JP7495469B2 - Semiconductor Wafer Cleaning System - Google Patents

Description

The present invention relates to semiconductor wafer cleaning, and more particularly to a wet cleaning method and apparatus using controlled sonic energy.

Semiconductor devices are manufactured or fabricated on semiconductor wafers using a series of processing steps that produce transistors and interconnect elements. These transistors are traditionally configured in two dimensions, but more recently, in three dimensions, such as FinFET transistors. The interconnect elements include conductive (e.g., metal) trenches, via holes, etc., formed in insulating materials.

In forming these transistors and interconnect elements, the semiconductor wafer undergoes multiple masking, etching, and deposition processes to form the desired structures for the semiconductor device. In particular, multiple masking and plasma etching steps can be performed to form recessed regions in the dielectric layer of the semiconductor wafer that serve as the fins and trenches and vias of the interconnect elements of the finFET transistors. A wet cleaning step is required to remove particles and foreign matter in the fin structures and/or trenches and vias by post-etching or photoresist ashing. However, wet cleaning with chemicals can result in sidewall loss. Sidewall loss of fins, trenches, and vias is a critical issue in maintaining critical dimensions, especially as device manufacturing nodes shrink to 14 nm or 16 nm and beyond. To reduce or eliminate sidewall loss, it is important to use only mild or diluted chemicals, or in some cases, deionized water. However, mild or diluted chemicals and deionized water typically cannot effectively remove particles in the fin structures and/or trenches and vias. Therefore, to effectively remove these particles, mechanical forces generated, for example, by ultrasonic (ultra sonic) or mega sonic energy are required. Ultrasonic or mega sonic energy generates cavitation bubbles to apply mechanical forces to the wafer structures during cleaning.

However, cavitation is a chaotic phenomenon. The occurrence of cavitation bubble generation and its collapse is influenced by many physical parameters. Severe cavitation, such as transit cavitation or micro jets, can damage pattern structures (fins, trenches and via holes). In conventional ultrasonic or high frequency ultrasonic cleaning processes, significant particle removal efficiency (PRE) can be obtained only when the power is high enough, for example, greater than 5-10 watts. However, significant wafer damage starts to occur when the power exceeds about 2 watts. Therefore, it is difficult to find a window region of power that can efficiently clean the wafer without causing significant damage. That is, maintaining stable or controlled cavitation plays an important role in controlling the sonic mechanical force below the damage limit while still allowing efficient removal of foreign particles from the pattern structures.

It is therefore desirable to provide a system and method for controlling bubble cavitation generated by ultrasonic or high frequency ultrasonic devices during a wafer cleaning process that can efficiently remove microscopic foreign particles without damaging pattern structures on the wafer.

The semiconductor wafer cleaning system disclosed herein includes a wafer holder for temporarily restraining the semiconductor wafer during the cleaning process, an inlet for supplying a cleaning liquid to the surface of the semiconductor wafer, an acoustic generator configured to operate alternately between a first predetermined time period at a first predetermined setting and a second predetermined time period at a second predetermined setting, and a controller programmed to determine the first and second predetermined settings, the first and second predetermined time periods, and the number of alternations between the first and second predetermined settings by the acoustic generator. After the ratio of the total bubble volume to the volume of the via hole, trench or recess space in the semiconductor wafer increases to a first predetermined value, the power is set to the second frequency and the second power level for the second predetermined time period, and after the ratio of the total bubble volume to the volume of the via hole, trench or recess space in the semiconductor wafer decreases to a second predetermined value, the power is set to the first frequency and the first power level for the first predetermined time period.

Other aspects, features and techniques will become apparent to those skilled in the art upon consideration of the detailed description of the embodiments below.

The accompanying drawings, which form a part of this specification, are included to illustrate certain aspects of the present disclosure. A clearer concept of the present disclosure, and of the components and operation of the systems provided therein, will be more readily apparent by reference to the exemplary, and therefore non-limiting, embodiments illustrated in the drawings, in which the same reference numerals (when present in more than one drawing) refer to the same elements. The present disclosure may be better understood by reference to one or more of these drawings in combination with the description presented below. It should be noted that the features illustrated in the drawings are not necessarily drawn to scale.

FIG. 1 illustrates a wafer cleaning apparatus using an ultrasonic or high frequency ultrasonic device according to one embodiment of the present invention. FIG. 1 illustrates a wafer cleaning apparatus using an ultrasonic or high frequency ultrasonic device according to one embodiment of the present invention. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. 1A-1C are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. FIG. 1 illustrates bubble implosion during a wafer cleaning process. FIG. 1 illustrates transit cavitation damaging pattern structures on a wafer during a wafer cleaning process. FIG. 1 illustrates transit cavitation damaging pattern structures on a wafer during a wafer cleaning process. FIG. 1 illustrates the change in thermal energy inside a bubble during the sonic wafer cleaning process. FIG. 1 illustrates an acoustic wafer cleaning process that ultimately results in microjet generation. FIG. 1 illustrates a sonic wafer cleaning process according to one embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to one embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to one embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 13 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates a sonic wafer cleaning process according to yet another embodiment of the present invention. FIG. 1 illustrates stable cavitation damaging pattern features on a wafer in a sonic wafer cleaning process. FIG. 1 illustrates stable cavitation damaging pattern features on a wafer in a sonic wafer cleaning process. FIG. 1 illustrates stable cavitation damaging pattern features on a wafer in a sonic wafer cleaning process. 4 is a flow chart illustrating an alternative wafer cleaning process according to one embodiment of the present invention. FIG. 2 illustrates a wafer cleaning process according to one embodiment of the present invention. FIG. 2 illustrates a wafer cleaning process according to one embodiment of the present invention. FIG. 2 illustrates a wafer cleaning process according to one embodiment of the present invention. FIG. 1 illustrates a wafer cleaning process according to another embodiment of the present invention. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 illustrates bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 1 shows the change in bubble volume as a function of sonic energy. FIG. 1 shows the change in bubble volume as a function of sonic energy. 1A-1D illustrate a sonic wafer cleaning process that effectively cleans pattern features such as high aspect ratio via holes and trenches in accordance with one embodiment of the present invention. FIG. 1 illustrates another cleaning process according to one embodiment of the present invention. FIG. 1 illustrates a wafer cleaning process using sonic energy according to another embodiment of the present invention. FIG. 23 illustrates an example of a wafer cleaning apparatus for performing the wafer cleaning process illustrated in FIGS. 7-22 in accordance with one embodiment of the present invention. 7-23 is a cross-sectional view of another wafer cleaning apparatus for performing the wafer cleaning process illustrated in FIGS. 7-22 in accordance with one embodiment of the present invention. FIG. 1 illustrates a control system for monitoring operating parameters of a wafer cleaning process employing sonic energy in accordance with one embodiment of the present invention. FIG. 26 is a block diagram of the detection system shown in FIG. 25 in accordance with one embodiment of the present invention. FIG. 26 is a block diagram of the detection system shown in FIG. 25 according to another embodiment of the present invention. FIG. 27 illustrates an example of the voltage attenuation circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 27 illustrates an example of the voltage attenuation circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 27 illustrates an example of the voltage attenuation circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 27 illustrates an example of the shaping circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 27 illustrates an example of the shaping circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 27 illustrates an example of the shaping circuit shown in FIG. 26 in accordance with one embodiment of the present invention. FIG. 28 illustrates an example of the main controller of FIGS. 26 and 27 according to one embodiment of the present invention. FIG. 28 illustrates an example of the main controller of FIGS. 26 and 27 according to one embodiment of the present invention. FIG. 28 illustrates an example of the main controller of FIGS. 26 and 27 according to one embodiment of the present invention. FIG. 13 is a diagram showing the sonic power source vibrating several times after the host computer has stopped the sonic power source. FIG. 28 illustrates an example of the amplitude detection circuit shown in FIG. 27 according to one embodiment of the present invention. FIG. 28 illustrates an example of the amplitude detection circuit shown in FIG. 27 according to one embodiment of the present invention. FIG. 28 illustrates an example of the amplitude detection circuit shown in FIG. 27 according to one embodiment of the present invention. 1 is a flow chart illustrating a wafer cleaning process according to one embodiment of the present invention. 4 is a flow chart illustrating a wafer cleaning process according to another embodiment of the present invention.

One aspect of the present disclosure relates to controlling bubble cavitation in semiconductor wafer cleaning using sonic energy. A preferred embodiment of the present invention will be described in detail below with reference to the drawings.

1A and 1B are diagrams illustrating a wafer cleaning apparatus using an ultrasonic or high frequency ultrasonic device according to an embodiment of the present invention. FIG. 1A is a cross-sectional view of a wafer cleaning apparatus including a wafer chuck 1014 for holding a wafer 1010, a rotary drive module 1016 for driving the wafer chuck 1014, and a nozzle 1012 for supplying a cleaning liquid 1032 to the surface of the wafer 1010. The cleaning liquid 1032 may be a cleaning chemical or deionized water. The wafer cleaning apparatus also includes an ultrasonic or high frequency ultrasonic device 1003 disposed above the wafer 1010. Then, with the rotation of the wafer 1010 and the constant flow of the cleaning liquid 1032 from the nozzle 1012, a film of the cleaning liquid 1032 with a thickness d is maintained between the wafer 1010 and the sonic device 1003. The ultrasonic or high frequency ultrasonic device 1003 further includes a piezoelectric transducer 1004 acoustically coupled to a resonator 1008 in contact with the cleaning liquid. The piezoelectric transducer 1004 is electrically excited to vibrate, and the resonator 1008 transmits high frequency acoustic energy to the cleaning liquid 1032. The cavitation of bubbles generated by the high frequency acoustic energy vibrates foreign particles, i.e., contaminants, on the surface of the wafer 1010, causing them to be removed therefrom.

Referring again to FIG. 1A, the wafer cleaning apparatus also includes an arm 1007. The arm 1007 is coupled to the sonic device 1003 to vary the liquid film thickness d by moving the sonic device 1003 in a vertical direction Z. A vertical drive module 1006 moves the arm 1007 vertically. Both the vertical drive module 1006 and the rotational drive module 1016 are controlled by a controller 1088.

As shown in FIG. 1B, which is a top view of the wafer cleaning apparatus shown in FIG. 1A, the acoustic device 1003 covers only a small area of the wafer 1010. Therefore, the wafer 1010 must be rotated to receive uniform acoustic energy across it. Although only one acoustic device 1003 is shown in FIGS. 1A and 1B, in other embodiments, multiple acoustic devices may be used simultaneously and intermittently. Similarly, two or more nozzles 1012 may be used to more uniformly spray the cleaning liquid 1032.

2A-2G are diagrams showing various shapes of ultrasonic or high frequency ultrasonic transducers. FIG. 2A shows a triangular or pie shape, FIG. 2B shows a rectangular shape, FIG. 2C shows an octagonal shape, FIG. 2D shows an elliptical shape, FIG. 2E shows a semicircular shape, FIG. 2F shows a quarter circle shape, and FIG. 2G shows a perfect circle shape. Sonic transducers in each of these shapes may be used in place of the piezoelectric transducer 1004 of the acoustic device 1003 shown in FIG. 1.

Figure 3 illustrates the implosion of a bubble during a wafer cleaning process. The shape of the bubble 3012 gradually compresses from a spherical A to an apple G as sonic energy is applied to the bubble 3012. Eventually, the bubble 3012 reaches an implosion state I and forms a microjet. As shown in Figures 4A and 4B, the microjet can cause very violent (potentially thousands of atmospheres and thousands of degrees Celsius) damage to the fine pattern features 4034 on the semiconductor wafer 4010, especially when the size t of the pattern feature elements shrinks to 70 nm or less.

4A and 4B are diagrams illustrating transit cavitation damaging patterned structures on a wafer during a wafer cleaning process. As shown in FIG. 4A, bubbles 4040, 4042, and 4044 are formed by acoustic cavitation above a patterned structure 4034 on a semiconductor wafer 4010. The patterned structure 4034 includes multiple features that need to be cleaned, including but not limited to fins, via holes, trenches, etc. The bubbles 4044 become very violent microjets that can reach several thousand atmospheres and several thousand degrees Celsius. As shown in FIG. 4B, when the microjets are generated, a part of the patterned structure 4034 is blown away. Such damage is more severe for wafers with device feature sizes of 70 nm or less.

5A to 5C are diagrams showing the change in thermal energy inside a bubble 5016 during a wafer cleaning process. When a positive sonic pressure acts on the bubble 5016, the bubble 5016 reduces its volume as shown in FIG. 5A. During this volume contraction process, the sonic pressure P _M acts on the bubble 5016, and mechanical work is converted into thermal energy within the bubble 5016. Therefore, the temperature T of the gas and/or vapor inside the bubble 5016 increases as shown in FIG. 5B. The relationship between the various parameters can be expressed by the following equation:

_p0v0 / _{T0 =} pv/T (1)

where _p is the pressure inside the bubble before compression, _v is the initial volume of the bubble 5016 before compression, _T is the temperature of the gas inside the bubble before compression, p is the pressure inside the bubble during compression, v is the volume of the bubble during compression, and T is the temperature of the gas inside the bubble during compression.

To simplify the calculations, we assume that the temperature of the gas does not change during compression, or that the compression is so slow that the temperature increase is offset by the liquid around the bubble. In this case, the mechanical work _wm due to the acoustic pressure _Pm during one bubble compression (from N units of volume to 1 unit of volume, i.e., compression ratio = N) can be expressed as follows:

where S is the cross-sectional area of the cylinder, _X0 is the length of the cylinder, and _p0 is the gas pressure inside the cylinder before compression. The above formula (2) does not take into account the temperature rise factor during compression, so the actual pressure inside the bubble will be higher due to the temperature rise. Therefore, the actual mechanical work due to the sound pressure will be larger than that calculated by formula (2).

Assuming that part of the mechanical work due to the sonic pressure is converted into thermal energy and partially mechanical energy of the high-pressure gas and/or vapor inside the bubble, which thermal energy contributes entirely to the temperature rise inside the bubble (no energy is transferred to the liquid molecules surrounding the bubble), and further assuming that the mass of the gas inside the bubble remains constant before and after compression, the temperature rise ΔT due to one compression of the bubble can be expressed by the following equation:

ΔT = Q/(mc) = βwm/(mc) = βSx ₀ p ₀ ln(x ₀ )/(mc)
(3)

In the above formula, Q is the heat energy converted from mechanical work, β is the ratio of heat energy to the total mechanical action due to sound pressure, m is the mass of gas inside the bubble, and c is the gas specific heat coefficient. Substituting β=0.65, S=1E ^{-12 m2} , _x0 =1000μm=1E ^-3 m (compression ratio N=1000), _p0 =1kg/ ^cm2 = ^1E4 kg/ ^m2 , m=8.9E ^-17 kg (for hydrogen gas), and c= ^9.9E3 J/( ^kg0k ) into the above formula (3), ΔT=50.9°C.

When the bubble reaches its minimum size of 1 micron, as shown in FIG. 5B, the gas temperature T ₁ within the bubble after the initial compression is calculated as follows:

T ₁ = T ₀ + ΔT = 20° C. + 50.9° C. = 70.9° C. (4)

Under such high temperature, some of the liquid molecules around the bubble will evaporate. Then, the sonic pressure becomes negative and the bubble size starts to expand. In the reverse process, the hot gas and/or vapor with pressure P _G will do work on the surrounding liquid surface. At the same time, as shown in FIG. 5C, the sonic pressure P _M also pulls the bubble in the expansion direction, so the negative sonic pressure P _M also partially does work on the surrounding liquid. As a result of these actions working together, the thermal energy inside the bubble cannot be completely released or converted into mechanical energy, so the temperature of the gas inside the bubble cannot be cooled down to the original gas temperature T ₀ or liquid temperature. After the first cycle of cavitation, the temperature T ₂ of the gas and/or vapor inside the bubble will be somewhere between T ₀ and T ₁ , as shown in FIG. 6B. Here, T ₂ can be expressed as follows:

T ₂ = T ₁ - δT = T ₀ + ΔT - δT (5)

Here, δT is the temperature drop width after the bubble expands once, and δT is smaller than ΔΔT.

When the second cycle of bubble cavitation reaches the minimum bubble size, the temperature _T3 of the gas and/or vapor within the bubble is:

T ₃ = T ₂ + ΔT = T ₀ + ΔT - δT + ΔT = T ₀ + 2ΔT - δT (6)

At the end of the second cycle of bubble cavitation, the temperature _T4 of the gas and/or vapor inside the bubble is:

T ₄ = T ₃ - δT = T ₀ + 2ΔT - δT - δT = T ₀ + 2ΔT - 2δT (7)

Similarly, when the nth cycle of bubble cavitation reaches the minimum bubble size, the temperature T _2n-1 of the gas and/or vapor inside the bubble is:
T _2n-1 = T ₀ + nΔT - (n-1) δT (8)

At the end of the nth cycle of bubble cavitation, the temperature T _2n of the gas and/or vapor within the bubble is:

T _2n = T ₀ + nΔT - nδT = T ₀ + n (ΔT - δT) (9)

From equation (8), the number of implosion cycles, n _i , can be expressed as follows:

n _i =(T _i -T ₀ -ΔT)/(ΔT-δT)+1 (10)

From equation (10), the implosion time τ _i can be expressed as follows:

τ _i = n _i t ₁ = t ₁ ((T _i - _{T 0} - ΔT)/(ΔT-δT)+1)

= n _i / f ₁ = ( ( T _i - _{T 0} - ΔT ) / ( ΔT - δT ) + 1 ) / f ₁ (11)

where _t1 is the cycle period and _f1 is the frequency of the ultrasound/high frequency ultrasound.

The implosion cycle number n _i and implosion time τ _i can be calculated by formulas (10) and (11). The calculated relationship between the implosion cycle number n _i , implosion time τ _i , and ( _ΔT -δT) is shown in Table 1, assuming that T _i =3000°C, ΔT=50.9°C, T ₀ =20°C, f 1 =500KHz, 1MHz, and 2MHz.

6A-6C show the sonic wafer cleaning process where microjets are eventually generated and the process parameters follow equations (1)-(11). As shown in FIG. 6A, power (P) is continuously supplied to the sonic device to generate bubble cavitation in the cleaning liquid. As the number of cycles n of bubble cavitation increases, the temperature of the gas and/or vapor increases as shown in FIG. 6B. Therefore, more molecules on the bubble surface evaporate towards the inside of the bubble 6082, which results in its size increasing over time as shown in FIG. 6C. Eventually, the temperature inside the bubble 6082 during compression reaches an implosion temperature T _i (T _i is typically as high as several thousand degrees Celsius), and a violent microjet 6080 is generated as shown in FIG. 6C. Therefore, to avoid damage to the patterned features of the wafer during cleaning, stable cavitation must be maintained and bubble implosion, i.e., microjet, must be avoided.

Figures 7A-7E illustrate a sonic wafer cleaning process according to one embodiment of the present invention. Figure 7A shows a waveform of a power output intermittently supplied to a sonic device to generate bubble cavitation in a cleaning liquid. Figure 7B shows the temperature curves corresponding to each cycle of cavitation. Figure 7C shows that during each cycle of cavitation, the bubble size increases during period τ ₁ and decreases when the power supply is terminated during period τ ₂ .

The detailed process steps for avoiding bubble implosion according to the first embodiment of the present invention are shown in FIG. 7D. The process starts with step 7010, in which an ultrasonic or high frequency ultrasonic device is placed near the top surface of the wafer being cleaned. In step 7020, a cleaning liquid, either water doped with a chemical or gas, is introduced onto the wafer, filling the gap between the wafer and the sonic device with the cleaning liquid. In step 7030, the wafer held by the chuck is started to rotate or vibrate. In step 7040, a power source having a frequency _f1 and a power level P1 is applied to the sonic device. In step 7050, the power output is set to zero before the temperature of the gas and/or vapor inside the bubble reaches the implosion temperature _T1 or the time period _τ1 reaches the time _τ1 calculated by equation (11). Then, the temperature of the gas and/or vapor inside the bubble starts to decrease because the temperature of the cleaning liquid is much lower than the gas temperature. In step 7060, after the temperature of the gas and/or vapor inside the bubble has fallen to room temperature _T0 or after the duration _τ2 has been reached (the power supply output is set to 0 for the duration _τ2 ), the power supply output is restored to frequency _f1 and power level P1. In step 7070, the cleanliness of the wafer is checked, and if the wafer is still not as clean as desired, steps 7010 to 7060 are repeated. Alternatively, the cleanliness check need not be performed every cycle. The number of cycles to be used may be determined in advance by experimentation using sample wafers.

7D, in step 7050, the period τ ₁ must be shorter than τ _i calculated using equation (11) to avoid implosion of the bubble. In step 7060, the temperature of the gas and/or vapor inside the bubble does not need to be cooled to room temperature or the cleaning liquid temperature. Rather, it can be at a temperature higher than room temperature or the cleaning liquid temperature. Preferably, this temperature is well below the implosion temperature T _i .

According to equations (8) and (9), if (ΔT-δT) is known, the implosion time τ _i can be calculated. However, usually, (ΔT-δT) cannot be calculated and cannot be easily measured directly. On the other hand, τ _i can be determined by experiment.

FIG. 7E is a flow chart showing a process for experimentally determining the implosion time τ _i . First, in step 7210, five different periods τ ₁ are exemplarily selected as design of experiments (DOE) conditions based on Table 1. In step 7220, the period τ ₂ is set to be at least 10 times longer than the selected period τ ₁ , and preferably 100 times longer for the first screening test. In step 7230, the power level is fixed at P ₀ and five different wafers having the same specific pattern structure are cleaned separately under the above five DOE conditions, where P ₀ is the power level at which the pattern structure is surely damaged when performing a continuous mode (non-pulsed mode) as shown in FIG. 6A. In step 7240, the damage state of the five wafers is inspected by a scanning electron microscope (SEM) or a wafer pattern damage survey tool such as AMAT SEM Vision or Hitachi IS3000. This allows the implosion time τ _i to be narrowed down to a specific range. The percentage of damaged pattern features can be calculated by dividing the total number of damaged pattern features inspected by the SEM by the total number of pattern features. There may be other ways to determine the percentage of damaged pattern features. For example, the final wafer yield may be used as an indicator of the proportion of damaged pattern features.

The range of implosion time τ _i can be narrowed by repeating the above steps 7210-7240. After obtaining the implosion time τ _i , the period τ ₁ may be set to a value smaller than 0.5*τ _i for a safety margin. The following paragraphs describe an example of such an experiment.

Assuming that the pattern structure is formed by 55 nm polysilicon gate lines, the ultrasonic frequency is 1 MHz, and the ultrasonic waves are generated by an ultrasonic/high frequency ultrasonic device manufactured by Prosys, operating in gap oscillation mode (disclosed in PCT Application No. PCT/CN2008/073471) to achieve uniform energy deposition within the wafer and from wafer to wafer. Other experimental parameters and final pattern damage data are summarized in Table 2 below.

In experiments, when τ ₁ =2 ms (2000 cycles), the above-mentioned ultrasonic cleaning process generates 1216 damage sites for a pattern feature having a feature size of 55 nm. When τ ₁ =0.1 ms (100 cycles), the ultrasonic cleaning process generates zero (0) damage sites for the same pattern feature. Thus, the implosion time τ _i is between 0.1 ms and 2 ms. Additional testing with a narrower range of τ ₁ can narrow the range of τ _i .

In the above experiments, the number of cycles depends on the power density and frequency of the ultrasound or high frequency ultrasound, with higher power density resulting in fewer cycles and lower frequency resulting in fewer cycles. From the above experiments, it can be predicted that when the power density of the ultrasound or high frequency ultrasound is greater than 0.1 watts/ ^cm2 and the frequency of the ultrasound or high frequency ultrasound is less than 1 MHz, the number of cycles without damage is less than 2,000. If the frequency increases to a range greater than 1 MHz or the power density is less than 0.1 watts/ ^cm2 , the number of cycles is predicted to increase.

After obtaining the period τ ₁ , the period τ ₂ can be obtained by experimentation based on the same DOE method as described above. In this case, τ ₁ is fixed at a predetermined value, and τ ₂ is gradually shortened in each DOE until damage is observed in the pattern structure. As the period τ ₂ becomes shorter, the temperature of the gas and/or vapor inside the bubble cannot be sufficiently reduced, and the average temperature of the gas and/or vapor inside the bubble gradually increases, eventually causing the implosion of the bubble. The time that causes this implosion is called the critical cooling time τ _c . By knowing the critical cooling time τ _c , the period τ ₂ can be set to a value larger than 2*τ _c with a safety margin.

Thus, multiple parameters of the cleaning process may be determined such that the yield improvement caused by the cleaning effect of applying sonic energy is greater than the yield loss caused by the damage caused by applying said sonic energy. A predetermined threshold for the damage rate may be specified, for example, by a customer. The parameters of the cleaning process may be determined such that the damage rate is less than the predetermined threshold, or substantially zero or exactly zero. The predetermined threshold may be, for example, 10%, 5%, 2% or 1%. The damage rate is substantially zero when the final yield of wafer production is substantially unaffected by any damage caused by the cleaning process. In other words, the damage caused by the cleaning process is acceptable from the perspective of the overall manufacturing process. The damage rate may be determined by inspecting sample wafers using an electron microscope, as described above.

Figures 8A-8D illustrate a sonic wafer cleaning process according to another embodiment of the present invention, in which the amplitude of the power source P is varied over time, rather than maintained at a constant level P1 as in step 7040 of Figures 7A and 7D, but otherwise remains the same as that shown in Figures 7A-7D. In one embodiment, the power source amplitude P increases for a period τ ₁ , as shown in Figure 8A. In another embodiment, the power source amplitude P decreases for a period τ ₁ , as shown in Figure 8B. In yet another embodiment, the power source amplitude P decreases at the beginning of the period τ ₁ and then increases, as shown in Figure 8C. In the embodiment shown in Figure 8D, the power source amplitude P increases at the beginning of the period τ ₁ and then decreases.

9A-9D illustrate a sonic wafer cleaning process according to yet another embodiment of the present invention, in which the frequency of the power source is varied over time, rather than maintained at a constant value _f1 as in step 7040 of FIGS. 7A and 7D, but otherwise remains the same as that shown in FIGS. 7A-7D. In one embodiment, as shown in FIG. 9A, the frequency of the power source is _f1 at the beginning of a period _τ1 and then _f3 , where _f1 is a higher frequency than _f3 . As shown in FIG. 9B, in one embodiment, the frequency of the power source is _f3 at the beginning of a period _τ1 and then increases to _f1 . As shown in FIG. 9C, in one embodiment, the frequency of the power source changes from _f3 to _f1 during a period _τ1 and then back to _f3 . As shown in FIG. 9D, in one embodiment, the frequency of the power source changes from _f1 to _f3 during a period _τ1 and then back to _f1 .

9C, in one embodiment, the power supply frequency is first set _{to f1 ,} then to _f3 , and finally to _f4 for a period τ1, where _f4 is less than _f3 , and _f3 is less than _f1 .

9C, in one embodiment, the power supply frequency is first set to _f4 , then to _f3 , and finally to _f1 for a period of time _τ1 , where _f4 is less than _f3 , _which is less than _f1 .

9C, in one embodiment, the power supply frequency is first set to _f1 , then to _f4 , and finally to _f3 for a period _τ1 , where _f4 is less than _f3 , and _f3 is less than _f1 .

9C, in one embodiment, the power supply frequency is first set to _f3 , then to _f4 , and finally to _f1 for a period of time τ1, where _f4 is less than _f3 , and _{f3 is} less than _f1 .

9C, in one embodiment, the power supply frequency is first set to _f3 , then to _f1 , and finally _{to f4 for} a period of time τ1, where _f4 is less than _f3 , and _f3 is less than _f1 .

9C, in one embodiment, the power supply frequency is first set to _f4 , then to _f1 , and finally to _f3 for a period of time _τ1 , where _f4 is less than _f3 , _which is less than _f1 .

Figures 10A-10C illustrate a sonic wafer cleaning process according to yet another embodiment of the present invention. Referring to Figure 10A, similar to the cleaning process shown in Figure 7A, during a period τ ₁ , a power source having a power level P1 and a frequency f ₁ is applied to the sonic device. However, during a period τ ₂ , the power is reduced to a level P2, rather than dropping to 0 as shown in Figure 7A. Thus, the temperature of the gas and/or vapor inside the bubble is reduced to T ₀ +ΔT ₂ , as shown in Figure 10B.

FIG. 10C is a flow chart showing the operation of the wafer cleaning process shown in FIG. 10A and FIG. 10B. In step 10010, an ultrasonic or high frequency ultrasonic device is placed near the top surface of the wafer being cleaned. In step 10020, a cleaning liquid, either water doped with a chemical or gas, is introduced onto the wafer, filling the gap between the wafer and the sonic device with the cleaning liquid. In step 10030, the chuck holding the wafer begins to rotate for the cleaning process. In step 10040, a power source having a frequency f ₁ and a power level P 1 is applied to the sonic device. In step 10050, the power level is reduced to P 2 while maintaining the frequency f ₁ before the temperature of the gas and/or vapor inside the bubble reaches the implosion temperature T _i or the period τ ₁ reaches the period τ _i calculated by equation (11). In step 10060, the power level is restored to P1 after the temperature of the gas and/or vapor inside the bubble has fallen to near room temperature _T0 or after the duration has reached _τ2 . In step 10070, the cleanliness of the wafer is checked, and steps 10010 through 10060 are repeated if the wafer is still not as clean as desired. Alternatively, the cleanliness check need not be performed every cycle. The number of cycles to be used may be determined in advance by experimentation using sample wafers.

11A-11B show a sonic wafer cleaning process according to yet another embodiment of the present invention. This sonic wafer cleaning process is similar to that shown in FIG. 10A-10C, with only the difference being step 10050. In the wafer cleaning process shown in FIG. 11A and FIG. 11B, instead of maintaining the power supply frequency at f ₁ , the frequency is reduced to f ₂ for a period of τ ₂ . The power level P2 should be sufficiently smaller than P1, preferably 1/5 or 1/10, so that the temperature of the gas and/or vapor inside the bubble can be reduced to near room temperature T ₀ .

Figures 12A-12B show a sonic wafer cleaning process according to yet another embodiment of the present invention. The only difference between this cleaning process and that shown in Figures 10A-10C is step 10050. In this wafer cleaning step, the power level P2 during period _τ2 is substantially the same as P1, while the power supply frequency is increased to _f2 .

Figures 13A-13B show a sonic wafer cleaning process according to yet another embodiment of the present invention. The only difference between this cleaning process and that shown in Figures 10A-10C is step 10050. In this wafer cleaning step, the power level is decreased from P1 to P2 during period _τ2 while the power supply frequency is increased to _f2 .

Figures 14A-14B show a sonic wafer cleaning process according to another embodiment of the present invention. The only difference between this cleaning process and that shown in Figures 10A-10C is step 10050. In this wafer cleaning step, the power supply frequency is increased from _f1 to _f2 while the power level is increased from P1 to P2 during period _τ2 . Since the frequency _f2 is higher than _f1 so that the sonic energy does not heat the bubbles as strongly, the power level P2 may be slightly higher than P1, but should not be too high to ensure that the temperature of the gas and/or vapor inside the bubbles decreases during period _τ2 as shown in Figure 14B.

Figures 15A-15C show stable cavitation damaging patterned structures on a wafer in a sonic wafer cleaning process. As shown in Figure 15A, a patterned structure 15034 having a spacing W is formed on a wafer 15010. Some bubbles 15046 formed in the cavitation process are in the gaps of the patterned structure 15034. As shown in Figure 15B, as the bubble cavitation continues, the temperature of the gas and/or vapor in the bubble 15048 increases, causing the size of the bubble 15048 to increase. When the size of the bubble 15048 becomes larger than the spacing W, the expansion force of the bubble cavitation can damage the patterned structure 15034, as shown in Figure 15C. Therefore, a new wafer cleaning process is required.

The damage site caused by bubble expansion shown in FIG. 15C may be smaller than the damage site from bubble implosion shown in FIG. 4B. For example, bubble expansion may result in a damage site on the order of 100 nm, while bubble implosion may result in a larger damage site on the order of 1 μm.

15D is a flow chart illustrating an alternative wafer cleaning process according to an embodiment of the present invention. The alternative wafer cleaning process begins at step 15210, where an ultrasonic or high frequency ultrasonic device is placed near the top surface of the wafer being cleaned. At step 15220, a cleaning liquid, either water doped with a chemical or gas, is introduced onto the wafer, filling the gap between the wafer and the sonic device with the cleaning liquid. At step 15230, the wafer held by the chuck is started to rotate or vibrate. At step 15240, a power source having a frequency f1 and a power level P1 is applied to the sonic device. At step 15250, the power output is reduced to zero before the size of the bubble reaches the value of the spacing W. The temperature of the gas and/or vapor inside the bubble then begins to decrease because the temperature of the cleaning liquid is much lower than the gas temperature. In step 15260, after the temperature of the gas and/or vapor inside the bubble has fallen to room temperature _T0 or after the duration _τ2 has been reached (the power supply output is zero for the duration _τ2 ), the power supply output is restored to frequency _f1 and power level P1. In step 15270, the cleanliness of the wafer is checked, and if the wafer is still not as clean as desired, steps 15210 through 15260 are repeated. Alternatively, the cleanliness check need not be performed every cycle. The number of cycles to be used may be determined in advance by experimentation using sample wafers.

15D , the temperature of the gas and/or vapor inside the bubbles does not need to be cooled to room temperature _T0 , but should preferably be cooled to a temperature much lower than the implosion temperature _T1 . In step 15250, the size of the bubbles can be slightly larger than the spacing W of the pattern features 15034, as long as the expansion force of the bubbles does not break or damage the pattern features 15034.

Referring again to Figure 15D, the duration of step 15240 can be experimentally obtained from the procedure shown in Figure 7E as τ _1. In some embodiments, the wafer cleaning process shown in Figures 7-14 can be combined with the wafer cleaning process shown in Figure 15.

16A-16C illustrate a wafer cleaning process according to an embodiment of the present invention. The wafer cleaning process is similar to that shown in FIG. 7A-7E, except for step 7050 in FIG. 7D. In the wafer cleaning process, the power supply output is set to a positive DC value as shown in FIG. 16A or a negative DC value as shown in FIG. 16B and 16C before the temperature of the gas and/or vapor inside the bubble reaches the implosion temperature _{T i} or the duration τ ₁ reaches τ _i calculated by Equation (11). As a result, the temperature of the gas and/or vapor inside the bubble starts to decrease because the temperature of the cleaning liquid is much lower than the temperature of the gas and/or vapor. In some embodiments, the amplitude of either the positive or negative DC output can be greater than (not shown), equal to (see FIG. _16A and FIG. 16B), or less than (see FIG. 16C) the amplitude of the power level P1 applied for the duration τ 1 to generate bubble cavitation in the cleaning liquid.

FIG 17 illustrates a wafer cleaning process according to another embodiment of the present invention. This wafer cleaning process is also similar to that shown in FIG 7A-7E, except for step 7050 in FIG 7D. This wafer cleaning step reverses the phase of the power supply output while maintaining the same frequency f ₁ applied for period τ ₁ , so that the bubble cavitation can be stopped quickly. As a result, the temperature of the gas and/or vapor inside the bubbles starts to decrease because the temperature of the cleaning liquid is much lower than the temperature of the gas and/or vapor.

17, the power level during period _τ2 is P2, which may be greater than, equal to, or less than the power level during period _τ1 , P1, in different embodiments. In one embodiment, the power frequency during period _τ2 may be different from _f1 if the phase is inverted. In some embodiments, the frequency _f1 of the ultrasonic or high frequency ultrasonic power source is between 0.1 MHz and 10 MHz.

FIGS. 18A-18J are diagrams illustrating bubble cavitation control to improve circulation of fresh cleaning solution within via holes or trenches of a semiconductor wafer. FIG. 18A is a cross-sectional view of multiple via holes 18034 formed in a wafer 18010, where the diameter of the via hole opening is shown as W1. The bubbles 18012 generated in the via holes 18034 by sonic energy facilitate removal of impurities such as residues and particles therefrom. FIG. 18B is a top view of the via holes shown in FIG. 18A.

FIG. 18C is a cross-sectional view of multiple trenches 18036 formed in wafer 18010. Similarly, bubbles 18012 generated in trenches 18036 by sonic energy facilitate removal of impurities such as residues and particles therefrom. FIG. 18D is a top view of trenches 18036 shown in FIG. 18C.

The saturation point _Rs is defined by the maximum amount of bubbles that can be contained within a pattern feature such as a via hole 18034, a trench 18036, or another recessed area. If the amount of bubbles exceeds the saturation point _Rs , the cleaning liquid will be blocked by the bubbles and will have difficulty reaching the bottom of the sidewall of the via hole 18034 or trench 18036, adversely affecting cleaning performance. If the amount of bubbles is below the saturation point _Rs , the cleaning liquid will be effective enough inside the pattern feature such as the via hole 18034 or trench 18036 to provide good cleaning performance.

Below the saturation point, the ratio R of the total bubble volume V _B to the volume of the via hole or trench or recess space V _VTR is:

R= _VB / _VVTR < _Rs

And the ratio R at the saturation point _Rs is:

R= _VB / _VVTR < _Rs

The volume of all bubbles in a pattern feature such as a via hole 18034, trench 18036 or other recessed space is:

_VB = N * _VB

where N is the number of bubbles in the pattern element and _VB is the average volume of a single bubble.

As shown in Figures 18E-18H, when ultrasonic or high frequency ultrasonic energy is applied to the cleaning solution, the size of the bubbles 18012 gradually expands to a certain volume, so that the ratio R of the total bubble volume _VB to the volume of the via hole or trench, i.e., the recessed space _VVTR , becomes close to or larger than the saturation point _Rs . The expanded bubbles 18012 block the path of replacement of the cleaning solution and removal of impurities in the via hole or trench. In such a case, the acoustic energy cannot efficiently transfer to the via hole or trench to reach their bottom and sidewalls, and the particles, residues and other impurities 18048 are trapped in the via hole or trench. This can easily occur in advanced semiconductor processes where the critical dimension W1 becomes smaller.

As shown in Figures 18I-18J, the size expansion of the bubbles 18012 by ultrasonic or high frequency ultrasonic energy is within limits, and the ratio R of the total bubble volume _VB to the volume of the via hole, trench or recess space _VVTR is much lower than the saturation point _Rs . Fresh cleaning solution 18047 circulates freely in the via hole or trench due to the small bubble cavitation inside the pattern structure elements, so that the impurities 18048, such as residues and particles, can be easily pushed out from the pattern structure elements to obtain good cleaning performance.

Because the total volume of bubbles in a pattern feature such as a via hole or trench is determined by the number and size of the bubbles, controlling the expansion of bubble size due to cavitation is important for the cleaning performance of wafers with pattern features with high aspect ratios.

19A-19D are diagrams showing the change in bubble volume according to sonic energy. During the first cycle of cavitation, the volume of the bubble is compressed from _V0 to _V1 through a positive sonic power cycle, and then expands to _V2 through a negative sonic power cycle. However, the temperature _T2 of the bubble corresponding to _V2 is higher than the temperature _T0 corresponding to _V0 , so the volume _V2 is larger than the volume _V0 , as shown in FIG. 19B. This volume increase is caused by the liquid molecules surrounding the bubble evaporating under the higher temperature. Similarly, the volume _V3 of the bubble after the second compression can be anywhere between _V1 and _V2 , as shown in FIG. 19B. _V1 , _V2 , and _V3 can be expressed as follows:

V ₁ =V ₀ -ΔV (12)

_V2 = _V1 + δV (13)

V ₃ = V ₂ - ΔV = V ₁ + δV - ΔV = V ₀ - ΔV + δV - ΔV = V ₀ + δV - 2ΔV (14)

Here, ΔV is the amount of volumetric compression of the bubble due to one compression caused by the positive pressure generated by the ultrasound/high frequency ultrasound, δV is the amount of volumetric increase of the bubble due to one expansion caused by the negative pressure generated by the ultrasound/high frequency ultrasound, and (δV-ΔV) is the amount of volumetric increase due to the temperature rise (ΔT-δT) resulting from one cycle calculated by equation (5).

Through the second cycle of bubble cavitation, the temperature continues to rise and the bubble expands to a larger size. The volume of gas and/or vapor inside the bubble, _V4 , is:

V ₄ = V ₃ + δV = V ₀ + δV - 2ΔV + δV = V ₀ + 2 (δV - ΔV) (15)

After the third compression, the volume of gas and/or vapor inside the bubble, _V5 , is:

V ₅ = V ₄ - ΔV = V ₀ + 2 (δV - ΔV ) - ΔV = V ₀ + 2 δV - 3 ΔV (16)

Similarly, when the nth cycle of bubble cavitation reaches the minimum bubble size, the volume of gas and/or vapor inside the bubble, V _2n-1 , is:

V _2n-1 = V ₀ + (n-1) δV - nΔV = V ₀ + (n-1) δV - nΔV (17)

At the end of the nth cycle of bubble cavitation, the volume of gas and/or vapor within the bubble, V _2n , is:

V _2n = V ₀ + n (δV - ΔV) (18)

To limit the bubble volume to a desired volume _{V that} is below a value with sufficient physical mobility or saturation point to prevent blocking of cleaning fluid replacement in pattern structure elements such as vias, trenches, or other recessed areas, the number of cycles _n can be expressed as:

n _i =( V _i - V ₀ - ΔV) / ( δV - ΔV) + 1 (19)

From equation (19), the desired time τ _i to achieve V _i can be expressed as:

τ _i = n _i t ₁ = t ₁ ((V _i - _{V 0} - ΔV)/(δV - ΔV) + 1)

= n _i / f ₁ = ( ( V _i - _{V 0} - ΔT ) / ( δV - ΔV ) + 1 ) / f ₁ (20)

where _t1 is the cycle period and _f1 is the frequency of the ultrasound/high frequency ultrasound. Therefore, the desired number of cycles _n1 and the desired time _τ1 to prevent the bubble size from reaching the cutoff level of the pattern structure element can be calculated from equations (19) and (20).

It should be noted that as the number of cycles n of bubble cavitation increases, the temperature of the gas and/or vapor inside the bubble increases, so that more molecules on the bubble surface evaporate into the bubble interior. Therefore, the size of the bubble 19082 will further increase and become larger than the value calculated by equation (18). In operation, the bubble size is determined by an experimental method disclosed below, so the bubble size affected by the evaporation of liquid or water to the bubble interior surface due to the temperature increase will not be discussed in theoretical detail here. As the average single bubble volume increases, the ratio _R of the total bubble volume VB to the volume _VVTR of the via hole, trench, or other recess space increases continuously from _R0 , as shown in FIG. 19D.

As the bubble volume increases, the bubble diameter will eventually reach the same size or the same order of size as the pattern structure element W1 of the via hole 18034 shown in Figures 18A and 18B or the trench 18036 shown in Figures 18C and 18D. And the bubbles inside the via hole 18034 and the trench 18036, especially when the aspect ratio (depth/width) is 3 or more, block the ultrasonic/high frequency ultrasonic energy from further entering into their bottoms. Therefore, the contaminants or particles at the bottom of such deep via holes or trenches cannot be effectively removed or cleaned. Therefore, a new cleaning process is proposed to prevent the bubbles from growing to a critical dimension that blocks the replacement path of the cleaning solution in the pattern structure element of the via hole or trench.

20A-20D illustrate an ultrasonic wafer cleaning process that effectively cleans pattern features such as high aspect ratio vias and trenches in accordance with an embodiment of the present invention. The wafer cleaning process limits the size of bubbles in the cavitation caused by ultrasonic energy. FIG. 20A shows a waveform of a power supply output at power level P1 during period τ ₁ and turned off during period τ _2. FIG. 20B shows the bubble volume curves corresponding to each cycle of cavitation. FIG. 20C shows the bubble size expansion during each cycle of cavitation. FIG. 20D shows the ratio R of the total bubble volume V _B to the volume V _VTR of the via hole, trench, or other recessed space.

R= _VB / _VVTR = _Nvb / _VVTR

According to the above formula, when the average single bubble volume expands by acoustic cavitation over a period τ ₁ over a given number of cycles n, the ratio R of the total bubble volume V _B to the volume V _VTR of the via hole, trench, or other recessed space increases from R ₀ to R _n , where R _n is controlled below the saturation point R _s as follows:

_Rn = _VB / _VVTR = _Nvb / _VVTR < _Rs

And when the average single bubble volume returns to its original size in the cooling process in period τ ₂ , the ratio R of the total bubble volume V _B to the volume of the via hole, trench or other recessed space V _VTR decreases from R _n to R ₀ .

Referring again to FIG. 20B, the gas bubble expands to a large volume V _n with ultrasonic/high frequency ultrasonic power applied to the cleaning liquid in period τ _1. In this state, the mass transfer pathway is partially blocked, and fresh cleaning liquid cannot flow sufficiently into the bottom and sidewall of the via hole or trench. Meanwhile, particles, residues and other impurities trapped in the via hole and trench cannot be effectively removed. However, this state transitions to the next state where the gas bubble shrinks when the ultra-high/high frequency ultrasonic power is turned off in period τ ₂ to cool the gas bubble, as shown in FIG. 20A. In this cooling state, fresh cleaning liquid can flow into the via hole and trench to clean its bottom and sidewall. When the ultrasonic/high frequency ultrasonic power is turned on again in the next cycle, particles, residues and other impurities can be removed from the via hole and trench by the pulling force generated by the increase in the gas bubble volume. When the two conditions are alternated in an ultrasonic/high frequency ultrasonic cleaning process, high aspect ratio pattern features in via holes, trenches, and other recessed areas on a wafer substrate can be effectively cleaned.

The cooling condition during period _τ2 plays an important role in this cleaning process. Also, the condition ( _τ1 < _τi ) is desirable to limit the size of the bubbles. In the following method, the period _τ2 for shrinking the bubble size during the cooling condition and the period _τ1 for limiting the bubble expansion to the path blocking size can be experimentally determined. The experiment is carried out using an ultrasonic/high frequency ultrasonic device coupled with chemical solutions to clean patterned substrates with small pattern feature elements such as via holes and trenches with traceable residues and evaluate the cleaning performance.

The first step is to select a τ ₁ long enough to block the pattern structure element, which may be to calculate τ _i based on equation (20). The second step is to select multiple different periods τ ₂ for running the DOE. The period τ ₂ is selected to be at least 10 times, preferably 100 times, of τ ₁ in the first screen test. The third step is to fix the period 1 and the power P ₀ and perform separate cleaning of substrates with specific pattern structures under at least five conditions, where P ₀ is the power at which the via hole or trench pattern structure element on the substrate is not reliably cleaned when operating in continuous mode (non-pulsed mode). The fourth step is to inspect the condition of traceable residues in the via hole or trench pattern structure element of the above five substrates by an element analysis tool such as SEM or EDX. The above first to fourth steps can be repeated several times to gradually shorten the period τ ₂ until the traceable residues in the via hole or trench pattern structure element are observed. As the period _τ2 becomes shorter, the volume of the bubble cannot be reduced sufficiently, which causes the pattern structure elements to be gradually blocked and affects the cleaning performance. This period is called the critical cooling period _τc . After the critical cooling period _τc is obtained, the period _τ2 is set to a value greater than _2τc to ensure a safety margin.

A more detailed example is given below. The first step is to select 10 different periods τ 1 such as τ ₁₀ , 2τ ₁₀ , 4τ ₁₀ , 8τ ₁₀ , 16τ ₁₀ , 32τ ₁₀ , 64τ ₁₀ , 128τ ₁₀ , 256τ ₁₀ , and 512τ ₁₀ as design of experiment (DOE) conditions, as shown in Table 3. The second step is to select a period τ ₂ at least 10 times, preferably 20 times, of 512τ ₁₀ in the first screen test, as shown in Table 3. The third step is to fix the power P ₀ and individually clean a substrate with a specific pattern structure under the above 10 conditions, where _{P 0} is the power at which the pattern structure elements of via holes or trenches on the substrate are not reliably cleaned when operating in continuous mode (non-pulsed mode). The fourth step is to process 10 substrates with pattern structure elements of via holes or trenches after plasma etching using the conditions shown in Table 3. The reason for choosing substrates after plasma etching is that polymers generated during the etching process are formed on the sidewalls of trenches and via holes. These polymers formed on the bottom or sidewalls of via holes are difficult to remove by conventional methods. The next step is to inspect the cleaning status of the pattern structure elements of via holes or trenches on the 10 substrates by SEM in the cross section of the substrate. The obtained data is shown in Table 3 below. From Table 3, it is clear that the cleaning effect is the best point in substrate #6, where τ ₁ =32τ _10. That is, the optimal period τ ₁ is 32τ ₁₀ .

If no peak is found, the above first to fourth steps are repeated again with a wider time range for τ ₁ to find the period τ _1. After finding the initial τ ₁ , the above first and fourth steps are repeated again with a narrower time range for τ ₁ to narrow the range for the period τ _1. After knowing the period τ _i , the period τ ₂ can be optimized by decreasing _it from 512τ ₂ to a value where the cleaning effect starts to decrease. The detailed procedure is disclosed in Table 4 below. From Table 4, the cleaning effect reaches the best point in substrate #5 where τ ₂ =256τ ₁₀ , so the optimal period τ ₂ is 256τ ₁₀ .

21A-21C show another cleaning process according to an embodiment of the present invention. This cleaning process is similar to the one shown in FIG. 20A-20D, and only differs in that the power of the current cleaning process remains on for a period of mτ ₁ even after the cavitation reaches the saturation point R _s , where m is any number between 0.1 and 100, preferably 2, which depends on the structure of the via hole and trench and the cleaning solution used. And the value of m needs to be optimized by experiment similar to the embodiment shown in FIG. 20A-20D.

22A and 22B show a wafer cleaning process using sonic energy according to another embodiment of the present invention. During the period τ ₁ when sonic power P1 is applied to the cleaning liquid, bubble implosion begins to occur when the temperature of the first bubble reaches the implosion temperature T _i , and then some bubble implosion continues to occur during the temperature increase from T _i to T n (during the period Δτ). After the sonic power is turned off during the period τ ₂ , the temperature of the bubble is cooled from T n back to T ₀ by the surrounding liquid. T _i is determined as the temperature threshold of bubble implosion in the pattern structure elements of via holes and trenches, which causes the first bubble implosion.

Since heat transfer is not exactly uniform within the pattern structure element, more bubbles may continue to form after the temperature T _i is reached. The bubble implosion intensity becomes higher and higher as the bubble temperature T increases. However, the bubble implosion must be controlled to be lower than the implosion intensity that would cause damage to the pattern structure. The bubble implosion can be controlled by controlling the temperature T n lower than the temperature T d by adjusting the time Δτ, where T n is the maximum temperature of the bubble due to the sonic power applied to the cleaning solution during n cycles, and T d is the temperature of a certain amount of bubble implosion accumulation at high intensity (or power) that causes damage to the pattern structure. In the present cleaning process, the bubble implosion intensity is controlled by adjusting the time Δτ after the first bubble implosion begins to achieve the desired cleaning performance and efficiency while avoiding an excessively high bubble implosion intensity that would cause damage to the pattern structure in the cleaning.

In order to increase the particle removal efficiency (PRE), it is desirable to have controlled transit cavitation in the ultrasonic or high frequency ultrasonic cleaning process, as shown in Figures 22A-22B. Controlled transit cavitation is achieved by turning the sonic power source with power P1 for a time interval shorter than _τ1 , then with power _P2 for a time interval longer than _τ2 , and repeating the above steps until the wafer is cleaned, where power _P2 is equal to 0 or much smaller than power P1, _τ1 is the time interval during which the temperature inside the bubble becomes higher than the critical implosion temperature, and _τ2 is the time interval during which the temperature inside the bubble drops below the critical implosion temperature. Controlled transit cavitation involves certain bubble implosion in the cleaning process, so controlled transit cavitation will provide higher PRE (particle removal efficiency) with minimal damage to pattern structures. The critical implosion temperature is the lowest temperature inside the bubble at which the first bubble implosion occurs. To further increase the PRE, the temperature of the bubble needs to be increased further, and therefore a longer period τ ₁ is required, and the temperature of the bubble can be increased by shortening the period τ ₂ .

Another parameter for controlling the implosion level is the frequency of the ultrasound or high frequency ultrasound. Controlled transit cavitation is achieved by switching the sonic power source to have a frequency _f1 for a time interval shorter than _τ1 , then to have a frequency _f2 for a time interval longer than _τ2 , and repeating the above steps until the wafer is cleaned, where _f2 is much higher than _f1 , preferably by a factor of two or four. Usually, the higher the frequency, the lower the implosion level or intensity. Again, _τ1 is the time interval during which the temperature inside the bubble rises above the critical implosion temperature, and _τ2 is the time interval during which the temperature inside the bubble drops to a temperature much lower than the critical implosion temperature. Controlled transit cavitation provides a higher PRE (particle removal efficiency) while minimizing damage to the pattern structure. The critical implosion temperature is the lowest temperature inside the bubble at which the first bubble implosion occurs. To further increase the PRE, the temperature of the bubble needs to be further increased, and therefore a longer period _τ1 is required. Also, the temperature of the bubbles can be increased by shortening the period τ _2. Generally, in the method disclosed in the present invention, ultrasonic waves or high frequency ultrasonic waves with a frequency of 0.1 MHz to 10 MHz may be applied to the wafer cleaning process.

FIG. 23 illustrates an example of a wafer cleaning apparatus for performing the wafer cleaning process shown in FIGS. 7-22 according to an embodiment of the present invention. The wafer cleaning apparatus includes a wafer chuck 23014 for mounting a wafer 23010. The wafer chuck 23014 rotates with the wafer 23010 in the cleaning process driven by a rotation drive mechanism 23016. The wafer cleaning apparatus also includes a nozzle 23064 for supplying a cleaning liquid, such as a cleaning chemical or deionized water 23060, to the wafer 23010. An ultrasonic or high frequency ultrasonic device 23062 is coupled to the nozzle 23064 for applying ultrasonic or high frequency ultrasonic energy to the cleaning liquid. The ultrasonic or high frequency ultrasonic generated by the ultrasonic or high frequency ultrasonic device 23062 propagates from the nozzle 23064 through the cleaning liquid 23060 to the wafer 23010.

Figure 24 is a cross-sectional view of another wafer cleaning apparatus for performing the wafer cleaning process shown in Figures 7-22 according to an embodiment of the present invention. The wafer cleaning apparatus includes a cleaning tank 24074 containing a majority of the cleaning liquid 24070 and a wafer cassette 24076 holding a plurality of wafers 24010 submerged in the cleaning liquid 24070. The wafer cleaning apparatus also includes an ultrasonic or high frequency ultrasonic device 24072 attached to a wall of the cleaning tank 24074 for applying ultrasonic or high frequency ultrasonic energy to the cleaning liquid. At least one inlet (not shown) is provided for filling the cleaning tank 24074 with the cleaning liquid 24070 so that the wafers 24010 are submerged in the cleaning liquid 24070 during the cleaning process.

In the above embodiment, if all the critical process parameters of the sonic power source, such as power level, frequency, power-on time (τ ₁ ) and power-off time (τ ₂ ), are preset in the power source controller without real-time monitoring during the wafer cleaning process, some abnormal condition during the wafer cleaning process may still cause damage to the pattern structure. Therefore, an apparatus and method for real-time monitoring of the working status of the sonic power source is required. If the parameters are not within the normal range, the sonic power source should be stopped and an alarm signal should be sent and reported.

25 illustrates a control system for monitoring operational parameters of a wafer cleaning process employing sonic energy according to one embodiment of the present invention. The control system includes a host computer 25080, a sonic generator 25082, a sonic transducer 1003, a detection system 25086, and a communication cable 25088. The host computer 25080 sends control commands to the sonic generator 25082, such as sonic parameter settings, such as power setting P1, power on duration setting τ ₁ , power setting P ₂ , power off duration setting τ ₂ , frequency settings, and power on commands. After receiving these commands, the sonic generator 25082 generates sonic waveforms and sends the sonic waveforms to the sonic transducer 1003 for cleaning the wafer 1010. Meanwhile, the parameter settings from the host computer 25080 and the actual output from the sonic generator 25082 are read by the detection system 25086. The detection system 25086 compares the actual output from the acoustic generator 25082 with the parameter settings from the host computer 25080. The comparison result is sent to the host computer 25080 via a communication cable 25088. If the output of the acoustic generator 25082 differs from the parameter settings sent by the host computer 25080, the detection system 25086 sends an alarm signal to the host computer 25080. Upon receiving the alarm signal, the host computer 25080 stops the acoustic generator 25082 to prevent further damage to the pattern structures of the wafer 1010.

26 is a block diagram of the detection system 25086 shown in FIG. 25 according to an embodiment of the present invention. The detection system 25086 includes, as an example, a voltage attenuation circuit 26090, a shaping circuit 26092, a main controller 26094, a communication circuit 26096, and a power circuit 26098. The main controller 26094 may be implemented with an FPGA. The communication circuit 26096 is established as an interface with the host computer 25080. The communication circuit 26096 performs RS232/RS485 serial communication with the host computer 25080 to read parameter settings from the host computer 25080 and send comparison results to the host computer 25080. The power circuit 26098 is designed to convert DC 15V to target voltages of DC 1.2V, DC 3.3V, and DC 5V for the entire system.

27 is a block diagram of a detection system 25086 according to another embodiment of the present invention. The detection system 25086 includes, by way of example, a voltage attenuation circuit 26090, an amplitude detection circuit 27092, a main controller 26094, a communication circuit 26096, and a power circuit 26098.

Figures 28A-C show an exemplary implementation of a voltage attenuation circuit 26090 according to one embodiment of the present invention. When the sound wave signal output from the sound wave generator 25082 is first read, the amplitude value is relatively high as shown in Figure 28B. The voltage attenuation circuit 26090 is designed using two op-amps 28102 and 28104 to reduce the amplitude value of the waveform as shown in Figure 28C. The attenuation factor of the voltage attenuation circuit 26090 is set in the range of 5 to 100, and is preferably set to 20. The voltage attenuation can be expressed by the following equation:

V _out = (R2/R1) * V _in

Assuming R1=200k and R2=R3=R4=10K, then V _out =(R2/R1)*V _in =V _in /20.

Here, V _out is the amplitude value output by the voltage attenuation circuit 26090, V _in is the amplitude value input to the voltage attenuation circuit 26090, and R1, R2, R3, and R4 are the resistance values of the two operational amplifiers 28102 and 28104.

29A-29C are diagrams illustrating an example of the shaping circuit 26092 shown in FIG. 26, according to one embodiment of the present invention. Referring again to FIG. 26, the output of the voltage attenuation circuit 26090 is connected to the shaping circuit 26092. The waveform output from the voltage attenuation circuit 26090 is input to the shaping circuit 26092 to convert the sine wave into a square wave so that it can be processed by the main controller (FPGA) 26094. The shaping circuit 26092 includes a window comparator 29102 and an OR gate 29104, as shown in FIG. 29A. If V _cal- < V _in < V _cal+ , then V _out = 0, otherwise V _out = 1. Here, V _cal- and V _cal+ are two thresholds, V _in is the input value of the shaping circuit, and V _out is the output value of the shaping circuit. The waveform passing through the voltage attenuation circuit 2190 is a sine wave as shown in FIG. 29B. The shaping circuit 26092 converts the sine wave into a square wave as shown in FIG. 29C.

30A-30C show an exemplary implementation of the main controller 26094 of FIG. 26 and FIG. 27 according to an embodiment of the present invention. As shown in FIG. 30A, the main controller 26094 includes a pulse conversion module 30102 and a period measurement module 30104. The pulse conversion module 30102 converts a pulse signal of period τ1 to a high level signal and keeps it low level signal for period τ2 as shown in FIG. 30B and FIG. 30C. The circuit symbol of the pulse conversion module 30102 shown in FIG. 30A is Clk_Sys is a 50 MHz clock signal, Pulse_In is an input signal, and Pulse_Out is an output signal. The period measurement module 30104 is used to measure the duration of high and low levels by a counter using the following formula:

τ ₁ = Counter_H * 20ns, τ ₂ = Counter_L * 20ns

Here Counter_H is the number of the high level and Counter_L is the number of the low level.

The main controller 26094 compares the calculated power-on time with a preset period τ _1. If the calculated power-on time is longer than the preset period τ ₁ , the main controller 26094 sends an alarm signal to the host computer 25080. The host computer 25080 that receives the alarm signal stops the sound wave generator 25082. The main controller 26094 compares the calculated power-off time with a preset period τ _2. If the calculated power-off time is shorter than the preset period τ ₂ , the main controller 26094 sends an alarm signal to the host computer 25080. The host computer 25080 that receives the alarm signal stops the sound wave generator 26082. In one embodiment, the main controller 26094 can be implemented using an Altera Cyclone IV FPGA model number EP4CE22F17C6N.

FIG. 31 shows that the sonic power source oscillates several times due to the characteristics of the sonic device after the host computer stops the sonic power source. The period τ ₃ during which the sonic generator 25082 oscillates several cycles after power down is measured by the main controller 26094. The period τ ₃ can be determined by experiment. Therefore, if the time calculated by the periodic measurement module 25104 is τ, the actual power-on time is equal to τ - τ _3. The main controller 26094 compares the calculated power-on time with a preset period τ _1. If the calculated power-on time is longer than the preset period τ ₁ , the main controller 26094 sends an alarm signal to the host computer 25080.

32A-32C are diagrams illustrating an example of the amplitude detection circuit 27092 shown in FIG. 27 according to an embodiment of the present invention. The amplitude detection circuit 27092 includes a reference voltage generation circuit and a comparison circuit. As shown in FIG. 32B, the reference voltage generation circuit is designed to use a D/A converter 32118 to convert the digital input from the main controller 26094 into analog DC reference voltages V _ref+ and V _ref- as shown in FIG. 27C. The comparison circuit is designed to compare the attenuated amplitude V _in , which is the output from the voltage attenuation circuit 26090, with the reference voltages V _ref+ and V _ref- using a window comparator 32114 and an AND gate 32116. If the attenuated amplitude V _in exceeds the reference voltages V _ref+ and/or V _ref- , the amplitude detection circuit 27092 will send an alarm signal to the host computer 25080. Upon receiving the alarm signal, the host computer 25080 shuts down the acoustic generator 25082 to prevent further damage to the pattern structures on the wafer 1010.

FIG. 33 is a flow chart showing a wafer cleaning process according to an embodiment of the present invention. The wafer cleaning process starts with step 33010, which applies a cleaning liquid to a space between the wafer and the ultrasonic/high frequency ultrasonic device. In step 33020, the ultrasonic/high frequency ultrasonic power supply is set to a frequency f ₁ and a power level P 1 to drive the ultrasonic/high frequency ultrasonic device. In step 33030, the detected power-on time is compared with a preset period τ _1. If the detected power-on time is longer than the period τ ₁ , the power supply is shut off and an alarm signal is sent. In step 33040, the ultrasonic/high frequency ultrasonic power supply is set to output 0 before the bubble cavitation in the cleaning liquid damages the pattern structure on the wafer. In step 33050, the sonic power supply is returned to the frequency f ₁ and the power level P 1 after the temperature inside the bubble drops to a certain level. In step 33060, the detected power-off time is compared with a preset period τ _2. If the detected power-off time is shorter than the period τ ₂ , the ultrasonic/high frequency ultrasonic power supply is shut off and an alarm signal is sent. In step 33070, the wafer is checked for cleanliness, and if the wafer is still not as clean as desired, steps 33010 through 33060 are repeated. Alternatively, the cleanliness check need not be performed every cycle. The number of cycles to be used may be determined in advance by experimentation using sample wafers.

FIG. 34 is a flow chart showing a wafer cleaning process according to another embodiment of the present invention. The wafer cleaning process starts with step 34010, which applies a cleaning liquid to a space between the wafer and an ultrasonic/high frequency ultrasonic device. In step 34020, an ultrasonic/high frequency ultrasonic power supply is set to a frequency _f1 and a power level P1 to drive the ultrasonic/high frequency ultrasonic device. In step 34030, the amplitude of the sonic output is detected and compared to a preset value. If the detected amplitude is greater than the preset value, the power supply is shut off and an alarm signal is sent. In step 34040, the sonic supply is set to an output of 0 before the bubble cavitation in the cleaning liquid damages the pattern structure on the wafer. In step 31050, the sonic power supply is returned to a frequency _f1 and a power level P1 after the temperature inside the bubbles has decreased to a certain level. In step 34060, the cleanliness of the wafer is checked, and if the wafer is not yet clean to the desired degree, steps 34010 to 34050 are repeated. Alternatively, the cleanliness check need not be performed every cycle, and the number of cycles to be used may be determined in advance by experimentation using sample wafers.

In some examples, multiple wafer cleaning processes depicted in various figures throughout this disclosure can be combined to produce a desired cleaning result. As one example, the amplitude detection of step 34030 of FIG. 34 can be incorporated into the wafer cleaning process shown in FIG. 33. As another example, the voltage attenuation 26090 and shaping circuit 26092 of FIG. 26 and the amplitude detection circuit 27092 of FIG. 27 can be applied to the wafer cleaning process shown in FIGS. 33 and 34.

The present invention provides an apparatus for cleaning a semiconductor substrate using an ultrasonic/high frequency ultrasonic device, comprising a chuck, an ultrasonic/high frequency ultrasonic device, at least one nozzle, an ultrasonic/high frequency ultrasonic power supply, a host computer, and a detection system. The chuck holds a semiconductor substrate. The ultrasonic/high frequency ultrasonic device is disposed adjacent to the semiconductor substrate. The at least one nozzle introduces a chemical liquid onto the semiconductor substrate and into a gap between the semiconductor substrate and the ultrasonic/high frequency ultrasonic device. The host computer sets the ultrasonic/high frequency ultrasonic power supply to a frequency f ₁ and a power P 1, drives the ultrasonic/high frequency ultrasonic device, sets the ultrasonic/high frequency ultrasonic power supply to zero output before bubble cavitation in the liquid destroys a pattern structure on the semiconductor substrate, and sets the ultrasonic/high frequency ultrasonic power supply to frequency f ₁ and power P 1 again after the temperature in the bubbles has cooled to a set temperature. The detection system detects the power-on time and the power-off time at the power P 1 and the frequency f ₁ separately, and compares the detected power-on time at the power P 1 and the frequency f ₁ with a preset period τ ₁ . If the detected power-on time is longer than the preset period τ ₁ , the detection system sends an alarm signal to the host computer, and the host computer receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power supply. The detection system also compares the detected power-off time with the preset period τ _2. If the detected power-off time is shorter than the preset period τ ₂ , the detection system sends an alarm signal to the host computer, and the host computer receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power supply.

In one embodiment, the ultrasonic/high frequency ultrasonic device is further coupled to a nozzle and positioned adjacent to the semiconductor substrate, and energy from the ultrasonic/high frequency ultrasonic device is transferred from the nozzle through a liquid column to the semiconductor substrate.

The present invention provides another apparatus for cleaning a semiconductor substrate using an ultrasonic/high frequency ultrasonic device, comprising a chuck, an ultrasonic/high frequency ultrasonic device, at least one nozzle, an ultrasonic/high frequency ultrasonic power supply, a host computer, and a detection system. The chuck holds a semiconductor substrate. The ultrasonic/high frequency ultrasonic device is disposed adjacent to the semiconductor substrate. The at least one nozzle introduces a chemical liquid onto the semiconductor substrate and into a gap between the semiconductor substrate and the ultrasonic/high frequency ultrasonic device. The host computer sets the ultrasonic/high frequency ultrasonic power supply to a frequency _f1 and a power P1, drives the ultrasonic/high frequency ultrasonic device, sets the ultrasonic/high frequency ultrasonic power supply to zero output before bubble cavitation in the liquid destroys a pattern structure on the semiconductor substrate, and sets the ultrasonic/high frequency ultrasonic power supply to frequency _f1 and power P1 again after the temperature in the bubbles has cooled to a set temperature. The detection system detects the amplitude of each waveform output from the ultrasonic/high frequency ultrasonic power supply, and compares the detected amplitude of each waveform with a specified value. If the detected amplitude of any waveform is greater than a preset value, the detection system outputs an alarm signal to the host computer, which receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power source, where the preset value is greater than the waveform amplitude during normal operation.

In one embodiment, the ultrasonic/high frequency ultrasonic device is further coupled to a nozzle and positioned adjacent to the semiconductor substrate, and the energy of the ultrasonic/high frequency ultrasonic device is transferred from the nozzle through a liquid column to the semiconductor substrate.

The present invention provides another apparatus for cleaning a semiconductor substrate using an ultrasonic/high frequency ultrasonic device, comprising a cassette, a tank, an ultrasonic/high frequency ultrasonic device, at least one inlet, an ultrasonic/high frequency ultrasonic power supply, a host computer, and a detection system. The cassette holds at least one semiconductor substrate. The tank holds the cassette. The ultrasonic/high frequency ultrasonic device is mounted on the outer wall of the tank. The at least one inlet fills a chemical liquid into the tank for immersing the semiconductor substrate. The host computer sets the ultrasonic/high frequency ultrasonic power supply to a frequency f ₁ and a power P 1, drives the ultrasonic/high frequency ultrasonic device, sets the ultrasonic/high frequency ultrasonic power supply to zero output before the bubble cavitation in the liquid destroys the pattern structure on the semiconductor substrate, and sets the ultrasonic/high frequency ultrasonic power supply to frequency f ₁ and power P 1 again after the temperature in the bubbles is cooled to a set temperature. The detection system detects the power-on time and the power-off time at the power P 1 and the frequency f ₁ separately, and compares the detected power-on time at the power P 1 and the frequency f ₁ with a preset period τ ₁ . If the detected power-on time is longer than the preset period τ ₁ , the detection system sends an alarm signal to the host computer, and the host computer receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power supply. The detection system also compares the detected power-off time with the preset period τ _2. If the detected power-off time is shorter than the preset period τ ₂ , the detection system sends an alarm signal to the host computer, and the host computer receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power supply.

The present invention provides another apparatus for cleaning a semiconductor substrate using an ultrasonic/high frequency ultrasonic device, comprising a cassette, a tank, an ultrasonic/high frequency ultrasonic device, at least one inlet, an ultrasonic/high frequency ultrasonic power supply, a host computer, and a detection system. The cassette holds at least one semiconductor substrate. The tank holds the cassette. The ultrasonic/high frequency ultrasonic device is mounted on the outer wall of the tank. The at least one inlet fills a chemical liquid into the tank for immersing the semiconductor substrate. The host computer sets the ultrasonic/high frequency ultrasonic power supply to a frequency _f1 and a power P1, drives the ultrasonic/high frequency ultrasonic device, sets the ultrasonic/high frequency ultrasonic power supply to zero output before the bubble cavitation in the liquid destroys the pattern structure on the semiconductor substrate, and sets the ultrasonic/high frequency ultrasonic power supply to a frequency _f1 and a power P1 again after the temperature in the bubbles is cooled to a set temperature. The detection system detects the amplitude of each waveform output from the ultrasonic/high frequency ultrasonic power supply, and compares the detected amplitude of each waveform with a specified value. If the detected amplitude of any waveform is greater than a preset value, the detection system outputs an alarm signal to the host computer, which receives the alarm signal and shuts down the ultrasonic/high frequency ultrasonic power supply, where the preset value is a value greater than the waveform amplitude during normal operation.

Although the present disclosure has been particularly shown and described with reference to illustrative embodiments thereof, those skilled in the art will understand that various changes in form and details may be made therein without departing from the spirit and scope of the claims.

Claims (53)

1. A system for controlling damage during cleaning of a semiconductor wafer having a plurality of pattern features, comprising:
a wafer holder for temporarily restraining the semiconductor wafer during the cleaning process;
an inlet for supplying a cleaning liquid to a surface of the semiconductor wafer;
an acoustic generator configured to apply sonic energy to the cleaning liquid, the sonic generator being configured to operate alternately between a first predetermined time period at a first frequency and a first power level and a second predetermined time period at a second frequency and a second power level, the first predetermined time period and the second predetermined time period being consecutively linked to each other, and the sonic energy causing an increase in bubble size in the cleaning liquid during the first predetermined time period and a decrease in bubble size during the second predetermined time period;
a controller programmed to provide the first and second frequencies, the first and second power levels, the first and second predetermined time periods, and a number of alternations of the first and second predetermined time periods by the acoustic wave generator;
a system in which power is set to the second frequency and the second power level for the second predetermined period after a ratio of total bubble volume to volume of via hole, trench or recess space in the semiconductor wafer increases to a first predetermined value, and power is set to the first frequency and the first power level for the first predetermined period after a ratio of total bubble volume to volume of via hole, trench or recess space in the semiconductor wafer decreases to a second predetermined value. The system of claim 1, wherein the wafer holder is a rotating chuck. The system of claim 1, wherein the wafer holder is a cassette submerged in a cleaning tank. The system of claim 1, wherein the inlet includes a nozzle. The system of claim 1, further comprising an acoustic transducer coupled to the acoustic generator. The system of claim 5, wherein the acoustic transducer is positioned across a gap on the semiconductor wafer, and the gap is filled with the cleaning fluid during a cleaning process. The system of claim 6, wherein the gap changes during the cleaning process. The system of claim 5, wherein the sonic transducer is connected to the inlet and applies sonic energy to the cleaning fluid flowing through the inlet. The system of claim 1 , wherein the cleaning solution is selected from the group consisting of a chemical solution, deionized water, and a combination of the chemical solution and the deionized water . The system of claim 1, wherein the second power level is 0. The system of claim 1, wherein the first frequency is equal to the second frequency and both frequencies remain constant during their respective periods, while the first power level is higher than the second power level and both power levels remain constant during their respective periods. The system of claim 1, wherein the first frequency is higher than the second frequency and both frequencies remain constant during their respective periods, while the first power level is higher than the second power level and both power levels remain constant during their respective periods. The system of claim 1, wherein the first frequency is lower than the second frequency and both frequencies remain constant during their respective periods, while the first power level is equal to the second power level and both power levels remain constant during their respective periods. The system of claim 1, wherein the first frequency is lower than the second frequency and both frequencies remain constant during their respective periods, while the first power level is higher than the second power level and both power levels remain constant during their respective periods. The system of claim 1, wherein the first frequency is lower than the second frequency and both frequencies remain constant during their respective periods, while the first power level is lower than the second power level and both power levels remain constant during their respective periods. The system of claim 1, wherein the first power level is increased during the first predetermined period. The system of claim 1, wherein the first power level is decreased during the first predetermined period. The system of claim 1, wherein the first power level both increases and decreases during the first predetermined period. The system of claim 1, wherein the second frequency is substantially near zero and the second power level is a constant positive value during the second predetermined period. The system of claim 1, wherein the second frequency is substantially near zero and the second power level is a constant negative value during the second predetermined period. The system of any one of claims 5 to 8, wherein the sound waves from the acoustic transducer in the first predetermined period and the sound waves from the acoustic transducer in the second predetermined period are in antiphase. The system of claim 1, wherein the first predetermined period is less than 2,000 times the cycle period of the first frequency. 2. The system of claim 1, wherein the first predetermined period is shorter than ((T _i - T ₀ - ΔT)/(ΔT-δT)+1)/f ₁ , where T _{i is} an implosion temperature, T 0 is a temperature of the cleaning liquid, ΔT is a temperature rise caused by one compression, δT is a temperature fall caused by one expansion, and f ₁ is the first frequency. The system of claim 1, wherein the first frequency varies from a high value to a low value during the first predetermined period. The system of claim 1, wherein the first frequency varies from a low value to a high value during the first predetermined period. The system of claim 1, wherein the first frequency changes from a low value to a high value and then returns to the low value during the first predetermined period. The system of claim 1, wherein the first frequency changes from a high value to a low value and then returns to the high value during the first predetermined period. 2. The system of claim 1 _, wherein the first frequency is first _f1 , then _f3 , and finally _f4 during the first predetermined period, where _f4 is less than f3, and _f3 is less than _f1 . 2. The system of claim 1, wherein the first frequency is first _f4 , then f3, and finally _f1 during the first predetermined period, where _f4 is less than _f3 , and _f3 is _less than _f1 . 2. The system of claim 1, wherein the first frequency is first _f1 , then _f4 , and finally _f3 during the first predetermined period, where _f4 is less than _f3 , and _f3 is less than _f1 . 2. The system of claim 1, wherein the first frequency is first f3, then _f4 , and finally _f1 during the first predetermined period, where _f4 is less than _f3 , and _f3 is _less than _f1 . 2. The system of claim 1 _{, wherein the first frequency is first f3, then f1, and finally f4 during the} first predetermined period, where _f4 is less than _f3 , and _f3 is less than _f1 . 2. The system of claim 1 _{, wherein the first frequency is first f4, then f1, and finally f3 during the} first predetermined period, where _f4 is less than _f3 , and _f3 is less than _f1 . The system of claim 1, further comprising a detection circuit coupled to the sound wave generator for detecting an output of the sound wave generator. The system of claim 34, wherein the detection circuit includes a voltage attenuation circuit for attenuating the input signal. 35. The system of claim 34, wherein the detection circuit includes a shaping circuit for converting a signal from a first waveform to a second waveform. The system of claim 36, wherein the first waveform is a sine wave and the second waveform is a square wave. The system of claim 34, wherein the detection circuit includes an amplitude detection circuit for detecting the amplitude of the input signal and comparing it to a reference value, and the detection circuit generates an alarm signal and turns off the sound generator when the detected amplitude exceeds the reference value. The system of claim 38, wherein the reference value is generated by a digital-to-analog converter (DAC). The system of claim 34, wherein the detection circuit detects the first predetermined period, compares it to a predetermined value, and generates an alarm signal and turns off the sound generator when the first predetermined period exceeds the predetermined value. The system of claim 34, wherein the detection circuit detects the second predetermined period, compares it to a predetermined value, and generates an alarm signal and turns off the sound generator when the second predetermined period is less than the predetermined value. The system of claim 1, wherein the first predetermined period is determined by experimentation to avoid implosion of gas bubbles in the cleaning liquid by inspecting the semiconductor wafer for damage to the patterned features. The system of claim 42, wherein the empirical determination includes selecting different values for the first predetermined period for different cleaning processes while maintaining the first and second frequencies and the first and second power levels at constant values and maintaining the second predetermined period at a constant value that is significantly longer than the first predetermined period. The system of claim 1, wherein the first predetermined period is determined by experimentation to allow limited bubble implosion without causing damage to the pattern features in the semiconductor wafer being cleaned. The system of claim 44, wherein the empirical determination includes selecting different values for the first predetermined period for different cleaning processes while maintaining the first and second frequencies and the first and second power levels constant and maintaining the second predetermined period at a constant value that is significantly longer than the first predetermined period. The system of claim 1, wherein the second predetermined period is determined by experimentation so that the temperature inside the bubbles in the cleaning liquid can be lowered to a predetermined temperature. The system of claim 46, wherein the predetermined temperature is substantially near room temperature. The system of claim 1, wherein the first frequency and the first power level are determined by experiment to avoid implosion of gas bubbles in the cleaning liquid by inspecting the semiconductor wafer for damage to the patterned features. The system of claim 1, wherein the first frequency and the first power level are determined by experimentation to allow limited bubble implosion without causing damage to the pattern features in the semiconductor wafer being cleaned. The system of claim 1, wherein the second frequency and the second power level are determined by experimentation so that the temperature inside the bubbles in the cleaning liquid can be reduced to a predetermined temperature. The system of claim 50, wherein the predetermined temperature is substantially near room temperature. The system of claim 1, wherein the number of rotations is determined empirically by inspecting the semiconductor wafer for damage. The system of claim 1, wherein the yield improvement caused by the cleaning effect of applying the sonic energy is greater than the yield loss caused by damage caused by applying the sonic energy.

Images