The invention relates to a method of fabricating an integrated circuit in a microelectronic device. More particularly, the present invention is directed to a method of trimming an organic layer in order to reduce a critical dimension below a feature size that can be achieved by lithographic methods.
One of the key steps in the manufacture of a Metal-Oxide-Semiconductor Field Effect Transistor (MOSFET) is formation of a gate electrode comprised of a conducing layer in which a gate length is typically one of the smallest dimensions in the device. To satisfy a constant demand for higher performance devices, the gate length is continually being reduced in each successive technology generation. For current technology, a gate length (LG) as small as 60 or 70 nm is required and LG will continue to shrink as sub-100 nm technology nodes are implemented in manufacturing. One shortcoming of state of the art lithography processes is that they are incapable of controllably printing features such as a gate in a photoresist layer with a LG smaller than about 100 nm. To overcome this limitation, many semiconductor fabs use a trimming process which laterally shrinks a photoresist feature such as a line with a plasma etch step.
A conventional MOSFET 1 is pictured in
The gate pattern which defines the gate length d is initially formed by patternwise exposing a photoresist layer (not shown) on the gate layer and developing with an aqueous base to selectively remove exposed or unexposed portions of the photoresist layer depending upon the tone of the photoresist. A positive tone photoresist undergoes a reaction in exposed regions that renders them soluble in an aqueous base developer solution while unexposed portions of the photoresist film remain insoluble in the developer. In a negative tone photoresist, exposed regions are typically crosslinked to become insoluble in a developer while the unexposed portions are washed away in the developer.
A photoresist may be applied as a single layer or as the top layer of a bilayer system. A single layer photoresist is usually coated over an anti-reflective coating (ARC) that helps to control a subsequent imaging process. In bilayer applications, a pattern is formed in a thin photoresist layer and is etch transferred through a thicker underlayer that is used for its planarization and anti-reflective properties.
In some cases, a single layer or bilayer photoresist is selected in which the photoresist (imaging) layer is very opaque to the incident exposing radiation such that only a top portion near the surface absorbs energy and undergoes a chemical change. Top surface imaging techniques are frequently combined with a silylation process which forms O—Si bonds selectively in either the exposed or unexposed regions. For example, a silicon containing gas may react with a phenol group in the polymer component of a photoresist layer to yield O—Si bonds. A subsequent plasma etch that includes oxygen chemistry removes portions of the photoresist that are not protected by the O—Si bonds and thereby produces a pattern without the need of a developer solution.
The lithography process that is used to pattern the photoresist above the gate layer generally involves exposure tools which use wavelengths that are selected from a range of about 450 nm (near UV) to approximately 13 nm for extreme UV (EUV) exposures. High throughput projection electron beam tools that have the capability of imaging 50 to 70 nm resist features may be used in manufacturing in the near future. Even with the most advanced exposures tools, phase shifted masks, and other resolution enhancement techniques, the minimum feature size that can be reliably printed in a photoresist layer is not small enough to meet the demand for sub-100 nm gate lengths for most new devices. As a result, the industry has resorted to other methods that involve trimming the photoresist pattern such as an isotropic plasma etch process.
A plasma etching process employs the use of a photoresist mask to selectively allow an etchant to remove an underlying layer that has been exposed through openings in the mask pattern. In an anisotropic etch, the etchant only removes the underlying layer uncovered by the photoresist pattern. On the other hand, an isotropic etch involves removing exposed portions of the underlying layer along with some of the photoresist along the sidewalls of the openings in the pattern. Ideally, the photoresist layer is not distorted during the etch and should retain a majority of its thickness in order to avoid the formation of edge roughness and sidewall striations that may be transferred into the underlying layer. However, the polymers in photoresists developed for 193 nm or 157 nm lithography applications in new technologies do not contain aromatic groups which have an inherently high absorbance below about 240 nm. Therefore, the 193 nm and 157 nm photoresists based on acrylate and cyclic olefin based polymers are not as robust during a plasma etch process as their DUV (248 nm) or i-line (365 nm) predecessors which contain aromatic groups for high etch resistance.
Additionally, as the exposure wavelength shrinks to print smaller features in a photoresist layer, the thickness of the photoresist must also decrease to maintain a good focus and exposure latitude. Generally, the height of a photoresist line should not be more than about four times its width in order to prevent a phenomenon called line collapse. Therefore, etching sub-100 nm features using a thin 193 nm or 157 nm based photoresist mask of about 3000 Angstroms or less that has minimal etch resistance is problematic for single layer imaging schemes. Not only does the 193 nm or 157 nm photoresist have a lower etch resistance than DUV or i-line photoresists, but a thinner etch mask is used than in conventional DUV or i-line applications. At best, the amount of trimming or CD reduction is limited to about 10 nm or less for 193 nm or 157 nm photoresist layers which does not satisfy the need for large scale trimming of about 30 nm or more in many technologies.
One concept that has been practiced to overcome photoresist etching issues is to etch a pattern in a photoresist layer into an underlying hard mask that has a much better selectivity towards the gate layer than the photoresist. Once the pattern is transferred, the photoresist layer is stripped and the hard mask serves as the template for the etch transfer step to define the gate length in the gate layer. However, this method also has drawbacks including a poor profile control of the hard mask and damage to the gate layer and a silicon substrate when a hardmask such as silicon nitride is removed by phosphoric acid, for example.
In U.S. Pat. No. 6,500,755, a photoresist is patterned and trimmed on an optional cap layer on a dielectric layer. The pattern is etched into the cap layer and the photoresist is removed. A hard mask is deposited on the cap layer and is planarized to leave a portion of the cap layer exposed. The exposed cap layer and underlying dielectric layer are selectively removed by an etch to generate openings above a substrate.
In U.S. Pat. No. 6,482,726, a photoresist layer is patterned and trimmed above a second hard mask layer. The pattern is anisotropically etched through the second hard mask which may be SiO2. Once the photoresist layer is removed, a wet etch with H3PO4 isotropically transfers the pattern through an underlying first hard mask layer that is silicon nitride and laterally shrinks the first hard mask to a width less than that for the second hard mask. After the second hard mask is removed, the pattern is etched into a gate layer.
A multilayer anti-reflective coating (ARC) process is described in U.S. Pat. No. 6,548,423 in which a photoresist layer is patterned and trimmed above a second ARC which is silicon nitride or SiON. The pattern is anisotropically transferred through the second ARC and a first ARC which is CVD deposited carbon. The photoresist layer is stripped and the pattern is etched into an underlying gate layer using the ARCs as a combined hard mask.
An etching method is described in U.S. Pat. No. 6,492,068 in which a photoresist layer is patterned over a bottom ARC (BARC). The pattern is anisotropically etched through the BARC by a gas mixture including Ar, O2, Cl2, and HBr and is then transferred into an underlying gate layer with a Cl2, O2, and HBr plasma. The photoresist layer, BARC, and gate layer are trimmed simultaneously with an O2 and HBr plasma etch.
A bilayer trim etch process is found in U.S. Pat. No. 6,541,360 where a photoresist layer is patterned above an organic layer. The pattern is isotropically etched with a plasma through the bottom organic layer so that the organic layer has sloped sidewalls and a top that is smaller than its bottom. After the top layer is removed, the pattern is etched into a gate layer to give a gate length that is smaller than the width of the initial photoresist feature. However, a reproducible gate length may be difficult since it depends on trimming a sloped sidewall in the bottom layer with a high degree of control.
In U.S. Patent Application Publication US2002/0164543A1, a bilayer photolithography process is described in which an imaging layer is patterned over an underlayer and the pattern is transferred through the underlayer with an O2/HBr plasma process. The method prevents residue from forming on the sidewalls of the etched pattern that normally occurs with an O2/SO2 based plasma.
Thus, a new method of trimming a photoresist feature and transferring the resulting pattern into an underlying layer is needed that overcomes the limitations presented by a 157 nm or 193 nm single layer process, an isotropic etch, or a hard mask transfer step.
One objective of the present invention is to provide good gate profile control during a method to trim a gate layer to afford a gate length that is smaller than can be generated by a lithography method.
A further objective of the present invention is to provide a method of trimming a gate layer where the organic masking layer has sufficient thickness to allow large scale trimming and prevents sidewall striations from forming in the trimmed gate layer.
A still further objective of the present invention is to provide a method of trimming a gate layer that does not require a hard mask to be removed after the trimming step.
Yet another objective of the present invention is to provide a trimming method that is extendable to forming gate lengths below 100 nm.
These objectives are achieved by providing a substrate on which a gate dielectric layer and gate layer have been sequentially formed. A bilayer resist consisting of a lower non-photoimageable organic layer and a top photoresist layer is coated on the gate layer. The lower layer also referred to as the underlayer serves as an anti-reflective layer and is also thicker than the photoresist layer since the underlayer will function as an etch mask during a subsequent pattern transfer into the gate layer. The underlayer is formed by coating and baking a commercially available BARC, i-line photoresist, or DUV photoresist. When an i-line or DUV photoresist is employed as the underlayer, the underlayer is hard baked to destroy the photosensitive component and prevent interaction with the top photoresist layer. The top photoresist layer has a silicon-containing composition and is preferably a positive tone photoresist. The top photoresist layer is preferably patternwise exposed with 193 nm or 157 nm radiation to enable the lithography process to print sub-100 nm features. Optionally, the top photoresist layer is imaged with a shorter wavelength than 157 nm such as 13 nm radiation from an EUV source.
A key feature of the present invention is that the photoresist pattern is anisotropically transferred through the underlayer with a plasma etch process based on N2/H2 and SO2 chemistry. It is important to achieve vertical sidewalls on the underlayer with no etch bias and no line collapse. Next, another crucial step is a trimming process that involves a plasma etch with Cl2, HBr, and O2 gases. The bilayer resist profile retains vertical sidewalls while the width of the feature is reduced considerably to a size that cannot be achieved by a lithography method. The pattern is then transferred through the underlying gate layer to produce a gate length that is equivalent to the width of the trimmed feature in the underlayer. The photoresist pattern is typically consumed during the gate etch and the underlayer may be removed in a following step by oxygen ashing which results in a patterned gate layer or gate electrode in a partially formed MOSFET.
In a second embodiment, a bilayer resist comprised of an upper photoresist layer on an underlayer is coated on a gate layer formed on a gate dielectric layer as in the first embodiment except that the photoresist layer does not contain silicon. The top photoresist layer is highly absorbing of the exposing wavelength so that a chemical reaction is induced only near the surface of exposed areas. The top surface imaging technique continues with a silylation procedure to selectively silylate regions of photoresist layer above portions of gate layer where a gate electrode is to be subsequently formed. In one embodiment, the exposed regions are silylated. Alternatively, the exposed regions are rendered inactive and the silylation occurs in unexposed regions of the top surface. Instead of a wet development to form a photoresist pattern as in the first embodiment, a plasma etch is performed to selectively remove portions of the photoresist that are not protected by silylated surface regions.
The remaining steps in the second embodiment including the anisotropic pattern transfer through the underlayer with N2/H2 and SO2 chemistry and the trim etch with HBr, Cl2 and O2 gases are the same as described previously in the first embodiment.
The present invention is a method that involves a plasma etch process to reduce the width of a feature in a photoresist pattern in a controlled manner and is particularly useful during the formation of a gate electrode having a gate length that is substantially smaller than can be achieved by a lithography process. Those skilled in the art will appreciate that the invention is not limited to the specific examples described herein and may be applied to trimming photoresist features in other applications. Furthermore, the relative sizes of the various elements shown in the drawings may be different than those in an actual device.
The first embodiment is depicted in
A gate stack that includes a gate layer 13 on a gate dielectric layer 12 is formed on the substrate 10. First, a gate dielectric layer 12 with a thickness of about 5 to 100 Angstroms is formed on the substrate 10 by an oxidation method, a chemical vapor deposition (CVD), or by a plasma enhanced CVD (PECVD) technique. In one embodiment, the gate dielectric layer 12 is silicon oxide. Alternatively, the gate dielectric layer 12 may include a high k dielectric layer comprised of one or more of ZrO2, HfO2, TiO2, Ta2O5, Al2O3, Y2O3, and La2O3 or a silicate, one or more of Zr, Hf, Ti, Ta, Al, Y, and La formed on a thin interfacial layer that is silicon nitride, silicon oxide, or silicon oxynitride. High k dielectric layers are formed by a CVD, metal organic CVD (MOCVD), or an atomic layer deposition (ALD) process.
In one embodiment, a gate layer 13 that is doped or undoped polysilicon is deposited by a CVD or PECVD method on the gate dielectric layer 12 and has a thickness between about 500 and 5000 Angstroms. Optionally, the gate layer 13 may be comprised of amorphous silicon, SiGe, or SiGeC.
Next, a bilayer stack comprised of a top photoresist layer 15 and an organic underlayer hereafter referred to as underlayer 14 is formed on the gate layer 13. An underlayer solution that is typically comprised of an organic polymer is commercially available. One source is Arch Chemicals, Inc. of E. Providence, R.I. The underlayer solution is spin coated and then baked at a temperature of up to 230° C. to remove organic solvent and form an underlayer 14 that is inactive toward a subsequently coated photoresist layer 15. Optionally, a commercially available i-line photoresist or Deep UV (DUV) photoresist is spin coated on the gate layer 13 and is baked at a temperature up to about 230° C. to remove organic solvents and to render the photosensitive components inactive. The resulting underlayer 14 has a thickness from about 1000 to 10000 Angstroms and is preferably thicker than the photoresist layer 15 in order to prevent edge roughness and striations in the photoresist pattern from being etch transferred into the gate layer in a later pattern transfer step. The underlayer 14 typically forms a crosslinked polymer network upon heating and is used to control reflectivity during a subsequent lithographic step in which a pattern is printed in the photoresist layer 15. Therefore, an underlayer 14 is preferably selected that has a refractive index (n and k values) which minimizes reflectivity of the exposing radiation (not shown) off the underlayer and gate layer 13 into the photoresist layer 15.
The photoresist layer 15 may have either a positive or a negative tone composition but is preferably a positive tone photoresist that contains silicon and has a thickness between about 100 to 3000 Angstroms. A silicon containing photoresist layer 15 is employed to provide a high selectivity of the etchant to the thicker underlayer 14 during the etch transfer of the resulting photoresist pattern through the underlayer in a subsequent step. In other words, there should be a minimal thickness loss in the photoresist pattern 15 during a pattern transfer through the underlayer 14. Preferably, the photoresist layer is patternwise exposed with 193 nm or 157 nm radiation and is developed in an aqueous base solution to generate a photoresist pattern 15 with features (lines) having a width w1. Note that the photoresist pattern 15 is aligned so that a line formed therein is centered approximately midway between the isolation regions 11. A 193 nm or 157 nm exposure wavelength is necessary to print lines where w1 is sub-100 nm for advanced technologies. Optionally, a shorter wavelength such as 13 nm radiation from an EUV source or a projection electron beam system may be used to generate a w1 of less than 100 nm. It should be understood that the present invention also applies to printing and trimming feature sizes greater than 100 nm. However, the requirement for a controllable etch process is more demanding as w1 shrinks and the present invention is more beneficial for sub-100 nm applications where prior art methods are limited.
Referring to
Referring to
The second plasma etch for trimming the feature size from w1 to w2 in the photoresist pattern 15 and underlayer 14 comprises the following conditions: a 10 to 500 sccm Cl2 flow rate, a 1 to 50 sccm O2 flow rate, a 10 to 500 sccm HBr flow rate, a chamber temperature of about 0° C. to 100° C., a RF power from about 100 to 1000 Watts, and a chamber pressure between about 3 and 500 mTorr for a period of about 5 to 200 seconds. Under these conditions, the trimming rate to reduce w1 to w2 is about 30 Angstroms per minute.
Referring to
Referring to
A second embodiment is illustrated in
In another embodiment, a pattern comprised of the lines 23, 24, 25 is formed by performing a first etch step with an oxygen based etch process such as one including O2 and Ar, for example. Lines 23, 24, 25 with a feature size w1 are generated which include a silylated region 21a and an underlying portion of the photoresist layer 20. The photoresist layer 20 that is not protected by an overlying silylated region 21a is removed during the etch process. This development process stops on the underlayer 14. Next, a second plasma etch step involving a 10 to 500 sccm flow rate of H2, a 10 to 500 sccm N2 flow rate, a 10 to 500 sccm SO2 flow rate, a chamber temperature of about 0° C. to 100° C., a RF power from about 100 to 1000 Watts, and a chamber pressure between about 3 mTorr and 500 mTorr may be used to transfer the pattern through the underlayer 14. The etch process through the photoresist layer 20 and underlayer 14 results in lines 23, 24, 25 that have straight sidewalls and a width w1 which is a key step of this embodiment. Alternatively, the underlayer 14 has sidewalls with a retrograde profile in which the top of the underlayer has a width w1 and the bottom has a width slightly less than w1. The first and second plasma etch steps may be performed in the same process chamber. It is understood that the resulting pattern may include other features (not shown) besides the lines 23, 24, 25 as appreciated by those skilled in the art.
Referring to
Referring to
Referring again to
Referring to
The underlayer 14 is used as an etch mask in the next plasma etch step to transfer the lines 23, 24, 25 having a width w2 through the gate layer 13. An anisotropic etch process is preferred so that vertical sidewalls are formed in the etched gate layer (gate electrode) 13 and in order to form a gate length w2 with excellent process control. When the gate layer 13 is polysilicon, an exemplary process for generating a gate length w2 is a Cl2 flow rate of from 10 to 500 sccm, a HBr flow rate of 10 to 500 sccm, an O2 flow rate of about 1 to 10 sccm, a chamber temperature from about 0° C. to 100° C., a RF power of about 100 to 1000 Watts, and a chamber pressure from about 3 to 500 mTorr. Under these conditions, the silylated layer 21a and underlying photoresist layer 20 are typically consumed and are completely removed. A small top portion of the underlayer 14 may be consumed but a sufficient thickness of the underlayer 14 remains to serve as a stable etch mask during the plasma etch through the gate layer 13.
Referring to
While this invention has been particularly shown and described with reference to, the preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of this invention.