Controlling Reflectivity and Thin Film Interference in Monochromatic Lithography

 

Monochromatic exposure of photoresist films (as opposed to broad spectrum or “broadband” exposure) brings several advantages to the lithography process in terms of process capability and control. Thin film interference however, caused by the interaction of incident exposure energy with reflected exposure energy, is a significant disadvantage of monochromatic exposure processes and must be mitigated effectively in order to maintain consistent pattern fidelity and critical dimension (CD) control. 

The amount of energy required to fully image a given photoresist depends largely on the thickness of the film. More bulk film requires more photochemical events which requires more photon energy. A predominately linear relationship between film thickness and dose may seem intuitively obvious, however adding thin film interference to the bulk effect causes the relationship between exposure dose and film thickness to become wildly non-linear. The characteristic curve generated by mapping the dose to clear (or dose to size for a given CD) vs. film thickness is known as the “swing curve”. 

 

g-line Swing Curve for AZ® 3312 Photoresist on Si (λ=435nm)

 

The amplitude and period of the typical swing curve sinusoid are dependent on the substrate reflectivity, the incident energy wavelength and the photoresist film thickness. Since the intensity of the reflected wave is reduced with increasing bulk film absorption, the amplitude of the signal gradually decreases with increasing resist thickness. In most DNQ type resists, the swing amplitude approaches zero at film thickness of about 7.0-8.0µm, above which the dose vs. thickness curve is essentially linear. In thinner films, the phase of the reflected energy relative to the incident energy shifts continuously as the resist thickness increases.

The swing amplitude (S) is described in good approximation by Brunner's formula:

 

 

Where:

Rb = Reflectivity at the resist/substrate interface (reduced by BARC)

Rt = Reflectivity at the resist/air interface (reduced by TARC)

α = Resist bulk absorbance (increased by dye addition)

d = Resist thickness

 

Local maxima on the swing curve correspond to film thicknesses that shift the reflected energy phase by 180º relative to the incident wave (i.e. peak destructive interference) while the local minima correspond to phase matched (constructive) interaction between the incident and reflected energy.                  

One detrimental effect of thin film interference is large variations in CD’s resulting from relatively small variations in photoresist thickness. In the example curve above, it is obvious that the magnitude of the energy delta coupled with the rate of change in energy required as the photoresist varies from 1.05 to 1.10µm (only 50nm!) has the potential to cause significant problems with CD control. Since spin coated photoresist is only partially planarizing, film thickness variations over substrate topography steps can easily span of full period or more of the swing curve.

 

 

      

Schematic representation of film thickness variation in photoresist spin coated over surface topography

 

 

The below schematic is one possible result of imaging a line over the topography step shown above without reflectivity control.

  

    

 

Top down view of one potential CD non-uniformity signature in resist imaged over a topography step

 

One simple method for suppressing swing amplitude and improving CD uniformity over topography (and across varying focal planes) is to use a Top Anti-Reflective Coating (TARC) such as AZ® Aquatar. TARCs are simple to process because they are coated on top of the photoresist after soft bake and are easily removed during the develop step. By reducing the amount of reflected light that gets back scattered into the resist film from the air/photoresist interface, TARCs can suppress swing ratios by 70-80%. Since the swing ratio is the cumulative effect of light "bouncing" many times between the substrate and air/resist interface, limiting these "back reflections" from the air significantly reduces the swing amplitude. A simple schematic of a TARCs working mechanism is shown below where:

 

I = incident exposure energy

C = primary reflection from substrate

D = back reflection from resist/TARC interface

G = back reflection from TARC/air interface

t = TARC thickness 

 

 

 

The refractive index and thickness of the TARC layer can be designed such that back reflections D and G will destructively interfere. This reduction in back scattered energy intensity eliminates subsequent reflections and hence all contributions to the swing ratio except that of the primary reflection. In order to maximize this effect, two conditions must be met. The intensities of reflections D and G must be equalized and their phases must be shifted by 180º. Since the energy that results in vector G makes two passes through the TARC layer, the TARC thickness t must be set to impart a 90º phase shift. The necessary optical conditions are as follows:

 

 

For i-line lithography (λ=365nm), DNQ type photoresists have a refractive index at the actinic wavelength around 1.7, hence an ideal TARC for i-line would have a refractive index at 356nm equal to about 1.3. The TARC AZ Aquatar has a nearly ideal refractive index of 1.43 @ 365nm. Applying the above thickness requirement for n=1.43 yields an optimum TARC film thickness of 63.8nm or 638Å.

The impact of a 640Å layer of AZ Aquatar on swing ratio is shown below where AZ 3312 photoresist was exposed at 365nm and swing curves generated with and without the TARC coating. Note that for 365nm lithography, the TARC also shifts the phase of the swing curve by about 180º. Also note the process becomes notably faster (lower dose required) as the reduction in destructive interference improves the incident energy's in-coupling to the photoresist system.   

 

 

 

Clearly this dramatic reduction in swing ratio should yield less CD variation in response to small changes in photoresist thickness and numerous case studies have verified this effect.

Below is the data from one such study where the CD uniformity of 0.29µm contact holes in a 248nm lithography process is studied and process capability statistics are computed with and without TARC.

 

 

    

 

 

Due to their tendency to improve substrate wettability during the develop process, TARCs have also been shown to significantly reduce pattern defect density, especially for small dark field features such as contact holes or trenches.

Despite their ease of use and effectiveness in improving CD uniformity however, there are a few significant reflectivity induced process issues that cannot be mitigated using TARCs alone. Standing waves in resist sidewalls, first order thin film interference effects (i.e. the residual 20-30% swing ratio) and reflective notching damage to resist features all result from the primary substrate reflection and therefore are not eliminated by a TARC layer. While standing waves are effectively controlled via the Post Exposure Bake in DNQ type resists, thin chemically amplified (CA) resist features typically do not smooth sufficiently via the PEB. In all of these cases, a Bottom Anti-Reflective Coating (BARC) layer is needed.

 

 

 

 

Unlike TARCs, BARC layers dramatically reduce or even eliminate the primary substrate reflection via a combination of absorbance and interference effects. Since BARC layers reside under the photoresist film, they also can resolve process issues caused by transparent substrates where light can pass through and reflect back from the wafer stage during exposure.

In the graph below, the swing ratio and substrate reflectivity (into photoresist) vs. BARC thickness for AZ BARLi II under 365nm exposure are shown. Note that at the optimal thickness of 200nm, the substrate reflectivity (R) is near zero and the residual swing amplitude (S) is less than 2% (as opposed to 20-30% for TARC). 

 

       

 

The effect of the BARC on the swing ratio for AZ MIR 701 photoresist can be seen in the optical simulation run below where the dose vs. film thickness relationship has essentially been reduced to bulk effect.

 

   

 

While BARCs are extremely effective for resolving process issues resulting from substrate reflectivity, they also introduce significantly more process complexity than TARCs. Since most BARC layers are not soluble in the photoresist developer, the resist pattern must be transferred to the substrate via an additional dry etch step. The BARC layer is very thin however, hence this "punch through" etch step is relatively quick and simple. Processing with organic BARCs also requires an extra bake step as the BARC layer must be stabilized before photoresist is coated on top. 

A significant non-optical advantage of organic BARC layers is their adhesion characteristics. Organic BARCs adhere extremely well to almost any substrate; so well in fact that oxidizing surfaces typically do not require pre-treatment with HMDS or other adhesion promoters. In extreme circumstances, thin BARCs are sometimes used as adhesion layers to prevent resist delamination during develop, etch or plating.     

 

I would like more information on anti-reflective coating materials and processing!