The present invention generally relates to method and apparatus for cleaning semiconductor wafer. More particularly, relates to controlling the bubble cavitation generated by ultra or mega sonic device during the cleaning process to achieve a stable or controlled cavitation on the entire wafer, which removes fine particles efficiently without damaging the device structure on the wafer.
Semiconductor devices are manufactured or fabricated on semiconductor wafers using a number of different processing steps to create transistor and interconnection elements. Recently, the transistors are built from two dimensions to three dimensions such as finFET transistors. To electrically connect transistor terminals associated with the semiconductor wafer, conductive (e.g., metal) trenches, vias, and the like are formed in dielectric materials as part of the semiconductor device. The trenches and vias couple electrical signals and power between transistors, internal circuit of the semiconductor devices, and circuits external to the semiconductor device.
In forming the finFET transistors and interconnection elements on the semiconductor wafer may undergo, for example, masking, etching, and deposition processes to form the desired electronic circuitry of the semiconductor devices. In particular, multiple masking and plasma etching step can be performed to form a pattern of finFET and or recessed areas in a dielectric layer on a semiconductor wafer that serve as fin for the transistor and or trenches and vias for the interconnection elements. In order to removal particles and contaminations in fin structure and or trench and via post etching or photo resist ashing, a wet cleaning step is necessary. Especially, when device manufacture node migrating to 14 or 16 nm and beyond, the side wall loss in fin and or trench and via is crucial for maintaining the critical dimension. In order to reduce or eliminate the side wall loss, it is important to use moderate, dilute chemicals, or sometime de-ionized water only. However, the dilute chemical or de-ionized water usually is not efficient to remove the particles in the fin structure and or trench and via. Therefore the mechanical force such as ultra or mega sonic is needed in order to remove those particles efficiently. Ultra sonic or mega sonic wave will generate bubble cavitation which applies mechanical force to wafer structure, the violent cavitation such as transit cavitation or micro jet will damage those patterned structures. To maintain a stable or controlled cavitation is key parameters to control the mechanical force within the damage limit and at the same time efficiently to remove the particles.
Mega sonic energy coupled with nozzle to clean semiconductor wafer is disclosed in U.S. Pat. No. 4,326,553. The fluid is pressurized and mega sonic energy is applied to the fluid by a mega sonic transducer. The nozzle is shaped to provide a ribbon-like jet of cleaning fluid vibrating at ultra/mega sonic frequencies for the impingement on the surface.
A source of energy vibrates an elongated probe which transmits the acoustic energy into the fluid is disclosed in U.S. Pat. No. 6,039,059. In one arrangement, fluid is sprayed onto both sides of a wafer while a probe is positioned close to an upper side. In another arrangement, a short probe is positioned with its end surface close to the surface, and the probe is moved over the surface as wafer rotates.
A source of energy vibrates a rod which rotates around it axis parallel to wafer surface is disclosed in U.S. Pat. No. 6,843,257 B2. The rod surface is etched to curve groves, such as spiral groove.
It is needed to have a better method for controlling the bubble cavitation generated by ultra or mega sonic device during the cleaning process to achieve a stable or controlled cavitation on the entire wafer, which removes fine particles efficiently without damaging the device structure on the wafer.
One method of the present invention is to achieve a damage free ultra/mega-sonic cleaning on a wafer with patterned structure by maintaining a stable bubble cavitation. The stable bubble cavitation is controlled by setting a sonic power supply with power P1 at a time interval shorter than τ1, and setting the sonic power supply with power P2 at a time interval longer than τ2, and repeat above steps till the wafer is cleaned, where power P2 is equal to zero or much smaller than power P1, τ1 is a time interval that the temperature inside bubble raises to a critical implosion temperature; and ρ2 is a time interval that the temperature inside bubble falls down to a temperature much lower than the critical implosion temperature.
Another method of the present invention is to achieve a damage free ultra/mega sonic cleaning on a wafer with patterned structure by maintaining a stable bubble cavitation. The stable bubble cavitation is controlled by setting a sonic power supply with frequency f1 at a time interval shorter than τ1, and setting the sonic power supply with frequency f2 at a time interval longer than τ2, and repeat above steps till the wafer is cleaned, where f2 is much higher than f1, better to be 2 times or 4 times higher, τ1 is a time interval that the temperature inside bubble raises to a critical implosion temperature; and τ2 is a time interval that the temperature inside bubble falls down to a temperature much lower than the critical implosion temperature.
Another method of the present invention is to achieve a damage free ultra/mega-sonic cleaning on a wafer with patterned structure by maintaining a stable bubble cavitation with bubble size less than space in patterned structure. The stable bubble cavitation with bubble size less than space in patterned structure is controlled by setting a sonic power supply at power P1 for a time interval shorter than τ1, and setting the sonic power supply at power P2 for a time interval longer than τ2, and repeat above steps till the wafer is cleaned, where P2 is equal to zero or much smaller than P1, τ1 is a time interval that the bubble size increases to a critical size equal to or larger than the space in patterned structures; and τ2 is a time interval that the bubble size decreases to a value much smaller than the space in patterned structure.
Another method of the present invention is to achieve a damage free ultra/mega-sonic cleaning on a wafer with patterned structure by maintaining a stable bubble cavitation with bubble size less than space in patterned structure. The stable bubble cavitation with bubble size less than space in patterned structure is controlled by setting a sonic power supply with frequency f1 for a time interval shorter than τ1, and setting the sonic power supply with frequency f2 for a time interval longer than τ2, and repeat above steps till the wafer is cleaned, where f2 is much higher than f1, better to be 2 times or 4 times higher, τ1 is a time interval that the bubble size increases to a critical size equal to or larger than the space in patterned structures; and τ2 is a time interval that the bubble size decreases to a value much smaller than the space in patterned structure.
The idea gas equation can be expressed as follows:
p
0
v
0
/T
0
=pv/T (1),
where, p0 is pressure inside bubbler before compression, v0 initial volume of bubble before compression, T0 temperature of gas inside bubbler before compression, p is pressure inside bubbler in compression, v volume of bubble in compression, T temperature of gas inside bubbler in compression.
In order to simplify the calculation, assuming the temperature of gas is no change during the compression or compression is very slow and temperature increase is cancelled by liquid surrounding the bubble. So that the mechanical work wm did by sonic pressure PM during one time of bubbler compression (from volume N unit to volume 1 unit or compression ratio=N) can be expressed as follows:
Where, S is area of cross section of cylinder, x0 the length of the cylinder, p0 pressure of gas inside cylinder before compression. The equation (2) does not consider the factor of temperature increase during the compression, so that the actual pressure inside bubble will be higher due to temperature increase. Therefore the actual mechanical work conducted by sonic pressure will be larger than that calculated by equation (2).
If assuming all mechanical work did by sonic pressure is partially converted to thermal energy and partially converted mechanical energy of high pressure gas and vapor inside bubble, and such thermal energy is fully contributed to temperature increase of gas inside of bubbler (no energy transferred to liquid molecules surrounding the bubble), and assuming the mass of gas inside bubble staying constant before and after compression, then temperature increase ΔT after one time of compression of bubble can be expressed in the following formula:
ΔT=Q/(mc)=βwm/(mc)=βSx0p0ln(x0)/(mc) (3)
where, Q is thermal energy converted from mechanical work, βratio of thermal energy to total mechanical works did by sonic pressure, m mass of gas inside the bubble, c gas specific heat coefficient. Substituting β=0.65, S=1E−12 m2, x0=1000 μm=1E−3 m (compression ratio N=1000), p0=1 kg/cm2=1E4 kg/m2, m=8.9E−17 kg for hydrogen gas, c=9.9E3 J/(kg ° k) into equation (3), then ΔT=50.9° k.
The temperature T1 of gas inside bubbler after first time compression can be calculated as
T
1
=T
0
+ΔT=20° C.+50.9° C.=70.9° C. (4)
When the bubble reaches the minimum size of 1 micron as shown in
T
2
=T1−δT=T0+ΔT−δT (5)
Where δT is temperature decrease after one time of expansion of the bubble, and δT is smaller than ΔT.
When the second cycle of bubble cavitation reaches the minimum bubble size, the temperature T3 of gas and or vapor inside bubbler will be
T3=T2+ΔT=T0+ΔT−δT+ΔT=T0+2ΔT−δT (6)
When the second cycle of bubble cavitation finishes, the temperature T4 of gas and/or vapor inside bubbler will be
T4=T3−δT=T0+2ΔT−δT−δT=T0+2ΔT−2δT (7)
Similarly, when the nth cycle of bubble cavitation reaches the minimum bubble size, the temperature T2n−1 of gas and or vapor inside bubbler will be
T
2n−1
=T
0
+nΔT−(n−1)δT (8)
When the nth cycle of bubble cavitation finishes, the temperature T2n of gas and/or vapor inside bubbler will be
T
2n
=T
0
+nΔT−nδT=T
0
+n(ΔT−δT) (9)
As cycle number n of bubble cavitation increase, the temperature of gas and vapor will increase, therefore more molecules on bubble surface will evaporate into inside of bubble 6082 and size of bubble 6082 will increase too, as shown in
From equation (8), implosion cycle number ni can be written as following:
n
i=(Ti−T0−ΔT)/(ΔT−δT)+1 (10)
From equation (10), implosion time τi can be written as following:
Where, t1 is cycle period, and f1 frequency of ultra/mega sonic wave.
According to formulas (10) and (11), implosion cycle number ni and implosion time τi can be calculated. Table 1 shows calculated relationships among implosion cycle number ni, implosion time τi and (ΔT−δT), assuming Ti=3000° C., ΔT=50.9° C., T0=20° C., f1=500 KHz, f1=1 MHz, and f1=2 MHz.
In order to avoid damage to patterned structure on wafer, a stable cavitation must be maintained, and the bubble implosion or micro jet must be avoided.
Step 1: Put ultra/mega sonic device adjacent to surface of wafer or substrate set on a chuck or tank;
Step 2: Fill chemical liquid or gas (hydrogen, nitrogen, oxygen, or CO2) doped water between wafer and the ultra/mega sonic device;
Step 3: Rotate chuck or oscillate wafer;
Step 4: Set power supply at frequency f1 and power P1;
Step 5: Before temperature of gas and vapor inside bubble reaches implosion temperature Ti, (or time reach τ1<τi as being calculated by equation (11)), set power supply output to zero watts, therefore the temperature of gas inside bubble start to cool down since the temperature of liquid or water is much lower than gas temperature.
Step 6: After temperature of gas inside bubble decreases to room temperature T0 or time (zero power time) reaches τ2, set power supply at frequency f1 and power P1 again.
Step 7: repeat Step 1 to Step 6 until wafer is cleaned.
In step 5, the time τ1 must be shorter than τi in order to avoid bubble implosion, and τi can be calculated by using equation (11).
In step 6, the temperature of gas inside bubble is not necessary to be cooled down to room temperature or liquid temperature; it can be certain temperature above room temperature or liquid temperature, but better to be significantly lower than implosion temperature Ti.
According to equations 8 and 9, if (ΔT−δT) can be known, the τi can be calculated. But in general, (ΔT−δT) is not easy to be calculated or measured directly. The following method can determine the implosion time τi experimentally.
Step 1: Based on Table 1, choosing 5 different time τ1 as design of experiment (DOE) conditions,
Step 2: choose time τ2 at least 10 times of τ1, better to be 100 times of τ1 at the first screen test
Step 3: fix certain power P0 to run above five conditions cleaning on specific patterned structure wafer separately. Here, P0 is the power at which the patterned structures on wafer will be surely damaged when running on continuous mode (non-pulse mode).
Step 4: Inspect the damage status of above five wafers by SEMS or wafer pattern damage review tool such as AMAT SEM vision or Hitachi IS3000, and then the implosion time τi can be located in certain range.
Step 1 to 4 can be repeated again to narrow down the range of implosion time τi. After knowing the implosion time τi, the time τ1 can be set at a value smaller than 0.5 τi for safety margin. One example of experimental data is described as following.
The patterned structures are 55 nm poly-silicon gate lines. Ultra/mega sonic wave frequency was 1 MHz, and ultra/mega-sonic device manufactured by Prosys was used and operated in a gap oscillation mode (disclosed by PCT/CN2008/073471) for achieving better uniform energy dose within wafer and wafer to wafer. Other experimental parameters and final pattern damage data are summarized in Table 2 as follows:
It was clear that the τ1=2 ms (or 2000 cycle number) introduced as many as 1216 damage sites to patterned structure with 55 nm feature size, but that the τ1=0.1 ms (or 100 cycle number) introduced zero (0) damage sites to patterned structure with 55 nm feature size. So that the ti is some number between 0.1 ms and 2 ms, more detail tests need to be done to narrow its range. Obviously, the cycle number related to ultra or mega sonic power density and frequency, the larger the power density, the less the cycle number; and the lower the frequency, the less the cycle number. From above experimental results, we can predict that the damage-free cycle number should be smaller than 2,000, assuming the power density of ultra or mega sonic wave is larger than 0.1 wattsorcm2, and frequency of ultra or mega sonic wave is equal to or less than 1 MHz. If the frequency increases to a range larger than 1 MHz or power density is less than 0.1 watts/cm2, it can be predicted that the cycle number will increase.
After knowing the time τ1, then the time τ2 can be shorten based on similar DEO method described above, i.e. fix time τ1, gradually shorten the time τ2 to run DOE till damage on patterned structure being observed. As the time τ2 is shorten, the temperature of gas and or vapor inside bubbler cannot be cooled down enough, which will gradually shift average temperature of gas and vapor inside bubbler up, eventually it will trigger implosion of bubble. This trigger time is called critical cooling time. After knowing critical cooling time τc, the time τ2 can be set at value larger than 2τc for the same reason to gain safety margin.
Similar to method shown in
Again similar to method shown in
Again similar to method shown in
Again similar to method shown in
Again similar to method shown in
Again similar to method shown in
Step 1: Put ultra/mega sonic device adjacent to surface of wafer or substrate set on a chuck or tank;
Step 2: Fill chemical liquid or gas doped water between wafer and the ultra/mega sonic device;
Step 3: Rotate chuck or oscillate wafer;
Step 4: Set power supply at frequency f1 and power P1;
Step 5: Before temperature of gas and vapor inside bubble reaches implosion temperature Ti, (total time τ1 elapes), set power supply output at frequency f1 and power P2, and P2 is smaller than P1. Therefore the temperature of gas inside bubble start to cool down since the temperature of liquid or water is much lower than gas temperature.
Step 6: After temperature of gas inside bubble decreases to certain temperature close to room temperature T0 or time (zero power time) reach τ2, set power supply at frequency f1 and power P1 again.
Step 7: repeat Step 1 to Step 6 until wafer is cleaned.
In step 6, the temperature of gas inside bubble can not be cooled down to room temperature due to power P2, there should be a temperature difference ΔT2 existing in later stage of τ2 time zone, as shown in
Step 1: Put ultra/mega sonic device adjacent to surface of wafer or substrate set on a chuck or tank;
Step 2: Fill chemical liquid or gas doped water between wafer and the ultra/mega sonic device;
Step 3: Rotate chuck or oscillate wafer;
Step 4: Set power supply at frequency f1 and power P1;
Step 5: Before size of bubble reaches the same dimension of space W in patterned structures (time τ1 elapse), set power supply output to zero watts, therefore the temperature of gas inside bubble starts to cool down since the temperature of liquid or water is much lower than gas temperature;
Step 6: After temperature of gas inside bubble continues to reduce (either it reaches room temperature T0 or time (zero power time) reach τ2, set power supply at frequency f1 power P1 again;
Step 7: repeat Step 1 to Step 6 until wafer is cleaned;
In step 6, the temperature of gas inside bubble is not necessary to be cooled down to room temperature, it can be any temperature, but better to be significantly lower than implosion temperature Ti. In the step 5, bubble size can be slightly larger than dimension of patterned structures as long as bubble expansion force does not break or damage the patterned structure. The time τ1 can be determined experimentally by using the following method:
Step 1: Similar to Table 1, choosing 5 different time τ1 as design of experiment (DOE) conditions,
Step 2: choose time τ2 at least 10 times of τ1, better to be 100 times of τ1 at the first screen test
Step 3: fix certain power P0 to run above five conditions cleaning on specific patterned structure wafer separately. Here, P0 is the power at which the patterned structures on wafer will be surely damaged when running on continuous mode (non-pulse mode).
Step 4: Inspect the damage status of above five wafers by SEMS or wafer pattern damage review tool such as AMAT SEM vision or Hitachi IS3000, and then the damage time τi can be located in certain range.
Step 1 to 4 can be repeated again to narrow down the range of damage time τd. After knowing the damage time τd, the time τ1 can be set at a value smaller than 0.5 τd for safety margin.
All cleaning methods described from
Generally speaking, an ultra/mega sonic wave with the frequency between 0.1 MHz˜10 MHz may be applied to the method disclosed in the present invention.
Although the present invention has been described with respect to certain embodiments, examples, and applications, it will be apparent to those skilled in the art that various modifications and changes may be made without departing from the invention.
This application is a divisional of U.S. patent application Ser. No. 15/575,793, filed Nov. 20, 2017, and titled “METHODS AND SYSTEM FOR CLEANING SEMICONDUCTOR WAFERS,” which is a U.S. National Stage Entry under 35 U.S.C. § 371 of International Application No. PCT/CN2015/079342, filed May 20, 2015, and titled “METHOD AND APPARATUS FOR CLEANING SEMICONDUCTOR WAFERS,” the entire disclosures of which are hereby incorporated by reference herein.
Number | Date | Country | |
---|---|---|---|
Parent | 15575793 | Nov 2017 | US |
Child | 17141097 | US |