Lithography
Optical, E-beam and x-ray
Concept of mask transfer for IC fabrication
Quartz plate
Chromium pattern to block light
The wafer is coated with photosensitive resist, exposed and then developed and rinsed
Process sequence of a typical lithographic process
Process sequence of the lithography process. Note that for a Pentium chip, about 80 iterations of this sequence must be undertaken.
Definition of positive and negative resist: Positive resist dissolves when exposed with radiation, whereas negative resist stays
Description of some of the terminology for alignment of masks. The critical dimension is the desired dimension on the sample.
Three commonly used methods for exposing resists: contact printing often wears out the masks, and both contact and proximity printing need mask features equal to the sample features
Projection printing allows a 5:1 or 10:1 reduction of the feature sizes from mask to sample
The light source can also be an excimer (fluoride) laser (ArF or KrF)
This lens may contain many (100) elements
U-V photolithography relies mainly on 5:1 or 10:1 steppers like this one. For research applications with odd shaped samples, contact aligners are sometimes used
Throughput = 20 wafers/hour
Diffraction effects will be different for proximity and projection exposure methods. Phase shift masks are used so the projection printing intensity profile becomes more square
Typical resist datasheet
Resolution test:
Resist coverage over nonplanar substrates:
The interference pattern which is built up at the reflecting wafer surface will result in high and low field regions in the resist. This results in the resist being non-uniformly exposed in the vertical direction. The photoresist develops in steps.
Typically, the resist is chosen for resolution and contrast
Critical dimension
Depth of focus dependence on CD
Adhesion promoter: This is a material which helps the surface bonding between photoresist and substrate. Typically, a good adhesion promoter minimizes delamination during further processing like wet etching etc.
Immersion Lithography for ever-smaller features
Vector-scanned electron beam lithography is the highest resolution lithography method presently available. Resolution below 10 nm is obtainable with this technique, since the wavelength of electrons is much shorter than the wavelength of light
Similar to a CRT tube, except with much smaller spot size
Claudius Ptolemy 2nd Century BC
Salvino DArmate 12th Century
1590 - Two Dutch spectacle makers, Zacharias Jansen and his father Hans started experimenting by mounting two lenses in a tube, the first compound microscope.
1609 - Galileo Galilei develops a compound microscope with a convex and a concave lens.
1665 - Robert Hooke's book called Micrographia officially documented a wide range of observations through the microscope. 1674 - Anton van Leeuwenhoek used his knowledge of grinding lenses to achieve greater magnification which he utilised to make a microscope, enabling detailed observations to be made of bacteria.
1826 - Joseph Jackson Lister created an achromatic lens to eradicating the chromatic effect caused by different wavelengths of light.
1860s - Ernst Abbe discovers the Abbe sine condition (a condition that must be fulfilled by a lens or other optical system in order for it to produce sharp images), a breakthrough in microscope design, which was until then largely based on trial and error. 1931 - Ernst Ruska starts to build the first electron microscope.
Comparison of Microscopes
Electron Microscopes
Ruska, Borries
Electron Lenses
Vector-scanned electron beam lithography
A focussed electron beam is scanned over the sample to define small lithographic patterns. Problems with low throughput
Polymethylmethacrylate or plexi-glass is a common high contrast positive e-beam resist
Difference between positive and negative electron beam resists: The dose, given in Coulombs/cm2 determines the ultimate speed at which the pattern can be written
20nm
40nm
60nm
80nm
100nm
150nm
200nm
17nm diameter 80nm c-c spacing
100Gbits/in2
Nanostructures
Comparison of the resolution and sensitivity of some electron beam resists. Notice that the linewidth has shrunk to 0.01 microns by now
Concept of proximity effect. Secondary electrons and backscattered electrons from a beam-written feature can expose features close to the beam. This leads to problems in dense patterns.
No secondary electrons expose the ends and the exposure dose is low
Too many secondary electrons expose inside corners of patterns Here is a result of the problems caused by proximity effects.
Schematic of the SCALPEL process (Lucent)
A high voltage electron beam is transmitted through a silicon nitride membrane with high mass/thickness contrast features to form a shaped electron beam.
SCALPEL projector design
Electromagnetic lenses project the features from the mask onto the sample. Lens distortion limits the maximum field of exposure. High voltages are needed to avoid chromatic aberration.
Simple high-energy xray lithography system. In many ways, this resembles an optical contact mask aligner
The Be window is used to allow x-rays to pass through from the Pd target in the vacuum
He is added between sample and window for minimum scattering
Geometric considerations for proximity x-ray exposure system
EUV lithography
A soft x-ray beam is projected through several reflective x-ray mirrors onto the sample. (4:1 reduction) A very high brightness x-ray beam is needed to compensate for absorption losses. Mask inspection and repair are still unresolved problems.
Beam path of EUV systems
Laser Produced Plasma LPP Source (vs. Discharge Produced Plasma (DPP))
Recent progress in high power lasers => LPP-based sources, At the SEMATECH workshop, 12 suppliers and R&D groups described the readiness and development plans of three types of high power laser systems designed to bombard tin to produce EUV light. Pulsed CO2 lasers generate the most power (7 kilowatts), but are still being developed and have not been subjected to continuous operation Nd:YAG laser modules with 1500 W power capable of continuous operation are already available as commercial products, with 3000 W modules in development. (Nd:YAG systems utilize a neodymium yttrium aluminum garnet crystal as the lasing medium.) Fiber-based lasers require less electricity, cost less and can be easily scaled but are also less developed than CO2 and Nd:YAG lasers. Although each of these LPP-based approaches falls short of the 10-20 kilowatts of laser power that will be needed for commercial EUV scanners, the necessary wattage can be obtained by multiplexing less powerful systems
ASML EUV tools in use or in development
EUV lithography development
Available Year 2006 2010 2012 EUV Tool ADT NXE:3100 NXE:3300B Resolution 32 nm 27 nm 22 nm Depends on resist diffusion length Throughput 4 WPH 60 WPH 125 WPH Dose, Source 5 mJ/cm2, ~8W 10 mJ/cm2, >100W 15 mJ/cm2, >250W 15 mJ/cm2, >350W
2013
NXE:3300C
150 WPH
Fugure EUV system
X-ray resists are very similar to electron beam resists. PMMA is a common x-ray resist, too.
It is desirable for the x-rays to be absorbed strongly in Au and PMMA, but not in the Be window and the atmosphere
Addition of organometallic materials to the resist make the xray resists more sensitive, since more x-rays get absorbed in these materials. In this schematic, the process sequence of a surface enhanced resist is shown. In such a resist, the surface is selectively made more resistant to a later oxygen etch. Another common method to do this is by surface silation or diffusion of silicon into the unexposed regions of the resist
The down side of proximity x-ray lithography: a very thin x-ray transparent mask has to be constructed with features which efficiently absorb x-rays. The feature height (0.6 microns) here severely limits the minimum lateral feature size which can be exposed.
Concept of multilevel resist which can be used for planarization. This becomes extremely important as the depth of focus is reduced in modern Deep-UV optical lithography
Concept of an inorganic resist. GeSe used to be a popular self-developing resist.
Application ranges of different lithography systems: Note that resolution and throughput will make you choose an exposure method
Working Principle of ELIPS
Focussing Condition:1 Bd 2mV = e Vo cos V
B and E fields quartz substrate u-v ewafer e emitter u-v opaque mask resist Figure 1. The ELIPS principle
-
3 kV
Specimen Holder Geometry
3 kV acceleration voltage applied to the top of the emitter. 1.5 mm gap between sample and emitter
alignment laser focussed with a diffractive optical lens on mask surface onto alignment mark
applied magnetic field
electron emission from the emitter surface
u-v light from a filtered Hg lamp
Figure 4. Sample and mask holder assembly of ELIPS
Electron Image Projection without Magnetic Field
Electron Image Projection Lithography
Projected pattern
Mask pattern Projected image of Caltech logo
ELIPS vacuum system and magnet
magnetically coupled feedthrough for specimen exchange sample load-lock Turbomechanical pump sample sandwich with emitter
Helium Dewar superconducting magnet
ion pump mechanical pump
The sample is transferred into the high magnetic field of a warm bore NMR magnet and the pattern is projected from below using a Hg light source
Hg light source
Figure 5. Prototype ELIPS vacuum system
Electron Image Projection System (ELIPS) operation
Magnetically coupled feedthrough Load-lock
Turbo-pump
High-field Magnet
Test Gratings of Au on Silicon Substrates.
We have demonstrated 90 nm resolution.
Ion Etched Test Structures on ELIPS patterned Si
Optical Gratings
Magnetic Media
Nanoimprint Lithography
A mold is pressed into a resist to form the pattern
Unresolved questions: Uniformity over large areas? Wear of die?
Epitaxial Deposition
I.
CVD deposition processes: Si epitaxy
CVD: Chemical Vapor Deposition Epitaxy: from Greek epi (upon) taxis (ordered)
Boundary layer formation in a horizontal Si CVD reactor
Two approaches to epitaxial growth: CVD and MBE. In CVD a gas is flowed over a crystalline seed substrate, whereas in MBE an evaporation is used. Discrete BJTs were made in bulk wafers, but but the resulting excessive collector resistance limited high-frequency response. Solved by epitaxial growth of high-R layer on low-R substrate. Two fundamental advantages of epitaxial over bulk wafers: 1. One or more epi layers allows control over doping profile. 2. Physical properties of epi layer are different from bulk (generally O2 and C- free.)
Influence of Reynolds number on the deposition uniformity Gas density Re = {Dr v } /
Diameter gas velocity viscosity
Isotherms in a CVD reactor
CVD reaction to deposit Si from silicon tetrachloride/hydrogen
However, the overall reaction is more complicated
Temperature variation of the equilibrium gas phase composition at 1 atm and Cl/H = 0.01
At different temperatures, different species are stable and the composition of the gas changes
Species detected by IR spectroscopy in a real reactor containing silicon tetrachloride and hydrogen
Note that the growth rate depends on temperature and the gas source used
CVD layer growth rate of Si versus temperature. Note that it is negative at low and high temperatures.
Schematic image of the CVD process showing the influence of the addition of arsine for doping Si.
Gas reaction for arsine gas used to dope Si layers
The influence of the growth rate on the Arsenic concentration in the layer
The layer can also be unintentionally doped by dopants which are in the substrate. These can volatilize and redeposit in the new layer which is grown.
Radiant-heated barrel reactor
Three common susceptor shapes
Susceptor geometries which are commonly used in CVD reactors
It does not pay to be too impatient
Maximum growth rate for which monocrystalline silicon can be obtained as a function of temperature
MOLECULAR BEAM EPITAXY
Pressure in Torr Mean free path in cm
MBE allows integration of analytical techniques into the system for insitu monitoring.
Flux of various dopant species versus oven temperature
The defect density of the MBE grown layers depends on the background pressure used
This means that the system has to be pumped down very well
The defect density also depends on the growth temperature used
A interesting fact about silicon is that the refractive index changes significantly when the material is heavily doped free carrier absorption?
II. Dielectric deposition techniques:
For poly-silicon films we can use this simple reaction (pyrolizing silane) at 600-650 degrees C in a low-pressure reactor
These are some geometries which are suitable for CVD deposition.
Sometimes it is desirable to deposit in a plasma, and this is then called PECVD. This method generates very high quality films.
The down side to CVD deposition of materials:
Arrhenius plot for polysilicon deposition shows the partial pressure dependence of the growth rate.
Deposition of Silicon dioxide by CVD
Note: This reaction can be done at a low enough temperature so that SiO2 can be deposited after aluminum metallization
Often, phospho-silicate glasses (PSG) are deposited to planarize devices and to protect circuits from degradation. This is because such glasses have a low melting temperature
Tetraethoxysilane (TEOS) deposition is normally done at 650-750 C (provides excellent uniformity and conformal step coverage)
Dichlorosilane deposition is done at 900 C
PSG (phosphosilicate glass) is a very popular material for semiconductor processing. It is a mixture of SiO2 and P2O5: Any amount of P2O5 can be incorporated into SiO2 Si becomes more hygroscopic as P2O5 content is increased The P2O5 content is usually 2-8 wt% For undoped SiO2, the built-in stress is 3x108 N/m2 (tensile) with 2 wt% P2O5, the stress is close to 0 The thermal expansion coefficient of PSG can also be altered with the composition of phosphorus. PSG can be designed to thermally match the underlying Si substrate PSG is sometimes used as a phosphorus doping source for semiconductor doping.
Oxygen can be used to surface-catalyze the deposition reaction
The deposition temperature strongly influences the deposition rate, since it determines how efficiently the source material is decomposed.
Good step coverage of the deposited oxide is very desirable to avoid short circuits and high electric field regions in the circuit. Mostly, step coverage of the sidewalls is used to characterize the deposition quality.
Scanning electron micrographs of 4.6 wt% P-glass annealed in steam at 1100 C for: 0 minutes 20 minutes 40 minutes 60 minutes
3200X magnification
The quality of the silicon dioxide depends on the deposition reaction, and the temperature. Notice that the dielectric strength and the etch resistivity are highest for hightemperature deposited oxides
Silicon nitride deposition by CVD Silane/ammonia reactions are commonly used between 700 C and 900 C. Silane can also be replaced by dichlorosilane.
Resistivity of the plasma silicon nitride versus composition. Note that the higher silicon content layers are more conductive.
Some plasma-assisted CVD reactions: SiO2 is formed from silane and nitrous oxide in an argon plasma Si3N4 is formed by reacting silane and ammonia in an argon plasma or by reacting silane in a nitrogen discharge
Summary of CVD deposition techniques and their advantages and problems
Summary: APCVD is a simple process, has relatively poor step coverage, and suffers from contamination problems. It is usually used to deposit low-T oxides. LPCVD provides excellent purity and uniformity oxides, can be used on large wafers, has a low deposition rate and requires high temperatures PECVD provides low-temperature, fast deposition and good step coverage. It sometimes suffers from chemical contamination, and often is used for lowtemperature insulator deposition.
Thermal oxidation of Silicon: This is one of the most thoroughly studied materials systems in the world.
Silicon can be oxidized either using molecular oxygen or water vapor:
Silicon dioxide: Eg = 8eV Density: 2.27 g/cm3 Resistivity:1014-1016 ohm.cm
Hf = 210 kcal/mole = 1eV/molecule
Etch rate in Buffered HF: 100 nm/min Refractive index: 1.46 Melting point: 1700 C
As the oxide layer grows, it consumes some of the silicon from the surface of the wafer. In fact, for an oxide of thickness d, a layer of silicon 0.44d thick is consumed. Volume expansion
Oxygen diffusion
Si oxidation fluxes F1, F2, F3. These should be equal at equilibrium, and the rate of oxidation is limited by the slowest of the three
The gas phase flux can be linearly approximated by assuming that the flux of oxidant from the gas phase at the gas/oxide interface is proportional to the difference between the oxidant concentration in the bulk of the gas CG and the oxidant concentration adjacent to the oxide surface Cs.
Gas-phase mass transfer coefficient
Henrys law tells us that the equilibrium oxidizing species concentration in the oxide to that in the gas phase is Partial pressure of oxygen adjacent to the oxide surface
Henrys law constant = H
Equilibrium concentration in the oxide at the outer surface
Co = Hps
and
The equilibrium bulk concentration of the oxide
C* = oxidant which would be in equilibrium with partial pressure of gas. The driving force for the reaction is C*-Co
Partial pressure of oxygen in the bulk of the gas
Derivation of F1:
F1: Reaction at the surface of the SiO2/gas interface We use Henrys law along with the ideal gas law to rewrite CG and Cs
F1
Therefore, combining all of the equations Where h = hG/HkT
Gas-phase mass transfer coefficient in terms of the concentration in the solid
F2: Diffusion of oxygen through the oxide layer
F2
This is just the diffusion flux resulting from the concentration gradient in the oxide
F3: Reaction at the SiO2/Si interface to form new oxide We can assume that the reaction speed for oxidation of Si is proportional to Ci F3
Chemical rate constant for reaction
When we set F1=F2=F3, we can solve simultaneous equations and obtain expressions for Ci and Co
Several possibilities could occur: a. When the diffusivity is very small, Ci~0 and Co~C*. This is the diffusioncontrolled case. F = DCoks/(D+ksd) b. When D is large, Ci=Co=C*/(1+ks/h) and this is the reaction-controlled case
Now let us assume that N1 is the number of oxidant molecules incorporated into a unit volume of the oxide layer. Since the oxide has 2.2x1022 SiO2 molecules/cm3 and one O2 is incorporated into each SiO2 molecule, whereas 2 H2O molecules are incorporated into each SiO2 molecule, N1 equals 2.2x1022cm-3 for dry oxygen and twice this number for water-vapor oxidation.
The flux of oxidants reaching the oxide-silicon interface:
Solution to this differential equation
The shift in the time coordinate to account for the presence of the initial oxide layer d.
The thickness of oxide as a function of time:
For t>>: (parabolic rate) For short oxidation times when (t+)<<A2/4B: (rate is linear)
The effect of temperature on the linear rate constant B/A in microns/hr.
Oxide thickness versus oxidation time for several temperatures in dry oxygen
Note that this is a log/log plot
Oxide thickness versus oxidation time for silicon in water vapor at 640 Torr (close to an atmosphere)
Typical values for A,B, for different temperatures and Si surface orientations are given here:
Effect of dopants on the oxidation rate: Typically, the reaction rates are increased when silicon is heavily doped Case for Boron doped Si:
Effect of Phosphorous dopants on the oxidation rate
Oxidation rate constants as a function of the phosphorous impurity concentration of the Si at 900C.
Boron and phosphorus increase the oxidation rate. Chlorine and HCl additions can change the oxidation rate: 2HCl + O2 => H2O + Cl2
High pressure also increases the oxidation rate: Note that this oxidation is done at 20 atmospheres steam.
Dependence of the oxidation rate constants on the temperature and pressure of steam
What happens when we oxidize a doped silicon surface? Dopants from the Si can redistribute themselves and either prefer the oxide or avoid it.
Note that the oxide is an excellent diffusion mask for B, but not for Gallium
Typical charges in a thermally oxidized silicon. Note that Na and K are positive charges which can be incorporated into the oxide if care is not take to clean the wafers before oxidation.
Note: HCl addition to oxides can passivate the oxide (I.e. traps and neutralizes positively charged ions)
Redistribution of impurities during oxidation:
We can define a segregation coefficient m, such that: m= equilibrium concentration of impurity in Si equilibrium concentration of impurity in oxide
When m>1, the oxide rejects the impurity When m<1, the oxide concentrates the impurity Even when m=1, redistribution can occur since the oxide/semiconductor interface is advancing into the Si substrate. B has a m < 1 P has a m > 1
Impurity segregation at the Si/SiO2 interface resulting from thermal oxidation. (a) Diffusion in oxide is slow (Boron), (b) Diffusion in oxide is fast (Boron) (c) Diffusion in oxide is slow (Phosphorus) (d) Diffusion in oxide is fast (Gallium) Notice that boron likes to go into the oxide, whereas P dopants want to avoid it.
The segregation coefficient m shows the equilibrium concentration of the impurity in silicon to that in the oxide at the interface. Most of the work on these coefficients has been done for Boron:
Metallization
Ohmic contact resistance
For low doping in a semiconductor, the contact resistance can be represented by:
A*: Richardson constant At higher doping concentrations, the barrier width decreases, and tunneling becomes important:
Measured and modeled contact resistance as a function of the donor concentration and barrier height
Barrier heights for several metal systems for both n- and p-type Silicon
Schematic view of an evaporator: The high vacuum pump may be a diffusion pump, an ion pump, a cryopump, or a turbopump
Note that substrates may be arranged on a hemispherical holder and rotated for even evaporation
Pump-down rate of a vacuum system: If we assume that a vacuum station has a volume V (in liters), and no leaks, and is equipped with a pump with speed S (in liter/sec). P, the pressure at time t, is related to the pumping time by:
Here, P0 is the initial pressure, and Q is the outgassing rate of the vacuum system Note that after the initial pumping time, the second term dominates the vacuum pressure
Thickness measurement and monitoring. If a microbalance is used to weigh the substrate before and after deposition, the increase in mass m can be related to the thickness t by:
Most modern vacuum deposition systems have quartz microbalances right in the evaporation path to measure thickness during evaporation
Density of the sample
Area of deposited sample
The Kinetic Theory of Gases: This theory yields two concepts that are useful in vapor deposition: The rate of bombardment of an exposed surface in a chamber is:
Diameter of gas molecules
Where N is the bombardment rate in molecules cm-2 s-1 for a gas of molecular mass m at temperature T in kelvins and pressure p; Another way of writing this relationship is:
The mean-free path can also be expressed
The fraction of total molecules no not suffering a collision while traveling distance d is:
Grammolecular mass
Pressure in Pa
The rate of evaporation of metal from a melt is estimated by the Hertz-Knudsen equation:
The total loss RT per unit time from the source may be found by integrating over the source area: RT = R dA
D is the deposition rate in g/cm2s from the angle between the receiving surface normal and the direction back from the source:
The deposition rate can be determined from the geometry of the system and the source/sample distance
Metallization sources:
1) Thermal evaporation from a W coil acting as support and heat source 2) E-beam evaporation source (note the magnetic field causes the electrons to curve 3) Inductively heated source
Problem with step coverage from a vapor deposition system with low surface mobility
Is is obviously very difficult to metallize onto structures which are undercut and still have good step coverage
Another method for metallization commonly used is sputter deposition: Here, we show a planar magnetron system,where the magnets are permanent or electromagnets. Note the selective erosion of the target. anode Sputter deposition often provides more conformal sidewall coverage
Views of lift-off crossections for (a) evaporated metal and (b) sputtered metal. Note that the high-shadowing features in evaporation and the undercut masks which must be used in sputtering.
Chemical Vapor Deposition for Metallization: Usually, we start with a gas source containing the metal (metalorganic), which breaks apart at the hot sample surface Example: WF6 can be broken down using three mechanisms:
Note that low-pressure CVD provides excellent step coverage
CVD reactions for typical refractory metals used in the semiconductor industry. Note the high temperatures needed for deposition:
Aluminum can be CVD deposited by using tri-isobutyl-aluminum:
Followed by:
CVD reactor
Simplified view of a low-pressure CVD reactor system. To obtain enhanced reactions, the furnace could be augmented with a plasma source or an intense light source
Schematic view of an aluminum film contacting a large window (10x10 micron). Note the Al spikes (pits) in the silicon and the precipitated Si.
A schematic of a VLSI aluminum contact to a window (1.5 x 1.5 micron). Note that the pit in the silicon can fill the window
The reason for Al spikes stem from the local dissolution of Si, which can occur generally in IC processing. The problem is very important in VLSI processing, where the spikes can alloy through shallow junctions of the order of 0.3 microns deep. One method to avoid spiking is to incorporate Si into the aluminum evaporant to match the equilibrium concentration at a followon process temperature
Another way to avoid spiking is the use of noble silicide contacts which act as diffusion barriers.
Failure modes of metal layers: The layer morphology, and in particular the grain size, is very important in determining the conductor lifetime. Here the median lifetime t(50%) is plotted
Electromigration failure of e-beam evaporated Al on polysilicon. (a) metal is on polysilicon (b) metal is etched to reveal the polysilicon
Electromigration failure (a) S-gun magnetron-deposited Al0.5wt%Cu and (b) in-source-evaporated Al-0.5% Cu
Median conductor lifetime versus linewidth for Al alloys deposited in three different ways
Silicides and intermetallics for gates and interconnections S-gun deposition rate vs. power for Ta and Si
Sputter sources (magnetrons) and planetary system for co-sputter deposition
Properties of some common silicides which are used for contact materials
Typical heat-treatment sequence used for decreasing the resistance of a TaSi2 contact. Notice the change in the tensile stress with treatment
And when everything is done correctly, we have
R-C time constant per unit length for three conductive materials as a function of feature size. Also shown is the delay per stage of ring oscillators as a function of feature size
Concept of multilayer metallization levels
Subtractive and additive processing sequences for creating a patterned structure on a substrate
Etch
Lift-off
Wet etching of SiO2: SiO2 + 6 HF => H2SiF6 + 2H2O (rate ~ 100nm/min) In practice, HF is diluted and buffered with NH4F. This replenishes the F and provides a more even etch. Phosphosilicate glass (PSG) is etched faster than SiO2, at about 250-300nm/min
Wet etching of Si3N4: Si3N4 is etched in H3PO4 (rate ~ 12 nm/min) Pure Si3N4 is more or less inert to HF (rate < 1-2nm/min), but if it is deposited in presence of oxygen, this rate dramatically increases
Example of why directional pattern transfer is desirable. Here, the mask width would have to be compensated for the undercut during an etch
It is much more desirable to avoid the undercut and use an anisotropic etch as shown on the left:
This could be photoresist mask into which a pattern was defined with sloppy lithography
Often, the mask itself is not anisotropic and may have beveled edges. In that case, even an anisotropic etch can be compromised
Glow-discharge tube (Ne-bulb)
Vacuum must be from 1-1000 mTorr
Capacitor added here
Note that in an r-f discharge, if symmetric, the field would not result in ions being attracted to either contact. To break this symmetry, either a capacitor is introduced on the RF side, or the relative areas of anode vs. cathode are changed. This establishes a bias potential which then attracts ions to the cathode, which is sputtered away
Schematic view of an r-f discharge. The potential is shown as a function of position in the discharge for the case where the area of the powered electrode is much less than the area of all grounded surfaces in contact with the discharge.
Sputter deposition is a popular metallization technique in the semiconductor industry
Al Al Ar Argon is ionized in a plasma and accelerated to the cathode, which in this case is a plate of aluminum Ar
Simple ionization
Dissociative ionization
Dissociative ionization with attachment
Simple plasma etch reactor: Gases are introduced and ionized. The wafers are on the cathode and are bombarded by the positive ions.
Vacuum of 1mTorr to 1 Torr
(760 Torr=1 atm)
Could be mechanically pumped or diffusion pumped depending on the vacuum required
Freons are ideal for Si and SiO2 etching since SiF4 is a gas
With photoresist mask
Sometimes the wafer is heated to increase the etch rate
Reactive ion etching or Plasma etching process
More complex reactive ion etch reactor design. In this geometry, the reactive gas is forced to flow over the surface of the wafers in a radial flow. Notice that the pressure, the distance from anode to cathode, and the r-f power will determine the bias voltage.
Note that positive and negative ions, atoms, radicals, and electrons are generated in typical plasmas. Examples are given here: There are also loss mechanisms for ions due to recombination: For example, dissociative recombination occurs: e + O2+ => 2O And dissociative attachment occurs: e + CF4 => CF3 + F-
Schematic of an ion milling system. In this case, a confined plasma is used to generate ions, which are then accelerated through a grid towards the sample
Chlorine or XeF2
In such a system, it is also possible to spray reactive gas onto the surface of the sample and to accelerate the etching process 30-fold
One of the first observations of chemically assisted ion beam etching (CAIBE)
Commonly used gases to enhance the etch rate of semiconductors: Note that the gases are usually selected to provide volatile reaction products (I.e. fluorine forms SiF4 which can be pumped away, and chlorine forms AlCl3 which can also easily be volatilized)
Note that silicon will only etch fast in mixtures of O2 and CF4, since without oxygen, there is a carbon buildup on the sample which will prevent further etching
Example of etch chemistry of CF4 in a plasma etcher:
Silicon dioxide can be etched by CF4
Silicon is not etched by pure CF4. To etch it, one needs to add oxygen, or hydrogen to volatilize the carbon
Relative etch rates of silicon and some resists in a CF4/H2 mixture. Note that oxygen would etch away the organic resists much more rapidly
Typical mask erosion occurs by beveling the resist first, since the maximum milling rate is typically at 60 degrees.
Etch rate angle 90 deg
Trenching is commonly observed during ion milling and is a result from ions deflecting from the edges of the etched structure
Such trenches can be avoided by adding reactive gas to the etch and avoiding overcut
Typical gas chemistries and etch rates for materials found in VLSI electronics