illustrated in
The gas may be ionized by coupling an RF source signal to a first RF electrode (e.g., the first RF electrode 106 in
As described above and indicated in box 206 in the flowchart of method 200, illustrated in
In the method 200, a pulsed DC magnetic field is generated to modulate the electron temperature profile of the steady state plasma. In other words, Te is varied not only with position but also with time by applying the pulsed DC magnetic field. The pulsed DC magnetic field is generated by coupling a magnetizing signal to an electromagnet, where the magnetizing signal is a periodic series of current pulses, as indicated in box 208 in the flowchart of method 200, illustrated in
As mentioned above, a low Te of the plasma above the substrate is desired to reduce the spreads in ion angle and energy distributions, but a high Te is desired to efficiently generate ions and radicals to sustain a desired plasma density. Both desires may be satisfied by a time-varying Te in the plasma above the substrate. Thus, embodiments of the present disclosure modulate Te in the plasma above the substrate to provide a low Te1 (an average Te in the central region during a time period when the bias power is present) and restore Te to a high Te2 (an average Te in the central region during a time period when the bias power is absent), as explained above. Hence, as mentioned above, prior to coupling the magnetizing signal, the magnetizing signal is synchronized to be in phase with the bias signal (box 210). In the example system 100 (in
In some other embodiment, the phase of the source pulses may be selected such that the RF pulses of the source signal are present only during the bias-OFF times, i.e., the pulses of the source signal are “180° out of phase” with the bias pulses. In other words, the source-ON time is coincident with the bias-OFF time and the source-OFF time is coincident with the bias-ON time. Of course, the current pulses of the magnetizing signal, the bias signal being in phase with the magnetizing signal.
It is understood that other set of waveforms may also be used with the method 200, summarized in the flowchart illustrated in
Example embodiments of the invention are described below. Other embodiments can also be understood from the entirety of the specification as well as the claims filed herein.
