1.
Introduction
CMP is the most widely used technique of planarization[1–4]. To manufacture semiconductor devices of high quality, it is important to guarantee the cleanliness of the polished surface[5, 6]. In the CMP process, the surface energy of the newly produced copper surface is relatively high. The abrasive silica particles in the polishing slurry are easily adsorbed on the surface[7]. The particles form “nail” holes in lithography and ion implantation, which causes the short circuit and open circuit of devices. This seriously affects the device performance and even leads to device failure[8, 9]. The effect of abrasive particle removal is very important for integrated circuit manufacturing[10, 11].
Jim et al.[12] proposed brush scrubber cleaning as the most effective method in CMP cleaner, which showed high cleaning performance by physical force. Venkatesh et al.[13]reported a cleaning solution consisting of TMAH as the cleaning agent and arginine as the complexing agent. It was reported that the solution had the ability to remove abrasive particles from copper surface by changing the charges of particles[14]. It is easy for TMAH to decompose and evaporate into the environment. In addition, TMAH will cause several health problems by breathing or sticking to the skin. Our previous work[15] has developed a type of alkaline FA/O cleaning solution consisting of FA/O II chelating agent and FA/O surfactant. FA/O II chelating agent has more than thirteen chelating rings with high stability and strong chelating ability. FA/O surfactant is a large molecular non-ionic surfactant with low surface tension. It has a significant effect on particle removal and protection against corrosion of a copper surface. Compared with TMAH, the FA/O cleaning solution has a better effect on contamination removal, and the corrosion rate for the wafer surface is lower. FA/O cleaning solution is nontoxic. However, the effect of cleaner parameters has not been comprehensively studied. In this work, all of the experiments were performed utilizing FA/O cleaning solution. The corresponding mechanism was revealed. The effects of brush rotational speed, cleaning time, and brush gap for abrasive particles were investigated.
2.
Experiment
Copper coupons (with diameter 10.16 mm) cut from 300 mm blanket copper wafers were used for all the experiments. The polishing process was carried out utilizing a CMP polisher (Alpsitec, E460). The polishing pad is a POLITEXTIM REG PAD made from Rohm & Haas Co. The prepared slurry was composed of 20 wt% colloidal silica abrasive, 0.15 wt% FA/O II chelating agent, 1.2 wt% BRJI 30 nonionic surfactant, and 1 mL/L H 2O2 developed by Hebei University of Technology. The working pressure was 2 psi, the head speed/the platen speed was 55/60 rpm, the slurry flow rate was 150 ml/min, and the polishing time was 10 s. Subsequently, CMP cleaner (G&P, 412S) was used to deal with the copper coupons. The cleaning solution consisted of FA/O II chelating agent (Tianjin Jingling, China) and FA/O I surfactant. The cleaning machine has two separate spray bars above the brushes and either the cleaning solution or DIW flows from the spray bar and flows through the brush core. The cleaning solution was sprayed at the speed of 1.13 L/min for 60 s. Then the test wafers are brushed with DIW and blow-dried by nitrogen. The schematic of the brush cleaning module is shown inFig. 1. The wafer is located horizontally in the brush cleaning box. The spray bars are above the brush. Either the cleaning solution or DIW flows from the spray bars and flows through the brush core. Rotation speed, cleaning time, and brush gap were set as variables for these experiments. The brush gap was defined as the distance between the wafer surface and the brush nodule. The smaller the gap is, the higher the pressure to the wafer. Atom force microscopy (AFM, Agilent 5600LS) was used to determine the cleaning performance. Scanning electron microscopy (SEM, Zeiss, Sigma 500) was adopted to identify the type of particles and verify the cleaning results.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-1_mini.jpg)
class="figure_img" id="Figure1"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure1.
(Color?online) Schematic of brush cleaner module.
3.
Mechanism for particle removal
The Van der Waals force between the wafer and abrasives makes these abrasive particles adsorbed on the wafers rapidly. Due to H2O2 in the barrier slurry, copper tends to be oxidized to form cupric/cupreous oxides (CuO or Cu2O) and hydroxides passivation on the wafer surface. The FA/O II chelating agent is a potent alkaline chelating agent with 13 chelating rings, which can be shortened as R(NH3)4. The FA/O II chelating agent has an effect on reacting with Cu2+/Cu+ generated from Cu–SiO2 and cleaving the Cu–SiO2 bonding, as shown in Fig. 2[16].
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-2_mini.jpg)
class="figure_img" id="Figure2"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure2.
Schematic illustration of the removal of abrasive silica particles.
$!!!{ m{C}}{{ m{u}}_2}{ m{O}} + {{ m{O}}_2} to { m{CuO}} + {{ m{H}}_2}{ m{O}}mathop to limits^{{ m{O}}{{ m{H}}^ - }} { m{Cu}}{left( {{ m{OH}}} ight)_2} Leftrightarrow { m{C}}{{ m{u}}^{2 + }} + 2{ m{O}}{{ m{H}}^ - },$ ![]() | (1) |
${ m{C}}{{ m{u}}^{2 + }} + { m{R}}{left( {{ m{N}}{{ m{H}}_3}} ight)_4} to [{ m{CuR}}{left( {{ m{N}}{{ m{H}}_3}{)_4}} ight]^{2 + }},$ ![]() | (2) |
$2{ m{C}}{{ m{u}}^ + } + { m R}{left( {{ m{N}}{{ m{H}}_3}} ight)_4} to { m{C}}{{ m{u}}_2}{ m{R}}{left( {{ m{N}}{{ m{H}}_3}} ight)_4}{]^{2 + }},$ ![]() | (3) |
${ m{Si}}{{ m{O}}_2} + 2{ m{O}}{{ m{H}}^ - } to { m{SiO}}_3^{2 - } + {{ m{H}}_2}{ m{O}}{ m{.}}$ ![]() | (4) |
The product is soluble which can be taken away easily from the copper surface. Due to the acid-base neutralization, SiO2 can react with OH– generated from R(NH3)4, as shown in reaction 4. The silica particles can be removed. The FA/O surfactant molecule can penetrate between the particles and the surface of the copper wafer and spread out on the surface of the wafer[17]. The formation of the molecular adsorption layer of the surfactant reduces the contact area between the particle and wafer surface, resulting in the desorption of silica particles[18].
4.
Results and discussion
Fig. 3 shows the AFM image of polished copper surface. Surface roughness reflects the fluctuation of surface morphology. The size of colloidal silica particles is about 90 nm. If dispersed on the surface, it will cause obvious changes in morphology. Because the residual silica particles can deteriorate the roughness, the less silica particles that remain on the wafer surface, the better the surface roughness. Further, it can be seen that lots of residual particles existed on the wafer surface before cleaning. The SEM image and EDX analysis of copper wafer verified that most were abrasive silica particles from polishing slurry, as is shown in Fig. 4. To remove the particles effectively, a series of experiments were conducted to research the parameters of CMP cleaner, as is shown in Table 1.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-4_mini.jpg)
class="figure_img" id="Figure4"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure4.
SEM picture and EDX analysis of copper wafer after polishing.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-3_mini.jpg)
class="figure_img" id="Figure3"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure3.
(Color?online) The surface morphology after polishing.
Group | Brush rotation (rpm) | Brush gap (mm) | DIW flow time (s) |
1 | 0 | ?0.75 | 150 |
2 | 100, 150, 200, 250 | ?0.75 | 150 |
3 | 200 | ?0.25, ?0.5, ?0.75, ?1.0, ?1.25 | 150 |
4 | 200 | ?0.75 | 30, 60, 90, 120, 150 |
5 | 200 | ?0.75 | 120 |
Table1.
Experimental conditions.
Table options
-->
![](http://www.jos.ac.cn:80//style/web/images/down-csv.png)
Download as CSV
Group | Brush rotation (rpm) | Brush gap (mm) | DIW flow time (s) |
1 | 0 | ?0.75 | 150 |
2 | 100, 150, 200, 250 | ?0.75 | 150 |
3 | 200 | ?0.25, ?0.5, ?0.75, ?1.0, ?1.25 | 150 |
4 | 200 | ?0.75 | 30, 60, 90, 120, 150 |
5 | 200 | ?0.75 | 120 |
4.1
The effect of brush rotation speed
The AFM image of wafer surface treated with different rotational speeds (100, 150, 200, 250 rpm respectively) are presented in Fig. 5. It can be seen from Fig. 6 that at low brush rotation (100 rpm), there were still significant amounts of silica particles on the copper surface. As rotation speed increased, particle contaminations dropped quickly and it had the minimum surface roughness value under the condition of 200 rpm. The reason may be that the low rotational speed could not take the particles away completely. When the rotating speed was increased, the removal for silica particles was working more efficiently. The surface roughness value decreased to 1.16 nm at over 200 rpm condition. For the contact time became longer as the brush rotating speed was increased to 250 rpm, partial particles were adsorbed on the copper surface again. The result also suggests longer brush contact is more vulnerable to particle contamination rather than higher brush contact frequency.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-6_mini.jpg)
class="figure_img" id="Figure6"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure6.
(Color?online) Wafer surface morphology with respect to brush rotation speed. (a) RPM = 100. (b) RPM = 150. (c) RPM = 200. (d) RPM = 250.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-5_mini.jpg)
class="figure_img" id="Figure5"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure5.
The wafer surface roughness value with respect to brush rotation speed.
4.2
The effect of brush gap
It can be seen from Fig. 7 that when the brush gap is smaller than 0.75 mm, surface roughness decreases as brush gap increases. As mentioned below, the smaller the brush gap indicates the higher contact pressure. The friction between wafer and brush is also increased, which increases the drag force, making it easier for the abrasive silica particles to be removed from the wafers. When the brush gap is over 0.75 mm, as shown in Fig. 8, the surface roughness value of copper is not improved. The reason may be that if high pressure is applied, particles will penetrate deeper into the wafer surface and become more difficult to remove.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-8_mini.jpg)
class="figure_img" id="Figure8"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure8.
(Color?online)?Wafer surface morphology with respect to brush gap. (a) Gap = ?0.25 mm. (b) Gap = ?0.5 mm. (c) Gap = ?0.75 mm. (d) Gap = ?1.0 mm. (e) Gap = ?1.25 mm.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-7_mini.jpg)
class="figure_img" id="Figure7"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure7.
The wafer surface roughness value with respect to brush gap.
4.3
The effect of DIW flow time
After the brushing process of scrubbing with the cleaning solution, DIW flows through the brush core. The hydrodynamics of the fluid is one of the key parameters to affect the particle removal. As is shown in Fig. 9, when the DIW flow time is short, there are still significant amounts of particle residuals on the wafer surface. During the DIW cleaning process, some of the removed particles may be absorbed on the brush. While the brush is being compressed to the wafer, those particles contaminate the wafer surface again. Thus, the particles will remain on the wafer surface if the DIW flow time is short. It was found that the longer time of DIW flow is favorable for the removal of particle residuals, as shown in Fig. 10. However, when the flow time is longer than 120 s, the improvement in the surface morphology is not obvious. It can be concluded that 120 s is enough for the particle removal.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-10_mini.jpg)
class="figure_img" id="Figure10"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure10.
(Color?online)?Wafer surface morphology with respect to DIW flow time. (a) Flow time = 30 s. (b) Flow time = 60 s. (c) Flow time = 90 s. (d) Flow time = 120 s. (e) Flow time = 150 s.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-9_mini.jpg)
class="figure_img" id="Figure9"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure9.
The wafer surface roughness value with respect to DIW flow time.
Fig. 11 shows the AFM image of the wafer surface after being cleaned under condition 5. It can be seen that the surface morphology is smooth and clean. The wafer surface has been greatly improved compared with Fig. 3. The SEM image and EDX analysis verified that the abrasive silica particles were removed efficiently, as is shown in Fig. 12. Many papers in the literatures have proposed particle removal after post CMP cleaning on the selection of cleaning solution. However, most of them have not paid attention to the process parameters of CMP cleaner. A traditional cleaning method cannot guarantee the same condition each time. This paper researches the optima process condition for the particle removal, which can be treated as the basis parameters for the subsequent study to ensure the accuracy of experiments.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-12_mini.jpg)
class="figure_img" id="Figure12"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure12.
The SEM picture and EDX analysis of copper wafer after cleaning.
![](http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/12/PIC/18050022-11_mini.jpg)
class="figure_img" id="Figure11"/>
Download
![](http://www.jos.ac.cn:80//style/web/images/down-img.png)
Larger image
![](http://www.jos.ac.cn:80//style/web/images/down-ppt.png)
PowerPoint slide
Figure11.
(Color?online) The surface morphology of copper wafer after cleaning.
5.
Conclusion
Most of abrasive silica particles remain on the copper surface after CMP, which seriously affects the performance and yield of semiconductor devices. This paper researched the parameters of CMP cleaner for particle removal. The experimental results show that when the brush rotational speed was 200 rpm, the brush gap was ?0.75 mm, and the DIW flow time was 120 s, the abrasive silica particles were removed efficiently. Under that condition, the AFM results showed that the surface roughness value was 1.08 nm. A better surface morphology was obtained. The SEM measurement further verified this conclusion. This study is of significance for the post CMP cleaning process.